|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 HYDROLOGIC CONTROL ON ECOSYSTEM METABOLISM: LOCAL PROCESSES AND LANDSCAPE DYNAMICS IN THE EVERGLADES RIDGE SLOUGH By DANIELLE LEYANNE WATTS A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2013
2 2013 Danielle Leyanne Watts
3 To Science!
4 ACKNOWLEDGMENTS No dissertation research is a singular enterprise, and I have had extraordinary assistance by my compatriots and committee members. This work involved more than a few battles with large insects, leaches, large, touchy equipmen t, strange hours, and tropical sunshine and I would like to express my appreciation to Mike Camp, rest of the pirates and deputy pirates who have spent a day or so in the fi eld with me. Equally important has been the intellectual contributions of Daniel McLaughlin and David Kaplan, with my thanks. As priceless as this dissertation is to me, more priceless yet are my memories of numerous adventures. To Everglades pirates every I have a wonderful committee who has helped me become not just a scientist, but also a professional: Matthew Cohen, who has done much to mold my mind into that of a scientist; Todd Osborne, to whom I att ribute any skills I may have as a field ecologist and leader; Peter Frederick, who has been both a mentor and friend; Ted Schuur, whose brilliance is always inspiring; and Wendell Cropper, whose rigorous intellect is matched by his compassion to students. My admiration for each of them has pushed me to always work harder, question more deeply, and do just one more analysis. Wilson, Vanessa Doyle, an d importantly, my husband, Not only did he provide field assistance at many a twelfth hour, he also patched my field injuries, solved more than a few problems, listened to me whine, celebrated my successes, and has been my best frien d since we first met.
5 Finally, I have been lucky in my funding. Many thanks are owed to the Alumni Graduate Fellowships Fund, The Everglades Foundation, The University of Florida Supplemental Retention Scholarship Program, and the Doris Lowe and Earl and Verna Lowe Scholarship Fund. Accomplishing the work presented in this dissertation would have been much more difficult without the financial support of each of these programs.
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ .......... 9 ABSTRACT ................................ ................................ ................................ ................... 11 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 13 Patterned Peatlands ................................ ................................ ............................... 13 Research Questions ................................ ................................ ............................... 15 Ridges and Sloughs as Alternative Carbon Accretion Equilibria ...................... 15 Recent Hydrologic Modification and the Ridge Slough Landscape .................. 18 Research Aim ................................ ................................ ................................ ... 20 Research Questions ................................ ................................ ......................... 20 Research Approach ................................ ................................ ................................ 21 Synergy Through Model and Measurement ................................ ..................... 21 Manuscript Outline ................................ ................................ ........................... 22 2 HYDROLOGIC CONTROLS ON CO 2 RESPIRATION IN A SUBTROPICAL PATTERNED PEATLAND ................................ ................................ ...................... 25 Methods ................................ ................................ ................................ .................. 28 Study Sites ................................ ................................ ................................ ....... 28 Data Collection ................................ ................................ ................................ 30 Data Analysis ................................ ................................ ................................ ... 32 Results ................................ ................................ ................................ .................... 32 Instantaneous Fluxes ................................ ................................ ....................... 32 Model Extrapolation ................................ ................................ .......................... 34 Discussion ................................ ................................ ................................ .............. 35 3 NET ECOSYSTEM PRODUCTIVITY IN THE EVERGLADES RIDGE SLOUGH ... 49 Methods ................................ ................................ ................................ .................. 52 Site Description ................................ ................................ ................................ 52 Environmental Data ................................ ................................ .......................... 54 CO 2 Flux Measu rements ................................ ................................ .................. 54 Data Analyses ................................ ................................ ................................ .. 56 Results ................................ ................................ ................................ .................... 58 Climate and Environment ................................ ................................ ................. 58 Instanta neous Fluxes ................................ ................................ ....................... 59
7 Models ................................ ................................ ................................ .............. 60 Annual Fluxes ................................ ................................ ................................ ... 61 Discussion ................................ ................................ ................................ .............. 62 Conclusions ................................ ................................ ................................ ............ 69 4 EVIDENCE FOR TRANS PIRATION INDUCED NUTRIENT ACCUMULATION ..... 81 Methods ................................ ................................ ................................ .................. 84 Data Collection ................................ ................................ ................................ 84 Data Analysis ................................ ................................ ................................ ... 86 Subsidy Calculations ................................ ................................ ........................ 90 Results ................................ ................................ ................................ .................... 92 Discussion ................................ ................................ ................................ .............. 95 5 LONG TERM IMPLICATIONS OF THE LOCAL CARBON BUDGET ................... 110 Model Overview ................................ ................................ ................................ .... 110 The Model ................................ ................................ ................................ ............. 117 Initial Conditions ................................ ................................ ............................. 117 Model Soluti ons ................................ ................................ .............................. 118 Model Results ................................ ................................ ................................ ....... 122 Model Parameters ................................ ................................ .......................... 122 Peat Development ................................ ................................ .......................... 123 Peat Maintenan ce ................................ ................................ .......................... 123 Discussion ................................ ................................ ................................ ............ 126 6 SYNTHESIS ................................ ................................ ................................ ......... 142 APPENDIX A MODEL PARAMETERIZATION ................................ ................................ ............ 147 Carbon Losses ................................ ................................ ................................ ...... 147 Methane ................................ ................................ ................................ ......... 147 Dissolved Organic Carbon ................................ ................................ .............. 147 Carbon Dioxide ................................ ................................ ............................... 148 GPP Parameters ................................ ................................ ...................... 148 Peat Parameters ................................ ................................ ...................... 148 Productivity ................................ ................................ ................................ ........... 149 GPP Parameters ................................ ................................ ............................ 149 Peat Parameters ................................ ................................ ............................ 150 B SENSITIVITY ANALYSIS OF CARBON PARAMETERS ................................ ...... 154 LIST OF REFERENCES ................................ ................................ ............................. 157 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 170
8 LIST OF TABLES Table page 2 1 General attributes for each landscape block. Metrics are from Watts and others (2010) ................................ ................................ ................................ ..... 42 2 2 Parameters for respiration models. ................................ ................................ ..... 43 2 3 Extrapolated annual R aq CO 2 C and hydrologic characteristics for the median elevation for the dominant communities in each landscape block over the three years of measureme nt ................................ ................................ ............... 44 3 1 Hydrologic characteristics for each sampling block from 7/20/2011 7/19/2012. ................................ ................................ ......................... 71 3 2 Parameters and model criteri a for the rectangular hyperbola function (in mg CO 2 C m 2 min 1 ) ................................ ................................ ........................... 75 3 3 Summed, modeled NEP and respiration for hydrologic year 2011 2012 (g CO 2 C m 2 yr 1 ) ................................ ................................ ............................... 78 3 4 Literature values for ridge (or sawgrass community) and slough annual net aboveground primary productivity for the Everglades. ................................ ........ 80 4 1 Locations of the ground water wells. ................................ ................................ 101 4 2 Community area and water subsidies (averaged across days) to ridges (averaged across the landscape). ................................ ................................ .... 107 4 3 Subsidy of water and phosphorus to ridges for 2002 2011 ............................. 109 5 1 Model symbols and definitions ................................ ................................ .......... 131 5 2 Parameters for the state transition functions ................................ .................... 132 A 1 Default model parameters ................................ ................................ ............... 152 A 2 Ridge and slough accretion rates described by Bernardt and Willard (2009), and the average accretion rate for all positive rates for the Peat parameterization ................................ ................................ .............................. 153
9 LIST OF FIGURES Figure page 1 1 Patterned peatlands are a matrix of communities with corresponding differing peat elevations ................................ ................................ ................................ ... 23 1 2 Hypothesized relationship between hydrologic conditions and carbon in the Everglades ridge slough landscape ................................ ................................ .... 24 2 1 Location of 2 x 4 km landscape blocks in Water Conservation Area 3A (WCA 3A) for paired ridge slough sampling. ................................ ................................ 45 2 2 Water depths at the median ridge peat elevation (from Watts and others 2010) for 2009 2011 ................................ ................................ ........................... 46 2 3 Soil temperature (water temperature when inundated) is a poor predictor of respiration whereas water depths are the better predictor of respiration ............ 47 2 4 11 year average CO 2 C flux for soil elevations from Watts and others (2010) extrapolated from the asymptotic mode l ................................ ............................. 48 3 1 Environmental conditions during the period of study ................................ .......... 70 3 2 At the end of the 2011 dry season, sloughs w ere nearly devoid of vegetation .. 72 3 3 Daily average ( s.d. for each day/location) of daytime net CO 2 exchange and total respiration. ................................ ................................ ........................... 73 3 4 The variation in daily measured fluxes as a function of PAR .............................. 74 3 5 Ridge NEP and fitted model ................................ ................................ ............... 76 3 6 Measured ecosystem respiration (daily average, R eco ) and modeled soil respiration (R aq ) with relation to local water depths. ................................ ........... 77 4 1 Locations of the wells in WCA 3AN (Drained) and 3AS (Conserved) ............... 100 4 2 Daily average water depths and rainfall ................................ ............................ 102 4 3 Schematic of evapotranspiration calculations ................................ ................... 103 4 4 Water table position for Drained ridge and slough (relative to the ridge well) and the difference between the water tables for time periods .......................... 104 4 5 Water table position for Conserved ridge and slough (relative to the ridge well) and the difference between the water tables for time periods .................. 105 4 6 Evaptoranspiration/PET relative to the water t able position in the ridge well .... 106
10 4 7 The subsidy of ET induced convergent flow of phosphorus for the 2011 dry season into ridge edges in relation to ridge size ................................ ............... 108 5 1 Parameterized model forms for carbon uptake and losses for ridge and slough states ................................ ................................ ................................ .... 133 5 2 Change to peat accretion rates due to GPP model parameter perturbations ... 134 5 3 Change to peat accretion rates due to Peat model paramter petrubations. ...... 135 5 4 Soil elevations after 1000 years commencing from a slough, undefferentiated landscape ................................ ................................ ................................ ......... 136 5 5 Soil elevations after 1000 years commencing from a bimodal landscape. ....... 137 5 6 Proportion of points that become ridges after 100 times steps commencing as an undiffe rentiated slough landscape when transitions are manipulated ..... 138 5 7 The bimodal landscape after 100 time steps with altered transiti on probabilities ( GPP parameterization) ................................ ................................ 139 5 8 Bimodal soil elevation configurations after 100 time steps of subjections to modern hydrologic conditions. ................................ ................................ .......... 140 5 9 The bimodal landscape after 1000 years under differing fire return intervals in the GPP model parame terization. ................................ ................................ ..... 141 A 1 Parameterization of respiration values ................................ ............................. 151 B 1 Sensitivity of landscape conditions to perturbations to the GPP model parameters relative to baseline conditions. ................................ ...................... 155 B 2 Sensitivity of landscape conditions to perturbations to the Peat model parameters relative to baseline conditions ................................ ....................... 156
11 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy HYDROLOGIC CONTROL ON ECOSYSTEM METABOLISM: LOCAL PROCESSES AND LANDSCAPE DYNAMICS IN THE EVERGLADES RIDGE SLOUGH By Danielle Leyanne Watts May 2013 Chair: Matthew Cohen Major: Interdisciplinary Ecology Tropical and sub tropical peatlands represent a g lobally important carbon store where persistent higher temperatures release carbon budgets from this control and render other environmental factors more influential to ecosystem carbon budgets. The ridge slough mosaic in the Everglades is a lotic subtropic al peatland that has undergone significant ecological change in the 20 th century response to both hydrologic drainage and impoundment. These changes are reflected in a degradation of the metrics associated with patch stability and self organized patterning primarily the vertical differentiation of peat elevations. The de coupling of the peat topography from the limestone suggests autogenic controls maintain ridge and slough differentiation, although it does not speak to the particular geometry of ridge slo ugh patterning. Here, I investigate local and regional variation in hydrologic controls on carbon dynamics, and hence peat accretion dynamics, in this landscape. I do so via investigations into soil and ecosystem carbon exchange using closed chamber method s across a hydrologic gradient in Water Conservation Area 3 I further investigate the role ridge vegetation plays in inducing evapotranspiration scale dependent feedbacks to
12 nutrient availability. The culmination is a model that explores what is known abo ut the carbon budget and evaluates the conditions for vertical differentiation between ridge and slough configurations Results indicate that 1) the landscape scale modifications to hydrology have altered carbon cycling in the region; 2) the central portio ns of the Everglades retain a potential for long term carbon accretion; 3) peat elevation differences and a highly connected water table create conditions for convergent flow of water, and hence a nutrient subsidy to ridges that is scaled to patch size ; and 4) despite a higher peat accretion potential in ridges, the long term maintenance of peat elevation differences is induced by interannual hydrologic variability
13 CHAPTER 1 INTRODUCTION Patterned Peatlands Peatlands are estimated to cover arou terrestrial carbon (Wigley and Schimel 2000). These peatlands are found in nearly every part of the world, from the extensive peatlands in the arctic and boreal regions, to tropical peatlands in Indonesia, Latin America, Africa, and the Caribbean. Peatlands are of increasing interest due to both their potential for long term carbon storage (Page and others 2002; Jaenicke and others 2008) and their role in our understanding of long term ec osystem equilibria (Clymo and others 1998; Rietkerk and others 2004 ). The value of peatlands derives from the burial of incompletely decomposed of plant material under water logged conditions leading to ca rbon accumulation last ing for thousands of years For example radio carbon dating of sub coastal peatlands in Borneo suggests they started accumulating peat aroun d 27,000 years ago ( Page and others 2004). The biological and hydrological interactions that l ead to peat accretion where productivity (P) that exceeds decomposition (or respiration, R) lead to myriad types of peatlands and peatland configurations which can have substantial influence on regional water patterns (e.g. Watters and Stanley 2006), b iodiversity, and large scale C fixation. Many peatlands are a matrix of multiple ecosystems, forming a regular spatial pattern in communities and peat characteristics; examples include the Great Vasyugan Bog in Siberia, Glacial Lake Agassiz Peatland Minne sota, and the Everglades, Florida (Figure 1 1 ). Each patch type is a stable ecosystem arrangement given local environmental conditions. Patterned peatlands occur where conditions allow for a single
14 regional peat accretion rate to be achieved under differen t ecosystem configurations and where each configuration is self inhibitory at a larger distance. Importantly, the regional peat accretion rate may be zero (i.e., the peatland complex is at neutral long term carbon equilibrium), or may be positive (i.e., th e peatland eleva tion is increasing; Clymo and others 1998; Brady 2002). Regardless of the accretion rate, the different ecosystem configurations generally consist of a higher production, high er respiration state and a lower production, lower respiration st ate that form distinct microtopography with each patch type occupying different elevation modes. The ridge slough patterned peatland in the Everglades provides an excellent model system for testing the hypothesis that the dominant ecosystems in a patterned peatland represent alternative configurations to achieve carbon equilibr ium with the regional landscape Further, the presence of a hydrologic gradient from modern hydrologic modification of large portions of the Everglades allows for tests of deviations from this equilibrium. The existence of multiple peat accretion equilibria has been pr oposed (SCT 2003; Larsen and others 2007) but never explicitly tested in this landscape. Successful restoration of the Everglades ridge slough will require a quantificat ion of the hydrologic conditions under which elevation differences and patterning between ridges and sloughs can be maintained. The recent (~100 years) drainage and impoundment of the ridge slough landscape provides a living laboratory for the control hydr ology has on carbon dynamics with some areas being much drier than historical
15 op portunity for understanding alternative state stability in a subtropical wetland and a unique opportunity to study the drivers behind state stability loss. Enumerating the rates of productivity and respiration as a function of local and regional hydrology is of intrinsic restoration value. The Comprehensive Everglades Restoration Plan (CERP) has several performance measures related to the ridge slough landscape, including the stability of the ridge slough veget ative patterning, inundation patterns, with differing inundations of ridges and sloughs explicitly considered, and sheetflow measures, including timing and distribution of flows and flow volume and continuity. All of these performance measures necessarily consider the maintenance of elevation differences between ridges and sloughs as a component of successful restoration. The overall objective of this dissertation is to investigate the carbon dynamics of the two dominant patch types in the ridge slough pat terned landscape along a gradient of hydrologic modification. To achieve this objective, I plan to test whether the two states in the best conserved Everglades are in equilibrium, and investigate the degree to which net autotrophy has been lost with hydrol ogic modification. I will do this via investigations into how respiration and primary productivity respond to hydrologic drivers, both in observation and by simulation modeling. The combined outcomes of these studies will be able to provide guidance about the hydrologic conditions necessary for the maintenance of elevation differences between ridges and sloughs. Research Questions Ridges and Sloughs as Alternative Carbon Accretion Equilibria There are multi ple lines of evidence that lead to the hypothesis that ridges and sloughs represent alternative carbon accretion equilibria. Ridges and sloughs appear to have existed as distinct communities for around 2,700 years (Bernhardt and Willard
16 2009). That the two communi ties have differing hyd roperiod requirements and water depths suggests that the communities have also had differing peat elevations for the same length of time. Further, historic accounts of ridge slough elevation differences suggest that ridges were 1 to 3 feet (30 to 91 cm) hi gher than sloughs (McVoy and others 2011) with an abrupt transition zone between patches. In order to maintain long term positive peat accretion rates and elevation differences, there are three hypothesized solutions to the carbon budget (Figure 1 2). Un der historic hydrologic conditions, locations with elevations different from an equilibrium level are driven towards one of those equilibria by changes in peat accretion rates. For example, shallow water conditions will generally increase productivity on r idges in response to increased oxygen availability. However, oxygen availability also increases respiration rates, canceling out or surpassing the increm ental increases in productivity. With lowered water levels, peat oxidation brings the soil elevation ba ck to equilibrium. Similarly, sloughs at lower elevation than the equilibrium slough elevation will tend to accumulate peat more rapidly due to reduced exposure probability. The third intermediate configuration is unstable, because small environmental pert urbations change the ratios of productivity to respiration ( P:R ) such that the system ends up driven to one of the stable configurations (i.e., ridge or slough). Note that while R is drawn in Figure 1 2 as a linear relationship with hydrologic conditions, the relationship need not be linear as long as it is monotonic. The hypothesis proposing ridges and sloughs as alternative solutions to landscape peat accretion equilibrium invokes a homeostatic feedback between soil accretion, hydroperiod and soil redox. Increased peat accretion shortens hydroperiod
17 and increases soil redox potential, in turn inhibiting peat accretion via accelerated respiration. Higher peat accretion potential in ridges is ultimately offset by increased peat exposure caused by being situa ted higher in the water column. As ridge elevation increases, hydroperiod decreases (approx. 310 days in the best conserved areas ; Watts and others 2010 ), exposing peat in ridges to longer and more frequent periods of oxidation. Sloughs in contrast are rar ely exposed (approx. 350+ day hydroperiod), at least where hydrologic conditions are thought to best approximate pre dra inage conditions (Givnish and others 2007; Watts and others 2010). The balance of these positive and negative feedbacks produces two alt ernative attractors (high productivity high respiration and low productivity low respiration; Figure 1 2) whose similar net peat accretion explains the extended stability (>1000 years; Bernhardt and Willard 2009) of landscape patch configuration. Nutrients reinforce and even induce patch differentiation in some patterned peatlands (e.g. Rietkerk and others 2004), and are inexorably linked to productivity differences between adjacent patches. The mechanism leading to higher nutrient availability in the highe r elevation patch is the convective transport of nutrients, induced by belowground hydraulic flow towards areas with greater transpiration. This mechanism results in a positive feedback between nutrients and plant biomass. A two patch condition arises beca use the convergent flow of nutrients leads to nutrient limitation at some critical scale. The convective transport of nutrients, specifically phosphorus, in the Everglades has been modeled as an autogenic process helping to explain patch differentiation fo r tree islands (Ross and others 2006). A similar process may occur between ridges and sloughs, as suggested by the observation of higher soil
18 nutrient concentrations coinciding with the highest elevations where soil elevation bimodality is observed (Cohen and others 2009). Demonstrating multiple peat accretion equilibria would provide mechanistic evidence for the existence of alternative stable states in the ridge slough region, and aid in understanding the hydrologic requirements for landscape maintenance. The hydrologic gradient present in WCA 3A allows for an excellent natural experiment for these hypotheses. That is, directionality at the landscape scale can be inferred from carbon budgets spanning the drained areas to the impounded areas. Further, such a study would give us the hydrologic regime within which both ridge and slough patches are stable at the point scale. This information would be vital to management and restoration strategies where the goal is to maintain the historic ridge slough patterni ng. It is evident that our understanding of the processes that create and maintain patterning in the central Everglades is incomplete. Regardless of the mechanisms consideration of the scale of their action and interaction is an important unknown. Recent H ydrologic Modification and the Ridge Slough Landscape The remarkable leveling of the peat surface under both drained and impounded conditions (Watts and others 2010) in the Everglades is clear evidence that recent (100 year) anthropogenic alterations in th e regional hydrology have changed the underlying processes governing ridges and sloughs. The strong control hydrology has over peat accretion suggests that some of the leveling can be ascribed to altered carbon budgets. The draining of peatlands, where mor e of the peat experiences oxidation, can lead to a rate of subsidence that vegetative productivity cannot compensate for (examples include Galloway and others 1999; Gambolati and others 2006; Schipper and McLeod 2006). The loss of elevation differences bet ween ridges and sloughs may be further
19 exacerbated by observed changes in community structure in the Everglades; deep water sloughs are replaced by emergent prairie vegetation, which has higher productivity. Drained portions of the landscape are therefore experiencing the stress of excess respiration on ridges at the same times as increased productivity in the former sloughs. The loss of bimodality in peat elevations with hydrologic impoundment likely has more to do with changes in the carbon balance on ri dges than with changes in sloughs. Slough vegetation does not change markedly with impoundment, presumably because most of the species are adapted to permanently inundated conditions. In contrast, sawgrass is not well adapted to permanent deep inundation, and exhibits evidence of physiological stress and eventually patches dieback under deeper water conditions. Localized persistence of sawgrass in impounded areas is presumed to be due to observed upward extension of rhizome, a strategy that places the meris tem at a depth that may periodically become exposed, but limits the lateral expansion of each individual. (Figure 1 1D ), which can be linked to the loss of the productive sawgrass. This upward extension can not compensate for deeper and longer inundation, and bimodality in peat elevations is lost. We lack both the historic (pre 20 th century hydrologic modification) and modern carbon budgets necessary to evaluate the degree to which the landscape has degraded. The loss of landscape pattern has been inferred on many occasions by observing changes in community structure and vegetative patterns over time (examples include Wu and others 1997; Givnish and others 2007; Zweig and Kitchens 2008 ). State stability is inf erred by robust vegetative patterning in these studies. The limitation of this
20 approach, however, is the implicit over emphasizing of the role of productivity in this landscape; the role of respiration in regulating the ridge slough pattern is as yet poorl y understood (although see Debusk and Reddy 2003 for a laboratory study). An explicit consideration of the carbon budgets of ridges and sloughs along a gradient of hydrologic modification is the next step to understanding the underlying causes of pattern l oss in the Everglades. Research Aim In order to address the current gap in knowledge in the role of ecosystem carbon exchange in regulating microtopography in the Everglades, the following aim is formulated: The aim of the dissertation is to investigate w hether locally endogenous carbon budgets are sufficient to explain the divergence of elevations between ridges and sloughs. Research Questions In order the reach the aim of the dissertation, the research can be divided into investigations of the following research questions: 1. Does respiration demonstrate a monotonic and inverse relationship to water depths, as described in Figure 1 2? 2. Do observations of ecosystem productivity corroborate the predictions from the s curve relationship described in Figure 1 2? 3. Can productivity induced differences in evapotranspiration reinforce patch differentiation in the ridge slough? 4. What are the long term implications of considering patches from the point scale of Figure 1 2?
2 1 Research Approach Synergy Through Model and Measurement In this dissertation, a combination of empirical and modeling approaches is adopted. The empirical parts consist of field investigations of ecosystem scale carbon and water dynamics along a hydrologic gradient of regional drained thr ough impounded hydrologic conditions. I use the a priori hypotheses generated by Figure 1 2, which differs fundamentally from hypotheses generated by data interpretation (Belyea and Lancaster 2002). Strong inference in ecology is criticized as assuming mut ually exclusive competing hypotheses, where real ecosystems often exhibit properties of multiple competing hypotheses (i.e. non exclusionary mechanisms; e.g. Quinn and Dunham 1983; Roughgarden 1983; Scheffer 1999). I am therefore cognizant throughout this dissertation that pattern mechanisms in the Everglades may not be exclusionary and that competing hypotheses may act at different times, scales, and further, that investigations into what maintains the Everglades ridge slough may not be the same phenomena that gave rise to the Everglades ridge slough initially. The investigations in point scale phenomena comprise of field measurements of instantaneous carbon fluxes and small scale variability in diel water table in Water Conservation Area 3A. The spatial an d temporal scales of Everglades patch dynamics preclude the sorts of manipulative experiments that can identify direct cause effect relationships. Therefore, field measurements along a hydrologic gradient are chosen instead, with the aim of identifying whe ther variation in these variables is consistent with theoretical models. The sampling is along the hydrologic gradient described in Watts and others (2010), using 2x4km landscape blocks identified as Drained, Conserved 1
22 and 2, and Impounded. Each landscap e block has distinct characteristics with regard to hydrograph, vegetative prevalence, and microtopography. Modeling, both statistical and simulation, is used throughout this manuscript. The aims of all of the models are to investigate the implications of measured relationships to larger scales, primarily temporal. The integration of the simulations throughout provides synergy between the measured data and the larger scale of the research questions. Manuscript Outline In Chapters 2 and 3, I investigate th e ecosystem carbon dynamics using closed chambers equipped with an infra red gas analyzer. Chapters 2 and 3 therefore answer the first two research questions about the relationships of ecosystem carbon dynamics with water levels. In Chapter 4 I investigate the hypothesis that higher elevation community (here, ridges) can induce convergent flow of water and nutrients to that patch. Until now, nutrient accumulation induced by patch differences in evapotranspiration has only been demonstrated in northern peatl ands. I infer the potential of a subsidy of nutrients to ridges by investigating the connectivity of the water table as it recedes below the soil surface on ridges. In Chapter 5 I test the theoretical predictions of Figure 1 2 at longer time scales. The model asks whether point level processes can explain the microtopographic properties of the modern Everglades. The implication is if they cannot, then contagion properties induced by the spatial scale of patterning in the Everglades is necessary to re crea te microtopographic separation between ridges and sloughs. Finally, in Chapter 6 I synthesize the findings and discuss their implications for the future of the ridge slough Everglades
23 A B C D Figure 1 1 Patterned peatlands are a matrix of communities with corresponding differing peat elevations. Examples include A ) Lake Agas siz bogs in Minnesota (USA), B ) string bogs in the Yug anskiy Nature Reserve, Siberia and C ) the ridge slough tree island complex in the greater Everglades area, Florida (WCA 3AS shown). D) Hydrologic impoundment is associated with a teristic of ridges (photo courtesy of UF Small UAV progra m). Ridge vegetation is pale brown, slough is green ( N. odorata leaves) and black is open water (altitude 100 m). Figures A) B) and C) generated with SPOT imagery viewed with Google Earth (Google, Inc., Mountain view, CA, USA). 1km 50m 500m
24 Figure 1 2 Hypothesized relationship between hydrologic conditions and carbon in the Everglades ridge slough landscape. There is a range of water depths ideal for ridges and sloughs, where the ecosystem carbon balance will move towards th e equilibrium state over any time period. At some intermediate level of hydrologic conditions, the vulnerability of the state to even small changes in hydrologic conditions leads to a lack of long term stability; a point falling in this region will likely shift to either a ridge or slough state. The arrows between R and P represent three configurations to achieve a nominal landscape peat accretion rate. Images courtesy of D. Watts.
25 CHAPTER 2 HYDROLOGIC CONTROLS ON CO 2 RESPIRATION IN A SUBTROPICAL PATTERNED PEATLAND Peatlands are a globally important terrestrial carbon store ( Wigley and Schimel 2000) and are under increasing risk due to anthropoge nic change These anthropogenic changes may be in the form of increased te mperatures, artificially modified water tables, increased drought and fire risk, and altered timing and frequency of precipitation (1991; Oechel and others 1993; Maltby and Immirzi 1993; Hooijer and others 2006 ; Limpens and others 2008; Ise and others 2008 Gorham ). Although many studies have qualitatively considered the role water levels have on peatland carbon fluxes, only recently have water table dynamics been explicitly considered in modeling carbon fluxes ( Dimitrov and others 2010; Sulman and others 201 2 ) Relatively few studies have evaluated how soil carbon efflux changes with anthropogenically altered regional water tables ( Jauhiainen and others 2008). Peatland carbon dynamics in both northern and tropical peatlands have been shown to be sensitive to water table variation ( Alm and others 1999; Vasander and Jauhiainen 2001; Chimner 2004 ), although the magnitude of the effect has varied by ecosystem. While lower water table generally increases respiration rates, some experimental studies have found eithe r no change with altered water levels ( Updegraff and others 2001) or reduced carbon flux under saturated conditions compared to both flooded and non flooded conditions ( Debusk and Reddy 2003). A seasonal shift between temperature and water table control on respiration further complicates the water table response ( Bubier and others 2003). Moreover, the effect of water table likely depends on the characteristics of the peat such as degree of decomposition and fiber content ( Dimitrov and others 2010), potentia lly indicating community level differences
26 independent of exogenous controls. Subtropical and tropical peatlands are subjected to higher annual temperatures and thus have the potential for greater carbon effluxes ( Chimner and others 2004); p ersistent high temperatures may be associated with reduced sensitivity to variance in temperature and possibly greater sensitivity to altered water tables ( Jauhiainen and others 2008 ). The Everglades ridge slough landscape is one such subtropical peatland. Long term p ea t accretion rates in the Everglades prior to the modern era of severe hydrologic modification were relatively low ( 0. 1 to 0. 6 mm yr 1 accretion; Bernhardt and Willard 2009), lower than rates observed in other tropical peatlands ( Sorensen 1993; Page and oth ers 2004) but similar to those of temperate systems (0.2 to 1 mm yr 1 ; Aaby and Tauber 1975) and boreal and subarctic peatlands (0.2 to 0.8 mm yr 1 ; Gorham 1991). The peat micro topography is distinctly patterned in the well conserved portions of the landscape, hypothesized to be the result of feedback relationships between carbon dynamics and hydrology ( Larsen and others 2007; Watts and others 2010; Cohen and others 2011 ). The remarkable loss of the peat microtopography under both drained and impounde d conditions ( Watts and others 2010) in the Everglades is evidence that recent (100 year) anthropogenic alterations in the regional hydrology have changed the underlying processes governing landscape topographic variation One hypothesized mechanism govern ing the degradation of the landscape involves the loss of soil carbon from oxidation due to lowered water tables ( Larsen and others 2007; Watts and others 2010 ). In particular, Watts and others (20 10) hypothesize that micro topographic differences between the two predominant ecotypes (ridge, dominated
27 by Cladium jamaicense and a deeper water slough), is maintained in part by an inverse but monotonic relationship between ecosystem respiration and hydrology. The hypothesis of a monotonic relationship between ecosystem respiration and hydrology predicts a predominant control of hydrology over soil carbon mineralization. To test this hypothesis, my goals in this study were two fold. First, I sought to determine controls on peat respiration, explicitly consideri ng the dynamic water table, and contrasting the two dominant community types of the landscape. Second, I sought to understand these effects on respiration rates along a gradient of hydrologic modification from drained to impou nded, encompassing those areas where the patterning and ridge slough elevation diff erences are best conserved. N either soil nor ecosystem respiration are accurate descriptions of measured fluxes for the purposes of this study The presence of a biologically active water table means the carbon flux measured here contains aspects of both. In a wetland with primarily inundated conditions, the flux of interest for large scale carbon dynamics is the amount of CO 2 that ultimately m akes it out of the water column Thus respiration here is denoted R aq quantified as the CO 2 C flux out of the combined soil and water portions of the ecosystem. Although organic matter mineralization in wetland soils involves multiple electron donor sources, the contribution of methane to total respiration is ge nerally very low, as much as one to two orders of magnitude lower than CO 2 and with no apparent relationship to differences in water depths in the Everglades ( Debusk and Reddy 2003). The relatively low methane fluxes are a ttributed to methane oxidation; me thanotrophs can consume as much as 91% of maximum methane flux in these soils ( King and others 1990). I therefore
28 focused measurements on soil /water CO 2 fluxes in Water Conservation Area (WCA) 3A in the central Everglades along a hydrologic gradient. Meth ods Study Sites Most of the historical Everglades was a subtropical peatland overlaying limestone bedrock, with a maximum peat thickness of 2.7 to 3.3 m near Lake Okeechobee, thinning from north to south to an average ranging 0.3 to 1.5 m in the ridge slou gh mosaic of the central and southern Everglades. The Everglades has a relatively low topographic relief (ca. 3 cm per km throughout Water Conservation Area (WCA) 3A), and local increases in elevation as low as 0.1 m can result in a 45% reduction in water depth and a 20% reduction in hydroperiod in some areas ( David 1996). Precipitation is seasonal, with ca 70% of the rainfall between May and October and an average annual rainfall of 1.22 m (2002 to 2010 at site W11 from EDEN; http://sofia.usgs.gov/eden/). The rainfall is tracked by annual water levels, where the majority of outflow occurs in the rainy season. In order to capture the effects of regional hydrology on soil respiration, four 2x4 km landscape sampling blocks were located throughout the southern Everglades (WCA 3A, oriented along historical flow lines; Figure 2 1). These landscape blocks are a part of a larger effort to monitor restoration activities as a part of the Central Everg lades Restoration Project More information on these landscape bloc ks can be found in Watts and others (2010); for the sake of brevity, I will describe only their key features here. WCA 3A spans hydrologic conditions from drained in the north to impounded in the south; its center is widely viewed as the best conserved rid ge slough landscape, which I
29 assume implies hydrologic drivers most similar to historical conditions ( SCT 2003; Lars en and others 2007; Watts and others 2010). The Drained s ite located in WCA 3AN (Figure 2 1 ) has a reduced median annual water depth as com pared to the remaining sites in WCA 3AS (Table 2 1). This landscape block has lost nearly all vertical differentiati on between ridge and slough soils with replacement of the deep water slough community by emergent freshwater marsh vegetation. The Conserved 1 landscape block (Figure 2 1) has evidence of some recent changes in ridges and slough s including increased wet prairie prevalence, divergent land and water slope over the landscape block, and lower annual median water depths (Table 2 1). In contrast, the Conserved 2 block retains the strongest evidence of landscape patterning, with fidelity of communities to differing soil elevations (approx. 22 cm difference between community elevation modes). Wet prairies, likely transitional communities b etween deep water sloughs and sawgrass ridges, are absent from this block, suggesting hydrologic conditions remain favorable for deep water sloughs dominated by Utricularia spp and Nymphaea odorata The Impounded block has both the highest median water dep th (Table 2 1) and nearly permanent inundation in all communities. A levee to the east of this landscape block has resulted in a re routing of water from east to west that is then impounded against the east west road corridor of Tamiami Trail. Elevation di fferences between ridges and sloughs are no longer as pronounced as in the conserved areas, and sawgrass ridges show marked signs of patchy local dieback, presumably from inundation stress.
30 Data Collection Respiration of the combined soil and water column was measured with a Li 6400 portable gas exchange system fitted with a 6400 09 soil CO 2 flux chamber (LiCor, Inc., Lincoln, NE). Sites were generally 8 paired ridges and sloughs in each landscape block. M ore measurements were taken in Drained and Conserved 1 when water tables were below the surface in both ridges and sloughs in order to obtain higher power for this critical period. Respiration measurements (total n = 544) were taken approximately bi monthly between January 2009 and June 2011, with the timin g of measurements primarily motivated by observing the full range of water levels observed in the Everglades. Due to the presence of water at all sites for at least part of the year, a modular collar was used to extend the CO 2 flux chamber above the water column, using the water present to form a seal. Measurements in 2009 were located on a soil base inserted 15 cm into the soil (installed in November of 2008) with the modular collar fitted overtop. Measurements from 2010 through 2011 were undertaken with t he modular collar set atop the soil surface without insertion into the soil. M easurements of instantaneous flux rates were similar at the same water depth over all the years and a variance components analysis for 2009 respiration values showed little site effec t (<0.1% of the total variance). I therefore infer the methodological change did not significantly alter measured flux rates, and prevented the trampling impacts of repeated site visitation that were already evident after t he first year of measurement s. Further, randomly selected sites in each of the landscape blocks allowed us to capture some of the spatial variability in respiration fluxes as a result of micro topography Water column depth was noted for each sample period, and head volume space calc ulated accordingly. Between 3 and 10 independent measurements were made at each site,
31 with the averages across these measurement s reported. Chamber fluxes were sometimes suppressed because of high humidity or large changes in chamber temperature, or increa sed dramatically due to ebullition events. A linear function was used to extrapolate chamber CO 2 concentrations to rates, with an R 2 <0.98 used to remove data altered by unfavorable conditions. Simultaneous measurements of water column and soil temperature (at 10 cm below the water surface; separate soil and water temperature measurements commenced in 2010), and measurements of vegetative community species and cover within 1 m 2 were also performe d. The vegetative community was then used to distinguish between slough and ridge communities in the same manner described by Watts and others (2010). As calcareous periphyton is present throughout the ecosystems sampled, the potential for changes in pH du e to calcite dynamics to affect CO 2 efflux was a concern. Theoretically, high CO 2 in the chamber headspace could induce CO 2 to dissolve back into the water column. If this occurred, the pH of the water column could be sufficiently reduced to dissolve calci um carbonate that in turn would result in higher CO 2 respiration values. Increases of headspace CO 2 were limited to no more than 5 ppm above ambient CO 2 to reduce the likelihood of induced calcium dissolution To further enumerate any such effect, pH was m easured with an Accumet AP63 (Fisher Scientific, Pittsburg, PA) both within and adjacent (within 1 m) to the respiration chamber. T o ensure calcium dissolution was not induced via higher partial pressure of CO 2 the change in pH (ambient minus post measurem ent pH) was compared to water depth and soil respiration.
32 Data Analysis A series of non linear regression models were fitted to predict respiration as a function of pH, water depths, water column and soil temper ature and ecological community (ridge vs. s lough as a fixed effect) Lacking a mechanistic reason for a par ticular model structure, both a negative exponential and a monomolecular (which approaches the lower asymptote, here K) models were evaluated for water depths. I used a non linear least square s fitting method to test which environmental factors most strongly correlated with soil respiration in this system. Akaike Information Criterion (AIC) was used to discriminate between predictive models. The best fit model was then used to estimate annual flux rates for 2009 2011. Incorporating water elevations provided by the Everglades Depth Estimation Network (EDEN sites shown in Table 2 1; http://sofia.usgs.gov/eden/) and soil elevation s Watts and others (2010) allowed us to estimate daily water depths for 473 points among the four landscape blocks. These daily water depths were used to extrapolate to daily fluxes and then summed for an estimated annual carbon ef flux. T o demonstrate the change in long term soil respiration over the ra nge of extant soil elevations I used this technique to create an ave rage annual flux rate from 2000 to 2011 for the points in each landscape block (Figure 2 4). Results Instantaneous Fluxes Hydrologic conditions varied substantially over the c ourse of this study. In 2009 mean water levels fell only slightly below the 20 year average, 2010 was a chronically wet year, where the minimum water levels were in the 90 th percentile, and 2011 was the driest year in the recen t record (20 years) with a prolonged dry s eason (Figure 2 2). Mean annual air temperat ures in South Florida ranged 23 to 24.5 C from 2009 to 2011
33 (FAWN; Fort Lauderdale station, 2 m above ground surface; http://fawn.ifas.ufl.edu) with only a single day over that time period where temperatures drop ped to freezing (January 2011). Carbon efflux range d from 0.24 to 6.26 g CO 2 C m 2 d 1 for sloughs and 0.18 to 6.72 g CO 2 C m 2 d 1 for ridges. A notched box plot ( McGill and others 1978) of instantaneous respiration rates revealed no landscape block effect in instantaneous measured fluxes, nor did the residuals when accounting for water level and community effects. I infer that any block effects that may exist are small when compared to the temporal variation in instantaneous rates, and therefor e used models without a block effect. Soil temperature was a poor predictor of carbon efflux (R 2 = 0.01, Figure 2 3A ) with a negative fitted exponent. Carbon efflux was best predicted by water depth (R 2 = 0.51 p < 0.01, Table 2, Figure 2 3B ). Based on AIC carbon efflux was best predicted by a model with community specific parameterization. B oth the exponential decay model and the model with an asymptotic structure h ad very similar goodness of fit. I used the asymptotic model for extrapolations as this mo del incorporated fewer parameters Including soil/water temperature did not improve the explanatory power of the models (Table 2 2, with water temperature shown); the same result was obtained using air and soil temperature (not shown). Residuals of the sel ected model did not deviate significantly from normality. I explored whether temperature effects emerge below a water depth threshold, but no evidence for any threshold behavior was found For example, I continued to observe a much stronger relationship wi th water depth when water levels were less than 5 cm (R 2 = 0.44) than temperature (R 2 = 0.2), which remained true with water levels below 0 and 5 cm the soil surface. Notably, however,
34 the temperature exponent at shallow depths was positive, consistent wi th expectations. Soil and water temperature were weakly but positively related (r = 0.17) for the times they were measured separately, and each were more weakly related to water depth (r = 0.10 and 0.04, respectively), to which I infer soil temperatures ma y be buffered by water depths but are essentially independent from surface water temperatures. Water column pH increased on average during the measurements (ambient pH mean = 7.38, var. = 0.07; post measurement mean = 7.41, var. = 0.09; n = 157), but was n ot normally distributed. A t test performed on log transformed ambient and post measurement pH showed they were significantly different (p < 0.05) but positively correlated (r = 0.77). I was unable to establish a relationship between pH and soil CO 2 respiration rates untransformed data. The residual respiration rates (de trended for water depth and community effect) did have a weak but significant positive relationship with the ambient pH (linear model p < 0.01, R 2 = 0.084) and post measurement pH (l inear model p < 0.01, R 2 = 0.1), but had no significant relationship to the change in pH (linear model p > 0.1, R 2 = 0.0). This suggests that while pH may affect respiration rates, capping of the water column did not significantly alter CO 2 fluxes. Model E xtrapolation I estimated annual respiration for the three years of measurements (Table 2 3) modeling carbon effluxes as a function of water depth The largest estimated flux rate (825 g CO 2 C m 2 yr 1 ) was for Drained ridges in 2011 the record dry year Th is year was 36% higher than in 2009 ( 607 g CO 2 C m 2 yr 1 ), the year of average precipitation, and 79% higher than for the wet year of 2010 ( 460 g CO 2 C m 2 yr 1 ), illustrating the dramatic effects of varying hydrologic conditions (Table 2 3). The magnitude of annual differences between communities was small with annu al ridge respiration ranging 96
35 117% that of sloughs The mean annual hydroperiods across the 3 year measurement period were longer for sloughs than ridges, ca. 77 days longer in Cons erved 1, 60 days longer in Conserved 2, and 20 days longer in Impounded. The hydroperiod range was dramatic, with no difference in hydroperiods between communities for Impounded in 2010 to 136 days longer for sloughs than ridges in Conserved 2, 2011. Figu re 2 4 demonstrates the relationship between local soil elevation and extrapolated carbon efflux. The dashed lines indicate the long term averaged carbon efflux (2000 2011) for the median soil elevations for ridges and sloughs in each landscape block. A la rger difference in the averaged annual effluxes is observed where microtopography is also observed. The a veraged extrapolation from 2000 to 2011 showed difference in long term fluxes for median ridge and slough fluxes of 9 50 35 and 5 g CO 2 C m 2 yr 1 f rom Drained through Impounded. Discussion The estimated annual respiration rates I found in the Everglades peatland (Table 2 3) were substantially higher than northern bogs and mires (average 94 g CO 2 C m 2 yr 1 ; Raich and others 1992) but lower than tropical moist forests (average 1260 g CO 2 C m 2 yr 1 ; Raich and others 1992), secondary forests in South Kalimantan (1200 430 g CO 2 C m 2 yr 1 ; Inubushi and others 2003), and hummock and hollow swamps in Micronesia (with a 50/50 ratio of each, 1061 83 g CO 2 C m 2 yr 1 ; Jauhiainen and others 2005). Even the drained site di d not reach the annual flux rates of other tropical systems that experience prolonged periods without inundation such as Sarawak (2130 g CO 2 C m 2 yr 1 ; Melling and others 2005), and Indonesia (1993 to 2031 g CO 2 C m 2 yr 1 ; Jauhiainen and others 2008)
36 Water table variation appears to exert a stronger control on respiration than does temperature. The marked increase observed in flux rates as the water table dropped to and below the soil surface has been observed in other studies ( Bubier and others 1998; Alm and others 1999; Vasaner and Jauhiainen 2001; Chimner 2004 ). Despite this precedent, I expected to also observe some temperature effect, as observed by Inglett and others (2011) i n an incuba tion study of Everglades soils. The lack of a statistically discernible effect does not obviate the well established influence of temperature on respiration rates ( Raich and others 1992), but does underscore the growing recognition that tropical and subtropical peat systems respond more strongly to water table ( Chimner 200 4; Jauhianen and others 2005; Jauhianen and others 2008). The absence of a significant temperature effect may be attributed to low temporal variability in air temperature and th e thermal buffering capacity of standing water, as inferred by the low correlation between soil and water temperatures. Peatlands with relatively small water level fluctuations appear to be less controlled by hydrology (North Ca rolina: Bridgham and Richard son 1992; Micronesia and Hawaii: Chimner and others 2004; Sarawak, Malasia: Melling and others 2005) than those with larg er fluctuations (Borneo swamps: Jauhiainen and others 2005; drained and restored peatlands in Indonesia : Jauhiainen and others 2008; Ev erglades ridge slough: this study). This reinforces the emerging contention ( Sulman and others 2012 ) that the carbon cycle in peatlands should be studied and modeled as tightly coupled with hydrologic dynamics and, as such, mediated by both exogenous flows and also small scale local topographic variation. If the peat soils in the Everglades were persistently moist rather than waterlogged, a temperature response may become prominent, as indicated by the
37 stronger control of temperature observed when water le vels dropped below the soil surface. It is important to note, however, that low water table conditions in the Everglades generally occur during the coolest months. Were a low water table to occur during the warmest times of the year, the steep increase I o bserved in respiration may be larger. Under such conditions the vertical peat accumulation balance would be disrupted, and the vertical differentiation between ridges and sloughs would likely disappear as soil respiration rates eclipse productivity. Even a ssuming that the higher standing biomass stocks of sawgrass in the Everglades completely turns over annually to become source material for peat, extrapolated flux rates under drained hydrologic conditions suggest a rapid loss of peat with year round draina ge. For example, offsetting the hydrograph for 2011 in the driest site so that the water is at or below the soil surface year round leads to estimated respiration rates of approximately 2262 g CO 2 C m 2 yr 1 for ridges. This respiration would be far in excess of any published primary productivity values in the literature (e.g., 1436 427 g C m 2 yr 1 converted from a dry weight using an estimated leaf C of 48% ; Daoust and Childers 1998) It is in this context that water management and climate c hange (reduced summer rainfall, increased potential evapotranspiration) need to be considered. I observed high instantaneous flux rates that were comparable to other tropical peatlands when the water table dropped more than 50 cm below the soil surface. Th e maximum measured rate for Drained of 280 mg CO 2 C m 2 h 1 in 2011 was akin to ~132 to 166 mg CO 2 C m 2 h 1 in hummocks and 38 to 188 mg CO 2 C m 2 h 1 in hollows of a tropical peat swamp ( Jauhiainen and others 200 5) and 100 to 335 mg C m 2 h 1 forested peatland in Malaysia ( Melling and others 2005) The large water level drop and
38 resulting longer aeration of the typically saturated soils likely explains the large rise in soil carbon efflux in 2011. In contrast to the ca. 187 day hydroperiod in 2011 for this landscape block, t he long term average hydroperiod for ridges under hydrologically conserved conditions has been estimated to be ca 310 days and ca 292 days in the Drained area (2000 to 2008; Watts and others 2010, comparabl e to pre estimates of ~9 to 10 months; McVoy and others 2011) As described earlier, dramatic leveling of the peat surface has occurred in this part of the Everglades, with nearly a complete loss of vertical differentiation between communities ( Watts and oth ers 2010). There are several possible explanations for the small but (statistically) significant differences in carbon flux rates. In wetlands with complex microtopography, differing water tables can lead to offsetting respiration responses ( Dimitrov and o thers 2010) V egetative source material can also lead to offset respiration rates although the differences in community respiration rates resulted in annual fluxes differing by less than 20 g C for many year/hydrologic combinations (Table 2 3). This small difference is despite the higher recalcitrance of sawgrass vegetation compared to slough species ( Amador and Jones 1997; Debusk and Reddy 1998; Vaithiyanathan and Richardson 1998; Lewis 2005; Osborne and others 2007). Several factors may be at play result ing in such small community differences First, i t may be that soil material is made more of roots than leaf detritus and that root material betwee n communities is similar in decomposition rate Second, r idges also generally have higher concentrations of s oil phosphorus in those areas where soil elevation bimodality has been maintained ( Cohen and others 2009). Since phosphorus is the key limiting nutrient in the Everglades ( Noe and others 2001), C mineralization may be releasing organically bound phosphorus and
39 acting as a positive feedback, enhancing soil respiration ( Debusk and Reddy 2003) in ridges. Third there may be seasonal influences over soil respiration which this study did not cover. Community specific phenology may play a role during dry down not covered here, particularly in terms of the contribution of root respi ration. For example, the deeper water vegetation senesces during the dry season and may reduce root respiration relative to the ridge vegetation. There may also be lagged effects during dry downs caused by the length of time the soil was exposed, inducing a qualitative difference in fluxes I was unable to observe in my point measurements. The logistical challenge of partitioning soil and water column respiration resulted in measurements that were integrations of the two. The contribution from the water column to total respiration may be important in sloughs, giving the appearance of similar R aq C efflux between communities when there is standing water. T here is far greater abundance of su bmerged aquatic vegetation and algal communities in sloughs than in ridges. The ubiquitous nature of submerged vegetation and extensive algal and periphyton mats in sloughs means my measurements at least partially incorporated respiratory losses from these communities. Periphyton assemblages can dominate slough productivity in the Everglades ( McCormick and Stevenson 1998; Iwaniec and others 2006), although these assemblages have been measured as a net carbon sink ( Iwaniec and others 2006 ). Total aquatic met abolism indicates that the sloughs are a net carbon source across a variety of environmental and community factors ( Hagerthey and others 2010). However, much of the respiration from the water column appears driven by sediment carbon mineralization rather t han aquatic vegetation and periphyton respiration ( Belanger and others 1989; Hagerthey and others 2010).
40 Although CH 4 is an important green house gas emitted by wetlands, it is not a large part of carbon fluxes in the ridge slough Everglades. The Everglad es i s a weak source of atmospheric methane, with total system releases estimated below 0.5 Tg CH 4 yr 1 (Burke and others 1988) Relatively low rates of CH 4 production ( Bachoon and Jones 1992), combined with little to no relationship with temperature ( Harriss and others 1988; Bachoon and Jones 1992) and water depth ( Debusk and Reddy 1992), along with high rates of methane oxidation ( King and others 1990) suggests that this pathway is not significant for carbon mineralization. Although based on laborator y studies of Everglades peat from central WCA 2A potential methane fluxes appears to be between 0 and 8% of total carbon efflux, with an average of 2% across a range of water depths ( 15 to 10 cm ; Debusk and Reddy 1992). Further, CH 4 release appears to de crease with deeper water levels (Debusk and Reddy 1992) although it is possible that the release of gaseous carbon becomes dominated by ebullition events with deeper water (Burke and others 1988) Linking hydrology to carbon budgets in the Everglades rem ains an important step in understanding landscape degradation. Both the Drained and Conserved 1 landscape blocks have lost significant portions of their peat profile (peat depths average 56.9 and 57.8 cm, respectively) compared to Conserved 2 and Impounded (118.8 and 113.8 cm, respectively; Watts unpublished data) suggesting much of the soil legacy carbon has been mineralized in northern areas. The draining of peatlands can lead to rates of oxidation and subsidence that vegetative productivity cannot compe nsate for ( Wsten and others 1997; Galloway and others 1999; Furukawa and others 2005; Gambolati and others 2006; Schipper and McLeod 2006 ). Indeed, oxidized peat throughout the
41 Everglades is estimated to have released 12,500 kg CO 2 per ha of drained soil, with total C outflow on order of 10 million tons ( Smil 1985). The higher peat bulk density in this region suggests some amount of subsidence is occurring along side oxidation ( Bruland and others 2006 ). Certainly the control hydrolog y has over peat accretion suggests the leveling of the peat surface can be ascribed to altered carbon budgets. This study supports the hypothesis of monotonic changes in respiration with water depth, as outlined by Watts and others (2010). This hypothesis suggests that the long term stability of landscape rests in nearly equivalent peat accretion rates between ridges and sloughs a condition only possible if discontinuous productivity is balanced by an inverse but monotonic relationship between respiration and hydrologic conditions (presumable a balance of water levels and hydroperiod). Community differences in annual flux rates induced by differences in soil elevations (and hence, hydrologic conditions) between ridges and sloughs evaporates with hydrologic modification (Figure 2 4), with over all respiration rates higher with drainage and lower with hydrologic impoundment. While the magnitude of these changes is small when compared to instantaneous variation in CO 2 ef flux, the summed values suggest large di fferences in peat accumulation and/or deflation rates. This study also aids in predicting the amount of net primary productivity necessary for peat accretion in the region. Enumerating the rates of carbon uptake and release in relation to hydrologic modifi cation is clearly important for establishing the hydrologic conditions necessary for landscape peat accretion.
42 Table 2 1. General attributes for each landscape block Metrics are from Watts and others (2010). Coordinates are of the center of the la ndscape block. % Slough % WP % Ridge N umber of points Median soil elevation (cm) Median water depth (cm) Latitude Longitude EDEN water level gage Drained -52.3 47.7 130 280.8 12.8 26.21788 80.7380 Site 62 Conserved 1 28.9 17.0 54.1 135 249.9 18.1 26.08815 80.7355 Eden5 Conserved 2 47.4 -52.6 95 227.1 32.0 25.98338 80.6970 W11 Impounded 69.9 2.7 27.4 113 190.4 54.2 25.82245 80.7283 Site 65 Everglades Depth Estimation Network (EDEN) water level gages are from http://sofia.usgs.gov/eden.
43 Table 2 2. Parameters for respiration mo dels Model Community § K § a § b § c R 2 AIC NA 2.17 *** 0.017 ** 0.011 NA 0.876 *** 0.025 *** 0.77 *** 0.494 1239 1 Ridge 0.791 *** 0.029 0. 513 1220 Slough 0.826 *** 0.022 Ridge 0.744 *** 0.031 *** 0.83 *** 0. 516 1221 Slough 1.23 *** 0.012 *** 0.51 Ridge 1.23 ** 0.034 *** 0.029 0.89 *** 0. 519 1222 Slough 0.936 0.017 *** 0.011 0.51 § Significance codes: ***p<0.001, **p<0.01, *p<0.05 1 The selected model for subsequent ext rapolation R 2 shown is Cflux observed ~ f (Cflux extrapolated ).
44 Table 2 3 Extrapolated annual R aq CO 2 C and hydrologic characteristics for the median elevation for the dominant communities in each landscape block over the three years of measurement Annual e stimates were generated using soi l elevations from Watts and others (2010) and water elevations in Figure 2 2. 2009 2010 2011 Community Median Water Depth (cm) Hydro Period (d) gCO 2 C Median Water Depth (cm) Hydro Period (d) gCO 2 C Median Water Depth (cm) Hydro Period (d) gCO 2 C Drained R WP 10.2 14.5 248 266 607 581 20.0 24.3 347 362 460 487 1.0 3.3 187 207 825 705 Cons.1 R S /WP 13.9 29.4 277 357 513 470 19.4 34.9 344 363 467 445 3.8 19.3 206 337 567 500 Cons. 2 R S 30.2 49.3 315 360 452 421 24.6 43.8 365 365 427 410 6.3 25.5 209 345 553 473 Impounded R S 47.6 61.8 354 364 386 396 43.3 57.5 365 365 370 385 24.8 39.0 280 328 482 454 R = Ridge; S = Slough; WP = Wet p rairie
45 Figure 2 1. Location of 2 x 4 km landscape blocks in Water Conservation Area 3A (WCA 3A) for paired ridge slough sampling.
46 Figure 2 2. Water depths at the median ridge peat elevation (from Watts and others 2010) for 2009 2011. Mean slough water depths would be 6, 19, 22, and 10 cm for each of the respective landscape blocks (Drained to Impounded).
47 Figure 2 3. Soil temperature ( water temperature when inundated ) is a poor predictor of respiration wherea s water depths are the better predictor of respiration (b est fit model s show n). Solid lines are exponential models; dashed red lines are the asymptote models Parameter values are in Table 2 2.
48 A B C D Figure 2 4 11 year average CO 2 C flux for so il elevations from Watts and others ( 2010) extrapolated from the asymptotic model Panels correspond to A) Drained, B) Conserved 1, C) Conserved 2, and D) Impounded sites Dashed lines indicate the median elevations for a priori defined ridges and slough/wet prairies. Solid lines are the soil elevation probability density function, offset to the lowest landscape position. 0 0.02 0.04 0.06 0.08 0.1 0 100 200 300 400 500 600 700 0 10 20 30 gCO 2 C m 2 yr 1 0 0.01 0.02 0.03 0.04 0.05 0.06 0 100 200 300 400 500 600 700 0 10 20 30 Soil Elevation Propoportion 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0 100 200 300 400 500 600 700 0 10 20 30 40 50 gCO 2 C m 2 yr 1 Offset Soil Elevation (cm) 0.00 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0 100 200 300 400 500 600 700 0 20 40 60 Soil Elevation Proportion Offset Soil Elevation (cm) Ridge Slough
49 CHAPTER 3 NET ECOSYSTEM PRODUCTIVITY IN THE EVERGLADES RIDGE SLOUGH The carbon storage of peatlands is controlled by the presence of water and its impact on the balance between carbon uptake and loss Research on boreal to tropical peatlands has increasingly focused on incorporating water table dynamics into ecosystem carbon budgets (examples include Dimitrov and others 2010; Sulman and others 2012; Jauhiainen and others 2008). I previously demonstrate that water column and soil respirat ion fluxes in the Everglades a subtropical peatland, are predominantly controlled by water levels. In this work, I seek to expand on that finding to better understand the role of h ydrologic variation in controlling totally ecosystem respiration and primary production. The ridge slough Everglades is a component of the largest restoration project yet initiated globally but restoration efforts are constrained by insufficient informatio n on the feedbacks between hydrology and peat accretion The peat portions of the lands cape are patterned in both patch phenology and microtopography (e.g., Wu and others 2006; Watts and others 2010) This patterning must be driven at least in part by loca l carbon budgets as microtopographic variation in the peat surface is not reflected in the underlying limestone topography (Lewis 2005; Givnish and others 2007). Much of the original ridge slough landscape has been dramatically altered by 20th century cha nges to nutrient loading and hydrology, with changes to vegetative and periphyton communities, degradation of landscape patterning, an d a loss of microtopography. Large expanses of W ater C onservation A rea 3A N orth (WCA 3AN) and Everglades National Park dem onstrate such change in pattern metrics that the prognosis for rapid restoration remains in doubt (Watts and others 2010).
50 My overarching hypothesis is that ridges and sloughs represent two alternative ecosystem configurations that achieve the same long te rm carbon accretion rate. The principal evidence in support of this hypothesis is that peat surface differentiation between adjacent ridges and sloughs is mai ntained without continual divergence (SCT 2003; Ross and others 2006; Larsen and others 2007; Cohe n and others 2011). This hypothesis invokes a homeostatic feedback between soil accretion, hydroperiod and soil redox. Lowered water levels are associated with higher production and increased productivity in ridges that is ultimately inhibited by shortened hydroperiod and increased soil redox potential Any h igher peat accretion potential in ridges is ultimately offset by increased peat exposure. In contrast, sloughs sit uated lower in the water column have lowered peat accretion potential (via lowered produ ctivity) but are balanced by lowered respiratory potential under anoxic conditions. The balance of these positive and negative feedbacks produces two alternative attractors (high productivity high respiration and low produ ctivity low respiration ) whose similar net peat accretion could explain the extended stability (>1000 years; Bernhardt and Willard 2009) of landscape patch configuration. The remarkable leveling of the peat surface under both drained and impounded conditions (Watts and others 20 10) in the Everglades is clear evidence that modern changes to the regional hydrology have changed the underlying processes governing ridges and sloughs. The control hydrology has over peat accretion suggests much of the leveling can be ascribed to altered carbon budgets. Drainage where more of the peat experiences oxidation, can lead to a rate of subsidence that vegetative productivity cannot compensate for (examples include Galloway and others 1999; Gambolati and
51 others 2006; Schipper and McLeod 2006). T he loss of elevation differences between ridges and sloughs is further exacerbated by changes in community structure; deep water sloughs are replaced by more productive emergent prairie vegetation. Drained portions of the landscape are therefore experienci ng the stress of excess respira tion on ridges coinciding with increased productivity in the former sloughs. The hydrologic gradient in the central Everglades of drained to impounded hydrologic conditions provides an opportunity to quanitify carbon responses to changes in regional drivers As both drainage and impoundment are associated with changes in community structure and soil microtopography, presumably the ecosystem carbon balances have also been alte red such that equivalent carbon budgets are no longer plausible. In this study I measured net ecosystem productivity (NEP) and ecosystem respiration (R eco ) using large, closed system chambers for a year in ridges and sloughs across a range of hydrologic conditions in WCA 3A. I use the net CO 2 uptake as a proxy for net ecosystem productivity (NEP), but acknowledge there are alternative losses of carbon that are not explored in this study, including non biological oxidation (particularly photolysis) and carbon losses through DOC (dissolved organic carb on) transport. The purpose was to determine a) the role water levels plays in mediating C O 2 fluxes over the course of a year, b) compare NEP between ridge and slough communities to test for carbon budgets equivalency, and c) evaluate a) and b) across a lan dscape hydrologic gradient from drained to impounded conditions. This study adds to our understanding of CO 2 dynamics in subtropical wetlands as well as aids in evaluating the potential for long term peat accretion in the Everglades.
52 Methods Site Descripti on The South Florida ridge and slough landscape is a major component of the once contiguous Everglades system that started in the Kissimmee chain of lakes and Kissim m ee River and flowed to Florida Bay It is a predominantly surface flowing fen that consists of two dominant patch types elevate d ridges and deep water sloughs organized in strikingly non random pattern. Ridges comprise approximately 50% of the landscape and are distinctly elo ngated in the direction of flow. S loughs form the remaining la ndscape matrix are, in the best conserved areas, longitudinally connected to create flowpaths for water, solutes, and organisms through the landscape. Peat accumulat ion began roughly 3,000 years before present over a limestone basin. Peat accretion appears to have been discontinuous, linked to major climatic events in history and tracking sea level rise over that period (Bernhardt and Willard 2009). Peat thicknesses historically ranged from 2.7 to 3.3 m at the northern end near Lake Okeechobee, thinning to 0.3 to 1.5 m in the central and southern Everglades. Plant communities are dominated by Cladium jamaicense on ridges and Nymphaea odorata and Utricuclaria spp in sloughs. R idges are generally monotypic, with occasional small amounts of Justicia angustifo lia and Cephalanthus occidentalis S lough communities are considerably more variable, with assemblages specific to local water levels ( Busch and others 1998; Givnish and others 2007; Zweig and Kitchens 2008) C hronically lower water table in sloughs leads to abundant emergent vegetation such as Eleocharis spp, Panicum spp, Rhynchospora spp, Cyperus spp, among others -often referred to as wet prairies. Ubiquitous throughout this region is the presence of a periphyton community whose productivity and competition is altered by water levels,
53 water nutrient status, and light penetration (McCormick and Stevenson 1998; Hagerthey a nd others 2008) Microtopographic differentiation of soil elevation is consistent in the regions considered hydrologically conserved, with a mean 22 cm difference between ridges and sloughs (Watts and others 2010). A loss of this differentiation has been a ssociated with both drainage and impoundment, and co occurs with a loss of slough communities and expansion of ridges (drainage) or a lowered prevalence of ridge communities (impoundment) These changes are reflected in the metrics of patterning, including altered patch geometry, loss of patch anisotropy, reduced autocorrelation at near neighbor dis tances, and more ( Wu and others 2006; Watts and others 2010). Precipitation is seasonal, with 70% of the rainfall occurring between May and October and an avera ge annual rainfall of 1.22 meters (2002 2010 at site W11 from EDEN; http://sofia.usgs.gov/eden/). As rainfall is the primary source of water, rainfall is tracked by annual water levels with the majority of outflow occurring in the wet season. Although the hydrology is dominated by sheetflow, water velocities are generally below 5 cm s 1 (Leonard and others 2006). There is relatively little seasonal variation in air temperature; freezing temperatures are only rarely reached (Duever and others 1997), and summ ers temperatures are often within 10 C of winter temperatures The Everglades is now highly compartmentalized and the loss of ridge slough patterning that has ensued as a result of both drainage and impoundment is dramatic. The largest and best conserved remnant landscape is found in WCA 3A covering an area of roughly 2,370 km 2 This area has been drained by the construction of canals at the northern end and impounded by levees on the southern end Landscape sampling blocks had been established by Watts and others (2010) and were used in this study in
54 WCA 3A along a hydrologic gradient designated Drained, Conserved 1, Conserved 2, and Impounded. Selected sites within each landscape block were randomly chosen as paired ridges and sloughs. Random site selec tion was designed to avoid the incremental compaction and canopy disturbance associated with repeated measurements at the same site. Environmental Data Environmental data came from several sources. Precipitation and water elevation data was obtained from nearest EDEN made available by the Everglades Depth Estimation Network (EDEN; http://sofia.usgs.gov/eden/ ): EDEN Site 62 for Drained; Eden5 for Conserved 1; W11 for Con served 2; Site 65 for Impounded. Air temperature and photosynthetically active r adiation (PAR) for model extrapolations was obtained from site DBHYDRO (http://www.sfwmd.gov/), located approximately 12 km southwest of the Drained landscape block. CO 2 Flux Measurements Closed c hambers (non steady state) are widely used for quantifying carbon fluxes between ecosystems and the atmosphere due to portability and ease of operation in remote conditions. I built chambers similar to chambers constructed for use in other peatlands (Alm a nd others 1997; Tuittila and others 1999; Bubier and others 2002; Nykanen and others 2003; Burrows and others 2004; Laine and others 2006) Two Lexan chambers were constructed so they can be placed over sawgrass (in one case), and slough vegetation (in the other) with minimal canopy disturbance. Two chambers were needed to minimize hea dspace volume when assessing NEP in sloughs (volume~0.35 m 3 ), and to accommodate sawgrass height on dense ridges (volume~0.5
55 m 3 ). I used a dynamic chamber technique for CO 2 flux, using a Li 6400 portable infrared gas analyzer (LiCor, Inc., Lincoln, NE). CO 2 fluxes were measured pre dawn through 1 PM whenever possible to capture potential temperature controls on CO 2 flux Each measurement under ambient light conditions was fo llowed by covering the chamber with a shade clothe for a R e co (ecosystem respiration) under dark conditions. CO 2 concentrations in the chamber were monitored for 3 to 15 minutes. T he measurement length was dictated by the rate of increase in gas fluxes and intended to minimize chamber interior environmental changes such as changes in relative humidity and increases in temperature However, s torms, wind, and heavy condensation shortened or precluded measurements on several occasions. Measurements commenced o n 7/8/2011 and ended on 6/11/2012, with the start dates dictated by when all landscape blocks were accessible by airboat. Sampli ng was intended to be monthly in each landscape block although weather and logistics at times precluded such sampling. PAR ( in 1 m 2 ), water depths, and water surface and soil surface temperature were monitored during each sampling period. Species cover was noted both with in the chambers and outside to encompass a 1 m 2 plot for each sampling period. While temperature changes of the soil and atmosphere beneath the chamber are a potential source of error (Wagner and Reicosky 1992), the buffering effect of water on temperature changes due to standing water present during most of the year at nearly all si tes likely reduced this potential source of error. In particular, the warmest time of the year coincides with the warmest when temperature buffering would be most necessary.
56 Error due to the suppression of natural pressure fluctuations was reduced with th e inclusion of a vent as described by Hutchinson and Mosier (1981). The change in CO 2 concentrations with time during each measurement was regressed linearly (e.g., Xu and Qi 2001; Bubier and others 2002; Wang and others 2006; among many others) with the intention to reduce under estimation of CO 2 fluxes caused by headspace concentration increase s and reductions during measurement. Chamber conditions were occasionally altered by high humidity, heat and pressure changes induced by wind gusts. A standard of a regression fitting the data with an R 2 of at least 0.98 was used to remove data when conditions were not favorable for measurement. Data Analyses I use d the convention of uptake as positive and emission via respiration as negative. Using the rectangula r hyperbola (Thornley and Johnson 1990), the re lationship between NEP and PAR is described as the rectangular hyperbola, also called the apparent quantum yield, NEP max is the asymptotic approach to NEP, and R e co is the y axis intercept, or dark respiration. This relationship was used to fill measurement gaps and to estimate total annual CO 2 fluxes. I f urther compared the basic rectangular hyperbola fitted function t o ones that allows R e co NEP max and both flux parameters to be scaled by an influence of water depths such that full model looks like:
57 where b and c are water depth scalars for R e co and NEP max respectively. As the response of both uptake (NEP, as the response of both GPP and R eco ) and release (R eco ) may be expected to act differently with response to water, model fitting includes the addition of each scaling relationship separately and in combination. I tested for site level and community effects via an analysis of variance (ANOVA), testin g for differences in daytime NEP and ecosystem respiration. I did this both on the raw data and after de trending for the light response. To determine which ancillary environmental measurements are most important for capturing measured variance (including l andscape block, community, PAR, air and soil/water temperatur e and relative community ) I used a tree regression to inform how I divided the data for model extrapolations. I further evaluated the rectangular hyperbola models across seasons, dividing the dat a into rising water (wet season 2011, or 7/8/2011 to 11/2/2 01 1), falling water (dry season 2011, or 11/3/2011 to 5/17/2012), and the beginning of the f ollowing wet season (wet season 2012, or 5/18/2012 to 7/19/2012). I used Akai for whether treatin g the dynamic water table explicitly increases the predictive power of the model. A criterion of 4 AIC values was used to discriminate between models. As calculating an R 2 of direct model fits is inappropriate for non linear models, I use a proxy of fit by comparing the predicted values from the model to the measured values; the subsequent R 2 is a measure of how well the predicted data correlated to the measured data. The best fit model was then used to estimate annual flux rates for the period of the meas urements (incorporating water elevations ; each hydrograph was offset for the
58 center of the landscape block, Figure 3 2) and soil elevation from the sites measured in Watts and others (2010). Extrapolations for all m odels were done from 7/20/2011 to 7/19/20 12, the timing of which was dictated by when hydrologic data were available at all sites. By subtracting the median soil elevation for each community from the water elevation hydrographs, I constructed a water depth hydrograph for the most commonly observe d conditions for ridges and sloughs in each landscape block. To evaluate these fluxes in terms of their significance for peat accretion, I converted the summed annual estimated fluxes for each community/landscape block to a potential peat elevation increas e using soil values of an average bulk density of 0.13 g cm 3 and soil %C of 0.405 from Bruland and others (2006). Results Climate and Environment Rainfall was higher during the observation period than the long term average, with 1528 mm near Drained, 1528 mm near Conserved 1, 1707 mm near Conserved 2, and 1677 mm near Impounded (Figure 3 1A ). This hydrologic year was unusual, demonstrating both a recovery from a deep drought as well as an early onset to the 2012 wet season (Figure 3 1B Table 3 1). The beginning of the rise of the water table was a month or longer earlier than the long term (18 year) average across all sites One key effect of the deep drought was a long delay before the water table reached the soil surface in the central part of WCA 3A. Another key effect was in the near complete lack of vegetation in the sloug hs at the end of the dry season. V egetation often totaling less than 5% of measured plots (see Figure 3 2), recovering to 25 100% cover by the peak of the wet season. Drained water levels started and ended more than 20 cm below the long term average (18 year) A similar relationship was observed in Conserved 1,
59 although the ending water levels were closer to the long term median values. In contr ast, Conserved 2 started nearly 30 cm below the long term median value, and ended in the upper 75 90 th percentile of water levels, with a similar relationship observed in Impounded. Instantaneous Fluxes There were clear differences in the measured commun ity respiration (R e co indicated by positive values; Figure 3 4), with slough values averaging less than half those of ridge values. Although absolute CO 2 flux values differ when comparing community (ridge vs. slough), sawgrass cover ( C. jamiacense ; greater or less than 45%), or landscape sampling u nit, a clear seasonal response wa s not observed ( Figure 3 3). There were some general differences by season, where slo ugh respiration rates averaged 1.12 (wet season 2011) 0.67 (dry season 2011 2012) and 0.53 (wet season 2012) mg CO 2 C m 2 min 1 R idge respiration rates averaged 2.53, 1.57, and 1.75 mg CO 2 C m 2 min 1 for the wet season, dry season, and commencement of the following wet season. Thus R eco was still high early in the study, but thereafter se ttles into a relative constant value, where the similarity of which could have been induced by the unusually stable hydrologic conditions that persisted in the later portions of this study. Rates of respiratory loss across the hydrologic year were greatest on ridges and smallest in slough s (Figures 3 3, 3 4), with no significant difference among landscape units (p>0.25). The measured daytime NEP (PAR>125 mol s 1 m 2 ) ranged from 0.10 to 9 .15 and 1.10 to 5.26 mg CO 2 C m 2 min 1 for ridges and sloughs, respectively. Thus all the ridge sites showed net CO 2 uptake at PAR levels greater than 125 mol s 1 m 2 whereas slough sites demonstrated more variability across a broad range of PAR ( Figure 3 4A ). A tree regression suggested tha t after light levels and
60 community effects, the percent of ridge cover (at the 45% threshold) controlled the variance in flux measurements. I therefore used this threshold to compare flux rates (Figure 3 4 B C ; Figure 3 5A ), and observed that while there w as considerable overlap in flux rates, both R eco and NEP were generally lower for the lower sawgrass area coverage. I was unable to establish a landscape block effect, in respiration or in NEP, or once de trende d for light (ANOVA, p>0.05). T ree regression s also did not demonstrate a landscape block effect, thus landscape position was not included in subsequent analyses. Models Model fits for the rectangular hyperbola are shown in Table 3 2. Only models with significant parameters (p<0.05) are shown; where multiple models were significant, the model used for subsequent extrapolations is marked with an asterisk The inclusion of parameter s that permit inference of water level effects on both NEP max and R eco (water depth scalars b and c ) did not improve the model fit for ridges but did for sloughs in the global data set. Taking the residuals of global data set s there was no significant relationship (linear; p>0.05) for water depth (ridges only) temperature (air), or soil temperature (water temperature, if inundated). Dividing ridge measurements into two categories based on sawgrass cover improves model fit (as indicated by the proxy of an R 2 of the predicted as a function of measured values; model curve fit is demonstrated in Figure 3 5 ). T he observations of generally higher R eco and NEP for >45% sawgrass cover than for the lower is re flected in the model parameters The consequences for fitting the rectangular hyperbola independently to different sawgrass covers is significant, wi nearl y 1.9 times higher, NEP max n early 1.7 times higher, and R eco 1.6 times higher with the greater sawgrass
61 cover (>45% sawgrass cover; Table 3 2). Interestingly, the inclusion of water depth scaling parameters for both NEP max and R eco was significant for the lower cover data, but not for the higher (Table 3 2). Monthly extrapolations for the division of sawgrass are shown in Figure 3 7A Mo del fit s differed when dividing the data into hydrological seasons ( wet 2011, dry 2011, and wet 2012; Table 3 2). The addi tion of water depth scaling variables do not improve ridge models for seasonal responses, but does improve slough models across all seasons (Table 3 2). Notably, however, there is low sample density during the early post dawn hours across the entire data set due to condensation during this time. This low sampling density results in non significant quantum yield values for the wet 2012 time period (p>0.1). Monthly extrapolations for the seasonal models are shown in Figure 3 3B D for the Drained, Conserved 2 and Impounded sites (Conserved 1 not shown due to similarity to Conserved 2). T he potential for CO 2 uptake larger than the potential for loss was captured by the higher parameter values for NEP max relative to R eco (Table 3 3). To compare total ecosystem respiration to water column respiration I e valuated the R eco in relation to the estimated R aq ( Chapter 2 ). The calculated R aq ranged anywhere from an average of 31% of the R eco rates to 90% (Impounded and Conserved 1, respectively) in ridges, but were gene rally higher than R eco in sloughs (ranging from equal to nearly 2.5 times higher, Impounded and Conserved 2, respectively; Figure 3 6). Annual Fluxes Summed monthly slough NEP estimates ranged between net autotrophy and net heterotrophy across all the lan ds cape blocks. T he greatest net heterotrophy occurring during the dry season for the Drained site (Figure 3 7). Despite larger respiratory losses
62 on ridges, ridges were net autotrophic after September. Unfortunately I lacked sufficient data to divide sawgr ass into both seasonal responses and sawgrass cover, so I must evaluate model behavior for sawgrass density and seasonality on ridges separately. T he conclusion of large carbon uptake by Impounded ridge communities with low vegetative cover (Table 3 3) may not follow real patterns ob servations of sawgrass cover <50% were not observed in the field in the Impounded site. The greatest potential for carbon uptake (and respiratory loss) is in the high biomass ridge although all of the ridge models resulted in net annual carbon uptake across all of the hydrologic conditions. S loughs demonstrated variability of a net annual loss (Drained) to net annual uptake (Conserved through Impounded). The conversion of annual net carbon flux was converted to a potential peat accretion increase and is shown as parenthetical notation in Table 3 3. In general, the potential for incremental peat increase was higher for ridges than sloughs, ranging from 1.8 to 7.6 mm yr 1 in ridges and 0.4 to 1.6 mm yr 1 in sloughs. Discussion Th e central portion of the Everglades ridge slough landscape retains the capacity for carbon uptake to exceed c arbon losses. Community specific rates differed however in terms of productivity, respiration, and the balance of the two. S lough vegetative commu nities demonstrated variation with respect to hydrology, consistent with expectations in t erms of the effects of drainage. However, ridge communities did not demonstrate a corresponding hydrologic sensitivity Only low density sawgrass communities demonstr ated a hydrologic relationship with NEP The annual carbon uptake values I observe for ridges were not outside the range of carbon inputs for peat production in other tropical peat systems. For example, Chimner and others (2005)
63 estimated carbon accumulation to be closer to 300 g C m 2 yr 1 in peatlands of Hawaii and Micronesia similar to estimates in Thailand of 527 g C m 2 yr 1 (Suzuki and others 1999). However, estimated annual R e co values for ridges in the Everglades were considerably lower t han values for a tropical peat swamp in Indonesia (Hirano and others 2007; averaging 3866 g C m 2 yr 1 ) and all of the fluxes were substantially smaller than those observed in a drained tropical peat swamp of Indonesia (Hiraono and others 2007) The resul ts here are similar t o those found by Jimenez and others (2012) in a long hydroperiod peat portion of Everglades National Park from 2008 2009 However, the two studies are not directly comparable as the eddy covariance method incorporates both ridge and sl ough communities in the area footprint Their site in SRS had peat depths similar to my sites (0.6 1.0 m thick in SRS, 0.3 0.86 m thick in WCA 3AN, 0.5 1.5 m thick in WCA 3AS). Watts and others ( 2010 ) suggest this area of SRS is of intermediate hydrologic degradation due to low water levels; topographic variation between ridges and sloughs are reduced but not yet lost and much of the slough community has been replaced by shallower water marsh species. However, Jimenez and others (2012) estimate an annual net loss of carbon, in contrast to this study estimating an annual net uptake at all but the slough communities in the Drained landscape. Lowered water levels due to lowered precipitation the years they obs erved may further explain the gap between their measurements and this study ; precipitation during the sampling 2011 2012 period was roughly 50% higher than what they observed in 2009 The higher precipitation likely contributed to my observed lowered r espiration during the 2011 to 2012 sampling period.
64 The observed increased potential carbon uptake in sloughs from drained to imp ounded hydrologic conditions is consistent with the deeper water species responses. For example, Nymphaea odorata does not exh ibit physiological limitations up to water depths of 2 m (Sinden Hempstead and Killingbeck 1996). N. odorata (syn. N. tuberosa ) demonstrates a high rate of rhizome respiration under increasingly anaerobic conditions (Laing 1940) likely due to increased ox ygenation of the rhizosphere via a thermo osmotic tr ansport mechanism (Grosse and others 1991). As N. odorata is a dominant species in the deep water sloughs, it can be presumed that low water is a greater stressor for these communities than high water. An nual estimates of ridge soil /water respiration (R aq ) are roughly 1/3 to 1/2 of the modeled ecosystem respiration ( R eco ; Table 3 3) In contrast s lough R aq i s estimated to be larger than R eco The lack of emergent vegetation in sloughs means R aq may be nearly equal to R eco in this community There are three plausible explanations for this disparity. First, measured fluxes are far more variable than the modeled fluxes. The difference between the two model results may well not be significant with r espect to measured variance Secondly, it is possible that the difference in chamber sizes between the one deployed for R aq and the substantially larger one deployed in this study result ed in non comparable flux measurements. Further, both measured and mod eled R eco follow predictions that would be made based on community phenology and a bundance and seasonal responses. This suggesting whatever resulted in seemingly low respiratory fluxes in sloughs is systematic. Finally, I suggested earlier that this hydrol ogic year was unusual relative to others Since slough vegetation was severely affected by the 2010 2011 dry season drought, it is plausible that root respiration was
65 severely diminished in proportion to the loss of vegetative cover -an important consideration as root respiration can be 30 to 70% of total soil respiration (Schlesinger 19 77). It may be that the length and severity of the dry season dry down has a strong control on all respiration rates, as was observed in chapter 2 and has been obse rved in other tropical peatlands (e.g., Couwenberg and others 2010). I surmise I simply did not observe the high respiration conditions that may be present many other years in the Everglades, and the flux measurements presented here are representative of o ne of many annual hydrologic scenarios that occur in the Everglades As a relatively young peatland the Everglades landscape has not yet reached a steady state. Peat fo rmation started around 5000 years ago in the northern portions of the Everglades, with t he majority of the Everglades initiating peat formations 2000 3000 years ago (Gleason and Stone 1994). Pre hydrologic modification peat accretion rates in the Everglades were relatively low ( 0. 1 to 0. 6 mm yr 1 accreti on; Bernhardt and Willard 2009) T he re sults presented here demonstrate a significant potential for peat accret ion throughout WCA 3A under modern hydrologic conditions. The peat accretion estimates using NEP from this study are generally similar to other measurements of peat accretion rates in the Everglades Peat accretion rates in sawgrass communities using 137 Cs dating in WCA 3A (for years 1983 to 1989; Craft and Richardson 1993 ) ranged from 2.0 mm yr 1 for a core in the drained area to 3.2 mm yr 1 for a core in the southern impounded area. T he peat accretion rates I estimate for low density ridges matches these numbers, but estimates using the entire data set suggests much larger potential for peat accretion for ridges overall. However all of the peat accretion rates for sloughs were similar to values of 0.8 1.55 mm yr 1 measured by McDowell and others
66 (1969), although those values were for peat cores taken at the far northern end of the historic Everglades in what would have been a sawgrass plain. The conversion to a potential peat accretion is not a wholly accurate way of looking at peat accretion potential, because I lack dissolved and particulate organic carbon measurements, methane production, quantification of ebullition events, and known lags in carbon turnover, particularly via standin g dead matter in sawgrass. Although it is an important green house gas emitted by wetlands, CH 4 does not appear to be a large part of carbon fluxes in the ridge slough Everglades. Relatively low rates of CH 4 production ( Bachoon and Jones 1992), combined wi th little to no relationship with temperature ( Harriss and others 1988; Bachoon and Jones 1992) and water depth ( Debusk and Redd y 1992), and high rates of methane oxidation ( King and others 1990), suggest that CH 4 is not a significant carbon mineralization pathway with respect to carbon mass balance Although based on laboratory studies of Everglades peat, potential methane fluxes appears to be between 0 and 8% of total carbon efflux, with an average of 2% across a range of water levels ( 15 to 10 cm water l evel; un enriched peat cores; Debusk and Reddy 1992; also see Inglett and others 2012). Using that average to estimate CH 4 C would reduce net carbon u ptake estimates by roughly 0 to 27 g C m 2 yr 1 There is a further loss of dissolved organic carbon (DOC ) not covered by this study. A substantial portion of the DOC in the water column is photolysized (Qualls and Richardson 2003), which is likely included as a respiratory component of the NEP measurements under full light conditions. T he proportion of DOC m ineralized by microbes is fairly small (<10% over 6 months; Qualls and Richardson 2003), so the
67 respiration of DOC likely made up a very small proportion of the measured fluxes here. However, DOC production is also generally very low outside of the Everglades Agricultural Area, less than 6% of ecosystem productivity in WCA 2 in the un enriched areas (Qualls and Richardson 2003). I can therefore estimate combined losses from DOC and CH 4 of roughly 20 to 50 gC m 2 yr 1 for ridges and 9 to 14 gC m 2 yr 1 for sloughs. While this still leaves a small excess of production in ridges the balance for sloughs begins to approach carbon neutral. Ebullition may be the source of a large loss of carbon to the atmosphere in many peatlands, and remains difficult to quantify. Observed ebullition in peatland s can be quite high, with estimates as high as 35,000 mg CH 4 m 2 d 1 (Minnesota; Glaser and others 2004), 10 to 1666 mg CH 4 m 2 d 1 (Canadian fen; Strack and others 2004), and single releasing events between 100,000 and 172,000 mg CH 4 m 2 occurring within less than 4 hours (Comas and others 2008). Measurements in the Everglades peats suggest high potential carbon fl uxes via bubbling; Chanton and others (1988) determined bubble compositions between 10 to 33% CH 4 56 to 75% N 2 and 2 to 4% CO 2 Rare, l arge ebullition events in Everglades peats can be as much as an order of magnitude larger than daily ebullition, and are highly temporally heterogeneous (Comas and Wright 2012). It is also important to consider the role fire may play in regulating carbon dynamics Although natural return in tervals for the Evergla des are unknown (Lockwood and others 2003), sawgrass burns well and frequen tly, both via lightn ing strikes and fire management. The presence of standing dead, which can be as much as 3 yea rs old (Mark Clark, personal commu nication) means carbon losses are not just from a single
68 Presuming there is a proportionate loss of standing dead for each of the three years before becoming litter or peat the losses of aboveground biomass (living + dead) to fire could e clipse the annual carbon uptake for any given year. T his loss would be restricted primarily to ridges, as emergent biomass is generally too low to carry a fire across sloughs, and standing water during most years in sloughs would diminish the risk of peat fires. I can estimate a fire return interval using a very simplified set of conditions. Assuming 1/3 of the production in any year is converted to peat in each of three subsequent years. T he standing biomass would include production (t) + 0.67*production t 1 + 0.33*production t 2 Assuming fire would remove all the standing vegetative material resulting in a zeroed productivity for that year, a fire return interval of 0 11 years is sufficient to reduce all of the peat accretion values in Table 3 3 to a long te rm average of 0 mm. In reality, if the majority of the source mat erial for peat comes from roots then this calculation over estimates the control fire would have on peat production. This over estimation may be balanced by the occasional peat fire, which no t onl y consumes above ground biomass but also would likely kill roots and remove surface peat layers. Regardless of these considerations, it can be presumed that carbon loss to fire, although episodic, is a substantial regulator of long term carbon budgets Literature values for net aboveground primary productivity (NAPP) in the Everglades are shown in Table 3 4. Although NAPP and NEP are problematic to reconcile it is constructive to observe that the annual NEP values are not dissimilar to observed values of NAPP across the Everglades, both for ridges and sloughs for most studies As suggested by this study, the potential for peat accretion is highest in ridges
69 and varies based on sawgrass density an d stature. This provides some confidence in the estimates, although with the caution that different hydrologic years may result in dramatically different flux rates. Ultimately, interannual variation in hydrology m ay induce equivalency in the community ca rbon budgets. As ridges are situation higher in the water column, th ey are also more likely to experience prolonged periods of oxidation during drier years. The hydroperiod in ridges can vary substantially in length from year to year. For example, the hydr operiod was 160 days longer in 2010 than 2011 for ridges. Ecosystem respiration for the low density sawgrass demonstrated increases with decreased water levels (Table 3 3). It is within reason to expect that drier years would produce similar results across all ridges, ultimately lessening the high rate of potential peat accretion in ridges. Conclusions Despite substantial hydrologic modification to the Everglades, the central portion of the ridge slough landscape has a carbon accretion potential with declines in that potential with lowered water levels Carbon fluxes differ seasonally, and the rate for all carbon fluxes in ridges is related to sawgrass density. As expected, the potential for carbon accretion is substantially higher on ridges than sloughs. In order for ridges to have equivalent net carbon balances to sloughs requires a loss of carbon from sloughs that is not covered by this study, but does suggest that exogenous carbon inputs to maintain ridge elevations may not be necessary to mai ntain ridge and slough soil elevation differences
70 A B Figure 3 1. Environmental conditions during the period of study. A) Rainfall for the nearest meteorological stations, corresponding with Drained through Impounded and daily average air temperature from 3AS3WX. B) Water depths at the median ridge elevation (from Watts and others 2010) for hydrologic year 2011 2012. Mean slough water depths would be +6, +19, +22, and +10 cm for each of the respective landscape blocks. Shadings indicate (from left to right) wet season 2011, dry season, and the early commencement of the wet season, 2012. 0 5 10 15 20 25 30 35 0 20 40 60 80 100 120 140 160 Mean Air Temp C Precipitation mm SITE_62 EDEN_5 W11 SITE_65 3AS3WX
71 Table 3 1. Hydrologic characteristics for each sampling block from 7/20/2011 7/19/2012. Ridge water depths (cm) Slough water depths (cm) Min Med Max Hydro period Min Med Max Hydro period Drained 28.8 3.3 39.5 193 25.3 6.8 43.0 265 Conserved 1 9.4 13.7 48.7 310 7.1 30.2 65.0 365 Conserved 2* 20.9 15.6 38.9 269 0.0 36.3 60.0 365 Impounded 5.2 46.7 77.5 365 19.2 60.7 91.5 365 *There is a lack of data at the peak of the hydrograph, so data is interpolated linearly between points; this likely results in an underestimation of the peak water depth for this site.
72 Figure 3 2. At the end of the 2011 dry season, sloughs were nearly devoid of vegetation. This picture was taken less than a kilometer off of Tamiami Trail in WCA 3AS, oriented north, in what is commonly considered the wettest portion of the Everglades. The sinusoidal trail is an alligator track. Photo courtesy of D anielle Watts.
73 A B Figure 3 3. Daily average ( s.d. for each day/location) of daytime net CO 2 exchange (NEP, closed symbols) and total respiration (R eco open symbols) A) R idges and B) sloughs at sampled locations in WCA 3A. 10 8 6 4 2 0 2 4 6 8 7/6/11 8/25/11 10/14/11 12/3/11 1/22/12 3/12/12 5/1/12 6/20/12 mgCO 2 C m 2 min 1 Drained Conserved 1 Conserved 2 Impounded 6 5 4 3 2 1 0 1 2 3 7/6/11 8/25/11 10/14/11 12/3/11 1/22/12 3/12/12 5/1/12 6/20/12 mgCO 2 C m 2 min 1
74 A B C Figure 3 1 m 2 ) A ) R idges with sa wgrass cover less than 50% and B ) ridges with sawgrass cover equal to or greater than 50% and C) sloughs Colors indicate different sampling sites/days.
75 Ta ble 3 2. Parameters and model criteria for the rectangular hyperbola function (in mg CO 2 C m 2 min 1 ). Only models with significant parameters are shown. Parenthetical notation is the standard error of the estimate NEP max R eco R eco Scalar NEP Scalar AIC R 2§ Ridge Global Data 0.041 (0.007) 6.64 (0.27) 2.25 (0.11) 0.73 C. jam aicense <45 % cover 0.029 ( 0.006 ) 4.6 6 ( 0.23 ) 1.62 ( 0.0 9 ) 424 0.83 0.030 (0.006 ) 4.58 ( 0.25 ) 1.81 ( 0.09 ) 0.0067 ( 0.001 ) 416 0.84 0.030 ( 0.006 ) 4.18 ( 0.25 ) 1.62 ( 0.09 ) 0.0047 (0.001) 416 0.84 C. jamaicense >45 % cover 0.055 ( 0.014 ) 7.77 ( 0.41 ) 2.59 ( 0.17 ) 0.75 Wet 2011 0.038 ( 0.008) 6.18 ( 0.33 ) 2.66 ( 0.12 ) 0.74 Dry 2011 0.037 (0.012) 6.73 ( 0.56 ) 1.67 ( 0.19 ) 0.73 Wet 2012 0.173 ( 0.09 ) 6.6 8 ( 0.38 ) 2.1 2 ( 0.28 ) 0.82 Slough Global Data 0.011 ( 0.0024 ) 2.79 ( 0.17 ) 0.92 ( 0.05 ) 925 0.66 0.010 ( 0.002 ) 3.54 ( 0.34 ) 1.12 ( 0.10 ) 0.008 3 ( 0.003 ) 0.0091 ( 0.003 ) 917 0.67 Wet 2011 0.011 (0.002) 3.19 ( 0.21 ) 1.15 ( 0.06 ) 419 0.77 0.011 ( 0.003 ) 3.95 ( 0.47 ) 1.15 ( 0.06 ) 0.0086 ( 0.004 ) 416 0.77 Dry 2011 0.0048 ( 0.002 ) 2.01 ( 0.39 ) 0.69 ( 0.07 ) 216 0.57 0.0048 ( 0.002 ) 1.44 ( 0.30 ) 0.94 ( 0.09 ) 0.018 ( 0.005 ) 0.0064 ( 0.004 ) 191 0.66 Wet 2012 0.035 ( 0.031 ) 2.25 ( 0.25 ) 0.51 ( 0.15 ) 187 0.67 0.079 ( 0.14 ) 3.63 ( 0.42 ) 0.57 ( 0.14 ) 0.014 ( 0.004 ) 170 0.75 § R 2 is for predicted ~ f (measured) Where more than one model had significant fits, the model for extrap o lation is asterisked
76 Figure 3 5. Ridge NEP and fitted model ( model without water depth scalars; parameters in Table 3 2) for less than 45 % sawgrass cover (grey cirlces) and more than 45 % cover (black ciricles). 1 m 2
77 Figure 3 6. Measured ecosystem respiration (daily average, R eco ) and modeled soil respiration (R aq ) with relation to local water depths. 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 50.0 0.0 50.0 100.0 mgCO 2 C m 2 min 1 water depth (cm) Ridge Reco Ridge Raq Slough Reco Slough Raq
78 Table 3 3. Summed, modeled NEP and respiration for hydrologic year 2011 2012 (g CO 2 C m 2 yr 1 ). Numbers are for the median community elevation from Watts and others 2010 (sloughs and wet prairies combined as one community). Model parameters are in Table 3 2. R aq is based on the asymptotic model in Chapter 2. Parenthetical notation indicates conversion to an approximate peat accretion increase in mm yr 1 Ridge Slough NEP (mm yr 1 ) Global Sawgrass <50% Sawgrass >50% Seasonal Global Seasonal Drained 269 (5.1 ) 93 (1.8) 113 (2.1) 400 ( 7.6 ) 392 ( 7.4 ) 28 ( 0.5 ) 22 ( 0.4 ) Conserved 1 158 (3.0) 148 (2.8 ) 50 ( 0.9 ) 32 (0.6 ) Conserved 2 147 (2.8) 142 (2.7) 51 ( 1.0 ) 59 (1.1 ) Impounded 310 (5.9) 251 (4.8) 62 ( 1.2 ) 83 (1.6 ) GPP Drained 1453 1017 968 1766 1462 583 467 Conserved 1 1003 505 423 Conserved 2 997 498 427 Impounded 1016 429 399 Respiration Global Sawgrass <50% Sawgrass >50% Seasonal R aq Global Seasonal R aq Drained 1184 924 855 1366 1070 565 555 489 566 Conserved 1 859 486 455 391 456 Conserved 2 869 500 447 368 449 Impounded 707 376 367 316 389 Since 2 models appeared to have equal goodness of fit the first number is the model using a respiration scalar; the second is the model using the NEP scalar.
79 A B C D Figure 3 7. Seasonal variability in monthly C uptake and release. Extrapolations are for A) sawgrass density less than and greater than 45% (for Conserved 2), and seaonal models for B) Drained, C) Conserved 2, and D) Impounded. Parameters for the extrapolations are in Table 3 2. 100 50 0 50 100 150 gCO2 C m 2 mo 1 Low Density: Reco Scalar Low Density: NEP Scalar High Density NEP 100 50 0 50 100 150 gCO2 C m 2 mo 1 Slough Ridge 100 50 0 50 100 150 gCO2 C m 2 mo 1 100 50 0 50 100 150 gCO2 C m 2 mo 1
80 Table 3 4. Literature values for ridge (or sawgrass community) and slough annual net aboveground p rimary productivity for the Everglades. Community Location g C m 2 yr 1* Citation Sawgrass 1436427 Daoust and Childers 1998 Sawgrass 385 1457 Davis 1989 Sawgrass Northern ENP 144 408 Daoust and Childers 1999 Wet Prairie Northern ENP 14 65 Daoust and Childers 1999 Wet Prairie 19677 Daoust and Childers 1998 Sawgrass Shark River Slough 80 384 E we and others 2006 Tall Ridge Sawgrass North WCA 3A Central WCA 3A North ENP Central ENP 2062 9 1760 108 2715 48 1807 202 Lewis 2005 Short Ridge Sawgrass North WCA 3A North ENP Central ENP 1194 77 525 6 1674 135 Lewis 2005 Slough North WCA 3A Central WCA 3A North ENP Central ENP 808 128 406 195 319 103 728 76 Lewis 2005 *g DW m 2 yr 1 was converted to gC m 2 yr 1 using an estimate of 48% C.
81 CHAPTER 4 EVIDENCE FOR TRANSPIRATION INDUCED NUTRIENT ACCUMULATION The Everglades ridge slough patterning organizes along hydrologic flow lines but the mechanism s underlying pattern origin and maintenance remain uncertain (SCT 2003; Larsen and ot hers 2007; Cohen and others 2011 ) M ost attention has focused on surface flow phenomena with support for two mechanisms Substantial e mpirical and modeling effort s to date offers some support for the first, wherein preferential deposition of particulates from sloughs to ridges occurs due to velocity differences between the communities during high flow events ( sediment redistribution hypothesis; Larsen and other s 2007; Larsen and others 2009; Larsen and Harvey 2010 ) A second mechanism proposes that ridges and sloughs organize to maintain landscape discharge competence so that sloughs self organize to route water (self organizing canal hypothesis; Cohen and others 2011 ; Heffernan and others in press ). Here we provide support for a third mechanism, where subsurface flow induced by differences in evapotranspiration leads to a subsidy of nutrients to ridges (Ross and others 2006; Cheng and others 2010) Evapotranspiration gradients between communities can have significan t influences on ecosystem and landscape water budgets, nutrient movement, and carbon accr ual (Mc Laughlin and others 2012; Eppinga and others 2008; Mitsch and Gosslink 2011). Nutrient focusing induced by evapotranspiration differences between patch types ha s been demonstrated in both northern and subtropical peatlands ( for example, Eppinga and others 2008; Nachabe and others 2005). In these case s higher plant productivity in one patch coincides with higher ground elevation; higher evapotranspiration of this patch induces convergent flow of pore water with associated
82 nutrients towards the higher elevation from adjacent areas. This resource concentration supports the higher productivity in those patches but is critically limiting on a larger scale : As local fo cusing of water and nutrients is induced, there is a resulting distal inhibition where nutrient depletion in adjacent patches becomes limiting at larger scale s This combination of local activation and distal inhibition are referred to as scale dependent f eedbacks (Eppinga and others 2008) and results in regular peatland surface patterning, although it is important to note that this self organized patterning neither limited to peatlands nor is the only process demonstrated to lead to regular surf ace pattern ing ( e.g. Rie tkerk and others 2008). The contention that the Everglades i s a self organized patterned landscape is largely the result of pattern metrics, both in patch geometry (Wu and others 2006) and soil elevation (Watts and others 2010), which demonstrate predictable degradation with altered hydrology (primarily water depths and hydroperiod). Notably, patterning is found in both veg etative communities (e.g. patch configuration Wu and others 2006; spe cies assemblages Busch and oth ers 1998; patch p roductivity Daoust and Childers 1999; E we and others 2006, others) a nd soil elevation (Watts and others 2010). There is some evidence that both patterns are reflected in soil ph osphorus concentrations with ridge TP ca. 100 150 mg kg 1 high er in ridges than sloughs ( Cohen and others 2009) and tree islands as much as an order of magnitude hi gher than the surrounding marsh ( Ross and others 2006; Wetzel and others 2005 ). These observation s have led to hypoth eses of evapotranspiration gradient s between patch types concentrating nutrient to higher elevations (Wetzel and others 2005; Ross and others 2006) Indeed, the mechanism has proved likely on many of the Everglades tree islands with observations
83 of diurnal reduc tions in water tables in the 1 to 3 cm range dur ing the dry season and a predominance of groundwater recharge during the wet season (Ross and others 2006; Adamski 2009; Troxler and others 2009), although this resource concentration mechanism is likely only one of several working simulta neously on tree islands (Wetzel and others 2011). Applying t his hypothesis to pattern evolution in ridges and sloughs has been more pro blematic In contrast to tree islands, the more productive, higher elevation sawgrass patches (ridges) are inundated for most of the year (ca. 310 days in the central Everglades), with even longer hydroperiods in sloughs (350+ days). Precipitation exceeds evapotran spiration in most years (German 2000), so a persistent ly higher head in ridges is unlikely. Further t he window of time for ridge transpiration to obtain soil water pulled through the peat matrix rather than accessing shallow surface water is short ; in many years most ridges will be permanently inundated (Watts and others 2010 ; McVoy and others 2011 ). D eep and prolo nged inundation during the typical wet season, which is also the typical growing season, should result in a chronic diffusion of any bio available nutrients. There are, however, years with significantly longer and more severe dry seasons where an evapotran spiration mechanism could act more strongly E ven if a head gradient were not observed, equilibration of the water table between ridges and sloughs during the dry season would suggest a subsidy of water from sloughs to ridges. During those times when ridge soils are exposed, higher potential evapotranspiration must result in a subsidy of water to ridges from somewhere with the only likely source being inundated sloughs.
84 The objective of this study was to evaluate diurnal water table dynamics to assess an ev apotranspiration induced resource concentration on ridges. I do this via comparisons of diurnal water table dynamics between adjacent ridge s and slough s in Water Conservation Area 3A (WCA 3A) The hypothesis of a groundwater subsidy from sloughs to ridges implies amplified diurnal water table fluctuations in sloughs as the water table recedes from the soil surface in adjacent higher elevation ridges In this scenario the slough water table responds both to local evapotranspiration and demand for water from adjacent ridges. Under the scenario of a disconnected water table, the soil matrix amplifies diurnal changes to the water table in ridges but not sloughs A connected water table suggests an amplified diurnal response of the water table in sloughs and a di minished response in ridges as ridge water demand is subsidized by adjacent sloughs. The equilibration of the water table would result in the ridge and slough water table s responding as the average of above and belowground signals, with the magnitude propo rtional to the relative abundance of ridges and sloughs Methods Data Collectio n Two sites of adjacent ridge and slough were chosen in WCA 3A (Figure 4 1 ). The Drained site is located in WCA 3AN ( Figure 4 1 Table 4 1 ), designated as a drained area because of prolonged historically low water levels ( Drained; Watts and others 2010). This area has shallow peat depths of less than a meter (Table 4 1). Vegetation cover at the Drained ridge was Cladium jamaicense (saw grass) with epiphytic mats of periphyton. The slough was characteristic of a wet prairie, with periphyton, Sagitaria lancifolia Elechoaris spp., Bacopa caroliniana and Panicum spp.
85 The Conserved site (in the landscape unit referred t o as Conserved 2 in Watts a nd others 2010) is located in the central portion of WCA 3AS ( Figure 4 1 Table 4 1). Ridge vegetation was dense Cladium jamaicense with Cephalanthus occidentalis and sparse Justicia angustifolia Nymphaea odorata Utricularia spp., Nymphoides aquatica an d periphyton, with interspersed Crinum and Eleocharis spp, dominated the slough vegetation. The Conserved site has much deeper peats, averaging more than a meter deeper than those of the Drained site (Table 4 1). Diurnal water table changes were measured in surface water wells installed in a djacent ridges and slough s The wells were constructed of Sch edule 40 PVC pipe, screened above and below the soil surface. The installation depth was determined by the depth of peat to limestone bedrock (Table 4 1) ; anchoring to bedrock removes shrink swell effects of the peat, the magnitude of which remains unknown To tal pressure was measured at 15 minute intervals using a total pressure transducer (Solinst Gold Leveloggers, accuracy = 0.3 cm, resolution = 0.005 cm ). A third dry well, open to atmospheric pressure, was installed near each ridge well and equipped with a barometric pressure transducer hung below the soil surface t o account for temperature sensitivity described by McLaugh lin and Cohen (2011). T otal pres sure was corrected for barometric pressure variation by barometric pressure transducers (Solinst Barologgers, accuracy = 0 .1 cm, resolution = 0.003 cm). Water depths were noted at 3 locations around each well averaged and converted to soil surface elevati ons using water elevation hydrographs at nearby E verglades D epth E stimation N etwork (EDEN) gauging stations. P robes were changed at Conserved on 2/26/2011 and Drained on 2/27/2011; readings were adjusted for a small offset s between probes (<0.5 cm) The
86 wa ter table dropped below the sensors during the dry season a t the Drained site ( 3/20/11 to 6/28/11 for the ridge well, 5/5/2011 to 6/28/11 for the slough well ) and these data were excluded from analysis, as were days with rain (Fig ure 4 2 ) Data Analysis A small but persistent drift (2.5 cm over the period of record) between the ridge and slough water level loggers rendered an analysis of head gradients impossible for the Conserved wells. However, the change over any single day was very small (~0.005 cm). A s such, any analysis of diel variation was judged to be robust for subsequent analyses. This drift was not observed in the Drained site, so I evaluated water levels at this site for a persistent head difference between wells by comparing the difference in water stage In order to evaluate the presence of a lock step response in water ta bles between ridges and sloughs I used an evapotranspiration: potential eva potranspiration ( PET ) relationship to infer groundwater exchange between ridges and sloughs E vapotranspiration calculations are sensitive to the specific yield (Sy) of either the vegetation, when water levels are above the soil surface, or of the soil mat rix, when water levels a re below the soil surface, with substantial differences in the specific yield between the two conditions. Thus evapotranspiration/PET should be between 0 and 1 when the water table is above the soil surfa ce (where the specific yield approaches 1 due to the low di splacement by marsh vegetation). Evapotranspiration/PET should increase in proportion to soil specific yield when the water table is below the soil surface (in response to the much lower specific yield of soil, here peat). If ridges and sloughs are hydrologically disconnected (i.e., no or little equilibration of the water table in response to higher evapotranspiration on ridges) then the slough
87 evapotranspiration/PET should remain between 0 and 1 during the critical time when ridges are dry but sloughs remain inundated. However, should evapotranspiration/PET in both communities act as a composite of ridge and slough conditions during this critical time then I can infer nearly instantaneous equilibration between ridges and sloug s method (Hargreaves and Samani 1982) which incorporates length of day, air temperature, and monthly averaged incoming extraterrestrial radiation (http://www.fao.org) M eteorol ogical data used to calculate PET was accessed via DBHYRO (http://www.sfwmd.gov). Air temperature and solar radiation data was accessed from S140W, located approximately 10.5 km from the Drained wells and from 3AS3WX, which is located roughly 15 km southwest of the Conserved wells. I use this ratio method rather than directly measure evapotranspiration with specific yield for three reasons. First, the behavior of the groundwater is the variable of interest, and thus a relationship form is appropriate for this study. Moreover, there are conside rable difficulties in accurately measuring specific yield, the most careful measurement of which may still miss the effects of hysteresis, air entrapment, microbial generation of gases, dissolution of gases, shrinking and swelling of the soil matrix the e ffects of soil moisture, and capillary effe cts (examples include Duke 1972; Sophocleous 1985; Fayer and Hillel 1986a 1986b; Choi and others 2003). A number of studies provide estimates of specific yield in the Everglades. For example, Gesch and others (20 07) estimated porosities of 0.75 to 0.85 g cm 3 for organic histosols in the Everglades, the high porosities in lines with specific yield values of 0.2 used by regional hydrologic models ( Nair and others 2001; Wilsnack and others 2001), and strikingly
88 similar to the 0.22 calculated by Sumner (2007) for the Everglades. However, Sumner (2007) demonstrates out that estimating specific yield with simplified values a bove and below the soil surface creates large errors due to microtopography Microtopography significant ly infl uences realized specific yield (i.e. ESY) with a connected water table ( also see McLaughlin and others 2011 ), where the water table in lower elevation areas are partially controlled by the ground water response to specific yield of higher elevation areas. E vapotranspiration and exfiltration were calculated using two different methods using barometrically cor rected water level data (Figure 4 3 ). White (1932) developed a method of analyzing groundwater fluctuations to determine use by plant s, such that where h is the change in water elevation at midnight between one day and the next (or, h 1 h 2 ) and G is the slope of ordinary least squares best fit lines between nighttime water level and time and Sy is the specific yield (hereafter referred to as the White Method ; Figure 4 3 b ). The diurnal period for the fitting of the slope was extended to 22:00 5:00 from 0:00 4:00 to improve the fit due to noise during low flux times of the year. This method was used when the R 2 of the fitted s lope fell between 0.15 and 0.85. Below 0.15, ET is considered to be negligible and essentially 0. A non linear water table response occasionally caused the fit to be poor; on these occasions I instead used a method described by Hays (2003) as the d raw down recharge method (hereafter referred to as the Hays Method). This method calculates evapotranspiration as x Sy
89 where h is the water level high, l is the water level low, and t 1 and t 2 are the lengths of time for night time recharge and daytime recharge, respectively (Figure 4 3 a) The Hays method provides the average recharge rate during the diurnal cycle so that short term alterations in rates (which lead to intermediate R 2 values und er the White method) do not affect the calculations. This method does not compensate for nighttime transpiration as the White Method does, and thus is a more conservative estimate for water use. The White Method, in comparison, considers recharge and daily changes in the water table, but assumes that recharge remains constant over any 24 hr period. D ays where evapotranspiration values were artificially measured to be negative due to very low evapotranspiration fluxes were excluded. A ll such instances occurr ed during winter months when wate r was above the soil surface D ays with rainfall were also excluded; rainfall data came from gage adjusted Next Generation Radar (NEXRAD) data from the U.S. National Weather Service, made available by the Everglades Depth E stimation Network (EDEN) via http://sofia.usgs.gove/eden/nexrad.php. Rainfall data for the Drained wells was obtained from EDEN site ID 3AN1W1, located 2.45 km (ridge) to 2.11 km (slough) away from the well locations. Rainfall data for the Conserved wells came from EDEN site ID Site_64, located approximately 2.35 km from the wells. I observed differing evapotranspiration/PET response during the 2011 wet season from 2010 values for the same water levels, with evapotranspiration/PET well over 1.0 for all date s after 8/1/2011; during the 2011 rise in the water table the diurnal variation in water levels is as high as 1.5 to 3 cm even after the water table rose above the soil surface across all of the wells. As this is inconsistent with all theoretical models of evapotranspiration/PET when S y approaches 1, something systematic occurred either
90 to the sensors or to the local environment to dramatically increase the calculations. After careful evaluation of the barometric and temperature readings across all of the sensors, I conclude that the large diurnal signature is not consistent with errors due to sensor malfunction or drift nor is due to problematic calculations in evapotranspiration values. I am forced to conclude that there is an explanation not provided by available information (meteorological, stage, sensor, or vegetative) for these unusual readings I therefore use only the data prior to 8/1/2011 for further evapotranspiration analysis. W hat remains unk nown are regional water flows; m arsh surface water ve locity is not available but I infer from extended gate openings after 8/5/2011 at the southern end of WCA 3A that water flow may have been much higher during this time. E vapotranspiration calculations are generally problematic in flowing systems, thus accu rate evapotranspiration calculations in the Everglades may only be possible when flow is negligible Subsidy Calculations Under the scenario of a rapidly equilibrating water table between ridges and sloughs Sy and ESY (ecosystem spe cific yield) should div erge during the period of time when one ecosystem patch has the local Sy of peat (ridges) and the other has the Sy of open water (sloughs). Assuming that the full water demand is being met during this time, the observed change in water level in both ridges and sloughs should be PE T/ESY where ESY is approximated as A R and A S are the a reas of each ridges and sloughs and Sy o and Sy p are the specific yields of open water and peat, respectively. This presumes that the observed diurnal change in water levels in sloughs is PET plus some loss of water to ridges, and the
91 diurnal change in water levels in ridges is PET/Sy p minus some gain f rom sloughs. The implicit assumption is that all the atmospheric water d emand is being met, such that evapotranspiration x ESY = PET Therefore the subsidy of water from sloughs to ridges, Q, is then: which scale s the volume of flow, Q, to water demand (PET) by area of ridges (A R ), adjusted for the landscape response to water demand. Ridge patch geometry data (area, perimeter ) w as obtained from Jing Yuan. For the sake of convenience, sloughs were assumed to be pro portionate in size to the nearest ridges since sloughs are interconnected and therefor difficult to separate Sy p is estimated at 0.22, after Sumner (2009), and Sy o was approximated as 1 for convenience, which may slightly over approximate ESY. The total n um ber of days a year when induced lateral flow could occur were generated by subtracting the median soil elevation (from Watts and others 2010) for each patch type from the EDEN water elevation s from 2002 to 2011 The days of interest were those days where the ridge soil would be exposed but slough still inundated Individual points on the landscape may fall at the extremes of the variance in soil elevations, and thus experience differing hydr operiods. T his calculation captures the most representative but n ot all landscape conditions. It is important to note that this gives a potential subsidy, only, and consider s only the advective flow of phosphorus. To avoid concern about the disproportionate c hanges rainfall can create in diurnal water table fluctuations (e.g., Gerla 1992) calculations for ex trapolated ground water flow were done only on days without rainfall. Estimating the potential advective flow of
92 phosphorus (P) to ridges on a per area basis was accomplished using va lues for soluble reactive phosphorus in the porewater i.e., the bio available phosphorus, of 6.54 ppb P for the Drained landscape and 8.88 ppb P for the Conserved landscape ( Watts unpublished data ) Results I observed all the stage conditions of both com munities inundated, ridges dry but sloughs inundated, and both communities dry (Figure 4 2) Recorded water depths in ridges were consistently lower than in the sloughs (on average 9.7 0.54 cm in Drai ned, 18.0 0.95 cm in Conserved ), which aligned with t he difference in soil elevation. The trend in the Drained head between wells was not consistent through time towa rds either ridges or sloughs. F ocus ing specifically on times when the water table is below the ridge soil surface, but above the slough surface no persistent or enlarging head gradient was demonstrated, suggesting rapid later al equilibration Once the water table falls below the slough surface the difference in water table is generally less than 0.5 cm between wells The average difference in wa ter table between ridges and sloughs after ridge soil surface exposure is 0.07 ( 0.4) cm. Figures 4 4 (Drained) and 4 5 (Conserved) demonstrate water elevation s relative to an arbitrary datum defined by the ridge well, for 3 day periods The figures correspon d to both wells inundated (Figures 4 4&5A) ridge soil surface exposed but slough inundated ( Figures 4 4&5 B), and the water table below the soil surface for both ridges and sloughs ( Figures 4 4&5 C). The diurnal response is small when both communi ties are inundated but becomes exaggerated as the water table falls below the soil surface across the landscape ; this effect is larger for the Conserved site
93 There wa s good agreeme nt between the White and Hays me thods for calculating evapotranspiration ( R 2 = 0 .86 and 0.77 for the ridge and slough wells in Drained, and 0.93 and 0.94 for Conserved wells). E vapotranspiration/PET increased above 1 when the water dropped below the ridge soil surface for both sites (Figure 4 6 ). When both ridges and sloughs were dry, evapotranspiration/PET was consistently high for both the Drained (5.89 0.68 and 3.7 0.48, ridge and slough respectively) and Conserved locations (7.17 0.91, 8.75 1.77, ridge and slough, respectively). In comparison, evapotranspiration/PET was al ways significantly lower when both ridges and sloughs were inundated (0.79 0.48, 0.62 0.41 for Drained ridge and slough 0.7 0.44, 1.03 0.35 for Conserved ridge and slough ). The behavior of evapotranspiration/PET was very different between sites during the critical period when the ridge soil was exposed but sloughs remained inundated Consistent with expectations of instantaneous equilibration, the evapotranspiration/PET between the community types were correlated for both Conserved (0.7) and Drai ned (0.93) during the dr y ridge/wet slough time period (Figure 4 6 ): Conserved evapotranspiration/PET for both communities demonstrated a dramatic increase starting when water depths on ridges reached 2 cm, with no significant difference between values for until sloughs also became dry (p<0.05, 6.01 0.99 and 7.11 1.09 for the ridge and slough). In contrast, the evapotranspiration/PET for both ridges and sloughs in creased by roughly 0.37 per 1 cm drop in ridge daily water table until they reached the value s when both ridges and sloughs were dry. ESY was estimated as 0.5 7 for the Conserved landscape (Table 4 2) The Drained site had a larger ESY of 0.73 due to the lowered proportion of ridges to wet prairie (Table 4 2). Table 4 2 also shows values for Q, b oth in terms of total landscape
94 ridge area (Q area ) and as a function of sum of all of the ridge perimeters (Q perimeter ). Because these values are summed across the landscape, they represent the average ridge condition for all landscape patches What become s strikingly obvious is that the shape of the ridges has large implications for the total volume of flow to those patches. Although a lower specific yield and longer period of time when ridge soil surfaces are exposed but sloughs inundated would suggest la rger flow volumes to ridges at the Conserved site (as demonstrated by Q area ) the skew of larger patches in Drained (and resultant smaller perimeter: area ) results i n values similar to that of the Conserved landscape The influence of patch area on groundw ater subsidy fl ow is demonstrated in Figure 4 8 where the subsidy per unit ridge perimeter (Q perimeter ) is proport ional to the size of the patch and thus p ower function distributed as patch size is also power function distributed. Comparing the sum o f the subsidy across years when data was available (Table 4 3) shows variability in Q area and Q perimeter between Drained and Conserved The PET forcing differs by time of the year; PET forcing was quite low when the critical hydrologic conditions were met du ring winter months (e.g., Drained 2010, 2011) relative to years when the critical hydrologic conditions were met during summer months Averaging across all of the years, the total subsidy to ridge edges (Q perimeter ) was very similar between Drained and Con served, despite the observation that, PET being equal between the two sites, flow would be much higher into ridge edges in Drained because of the perimeter:area influence However, the annual subsidy of phosphorus, as a function of perimeter length, ultima tely differed because estimated pore water concentrations of P differed between the landscape units.
95 Discussion This study of diurnal variation in water levels supports the hypothesis that the underlying variability in soil phosphorus could be induced by e vapotranspiration induced convergent flow of water into ridges. Rather than demonstrate a persistent head towards ridges as has been proposed in this landscape and demonstrated in other ecosystems this study demonstrates groundwater surface water connect ivity can lead to convergent flow of groundwater to ridges during specific times of the year. Eppinga and others (2008) demonstrated a persistent head gradient from hollows towards ridges, driven by higher evapotranspiration rates there, which explained nutrient accumulation via pore water concentration below higher areas. Eppinga and others (2010) further demonstra ted that this mechanism varied climatically, where high annual evapotranspiration relative to precipitation (evapotranspiration: precipitation >1) was indicative of redistribution of nutrients towards higher elevations whereas higher precipitation relative to evapotranspiration (evapotranspiration: precipitation <1) would lead to a flow of nutrients to lower elevations I suggest there is a third alternative for climatic events, one that is specific to water table position. The Evergl ades has a precipitatio n : evapotranspiration greater than 1 for most years (German 2000 ), which is the major contributing factor to sheetflow hydrology across the historic Everglades. However, the exposure of the ridges while sloughs remain inundated leads to diurnal water respo nses that are clearly an intermediary between the conditions found in ridges and sloughs. The lack o f a head gradient is therefor e insufficient evidence to reject the hypothesis of a convergent ground water flow where there is strong hydrologic connectivit y, microtopography, and a period of the year when the water table drops below the soil surface for the higher elevation areas.
96 Water levels did not vary independently d uring the critical time when ridges were exp osed but sloughs were inundated Moreover t he water elevation in the slough at the Conserved site behaves as if it is a mixture of peat and open water, which would only occur if the slough is hydrologically connected to the adjacent ridges. The difference in response between the Drained and Conserv ed sites is likely due to the probability density function of the s oil elevations (see Watts and others 2010). Despite elevation differences between the adjacent ridge and slough at the Drained site the greater landscape is remarkably flat This suggests g radual transit ions between ridges and sloughs. A connected water table response would demonstrate a continuous change in realized specific yield until the entire landscape is dry. In contrast, the Conserved landscape is bimodal in soil elevations, with dis tinct and abrupt edges between the two communities. Under such a scenario a step function response in ecosystem specific yield would be expected consistent with Conserved observations. Both site responses are consistent with predictions posed by Sumner (2 007), where landscape position and microtopography has a significant influence over ESY for lower elevation regions. T he assumption of a step in elevation between ridge and slough edges is a limitation of the approach here The values presented in Table 4 3 assume all pat ches meet the median patch hydrologic conditions The subsidies are going to be exaggerated to some degree for both sites, but more so in the Drained landscape. That the variance of ridge elevations is smaller than the variance of the el eva tions overall alleviates some of this concern for both landscapes The observed lateral water subsidy could provide the feedback that gives rise to power function distributions in the Everglades (Yuan and others 2012) The
97 perimeter:area relationship with water demand leads to a larger subsidy to ridge edges with larger patch size (Figure 4 7). Power function distributions require a positive feedback that induces large patches to expand more rapdily than smaller patches In other words, the feedbacks that induce patterning give rise to regular patterning. Power functions of patch size needs a mechanism whose strength increases with patch size. The current hypotheses for the Everglades (sediment transport and self organized canals) could create regular, linea r features but lack a feedback that acts more strongly on larger patches I note it is unlikely that the evapotranspiration mechanism alone is sufficient to explain the Everglades patterning; t he length of time when con vergent flow is most likely ( Table 4 3) varies substantially from year to year However, the re remains power law pa tch geometry in both landscapes despite strongly altered hydrology (Yuan and others 2012). This geometry in Drained suggests the phosphorus subsidy is still in play to some degre e in that landscape As further evidence of convergent flow acting only as a reinforcing mechanism, the phosphorus availability on ridges is not as dramatically different from that of sloughs as compared to the third main Everglades patch, tree islands. T r ee islands may have as much as an order of magnitude higher P in both pore water and soil compared to both sloughs and ridges (Ross and others 2006) The greater concentration of phosphorus i n ridges compared to sloughs is far more subtle, and appears to b e more associated with elevation than patch desi gnation (Cohen and others 2009): T he significance of the correlation between soil elevation and soil phosphorus is reduced when vertical relief between patches is lost.
98 I approximate d the likely subsidy of ph osphorus to ridges by multiplying the flow (Q) by the concentration of phosphorus in marsh pore waters yielding the values presented in F igure 4 8 and Table 4 3. Although likely an overestimation it is useful to consider the scale of the subsidy to ridge edges, as the implication is this is the phosphorus that plays a role in driving ridge expansion. T he year to year subsidy can differ by as much as an order of magnitude, and the potential subsid y is actually relatively small on an area basis The values in Table 4 3 are essentially landscape averaged values; small and large ridges would have proportionately smaller and larger subsidies per length of ridge edge. However, if we consider soil phosph orus in the upper 10 cm (3AN = 452 mg/kg TP; 3AS = 402 mg/kg TP; Bruland and others 2006), with peat accretion approximate ly 3 mm yr 1 (Craft and Richardson 1993 measured 2 to 5 mm yr 1 for ridges ), then the annual subsidy of phosphorus to ridge edges is a s much as to 4% (Drained) to 14% (Conserved) of the annual soil P storage The question remains whether convergent flow in large ridges would result in a horizontal transport of water. Although the total flow volumes are large, the actual flow for any leng th of ridge is small and the flow path long from ridge edge to center. Weaver and Speir (1960) measured horizontal hydraulic conductivity ranging 11.3 to 35.5 cm h 1 for Everglades mucks in the upper 100 cm of peat with remarkable similar vertical saturated conductivity ranging 15.2 to 42.1 cm h 1 Presumably the shorter vertical flow path results in some water be sourced from deeper peat; this suggests that some but not all of the increased phosphorus in ridge soils can be attributed to a subsidy o f water from sloughs. A lternative mechanism s potentially generating greater phosphorus concentrations were not covered in this study and may complicate the phosphorus
99 narrative Carbon mineralization during soil exposure releases both bound nutrients and c arbon (Debusk and Reddy 2003), and the loss of carbon would lead to an apparent concentration of phosphorus without any exogenous subsidy. As this would be occurring during the same periods of the year when instantaneous ground/surface water equilibration would be significant for phosphorus movement, these two mechani sms may well act in concert. For phosphorus mass balance i t may not matter the exact source of phosphorus, however, as both phosphorus sources would converge to the highest locations Indeed, t he mixture of sources of phosphorus, as well as the local activities of microbes, may explain the difference in marsh surface dissolved organic phosphorus (DOP; 0.004 mL 1 ) to that of porewater (0.013 mL 1 ) observed in the nutrient un enriched interior por tions of WCA 2 (Qualls and Richardson 2003). In this study I provide evidence for an evapotranspiration induced convergent flow to ridges during critical times of the year when ridges are dry but sloughs remain inundated. There are several lines of eviden ce that would be predicted by the depth dependent convergent flow to ridges. First, phosphorus availability should be greater on larger ridges, because the exerted water demand is proportionate to area. Second, during ridge dry downs a phosphorus front sho uld be observed on ridge edges, induced by convergent flow. The magnitude of that front should be larger on the edges of big ridges than those of smaller ridges. Third, the expansion of ridges in a bi modal landscape should be faster than the expansion of small ridges. This would not be true in a drained landscape due to more favorable local hydrologic conditions at ridge edges; i.e. there is no need for a nutrient subsidy to overcome anoxic stress as there would where there are abrupt transitions between r idges and sloughs.
100 Figure 4 1 Locations of the wells in WCA 3AN (Drained) and 3AS (Conserved). Figure generated with SPOT imagery viewed with Google Earth (Google, Inc., Mountain view, CA, USA). Drained Ridge Slough Conserved Ridge Slough 145.75 m 110 m
101 Table 4 1. Locations of the ground water wells. Ridge Slough Baro Distance between R S (m) Date Length Peat Thickness (cm) Drained 80.72755 26.20422 80.72901 26.20419 80.72750 26.20420 151.75 11/2/2010 3/17/12 36.5 44.5 Conserved 80.69190 25.96972 80.6908 25.96974 80.69190 25.96972 109.98 11/3/10 3/18/12 110 124
102 A B Figure 4 2 Daily average water depths and rainfall A) Drained and B) Conserved for the period of record. The horizontal line repres ents the local soil elevation 40 30 20 10 0 10 20 30 40 50 60 0 2 4 6 8 10 12 14 11/6/10 12/6/10 1/6/11 2/6/11 3/6/11 4/6/11 5/6/11 6/6/11 7/6/11 8/6/11 9/6/11 10/6/11 11/6/11 12/6/11 1/6/12 2/6/12 3/6/12 water depth (cm) rainfall (cm) Rainfall Ridge Slough 80 60 40 20 0 20 40 60 80 100 0 1 2 3 4 5 6 7 8 9 11/6/ 12/6 1/6/11 2/6/11 3/6/11 4/6/11 5/6/11 6/6/11 7/6/11 8/6/11 9/6/11 10/6 11/6/ 12/6 1/6/12 2/6/12 3/6/12 water depth (cm) rainfall (cm)
103 Figure 4 3 Schematic of evapotranspiration calculations. A) The Hays Method and B) The White Method. 24 h H 2 H 1 H 2 L T 2 T 1 H 1
104 A B C Figure 4 4 Water table position for Drained ridge and slough (relative to the ridge well) and the difference between the water tables for time periods A) B oth ridge and slough are inundated, B) ridge soil surface is exposed but the slough is inundated, and C) both ridge and slough surfaces are exp osed. 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 0.05 0.1 0 0.5 1 1.5 2 2/1 0:00 2/1 12:00 2/2 0:00 2/2 12:00 2/3 0:00 2/3 12:00 2/4 0:00 difference ridge slough water depth (cm) Ridge Slough Difference 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 8 7.5 7 6.5 6 5.5 5 2/21 0:00 2/21 12:00 2/22 0:00 2/22 12:00 2/23 0:00 2/23 12:00 2/24 0:00 difference ridge slough water depth (cm) 0.5 0.3 0.1 0.1 0.3 0.5 0.7 0.9 18 17 16 15 14 13 12 11 4/2 0:00 4/2 12:00 4/3 0:00 4/3 12:00 4/4 0:00 4/4 12:00 4/5 0:00 difference ridge slough water depth (cm)
105 A B C Figure 4 5 Water table position for Conserved ridge and slough (relative to the ridge well) and the difference between the water tables for time periods A) B oth ridge and slough are inundated, B) ridge soil surface is exposed but the s lough is inundated, and C) both ridge and slough surfaces are exposed. 0.15 0.1 0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 7 7.5 8 8.5 9 2/1 0:00 2/1 12:00 2/2 0:00 2/2 12:00 2/3 0:00 2/3 12:00 2/4 0:00 difference ridge slough water elevation (cm) Ridge Slough Difference 0.6 0.5 0.4 0.3 0.2 0.1 0 0.1 0.2 0.3 0.4 19 18 17 16 15 14 13 4/1 0:00 4/1 12:00 4/2 0:00 4/2 12:00 4/3 0:00 4/3 12:00 4/4 0:00 difference ridge slough water elevation (cm) 0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 45 43 41 39 37 35 5/1 0:00 5/1 12:00 5/2 0:00 5/2 12:00 5/3 0:00 5/3 12:00 5/4 0:00 difference ridge slough water elevation (cm)
106 A B Figure 4 6 Evaptoranspiration/PET relative to the water table position in the ridge well A) Drained and B) Conserved sites. 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 30.00 25.00 20.00 15.00 10.00 5.00 0.00 5.00 10.00 ET/PET Ridge Slough ridge dry slough dry 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 55.00 35.00 15.00 5.00 25.00 45.00 ET/PET relative water level (cm) ridge dry slough dry
107 Table 4 2. Community area and water subsidies (av eraged across days) to ridges (averaged across the landscape). Ridge:Slough Ridge Perime ter:Area ESY Q Area m 3 m 2 d 1 Q Perimeter m 3 m 1 d 1 Drained 0.5 3 0.0 30 0.73 0.002 6 0.084 Conserved 1. 21 0.0 32 0.5 7 0.002 7 0.085 For the Drained landscape, sawgrass has expanded into slough/wet prairies such that the extent off sawgrass is qu ite a bit larger than the extent restricted to communities dominated by sawgrass.
108 A B Figure 4 7 The subsidy of ET induced convergent flow of phosphorus for the 2011 dry season into ridge edges in relation to ridge size A) Drained (21 days) and B) Conserved (26 days) 0 2 4 6 8 10 12 14 16 18 10 100 1000 10000 100000 1000000 10000000 Average mgP m length ridge 1 0 5 10 15 20 25 30 10 100 1000 10000 100000 1000000 10000000 mgP m length ridge 1 Ridge area m 2
109 Table 4 3. Subsidy of water and phosphorus to ridges for 2002 2011. Calculations are based on days without rainfall when hydrologic conditions are met and the sum of the ridge areas and perimeters, and thus are most appropriately considered an averaged value for the landscape Year No. Days § Avg PET (cm) Q perimeter m 3 m 1 yr 1 Q area m 3 m 2 yr 1 P Subsidy m g P m 1 yr 1 Drained 2002 -----2003 32 (59) 0.43 1.95 0.095 12.7 5 2004 25 (46) 0.42 1.4 9 0.073 9.71 2005 16 (19) 0.33 0.7 6 0.03 7 4.99 2006 20 (27) 0.32 0.91 0. 04 5 5.95 2007 20 (33) 0.27 0.7 7 0.038 5.03 200 8 24 (41) 0 .48 1.63 0.080 10.65 2009 13 (16) 0.21 0.40 0.019 2. 61 2010 26 (29) 0.17 0.61 0.0 30 4.01 2011 8 (4) 0.17 0. 20 0.010 1.2 9 Average 20 (30) 0.31 0.97 0.047 6.33 Conserved 2002 32 (49) 0.59 1. 39 0.115 12. 32 2003 0 (0) -0 0 0 2004 22 (65) 0.61 0.99 0.083 8.81 2005 0 (0) -0 0 0 2006 25 (71) 0.59 1.5 2 0.126 13. 51 2007 62 (127) 0.56 2. 57 0.214 2 2.82 200 8 1 (5) 0.64 0.05 0.004 0.4 2 2009 26 (50) 0.56 1.0 7 0.089 9. 48 2010 0 (0) -0 0 0 2011 43 (79) 0.46 1.47 0.123 13.08 Average 22 (41) 0.57 0.91 0.075 8. 04 § Days without rainfall (days with rainfall). The values are based on the median soil elevation from Watts and others (2010). Data not available for all of 2002 at 3AN1W1. Average of PET for the days of interest.
110 CHAPTER 5 LONG TERM IMPLICATIONS OF THE LOCAL CARBON BUDGET Model Overview (Gorham 1991). Carbon turnover in many peatlands has increased significantly due to anthropogenic nutrient increase s (Debusk and Reddy 2003), reduced water tables (Gorham 1991; Oechel and others 1993), peat mining (Gorham 1991), and climate change (Carroll and Crill 1997; Gorham 1991; Maltby and Immirzi 1993). Many of these wetlands have become a source rather than a sink of atmospheric carbon (Gorham 1991; Hooijer and others 2006; Maltby and Immirzi 1993; Oechel and others 1993; Pator and others 2002). Therefore, it has become particularly important to understand the mechanisms of p eat accretion as well as ecosystem t hresholds and feedbacks, as these can act as constraints on restoration activities (Suding and others 2004; Tuittila 2000). Many peatlands have a patterned microtopography where there are distinct and spatially organized higher and lower elevation communities. The higher elevation ; Daoust and Childers 1999; R ietkerk et al. 2004, Eppinga an d others 2008) Two key questions arise in understanding this patterning: First, what causes the divergence in peat elevations between patch types, and second, what causes the scale and configuration of the patches? The former deals with point scale processes that give rise to pattern for mation. The second deals with the landscape scale processes that give rise to pattern formation. While the two are doubtless linked, distribution metrics in peat elevations can act as indicators of
111 landscape stability and early indicators of landscape degr adation (Watts and others 2010). The carbon budget is the main process by which stable patterned topographic variation can occur That is, for peatland pattern to persist require s that each element of the landscape achieve the same long term net ecosystem carbon exchange or accretion rate. The core mechanism for maintenance of two distinct elevation modes lies in the landscape solving for two long term stable peat accretion equilbria: one with high production and high respiration on ridges creating a stabl e vegetative configuration with respect to long term average peat accretion at higher soil elevations; one with low production low respiration in sloughs which creates another at lower soil elevations (Larsen and others 2007; Cohen and others 2011; Heffern an and others in review). These configurations are bi stable in the sense that elevations markedly different from the two equilibrium levels (i.e., deep and shallow configurations with equal peat accretion rates) are, over time, unstable as they accrete mo re quickly or slowly than the long term landscape average, and thus converge over time to one or the other equilibria. Controls on carbon influx and efflux are therefore key components to patch stability in peatlands (examples include Nungesser 2003; Belye a and Baird 2006 ; Larsen and others 2007; Eppinga and others 2008). A necessary next step to understanding patterning and state stability in the patterned ridge slough Everglades lies in enumerating mechanisms behind peat accretion. A number of studies ha ve evaluated components of the C budget, both in situ and in laboratory studies. Parts of the Everglades have clearly become net carbon sources to the atmosphere. For example, Everglades Agricultural Area dissolved
112 organic carbon ( DOC ) is sourced from hist oric peat deposits, whereas DOC is mixed 2007, using 14 C). Chapter 3 demonstrates net annual carbon uptake throughout most of WCA 3 A, excepting the northern sloughs with low er water levels. A study of net carbon exchange in the ridge slough remnants of Everglades National Park demonstrated net annual carbon losses during a year of low rainfall (Jiminez and others 2012). Clearly linking carbon budgets to hydrology will allow u s to make inferences about long term stability of the peat landscape Ecophysiological attributes of the main community types, ridge and slough, reinforce the hypothesis of a discontinuity in patch productivity and hence peat accretion potential Each comm unity is associated with distinct elevatio ns and hydrology (water depths and hydroperiods; David 1996; Jordan 1997; Busch and others 1998; Givnish and others 2008; Zweig and Kitchens 2008; Watts and others 2010). Ridges are dominated by sawgrass, and exist at the highest elevations with the shortest hydroperiods of any of the peat marsh communities (approx imately 357 days, Givnish and others 2008). Physiological limitations to inundation restrict sawgrass to the highest elevation sites (Pezeshki and others 1996; Chabbi and others 2000; Lorenzen and others 2001 Wesiner and Miao 2004;). Highest densities and biomass of sawgrass are associated with the shortest hydroperiods (Toth 1987; Busch and others 1998), although sawgrass density appears resilient to inter annual (6 y ea r) variations in hydrology (Urban and others 1993). Increased annual water depths and hydroperiod reduce the productivity of sawgrass dominated commu nities, however (Childers and others 2006 ; Chapter 3). Sawgrass has long lived leaves (~400 weeks) with a low
113 turnover rate (Davis 1989), as well as a low growth rate, low capacity for phosphorus uptake, and relatively inflexible biomass partitioning (Lorenzen and others 2001). Sawgrass forms monocultures in part because of bel owground growth habit; sawgrass produces many ramets with low internode distances resulting in high densities, particularly as compared to similar species like cattail (Miao 2004). Further, sawgrass expansion lags hydrologic drainage as expansion is primar ily via rhizome ; seed propagation is generally low (Ponzio and others 1995; Lorenzen and others 2000). These physiological responses to environmental conditions induce time lags for sawgrass patch expansion and contraction and ridge prevalence Wet prairi es were relatively rare in the historic ridge slough Everglades, have variable species assemblages, and have largely replaced sloughs in the drained portions of the landscape (Loveless 1959; David 1996; Jordan and others 1997; Daoust and Childers 1999; Chi lders and others 2003; Zweig and Kitchens 2008; Bernhardt and Willard 2009). Many of the species of a wet prairie require a narrow range of hydrology (David 1996, Busch 1998), and are generally found at water depths intermediate between that of ridges and sloughs (Givnish and others 2008, Watts and others 2010), and may thus represent a lagging transitional condition between states. Many wet prairie species are found in the prolific Everglades seedbank (Leeds and others 2002), and appear in sloughs any time water levels are low enough to promote germination. Sloughs occupy a distinct elevation mode from ridges and thus are theorized to be the alternative stable peat accretion configuration to ridges Sloughs are the deepest water communities (Lov eless 1959 ; Jordan and others 1997; Givnish and others 2008; Watts and others 2010), where the dominant species are adapted to increased
114 hydroperiods (Wood and Tanner 1990; David 1996; Busch and others 1998; Zweig and Kitchens 2008). Vegetative productivity of sloug hs is low compared to ridges (Daoust and Childers 1 999; Chapter 3). Modern sloughs formed later than ridges in the historic Everglades the prevalence and composition of which appear driven by regional climate variability (i.e., precipitation; Bernhardt an d Willard 2009) The altered peat elevations and hydrology of the modern Everglades has led to increased species mixtures of deep water and emergent, shallower water vegetation ( i.e., wet prairie; Watts and others 2010; Zweig and Kitchens 2010). Thus, draw ing the distinction between wet prairie and slough community types may only be relevant in the modern Everglades as an indicator for incipient state transitions (David 1996; Jordan 1997; Watts and others 2010). Two Everglades specific peat accretion model s exist that attempt to draw inference of the role local carbon dynamics play in diverging peat elevations. The PeatAccrete 1.0 model (Larsen and others 2007) re create s several aspects of the modern landscape. In PeatAccrete, net peat accre tion is a funct ion of oxygen (primarily but not entirely controlled by water depths) and phosphorus availability in ridges; sloughs modeled to perpetually accrete using a time averaged value. The model adds sophistication using functions such as gravitational erosion (i. e., peat slumping) and drought responses (e.g., sawgrass wilting points, vadose zone water content). The model performs well in the evolution of a stable elevation difference between ridge and slough, but there are key characteristics of the soil elevation the model is not able to re create The model results in 2 sharply divided topographic modes without a variance around those nodes; the conditions for a in the soil elevation distribution between elevation modes is lacking Further, PeatAccrete s uggests that the modern
115 differences in ridge slough topography are unstable (~22 cm in the best conserved landscape). S table ridge slough topography is reached at elevation differences greater than 0.6 m. The robust pattern metrics (e.g., Wu and others), c ommunity structure (e.g., Zweig and Kitchens ), peat accretion potential (e.g., Chapter 3), and topography (e.g., Watts and others 2010) in central WCA 3A suggests this landscape is stable, at the very least in the time span of a management action. Further, the model is not structured to allow both communities to become net carbon losses, limiting inference that can be drawn in application to the modern Everglades. I therefore posit that the simplified carbon accounting of PeatAccrete is insufficiently respo nsive to explore peat accretion dynamics in the modern Everglades. A second peat accretion model for the Everglades takes a closed form approach to demonstrate the spontaneous divergence of elevations between patch forms (Heffernan and others in press) T he model uses long term relationships between local carbon balance, elevation, and discharge competence in (relatively) simple analytical equations that are solved for bi stability. However, the relationships presented are for long term means (timescales o f decades to centuries), precluding the ability to test modeled landscape responses on the scale of a management action. In this study, I co nstruct a simple model describing net carbon accretion in response to hydrologic forcing, parameterized by existing literature values This model presented here address es several limitations of previous Everglades carbon models. First, it enumerates relationships between topography and community by explicitly considering both carbon uptake and losses. Second, the model incorporates state transitions in a way to consider both stochasticity and water depth controls o n
116 community states. Third, the model is parameterized by field and laboratory observations of carbon dynamics, and thus provides a useful context in which to explore which components of Everglades carbon dynamics remain poorly constrained. The model tracks community type, hydrologic conditions, carbon balance, and transitions on annual time steps. The model allows me to test the sufficiency of point scale carbon dynamics for creating observed patterns of soil elevation (e.g., soil distribution convergence u nder drainage and impoundment; bimodality under moderate hydrologic conditions). If the model fails to r e create the observed patterns, I infer that spatial interactions, explicitly omitted from this model, may be necessary. There are several goals in expl oring this model. First, I wish to explore the carbon dynamics presented in Chapters 2 and 3 to evaluate longer time scales than evaluated there, particularly in terms of the theoretical model this work has been based on. In particular, I a m interested in 1) whether or not the relationships suggested in this manuscript are sufficient to give rise to diverging soil elevations between ecosystem s tates (emergence of bimodality); 2) if the relationships described here are sufficient to maintain the characterist ic microtopography found in the Evergla des (maintenance of bimodality); 3) if the model is able to maintain or produce statistical patterns such as variance around modes; and 4) if the modeled landscape s converge from a bimodal soil elevation s to a single mode with hydrologic modification (via both lowered and raised water levels). Results and insights gained from this model provide an improved mechanistic understanding of the role local carbon budgets play in the formation and persistence of ridge and slo ugh formations
117 The Model Initial Conditions This model uses a two simplifying assumptions to model local peat a ccretion. First, this model is non spatial; there are no neighborhood interactions Second, annual carbon budgets are modeled as a function of m edian water depths, which does not take into account drivers such as the depth and length of dry down The baseline model selects a water stage from a normal distribution The statistical property of the stage distribution is based on annual median stages from 1954 to 2012 for 31 WCA 3A marsh wells from DBHYDRO ( http:// www sfwmd.gov/ ; based on data availability ). The interannual variability (as a standard deviation) in median water depths range from 8 .7 to 27.5 cm, with a weighted average standard d eviation of 17.2 cm Consequently the water stage inputs are an input for the long term hydrologic trajectory, but with a standard deviation around the long term value of 17 cm. Solving for local water depths uses water stage a s a landscape average water d epth, such that th e water elevation is the stage plus the sum of the soil elevations, divided by the number of landscape points Local water depth is then simply the water elevation minus the local soil elevation. The landscape commences in one of two ways : either as a flat peat less entirely slough landscape ( approximating pre Everglades landscape initiation) or as a one with a bimodal distribution of soil elevations The bimodal landscape is set so that the first mode centers at 100 cm (sloughs ) and the second sets around 1 25 cm ( ridges; representative of modern conditions across Conserved 1, 2, and ENP sites in Watts and others 2010) with a standard deviation of 6 cm within each community Community proportions were 1:1
118 Model Solutions The model assum es that peat accretion is a l ocal process and thus begins by quantifying ecosystem respiration and productivity. Model notation is described in Table 5 1.The theoretical model (Figure 1 2) for peat accretion equilibria assumes monotonic increases in annual ecosystem respiration with decreasing water depths an assumption corroborated in Chapter 2. Therefore carbon losses are modeled as where a x and b x are the terms f or an exponential relationship of CO 2 respiration with water depths (annual median; h a ), m is the total contribution of losses from DOC and methane and the subscript x denotes community specific parameterization Parameterization for these terms is discus sed in Appendix A Due to differences in autecology among species and communities with respect to hydrology, ridge and slough productivity is modeled separately with an s curve where the sum of the relationship is approximated in Figure 1 2. Observations of variability in community productivity demonstrated in both Tables 3 3 and 3 4 suggests a simplified inverse sigmoid relationship of productivity to local water depth may approximate annual relationships. Community specific productivity ( P x ) is described as where A x and K x are the minimum and maximum productivity values, B x is the growth rate, and m x is the sh ift along the annual water depth axis. Parameterization comes from two sources: First, NEP/GPP relationships from Chapter 3 of this manuscript are used (called hereafter the GPP parameterization ). Because these numbers were generated from one year of measurements and thus
119 may not be representa tive of long term productivity rates a second parameteriza tion was done. The NEP budget was forced to balance to approximate long term peat accretion rates from Bernhardt and Willard (2009; hereafter referred to as the Peat parameterization ). The resultin g model forms are demonstrated in Figure 5 1. Discussions of parameterizat ions are presented in Appendix A In each case, the observations of soil elevation modes at roughly 25 cm apart ( Watts and others 2010 ) are presumed to be evidence that the maximum p eat accretion rate is equal to the difference in modes, and model maximum peat accretion is set to match. It has been argued that the historic elevation modes were farther apart, with differences between community soil elevations as high as ~90 cm (McVoy a nd others 2011) Although it is difficult to evaluate this claim under modern conditions, there should be no practical effect on the behavior of this model for ridges ; modern water depths and hydroperiods for this community are estimated to be approximatel y the same as historic (>100 ybp) The effect for sloughs is dramatically different. For this model, I infer that the modern soil elevation mode coincides with the peak in peat productivity (~22 cm lower than the peak for ridges) Deeper historic sloughs w ould suggest that the landscape was solving for a peak peat accretion at significantly higher water depths, for which we have scant evidence, as there are no landscape points with which to evaluate the carbon balance at such depths. The growth of peat elevation at each time step is the difference between carbon gains and losses ( P x R x ), converted to a height increase using average carbon content (54 % by mass) and bul k density estimates for WCA 3 ( 0.13 g C c m 3 ; Bruland and others 2006). I f the increme nt al change in peat elevation results in a loss of peat greater
120 than the peat profile at that poi nt (i.e., there is no peat to l ose but the net carbon balance is negative), then peat loss is equal to 0. Although the biomass inputs to peat are likely lagged i n the Everglades due to the presence of standing dead vegetation, I assume that at the longer time scales of this model (simulated for 100 and 1000 years) standing dead is integrated over time. As discussed in Chapter 3, fire may play an important role in regulating the carbon balance on ridges. The fire return interval for the marsh system is unknown, with speculated ranges quite large and the interaction between community, hydrology, and fire likely non linear (Lockwood and others 2003). I therefore eval uate model behavior with two fire return intervals: 3 to 15 years and 10 to 30 years, representing a high and low frequency fire return. This model lacks the temporal resolution to make fine scale decisions on the consumption of ecosystem carbon during fir e events, so whe n fire occurs, all ridges burn above ground productivity (estimated at 78% of the net carbon uptake using an average above to below ground biomass of 4.6 for sawgrass; Lorenzen and others 20 01) To approximate the consumption of litter and standing dead from previous years, fire events also net productivity. This is ad mittedly an over simplification as fire size is generally directly related to ridge patch s izes (Gunderson and Snyder 1997). These simplifying assumptions roughly account for biomass losses, but assume peat fires are very rare. changes in root productivity and the potential role fire recovery may play in the carbon dynamics, but do allow an investigation into trends we may observe with fire frequencies.
121 Temporal dynamics of patch state transitions are poorly constrained (although see Zweig and Kitchens (2009) for transitions during impoundment). T he effect s of changes in water level variance, mean water depth, and species composition also remains uncertain T he presence of two communities despite substantial hydrologic changes at the hydrologic end members observations of species assemblages over time (Arme ntano and others 2006; Zwieg and Kitchens 2008; Bernhardt and Willard 2010) suggests substantial lags in vegetative responses to hydrologic modification. Sawgrass is able to form tussocks under deeper water conditions (Snyder and Richards 2005), allowing it to persist even after the underlying substrate has broken apart. Similarly, sloughs and wet prairies have a number of sub community t ypes, with many species present must first compete with the established wet prairie species. These patch responses induce lagged responses to altered water levels Transition occurrences between ridge and slough communities appear to be generally low, and restricted primarily to points with more marginal hydrologic conditions to each community (Zweig and Kitchens 2009). In the model, t he probability of any point transitioning is defined as p(ridge transition) = If h a < w 1 p r.t. = p 1 ; If h a > w 2 p r.t. = p 2 ; If w 1 a 2 and p(slough transition) = If h a < w 3 p s.t. = p 4 ; If h a > w 4 p s.t. = p 3 If w 3 a 4
122 where w 1 and w 3 are the thresholds in water depths where transition probabilities begin to rise from the minimum p 1 &3 and w 2 and w 4 are the water depths where the transition probability becomes the maximum p 2 &4 Accounting for the observations of community shifts responding to multiple years of hydrology, the actual h a used in the simulation transitions is a 5 year averaged annual water depth (the current condition and 4 years pervious) In recognition of the deep uncertainty about when and how quickly community shifts occur, the form and thresholds (i.e., where the transition probabilities change) of the transition probabilities are explored using two model forms (Table 5 2). The minimum probability is non zero to account for stochastic community transition F urther, to maintain stochasticity in transitioning, a uniform probability distribution is generat ed ( z ) and as signed to each landscape point for each time step Where p < z for any point, no transition occurs. Model Results Model Parameters The GPP parameterization had positive annual peat accretion rates from 8 to 40 cm on ridges and 25 to 71 cm in sloughs (Figure 5 2) The maximum peat accretion rates were 2.68 mm yr 1 (ridge) and 0.908 mm yr 1 (slough) The Peat parameterization resulted i n maximum ridge peat accretion of 0.97 m m yr 1 at 16 cm median annual water depth s and was positive from 2 to 39 cm annual median water depth (Figure 5 3 ; average of positive peat accretion values was 0.063 cm yr 1 ) Sloughs had the highest peat accretion of 0.24 mm yr 1 at 41 cm median annual water depth and was positive from 26 to 62 cm (average of positive peat accretion values was 0.015 cm yr 1 ) The range of water depths for positive peat accretion as well as the maximum peat accretion was sensitive to small changes in the model parameters for both GPP and Peat
123 parameterizations particularly K x m x and CO 2 respiratory parameters Figures 5 2&3 demonstrate the changes to peat accretion rates when altering the parameters by 10% (see appendix B for further discussion). The modeled landscape response to perturbations to model parameters after 100 years are pres ented in Appendix B. In general, parameter perturbations to both the GPP and Peat parameterizations resulted in small or negligible changes to the proportion of ridges ( 9 to 10% change in proportion ridge) The sensitivity of the models with respect to so il elevations depended on whether the landscape was flat or bimodal on initiation. In general, the difference between median ridge and slough soil elevations for both models were fairly insensitive to the community growth parameter ( b x ), and more sensitive to K x m x and CO 2 loss parameters. Peat Development Under a variety of water depths, a bimodal peat distribution began to di fferentiate from the uniformly flat conditions for the GPP parameterization ( Figure 5 4 A B ). Further, by the end of 1000 time step s, under all hydrologic conditions slough elevations were rising most notably so under the deepest water levels (Figure 5 4C) The same was true to a lesser degree of the Peat parameterized simulations (Figure 5 4 D F) Under all scenarios, some proportion of the states beca me ridges, with the proportion relative to the water levels (i.e., predominance of ridge states with low water levels, vs. a predominance of slough states with high water levels) Peat Maintenance Und er initiating bimodal simulations the GPP parameterization was able to maintain elevation differences between ridges and sloughs across a broad range of long term mean hydrologic conditions, even though individual points may have
124 undergone collapse and re covery and state transitions (Figure 5 5 A C) However, the ridges converge d to an elevation representative of the maximum peat accretion rate (~16 cm median annual water depths) -something in nature that would not occur because of variability in productio n parameters. Although deeper interannual water depths led to a reduction in ridge proportions, some portion of the ridges maintained positive peat accretion, and thus distinct differences in state elevation modes was maintained (Figure 5 5C) The Peat mode l parameterization was also able to maintain differences between soil elevation modes, although with less separation between modes (Figure 5 5 D F). Low water led to a general decline in soil elevations and a collapse of the differences in state modes mode rate water levels led to a general maintenance of soil elevations, and deeper water led to behavior similar to that of the GPP parameterization The water depths for the state transitions had little influence the behavior of the model. I demonstrate the effect of the water depths in the transitions in Figure 5 6 using the slough landscape with undifferentiated topography Figures 5 6 A& B demonstrate that wi thout any water depth variance the proportion of the l andscape made of ridges traced the transition probabilities. The effect of the higher maximum transition probability for ridges ( p 1 =0.1) is seen in the lagging of the proportion of ridges to the functional form in Figure 5 6 A; reducing p 1 to equal that of p 3 results in a near perfect tracing of the proportion ridge to the functional form. Interannual v ariance in water depths essentially attenuates relationships, demonstrated in Figure 5 6 C and D, rendering the effect of differing transition structures on to tal community abundance negligible
125 The magnitude of the transition probabilities ( p ) also has consequences for the total number of state transitions across all points and years In order to illustrate that proportions and ridge slough elevation difference s can be maintained even with rapid transition shifts, I kept the transitions for water depths to the original configuration, but altered the probabil ities by an order of magnitude demonstrated in Figure 5 7 T he base model of non overlapping water depths was used and the manipulations are described in the figure caption In panels 5 7 A&C the ridge probabilit ies were manipulated (Figure 5 7 B representing the original functio nal forms) and in panel 5 7 D the slough probabilities were manipulated. Across all probability manipulations ridges and sloughs remained in roughly equal proportion and the consequences to the soil elevations was negligible after 100 time steps Since one of the motivating que stions was the recreation of the modern degraded landscapes, I highlight in Figure 5 8 examples from a bimodal landscape subjected to modern hydrologic conditions in drained, conserved, and impounded landscapes (using the GPP parameterization term water depths of 12.8 cm ( landscape ave rage; term water depths of 54.2 cm (s.d. of 22.1; hydrologic parameterization came from Watts and others 2010 ). After 100 time steps, th e slough elevation mode was maintained across all hydrologic conditions (although the Drained landscape was beginning to decline) C ommunity proportions shift towards ridges (drained) and sloughs (impounded) with the extremes in hydrologic conditions, but conditions Although the drained conditions did not lead to a single landscape mode
126 (Figure 8 8A), observing the peat elevation trajectory though time shows the ridges elevations were collapsing towards slo ughs, even as many sloughs were shifting to the ridge state. At 100 time steps, the tendency of the ridge mode to collapse to the highest elevation distribution was maintai ned (Figure 8 conditions best approximated modern conditions, where small proportions of points remained ridge, but an over all spreading of the global elevation distribution and convergence of means occurred (Figure 8 8C). There was a surprising lack of response of the landscape to the application of fire (Figure 5 9). Some ridge points demonstrated transition deflation in e levations in response to fire. Higher fire frequency resulted in a lower proportion of ridge states. W hen fire occurred, r idge el evations with marginal peat accretion underwent stage trans i tions to sloughs and rapidly deflated to the slough mode The result was an expanded separation between soil elevation modes accompanied with the landscape generally bei ng in an accretion tr ajectory. Because of the optimization of the peak productivity in the original parameterization, the long term large separation in peat modes is not stable under any simulations that do not include fire. Discussion Model results sugge st that although point scale processes may not be sufficient to completely explain the development of divergent ridge and slough elevations, they can induce the maintenance of ridge slough elevations. Further, responses to altered water levels can result i n precipitous declines and a rapid collapse of the ridge slough landscape. This collapse is induced solely by water levels, indicating that an interruption in local processes is sufficient to degrade a patterned landscape Indeed, all that is
127 necessary to re create modern degraded conditions is a perturbation to the local positive feedbacks ; i.e., the upstream water volumes that lead to ideal water depths Model outputs further suggest the modes of elevations in a bimodal landscape may remain fairly invariant but individual points of both ridges and sloughs demonstrate multiple and often rapid changes in elevations over time. Although this model was unable to re create the modern landscape for drained conditions it did demonstrate stability of ridge slough elevation differences and long term carbon storage using both sets of model parameters. A s the potential peat accretion i n ridges is more than twice that of sloughs under these parameters, it does suggest that the interactions between water le vels a nd carbon budgets and ultimately stable peat accretion, respond s to much longer time steps than a single year represent s The changes to state proportions and peat elevations under modern hydrologic conditions are consistent with observations of soil ele vations from Watts and others (2010) The results also support the conclusion from that pattern loss was likely due to an interruption of local scale processes. Notably, the drained simulations maintained distinct ridge slough elevation modes, suggesting t he point model of local carbon budgets was insufficient. In this landscape, both muck fires prior to the 1980 s ( Schortemeyer 1980; Zaffke 1983) and peat oxidations are generally blame d for the observed loss of soil elevation bimodality The application of severe fire in conjunction with the drainage may be necessary to more accurately simulate the modern landscape observations; although there has been a loss of soil elevation bimodal ity and a reduced
128 prevalence of ridges, there is no evidence that the landscape itself has collapsed through the loss of peat. Although the GPP parameterization is based on ecosystem carbon dynamics it is notable that the addition of fire amplified the differences in ridge slough peat elevations. The implication is that fire has a role in maintaining peat elevations between ridges and sloughs. The note of caution is in the loss of low accretion ridge s; areas with marginal carbon budgets may be sensitive and lack resilience to fire Since fire is currently used by natural resource management agencies throughout the region, further examination of this process on the long term dynamics of topographic pat terning is warranted. Spatial and temporal interactions in microtopography development have implications for elucidating some of the landscape features this model was unable to re create This model assumed local points were acting independently of other p oints, rather than as an assemblage of interacting patches. This model performed particularly poorly with respect to maintaining a variance around the ridge elevation mode -likely an artifact of the lack of variance in carbon budgets. We currently lack su fficient data to know how structured variance in the carbon dynamics may be. Although doubtless environmental conditions (rainfall, annual cloud cover, temperature ) induce noise to the carbon budgets, variance may also be spatially (one point influencing a neighboring point) or temporally (generation inheritance) correlated. State transitions may also demonstrate spatial structuring. A point in the center of patch is likely buffered from the conditions that lead to state transitions because of the spatial i nteractions in community structure and phenology (e.g., seed fall, root propagation) Spatially explicit models with
129 autocorrelated neighboring responses will be necessar y to ascertain whether patch configuration provides an additional control on the diver gence between and prevalence of ridges and sloughs. In addition to the carbon processes suggested here, patch elevation may adjust to flow velocities and nutrient dynamics (e.g., Ross and others 2006; Lar sen and others 2007). Further, c ommunity prevalence maybe equally dictated by the lateral expansion and contraction of patches. Lateral shifts in peat and vegetation patches may provide the system with resilience to changes in hydrology and facilitate the persistence of a community not currently favored by the hydrologic regime. As the landscape hydrology (in terms of velocity, volume, and depth) is tied to vegetative patterning, small but widespread changes in community elevations under marginal conditions may feedback to reduce the strength of the hydrolog ic control For example, incremental ridge contraction in response to deeper water results in a larger water storage, or conduit, capacity in sloughs, effectively reducing the water depths felt across all ridges (Heffernan and others in press ) Spatial an d temporal variations in peat growth have important implications for elucidating the response to 20 th century hydrologic management and the future of Everglades restoration. Laboratory and biomass studies fail to provide accurate estimates for the carbon e ntering the system to become peat because to date, they fail to consider the interactive roles between hydrology, microform, and vegetation. Moreover, our poor understanding of important patch level carbon dynamics compromise s estimating future responses t o hydrologic management as we have only recently begun to capture peat accretion responses to to hydrology.
130 Although the model presented here helps elucidate some of the patterns that may be expected over larger time scales, it also elucidates unknowns to carbon responses. Important areas of research remain in the ecosystem scale total carbon balance, particularly in terms of predicting what becomes peat (above ground vs. below ground vegetative components) and associated temporal lags from biomass to peat mass Although recent work has elucidated the link between multi year hydrologic conditions and shifts between ridges and sloughs under deeper water conditions (Zweig and Kitchens 2009), we still have little information on the over all probability of stat e shifts, the temporal link between those shifts and hydrologic drainage, and importantly, the link between those state shifts and the local development (or decline) in microtopography
131 Table 5 1. Model symbols and definitions. Symbol Definition Unit a x C efflux from CO 2 respiration when annual water depths = 0 g C m 2 yr 1 b x Decay coefficient for CO 2 respiration m Contribution of methane and DOC to ecosystem respiration g C m 2 yr 1 h e Annual average stage (water elevation) cm h a Annual average water depth at a point cm R Annual C losses g C m 2 yr 1 P x Sigmoid relationship of community productivity with water depths g C m 2 yr 1 A x Parameter representing the minimum productivity of a community g C m 2 yr 1 K x Parameter representing the maximum productivity of a community g C m 2 yr 1 B x Community specific growth rate m x Community specific parameter representing shift along the x axis cm p 1 p 3 Minimum transition probability p 2 p 4 Maximum transition probability w 1 w 3 Minimum h a for step function transition probability cm w 2 w 4 Maximum h a for step function transition probability cm
132 Table 5 2. Parameters for the state transition function s Non overlapping transitions (base model) Overlapping transitions Probabilities Upper w ater d epth ( w 1 ;w 3 ) Lower w ater d epth ( w 2 ;w 4 ) Upper w ater d epth ( w 1 ) Lower w ater d epth ( w 2 ) p 1 p 2 p 3 p 4 Ridge Slough 50 60 25 45 0.1 0.001 0.01 0.001 Ridge Slough 15 30 25 45
133 A B Figure 5 1. Parameterized model forms for carbon uptake and losses for ridge and slough states. A) GPP model parameterization; B) Peat model parameterization. 250 450 650 850 1050 1250 1450 1650 1850 10 0 10 20 30 40 50 60 70 80 (g C m 2 yr 1 ) Annual Median Water Depth (cm) Ridge Loss (g C m 2 yr 1) Slough Loss (g C m 2 yr 1) Ridge Production (g C m 2 yr 1) Slough Production (g C m 2 yr 1) 200 300 400 500 600 700 800 900 10 0 10 20 30 40 50 60 70 (g C m 2 yr 1 ) Annual Median Water Depth (cm)
134 A B Figure 5 2 Change to peat accretion rates due to GPP mode l parameter perturbations. A) P roductivity and B) carbon loss parameter perturbations. Changes to A x and methane paramters not shown because the shifts are very small. 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.1 0.2 0.3 0.4 0.5 10 10 30 50 70 Peat Accretion (cm yr 1 ) Original Model K + K b + b m 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.1 0.2 0.3 0.4 0.5 10 10 30 50 70 Peat Accretion (cm yr 1 ) Water Depth (cm) Original Model CO2 Decay CO2 +Decay CO2 Int CO2 + Int DOC 2% DOC 10%
135 A B Figure 5 3 Change to peat accretion rates due to P eat model paramter petrubations. A) P roductivity and B) carbon loss par ameter perturbations Changes to A x and methane paramters not shown because the shifts are very small. 0.5 0.4 0.3 0.2 0.1 0.0 0.1 0.2 0.3 10 10 30 50 70 Peat Accretion (cm yr 1 ) Original Model K + K b + b m 0.5 0.4 0.3 0.2 0.1 0.0 0.1 0.2 0.3 10 10 30 50 70 Peat Accretion (cm yr 1 ) Water Depth (cm) Original Model CO2 Decay CO2 +Decay CO2 Int CO2 + Int DOC 2% DOC 10%
136 A D B E C F Figure 5 4 Soil elevations after 1000 years commencing from a slough, undefferentiated landscape. A C ) GPP model parameters and D F ) Peat model paramaters. Hydrologic conditions are an interannual average of 10 cm (A, D), 30 cm (B E), and 50 cm (C, F), with a standard deviation of 17 cm.
137 A D B E C F Figure 5 5 Soil elevations after 1000 years commencing from a bimodal landscape A C) GPP model parameters and D F) Peat model paramaters. Hydrologic conditions are an interannual average of 10 cm (A, D), 30 cm (B E), and 50 cm (C, F), with a standard deviation of 20 cm. Inset: the starting landscape.
138 A B C D Figure 5 6 Proportion of points that become ridges after 100 times steps commencing as an undifferentiated slough landscape when transitions are manipulated. A) Overlapping and (B) non overlapping transition ranges across a range of interannual mean water depths. Introducing variance to the water depths (s.d. of 17 cm) smooths out the re lationships: C ) O verlapping transition ranges; D ) non overlapping transition ranges
139 A B C D Figure 5 7 The bimodal landscape after 10 0 time steps with altered transition probabilities ( GPP parameterization ). T ransition probabilities are A) p 1 = 0 .01 & p 2 =0.0001 B) p 1 =0. 1 & p 2 =0.001 C) p 1 =0.01 & p 2 =0.001, and D ) p 3 =0.1 & p 4 =0.001. The starting landscape is the same as in Figure 5 5
140 A B C Figure 5 8 Bimodal soil elevation configurations after 100 time steps of subjections to modern hydrologic conditions. A) D rained, B) conserved, and C) impounded hydrologic conditions. Green represents ridges, blue represents sloughs. The starting landscap e is the same as in Figure 5 5.
141 A B Figure 5 9 The bimodal landscape after 1000 years under differing fire return intervals in the GPP model parameterization. A) 3 15 year fire return intervals and B) 10 30 year fire return intervals. The starting landscape is the same as in Figure 5 5.
142 CHAPTER 6 SYNTHESIS The Everglades is a series of hydrologically interconnected ecosystems. The Everglades landscape has to adjust to short term seasonal variability, pronounced inter annual variabili ty, and long term climate variation in response to global phenomena (Bernhardt and Willard 2009; Obeysekera and others 2011). Uncertainty in these ecological conditions induces ecological resilience, manifested by the extended temporal scales of ridges and sloughs and the presence of many of the pre hydrologic modification features of the landscape. Self organized patchiness may indicate proximity to environmental thresholds (Rietkerk and others 2004), and it is generally thought that the same mechanism th at induces alternative stable states also plays a role in the self organized patterning. In the Everglades, self organized patterning has been variously hypothesized to be due to nutrient concentration (e.g., Ross and others), preferential deposition (Lars en and others), or landscape discharge competence (Cohen and others). While not mutually exclusive, each of these hypotheses suggests a locally positive feedback in peat accretion that is mediated by a landscape control that is scale dependent. Shifts from patterned to unpatterned conditions has been observed across the Everglades in response to nutrient additions and altered hydrology (Wu and others; Watts and others 2010). The aim of this dissertation was to explore the strengths of the local feedbacks as they relate to the third dime nsion of patterning in the Ever glades microtopography, and to explore the linkage between those patch dynamics and the greater landscape hydrology.
143 My research approach was a combination of empirical study and simulation. Th e four studies presented here are united in their examination to the mechanism behind observed patterns in vertical relief across the Everglades. In Chapters 2, 3, and 4, I extrapolated measured fluxes to larger temporal and spatial scales in order to expl ore the implications of the findings to observations of landscape pattern. In Chapter 5 I go a step further, and explore the implications of the model proposed in Chapter 1, in combination of the findings in Chapters 2 and 3, at significantly longer times steps. At this point it becomes appropriate to revisit the questions raised in Chapter 1. 1. Does respiration demonstrate a monotonic and inverse relationship to water depths, as described in Figure 1 2? In Chapter 2, I conclude that water depths represent th e strongest predictor of local water column respiration in a relationship that is indeed is both monotonic and inverse. Ultimately there is some control on the rate of CO 2 efflux that is community based, and is likely attributable to the differences in aqu atic productivity and community phenology between ridges and sloughs. Further, there is a non linearity to the relationship, wherein increased water depths do not have a proportionate increase in flux rates. Extrapolations to extant hydrologic conditions i n the landscape blocks across multiple years demonstrate the role drought and drainage have in dramatically increasing the source of CO 2 to the atmosphere. 2. Do observations of ecosystem productivity corroborate the predictions from the s curve relationship described in Figure 1 2? In Chapter 3 I explore ecosystem CO 2 carbon budgets and their responses to local water depths and community type. The results agreed qualitatively with the prediction that there is a discontinuity in the net ecosystem productivity between ridges and sloughs. The trend in productivity with water depths was sufficiently obfuscated that
144 the research question cannot be answered with a simple yes, however. When comparing annual extrapolations of the net ecosystem productivity to ecosyste m respiration, no equivalence in the potential peat accretion rate was observed. Ridges were persistently net autotrophic across a wide range of hydrologic conditions, with peat accretion rates as much as three times higher than those of sloughs. Further, there was the somewhat surprising finding of a large potential for peat accretion in ridges in the drained landscape block, where shallower peat depths and a loss of microtopography suggested previously this landscape would be in a net peat loss condition. I concluded that either the year of carbon exchange study was an unusual year, where hydrologic conditions allowed an unusually large carbon uptake or the landscape deflation was due to more severe drainage prior to the 1980s and increased water delivery has resulted in a shift towards peat accretion rather than loss. 3. Can productivity induced differences in evapotranspiration reinforce patch differentiation in the ridge slough? Although there are many pattern forming processes exhibited in nature, nutrients have been demonstrated to play a pivotal role in maintaining patch differences in some patterned peatlands. Indeed, the interactive effects of pattern forming processes have implications for th e type and dimension of the pattern formation (Eppinga and others 2009). The hypothesis of resource concentration in ridges induced by the convergent flow of pore water propelled by higher evapotranspiration on ridges gains credence by demonstration of a w ater subsidy to ridges. In Chapter 4 I provide such a demonstration by observing that there is a critical period of time when the diurnal water table in inundated sloughs is nearly exactly matching the diurnal water table of exposed ridges. In other words, the water table is in lock step, so that the amplified response of the
145 water table to evapotranspiration in the soil profile in ridges is matched closely in sloughs. The water table reachi ng this critical time period -ridge soil surfaces are exposed but sloughs remain inundated is pivotal, as this is the only time when we can presume that some portion of the demand of water on ridges is not answered locally; when there is standing water, the porous nature of peat means water would be supplied vertically, and when both communities are dry there is no proof of a horizontal subsidy. The size of the demand is patch size dependent, but results in a large potential movement of water from sloughs to ridges. Patch differentiation would therefore be reinforced by a nutrient flow into ridges, particularly at ridge edges, where increased nutrient availability would spur greater sawgrass growth. Further, this mechanism can only occur when there are elevation differences between ridges and slough. 4. What are the long term implications of considering patches from the point scale of Figure 1 2? In Chapter 5 I explore the longer time scale implications of the model presented in Chapter 1. Parameterization of the model using information from the literature and Chapters 2 and 3 allowed me to demonstrate that 1) much of our understanding of Everglades carbon accretion remains poorly cons trained and 2) modern carbon budgets result in a strong forcing of much of the landscape in favor of ridges, but is balanced at longer time scales. Parameterizing the model so that accretion dynamics conforms to long term peat accretion rates (described by Bernhardt and Willard) demonstrates that very small peat accretion rates end up being very sensitive to internan n ual variability. The results of the comparison of the two parameterizations suggest that a higher peat accretion potential in ridges may be ne cessary in the long term to provide resilience in the landscape to interannual variability.
146 However, much additional work remains to be done to understand the role hydrology has in maintaining the patterns observed in the Everglades. Much of our current k nowledge comes from laboratory analyses, which may not replicate actual environmental conditions, or from studies at sufficiently small time scales that seasonal and annual variability is not captured. Chapter 3 admittedly represents one such study, but is also the first study into whole ecosystem carbon dynamics that encompasses the ridge slough Everglades along a hydrologic gradient The idea that carbon retention in peatlands depends on hydrology (and vise versa) is hardly new. This dissertation does, ho wever, explore that balance as it relates to a system with substantial seasonal and internanual variability in water table. Further, this dissertation suggests that alternative stable states in this setting are the results of an averaging of conditions ove r much longer time periods than the common ecological study. Management implications also arise out of this study. For one, there is apparently a wide range of hydrologic conditions for peat accretion in the Everglades, barring extreme or catastrophic even ts. Thus the prognosis for long term carbon storage in the central portions of the Everglades remains positive. Model results starting from an undifferentiated landscape suggests t he spontaneous divergence in peat elevations in regions that have lost topog raphic variation will likely take significantly longer than the time scale of a management action, however. And although the spatial configuration of patches is a tantalizing avenue for understanding state stability, the research represented here suggests that the dynamics leading to vertical patterns are equally fruitful avenues.
147 APPENDIX A MODEL PARAMETERIZATION Order of magnitude realistic parameter values were taken directly from previous carbon studies, both from the literature and from the studies in this manuscript. Carbon Losses Methane As discussed in Chapter 2, no predictive relationship for CH 4 evolution has been established in the Everglades. In order to approximate reasonable values, I used data from Debusk and Re d dy (2003) laboratory study evaluating CH 4 and CO 2 fluxes in response to altered water levels. Cores C4 & C5 are considered to be from outside of the phosphorus enrichment front in WCA2, so only data from these cores are included. A linear relationship with CO 2 is used, such that CH 4 flux increases with increased CO 2 in relation to lowered water levels (Figure A 1A), and assumes that this relationship remains true across se asons. This was preferred over the measured relationship with water depth, as that regression caused a negative m ethane value for water depths over 6.75cm. Extrapolating the relationship in Figure A 1A alongside the CO 2 extrapolations from 2009 2011 in Figure A 1B forms the relationship of annual methane efflux equaling 21.18 (annual water depth) 0.2. While actua l fluxes may be more stochastic or controlled by variables not yet identified, this does allow the model to account for some methane carbon loss with increases in CH 4 efflux with increased water depths Dissolved Organic Carbon DOC is produced by leaching of plant material, detritus, and peat as well as produced by microbial breakdown of plant matter. Qualls and Richardson (2003) of DOC production in the central portion of the Everglades. They calculate a DOC
148 production by sawgra ss of 5.6% of NAPP (estimated from Table 2 in Qualls and Richardson 2003). Although slough material may produce more DOC, absent a similar relationship I default to using this value as a DOC loss term for all communities. I may therefore be underestimating the loss of carbon through this pathway. Carbon Dioxide In Chapter 2 I develop a relationship between soil/water column and local water depths. By using the extrapolations for the three years of study across the hydrologic gradient and among communities, I can estimate annual soil respiration as a function of annual hydrology (Figure A 1B). GPP Parameters When using the GPP values generated by Chapter 3 for this model, I am not accounting for the respiratory loss due to autotrophic respiration in the resp iration component of the model. In Chapter 3 I compare modeled ecosystem CO 2 efflux to that of modeled soli/water column CO 2 efflux and note that autotrophic CO 2 losses (the difference between net ecosystem and so il respiration) were between 0.3 and 0.56 t imes GPP in ridges (across all models). To include autotrophic respiration here, I added a respiratory loss (R auto ) of 0.48 (averaged ratio) times GPP to approximate the contribution by autotrophy in ridges. In comparison, sloughs ecosystem respiration and R aq were nearly equal, likely due the integration of the water column in my R aq methods. I used the base model for respiration of sloughs, assuming that the annual R aq estimate includes the autotrophic component. Peat Parameters The second parameter set f orces the system to accumulate peat equal to estimates from Bernhardt and Willard (2009), which I do by adjusting the NEP
149 relationship and allowing the respiration model to remain the same. Therefore, the base model is used. Productivity Lacking any data o n the growth rate and x axis shift parameters (B i and M i ), I drew inference from observations of community abundance with long term annual water depths from Watts and others (2010). My reasoning is thus: Reduced community abundance is likely do to reaching thresholds in community tolerance to environmental conditions, which is consistent with the hypothesis that self organized patterned systems exist in a narrow range of environmental conditions, between thresholds for conversion of the landscape to a homog enous condition of one or the other of the two states. The observation of reduced ridge abundance between 30 and 55 cm annual median water levels suggests that within this range ridges are no longer producing enough peat to overcome anoxic stress. Similarl y, increased wet prairie abundance is observed at 20 cm, but rarely at 30 cm or deeper. B i and M i are set to approximate observed relationships with community prevalence and productivity (Table A 1), but are admittedly poorly constrained. The remaining pa rameters are evaluated as two sets: GPP (set 1) uses values from Chapter 3 to estimate growth K i with a minimum value A i to allow some small but negligible amount of deep water productivity. Parameter set 2 ( Peat) forces the landscape carbon balance to ap proximate the long term peat accretion values described in Bernhardt and Willard (2009). GPP Parameter s The maximum GPP, K i was estimated using the global model developed in Chapter 3 (Table A 1). A R was simply set to 2 00 and 25 0 as arbitrary value s to keep
150 productivity from being equal to 0, but allowing it to go very low. The result is a NEP balance that goes positive again at a balance outside of the hydrologic conditions of the Everglades. NEP is positive (uptake) for al l ridge water depths betwe en 8 and 40 cm and positive for al l slough water depths between 25 and 71 cm Maximum p eat accretion for ridges is 0.27 cm yr 1 (~15 16 cm water depth ) and for sloughs is 0. 09 cm yr 1 (~45 cm water depth) Peat Parameter s As noted in Chapter 3, the estima ted accretion rates therein are higher compared to the long term peat accretion rates described in Bernhardt and Willard (2010; Table A 2). Zweig and Kitchens (2008) describe modern communities that are substantially different from older accounts of commun ity assemblages, evidenced by increases in deep water species in the impounded regions. Bernhardt and others (2004) describe inferred from pollen samples. These accounts sugges t that carbon balances prior to anthropogenic modification to hydrologic cycles may have been substantially different from the modern conditions. Reflecting this possibility, a Peat parameter set was developed to approximate the much lower long term peat a ccretion rates described by Bernardt and Willard (2010). As the base model was used for respiration, productivity parameters are more appropriately viewed as net primary production values. The resulting model has positive peat accretion rates for m edian wa ter conditions between 1 and 39 cm yr 1 on ridges and 26 to 62 cm yr 1 in sloughs. The accretion rate for the model in Table A 2 is the average accretion rate across all positive values.
151 A B Figure A 1. Parameterization of respiration values A) CH 4 and B) CO 2 CH 4 relationships are from cores C4 & C5 in Debusk and Reddy (2003). CO 2 model is based on results from Chapter 2 parameters are in Table A 1 CH 4 = 0.0128(CO 2 ) 0.0053 R = 0.45115 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0 1 2 3 4 5 mgCH 4 C cm 2 hr 1 mgCO 2 C cm 2 hr 1 200 300 400 500 600 700 800 900 5 15 35 55 gCO2 C m 2 yr 1 Water Depth (cm) Slough Ridge
152 Table A 1. Default model parameters. All carbon parameters are in gC m 2 yr 1 water parameters are median water depths in cm. Set 1 ( GPP ) Set 2 ( Peat ) Ridge A 200 100 K 1766 900 B 0.08 0.05 M 35 30 R auto 0.48 -Slough A 250 175 K 583 690 B 0.08 0.04 M 60 52 Losses (both GPP and Peat parameterization ) Ridge CO 2 a (intercept) 676.5 b (decay rate) 0.015 Slough CO 2 a (intercept) 662.6 b (decay rate) 0.010 DOC 0.056 CH 4 intercept 21 CH 4 slope 0.2
153 Table A 2. Ridge and slough accretion rates described by Bernardt and Willard (2009), and the average accretion rate for all positive rates for the Peat parameterization Peat accretion (cm yr 1 ) Pre MWP (>1500 BP) MWP LIA (1000 1500BP) LIA (400 100 BP) Peat average accretion Ridge 0.018 0.025 0.06 0.063 Slough 0.011 0.013 0.01 0.015
154 APPENDIX B SENSITIVITY ANALYSIS OF CARBON PARAMETERS Parameters were individually set to to +/ 10% of the original parameter value, except in the cases of x axis shift parameter (m x ; in cm) for the s curve and the DOC as a % of productivity. In these cases, the x axis shift was +/ 5cm for m x and reduced to 2% (0.02) or increased to 10% (0.1) from the base of 0.056 for the DOC parameter. Rather than adjust each of the intercept and slope terms for methane loss, the curve itself was shifted upwards and downwards by 10% by adjusting only the CH 4 intercept. The resultant effect of model perturbations to the difference between (median) ridge and slough elevations and proporti on of ridges of an average of 5 0 model runs for 100 time steps for 1000 points is provided in Figures B 1 ( GPP parameterizati on ) and B 2 ( Peat parameterization ). Two initial landscape configurations are shown. First, the landscape is set to emerge from an initially undifferentiated with no sawgrass (landscape initiation; Figure B 1A & 1B, B 2A & 2B). Secondly, the landscape comm ences as bimodal in peat elevations, with a separation between modes of 25 cm with a standard deviation of each of 6 cm and equal proportions 1:1 ridge and slough. In both cases the interannual wa ter levels are set at 30 17 cm. Parameter adjustments to pe at accretion rates are shown in Figures B 3 (undifferentiated landscape) and B 4 (bimodal landscape).
155 A B C D Figure B 1. Sensitivity of landscape conditions to perturbations to the GPP model parameters relative to baseline conditions. A) and B) initia te from a flat landscale, C) and D) from a bimodal one. 10 5 0 5 10 15 AR KR bR mR R.a AS KS bS mS CO2 int R CO2 decay R CO2 int S CO2 decay S DOC CH4 curve 10 30 years 3 15 years Ridge Productivity Slough Productivity Losses Fire Diference Ridge Slough (cm) 0.04 0.02 0.00 0.02 Proportion Ridge Parameter Decrease Parameter Increase 40 20 0 20 40 AR KR bR mR R.a AS KS bS mS CO2 int R CO2 decay R CO2 int S CO2 decay S DOC CH4 curve 10 30 years 3 15 years Ridge Productivity Slough Productivity Losses Fire 0.02 0.00 0.02 0.04
156 A B C D Figure B 2. Sensitivity of landscape conditions to perturbations to the P eat model parameters relative to baseline conditions. A) and B) initiate from a flat landscale, C and D) from a bimodal one. 6.0 4.0 2.0 0.0 2.0 4.0 6.0 AR KR bR mR AS KS bS mS CO2 int R CO2 decay R CO2 int S CO2 decay S DOC CH4 curve 10 30 years 3 15 years Ridge Productivity Slough Productivity Losses Fire Diference Ridge Slough (cm) 0.02 0.00 0.02 0.04 Proportion Ridge Parameter Decrease Parameter Increase 15.0 10.0 5.0 0.0 5.0 10.0 15.0 AR KR bR mR AS KS bS mS CO2 int R CO2 decay R CO2 int S CO2 decay S DOC CH4 curve 10 30 years 3 15 years Ridge Productivity Slough Productivity Losses Fire 0.01 0.00 0.01 0.02
157 LIST OF REFERENCES Aaby B, Tauber H. 1975. Rates of peat formation in relation to local environment as shown by studies of a raised bog in Denmark. Boreas 4 : 1 17. Adamski JC, Bevier C, Krupa S. 2009. Hydrologic and meteorologic factors affecting groundwater and surface water levels at tree islands in the Everglades of southern Florida. Geological Society of America Abstracts with Progr ams 41:3. Alm J, Talanov A, Saarnio S, Silvola J, Ikkonene E, Aaltonen H, Nykanen H, Martikainen PJ. 1997. Reconstruction of the carbon balance for microsites in a boreal oligotrophic pine fen, Finland. Oecologia 110: 423 431. Alm J, Schulman L, Walden J, Nyknen H, Martikainen PJ, Silvola J. 1999. Carbon balance of a boreal bog during a year with an exceptionally dry summer. Ecology 80:161 174. Amador JA, Jones RD.1997. Response of carbon mineralization to combined changes in soil moisture and carbon phosp horus ratio in a low phosphorus histosol Soil Sci 162:275 282. Armentano TV, Sah JP, Ross MS, Jones DT, Cooley JH, Smith CS. 2006. Rapid responses of vegetation to hydrological changes in Taylor Slough, Everglades National Park, Florida, USA. Hydrobiologi a 569:293 309. Bachoon D, Jones RD. 1992. Potential rates of methanogenesis in sawgrass marshes with peat and marl soils in the Everglades. Soil Biol Biogeochem 29:1157 1164. Belanger TV, Scheidt DJ, Platko Jr II. 1989. Effects of nutrient enrichment on th e Florida Everg lades. Lake Reservoir Manage 5:101 111. Belyea LR, Lancaster J. 2002. Inferring landscape dynamics of bog pools from scaling relationships and spatial patterns. J Ecol 90:223 234. Belyea LR, Baird AJ. 2006. Beyond the limits to peat bog gro wth: Cross scale feedback in peatland development. Ecol Monogr 76:299 322. Bernhardt CE, Willard DA. 2009. Response of the Everglades ridge and slough landscape to climate variability and 20th cent ury water management. Ecol Appl 19:1723 1738. Brady, MA. 20 JO, Pae SE, editors. Peatlands for People, Natural Resources Function and Sustainable Management Proceedings of the International Symposium on Tropical Peatland 22 21 August 2001, Jakarta, Indonesia BPPT and Indonesian Peat Association: Jakarta. p50 51.
158 Bridgham SD, Richardsom CJ. 1992. Mechanisms controlling soil respiration (CO 2 and CH 4 ) in southern peatlands. Soil Biol Biochem 24:1089 1099. Bruland GL, Grunwald S, Osborne TZ, Reddy KR, Newman S. 2006. Spatial distribution of soil properties in Water Conservation Area 3 of the Everglades. Soil Sci Soc Am J 70:1662 1676. Bubier JL, Crill PM, Moore TR, Savage K, Varner RK. 1998. Seasonal pattern s and controls on net ecosystem CO 2 exchange in a boreal peatlan d complex. Global Biogeochem Cy 12: 703 714. Bubier J, Crill P, Mosedale, A. 2002. Net ecosystem CO 2 exchange measured by autochambers during the snow co v ered season at a temperate peatland Hy drol Process 16:3667 3682. Bubier JL, Bhatia G, Moore TR, Roulet NT, Lafleur PM. 2003. Spatial and temporal variability in growing season net ecosystem carbon dioxide exchange at a large peatland in Ontario, Canada. Ecosystems 6: 353 367. Burke RA, Barber TR, Sackett WM. 1988. Methane flux and stable hydrogen and carbon isotope composition of sedimentary methane from the Florida Everglades. Global Biogeochemical Cycles 2:329 340. Burrows EH, Bubier JL, Mosedale, A, Covv W, Crill, PM. 2004. Net ecosystem exc hange of carbon dioxide in temperate poor fen; a comparison of automated and manual chamber techniques. Biogeochemistry 76: 21 45. Busch DE, Loftus WF, Bass Jr, OL. 1998. Long term hydrologic effects on marsh plant community structure in the southern Evergl ades. Wetlands 18:230 241. Carroll P, Crill PM. 1997. Carbon balance of a temperate poor fen. Global Biogeom Cy 11:349 356. Chabbi A, McKee KL, Mendelssohn IA. 2000. Fate of oxygen losses from Typha domingensis (Typhaceae) and Cladium jamaicense (Cyperacea e) and consequences for root metabolism. Am J Bot 87:1081 1090. Chanton JP, Pauly GG, Martens CS. 1988. Carbon isotopic composition of methane in Florida Everglades soils and fractionation during its transport to the troposphere. Global Biogeochem Cy 2:242 252. Childers DL, Doren RF, Jones R, Noe GB, Rugge M, Scinto LJ. 2003. Decadal changes in vegetation and soil phosphorus pattern across the Everglades landscape. J Env Qual 32:344 362. Chimner RA. 2004. Soil respiration rates of tropical peatlands in Micr onesia and Hawaii. Wetlands 24:51 56.
159 Choi J, Harvey J, Ne wlin JT. 2003. Significance of microtopography as a control on s urface water flow in w etlands. Proceedings of the Joint Conference on the Science and Restoration of the Greater Everglades and Florid a Bay Ecosystem April 13 18, 2003, Palm Harbor, Florida. Publication available online at: http:// www.sofia.usgs.gov /geer/2003/posters/ microtopo_cntrl/microcntrl pdf. Clymo RS, Turunen J, Tolonen K. 1998. Carbon accumulation in peatland. Oikos 81:368 388. Cohen MJ, Osborne TZ, Lamsal S, Clark MW. 2009. Regional distribution of soil nutrients: hierarchical soil nutrient mapping for improved ecosystem change detection. SFWMD Final Report. 91 p. Cohen M. Watts D, Heffernan J, Osborne T. 2011. Reciprocal biotic control on hydrology, nutrient gradients, and landform in t he greater Everglades. Crit Rev Environ Sci Technol 41:395 429. Comas X, Slater L, Reeve W. 2008. Seasonal geophysical monitoring of biogenic gases in a northern peatland: Implications for tempora l and spatial variability in free phase gas production rates. J Geophys Res 113:G01012. Comas X, Wright W. 2012. Heterogeneity of biogenic gas ebullition in subtropical peat soils is revealed using time lapse cameras. Water Resour Res 48:W04601. Couwenberg J, Dommain R, Joosten H. 2010. Greenhouse gas fluxes from tropical peatlands in Southeast Asia. Glob Change Biol 16:1715 1732. Daoust RJ, Childers DL. 1998. Quantifying aboveground biomass and estimating net aboveground primary production for wetland macr ophytes using a non destructive ph enometric technique. J Aquat Bot 62: 115 133. Daoust RJ, Childers DL. 1999. Controls on emergent macrophyte composition, abundance, and productivity in freshwater Everglades wetland communities. Wetlands 19:262 275. David PG. 1996. Changes in plant communities relative to hydrologic conditions in the Florida Everglades. Wetlands 16:15 23. Davis SM. 1989. Sawgrass and cattail production in relation to nutrient supply in the Everglades. Sharitz RR, Gibbons SW, editors. F reshw ater Wetlands and Wildlife Oakridge (TN): Office of Scientific and Technical Infor mation, US Department of Energy. p325 341. Deb usk WF, Reddy KR. 1998. Turnover of detrital organic carbon in a nutrient impacted Everglades marsh. Soil Sci Soc Am J 62 : 1460 1468. Debusk WF, Reddy KR. 2003. Nutrient and hydrology effects on soil respiration in a northern Everglades marsh. J Env Qual 32:702 710.
160 Dimitrov DD, Grant RF, Lafluer PM, Humphreys ER. 2010. Modeling the effects of hydroperiods on ecosystem respiration at Mer Bleue bo g. J Geophys Res 115:G04043. Duever MJ, Meeder JF, Meeder LC, McCollom JM. 1997. The climate of South Florida and its role in shaping the Everglades ecosystem. Davis SM, Ogden JC, editors. Everglades: The Ecosystem and Its Restoration. Boca Raton (FL): St. Lucie Press. 225 248p. Duke HR. 1972. Capillary properties of soils influence upon specific yield. T Am Soc Agr Eng 15 :688 779. Eppinga MB, Rietkerk M, Borren W, Lapshina ED, Bleuten W, Wassen, MJ. 2008. Regular surface patterning of peatl ands: confronting theory with field data. Ecosystems 11: 520 536. Eppinga, MB, Rietkerk M, Belyea LR, Nilsson MB, De Ruiter PC, and Wassen MJ. 2010. Resource contrast in patterned peatlands increases along a climatic gradient. Ecology 91 :2344 2355. Ewe SML, Gaiser EE, Childers DL, Iwaniec D, Rivera Monroy VH, Twilley RR. 2006. Spatial and temporal patterns of aboveground net primary productivity (ANPP) along two freshwater estuarine transects in the Florida Coastal Everglades. Hydrobiologia 569:459 474. Fayer MJ, Hillel D. 1986a. Air encapsulation: I. Me asurement in a field soil. Soil Sci Soc Am J 50:568 571. Fayer MJ, Hillel D. 1986b. Air encapsulation: II. Profile water storage and shallow water table fluctuations. Soil Sci Soc Am J 50:572 577. Furukawa Y, Inubushi K, Ali M, Itang AM, Tsuruta H. 2005. Effect of changing groundwater levels caused by land use changes on greenhouse gas fluxes from tropical peat lands. Nutrient Cycling in Agroecosystems 71:81 91. Galloway, D, DR Jones, SE Ingebritsen 1999. L and subsidence in the United States. US Geological Survey Circular 1182. 177 p. Gambolati G, Putti M, Teatini P, Gasparetto G. 2006. Subsidence due to peat oxidation and impact on drainage infrastructures in a farmland catchment south of the Venice Lagoon. Environ Geol 49:814 820. Gerla PJ. 1992. The relationship of water table changes to the capillary fringe, evapotranspiration, and precipitation to intermittent wetlands. Wetlands 12:91 98. German ER. 2000. Regional evaluation of evapotranspiration in the Everglades. USGS Water Resources Investigations Report 00 4217. Washington (DC): US Geological Survey. 48 p.
161 Gesch RW, Reicosky DC, Gilbert RA, Morris DR. 2007. Influence of tillage and plant residue management on respiration of a Florida Everglades. Hi st Soil Til l Res 92 156 166. Givnish TJ, Volin JC, Owen D, Volin VC, Muss JD, Glaser PH. 2007. Vegetation differentiation in the patterned landscape of the central Everglades: importance of local and lands cape drivers. Global Ecol Biogeogr 17:1 19. Glaser PH Chanton JP, Morin P, Rosenberry DO, Siegel DI, Ruud O, Chaser LI, Reeve AS. 2004. Surface deformations as indicators of deep ebullition fluxes in a large northern peatland. Global Biogeochem Cy 18:GP1003. Gleason PJ, Stone P, Davis SM, Ogden JC. 1997. Ag e, origin, and landscape evolution of the Everglades peatland Davis S, Ogden J, editors. Everglades: The Ecosystem and Its Restoration. Boca Raton (FL): St. Lucie Press. p149 197. Gorham E. 1991. Northern peatlands: role in the carbon cycle and probable re sponses to climatic warning Ecol Appl 1:182 195. Grosse W, Bernhard Bchel H, Tiebel H. 1991. Pressurized ventil ation in wetland plants. Aquat Bot 39:89 98. Hagerthey SE, Newman S, Rutchey K, Smith EP, Godin J. 2008. Multiple regime shifts in a subtropical peatland: community specific threshol ds to eutrophication. Ecol Monogr 78: 547 565. Hagerthey SE, Cole JJ, Kilbane D. 2010. Aquatic metabolism in the Everglades: dominance of wate r column heterotrophy. Limnol Oceanogr 55:653 666. Hargreaves GH, Samani ZA. 1982. Estimating potential evapotranspiration. J Irrig Drain E ASCE 108:225 230. Harriss RC, Sebacher DI, Bartlett KB, Bartlett DS, and Crill PM. 1988. Sources of atmospheric methane in the South Florida envi ronment. Global Biogeochem Cy 2:231 243. Hays KB. 2003. Water use by saltceder ( Tamarix sp) and associated vegetation in the Canadian, Colorado, and Pecos Rivers in Texas. Masters Thesis. College Station (TX): Texas A&M University. 132 p. Hef fernan JB, Watts DL, Cohen MJ. I n press Discharge competence as an ecohydrologic mechanism for pattern formation in peatlands: a meta ecosystem model of the Everglades ridge slough landscape. PLoS One. Hirano T, Segah H, Harada T, Limin S, June T, Hirata R, Osaki M. 2007. Carbon dioxide balance of a trop ical peat swamp forest in Kalimantan, Indonesia. Global Change Biology 13:412 425.
162 Hooijer A, Silbius M, W sten H, Page S. 2006. PEAT CO 2 Assessment of CO 2 emissions from drained peatlands in SE Asia. Deft Hydraulic Report Q3943. Hutchinson GL, Mosier AR. 1981. Improved soil cover method for field measurement of ni trous oxide fluxes. Soil Sci Soc Am J 45:311 316. Inglett K, Inglett P, Reddy K, Osborne T. 2011. Temperature sensitivity of greenhouse gas production in wetland soils of different vegetation. Bi ogeochemistry 108:77 90. Inubushi K, Furukawa Y, Hadi A, Purnomo E, Tsuruta H. 2003. Seasonal changes of CO 2 CH 4 and N 2 O fluxes in relation to land use changes in tropical peatlands located in coastal area of South Kalimantan. Chemosphere 52:603 608. Ise T, Dunn AL, Wofsy SC, Moorcroft PR. 2008. High sensitivity of peat decomposition to climate change through water table feedback. Nature Geoscience 1:736 766. Iwaniec DM, Childers DL, Rondeau DN, Madden CJ. 2006. Effects of hydrologic and water quality dri vers on periphyton in the southern Everglades. Hydrobiologia 569:223 235. Jaenicke J, Rieley JO, Mott C, Kimman P, Siegert F. 2008. Determination of the amount of carbon stored in Indonesian peatlands. Geoderma 147:151 158. Jauhiainen J, Takahashi H, Heikk inen JEP, Martikainen PJ, Vasanders H. 2005. Carbon fluxes from a tropical peat swamp forest floor. Glob Change Biol 11:1787 1797. Jauhiainen J, Limin S, Silvennoinen H, Vasander H. 2008. Carbon dioxide and methane fluxes in drained tropical peat before an d after hydrological restoration. Ecology 89:3503 3514. Jimenez KL, Starr G Staudhammer CL, Shedlbauer JL, Loescher HW, Malone SL, Obervauer, SF. 2012. Carbon dioxide exchange rates from short and long hydroperiod Everglades freshwater marsh. J Geophys Re s 117:G4. Jordan F, Jelks HL, Kitchens WM. 1997. Habitat structure and plant community composition in a northern Everglades wetland landscape. Wetlands 17:275 283. King GM, Roslev P, Skovgaard H. 1990. Distribution and rates of methane oxidation in sedimen ts of the Florida Everglades. Applied Env iron Microb 56:2902 2911. Kushlan JA. 1990. Freshwater Marshes. Myers RL, Ewel JJ, editors. Ecosystem of Florida. Orlando (FL): University of Central Florida Press. 325 362p. Laine A, Scottocornola M, Kiely G, Byrne KA, Wilson D, Tuittila, ES. 2006. Estimating net ecosystem exchange in a patterned ecosystem: Ex ample from blanket bog. Agr Forest Meteor 18:231 246.
163 Laing HE. 1940. Respiration of the rhizomes of Nuphar advenum a nd other water plants. Am J Bot 27:574 581 Larsen LG, Harvey JW, Crimaldi JP. 2007 A delicate balance: ecohydrological feedbacks governing landscape morphology in a lotic peatland. Ecol Monogr 77:591 614. Larsen LG, Harvey JW, Noe GB, Crimaldi JP. 2009. Predicting organic floc transport dynamics in shallow aquatic ecosystems: insights from the field, the laboratory, and numerical modeling. Water Resour Res 45:W01411. Larsen LG, Harvey JW. 2010. How vegetation and sediment transport feedbacks drive landscape change in the Everglades and wetlands wo rldwide. Am Nat 176:E66 E79. Leeds JA, Smith SM, Garrett PB. 2002. Seedbanks and their potential role in the vegetation dynamics of a northern Everglades marsh. Florida Scientist 65:16 34. Leonard L, Croft A, Childers DL, Mitchell Bruker S, Solo Gabriele H Ross MS. 2006. Characteristics of surface water flows in the ridge and slough landscape of Everglades National Park: implications for particulate transport. Hydrobiologia 569:5 22. Lewis CG. 2005. Linkages among vegetative substrate quality, biomass prod uction, and decomposition in maintaining Everglades ridge and slough vegetative communities. Gainesville (FL): Thesis, University of Florida. 58 p. Light SS, Dineen JW. 1997. Water control in the Everglades: a historical perspective. Davis SM, Ogden JC, editors. Everglades: The Ecosystem and Its Restoration. Boca Raton (FL): St. Lucie Press. 47 84p. Limpens J, Berendse F, Blodau C, Canadell GG, Freeman C, Holden J, Roulet N, Rydin H, Schaepman Strub G. 2008. Peatlands and the carbon cycle: from local proc esses to global implications A synthesis. Biogeosciences 5:1475 1491. Lockwood JL, Ross MS, Sah JP. 2003. Smoke on the water: the interplay of fire and water flow on E verglades restoration. Front Ecol Env 1:462 468. Lorenzen B, Brix H, McKee KL, Mendelss ohn IA, Miao SL. 2000. Seed germination of two Everglades species, Cladium jamaincense and Typha domingensis Aquat Bot 66:169 180. Lorenzen B, Brix H, Mendelssohn IA, McKee KL, Miao SL. 2001. Growth, biomass allocation and nutrient use efficiency in Cladi um jamaicense and Typha domingensis as affected by phosphorus and oxygen availability. Aquat Bot 70:117 133. Loveless CM. 1959. A study of the vegetation in the Florida Everglades. Ecology 40:1 9.
164 Maltby E, Immirzi P. 1993. Carbon dynamics in peatlands and other wetland soils: regional and global perspectives Chem o sphere 27:999 1023. McCormick PV, Stevenson RJ. 1998. Periphyton as a tool for ecological assessment and management in the Florida Everglades. J Phycol 34:726 733. McGill R, Tukeu JW, Larsen WA. 197 8. Variations of box plots. Am Stat 32:12 16. McLaughlin DL, Cohen MJ. 2011. Thermal artifacts in measurements of fine scale water level variation Water Resour Res 47 :W09601. McVoy CW, Said WP, Obeysekera J, VanArman JA, Dreschel TW. 2011. Landscapes an d hydrology of the predrainage Everglades. Gainesville (FL): University Press of Florida. 342 p. Melling L, Hatano R, Gog KJ. 2005. Soil CO 2 flux from three ecosystems in tropical peatland of Sarawak, Malaysia. Tellus B 57:1 11. Miao SL. Rhizome growth an d nutrient resorption: mechanisms underlying the replacement of two clonal species in Florida Everglades. Aquat Bot 78:55 66. Nachabe M, Shah N, Ross M, Vomacka J. 2005. Evapotranspiration of two vegetation covers in a shallow water table environment. Soil Sci Soc Am J 69 :492 499. Nair SK, Montoya AM, Wilsnack MM, Zamorano LM, Obeysekera J, Switanek M, Herr,J, Restrepo, JI. 2001. South Palm Beach County ground water flow model. Wast Palm Beach (FL): South Florida Water management District, Hydrologic System s Modeling Division. 43 p. Noe GB, Childers DL, Jones RD. 2001. Phosphorus biogeochemistry and the impact of phosphorus enrichment: why is the Everglades so unique? Ecosystems 4:603 624. Nungesser MK. 2011. Reading the landscape: temporal and spatial chang es in a patterned peatland. Wetl Ecol Manag19:475 493. Nykanen H, Heikkinen JEP, Pirinen L, Tiilikainen K Martikainen, PJ. 2003. Annual CO 2 exchange and CH 4 fluxes on a subarctic palsa mire during climatically different years. Global Biogeochem Cy 17:1 18. Oechel WC, Hastings SJ, Vourlitis G, Jenkins M, Riechers G, Grulke N. 1993. Recent change of Arctic tundra ecosystems from a net carbon dioxide sink to a source. Nature 371:520 523. Osborne TZ, Inglett PW, Reddy KR. 2007. The use of senescent plant biomas s to investigate relationships between point particulate and dissolved organic matter in a wetland ecosystem. Aquat Bo t 86:53 61.
165 Page SE, Siegert F, Rieley LO, Boehm HDV, Jaya A, Limin S. 2002. The amount of carbon released from peat and forest fires in I ndonesia during 1997 Nature 420:61 65. Page SE, Wst RA, Weiss D, Rieley JO, Shotyk W, Limin SH. 2004. A record of Late Pleistocene and Holocene carbon accumulation and climate change from an equatorial peat bog (Kalimantan, Indonesia): implications for pa st, present and fu ture carbon dynamics. J Quaternary Sc 19:625 635. Pastor J, Peckham B, Bridgham S, Weltzin J, Chen J. 2002. Plant community dynamics, nutrient cycling, and alternative stable equilibria in peatlands. Am Nat 160:553 568. Pezeshki SR, DeLau ne RD, Kludze HK, Choi HS. 1996. Photosynthetic and growth responses of cattail ( Typha domingensis ) and sawgrass ( Cladium jamaicense ) to soil redox conditions. Aquat Bot 54:25 35. Ponzio KJ, Miller SJ, Lee MA. 1995. Germination of sawgrass, Cladium jamaice nse Crantz, under varying hydrologic conditions. Aquat Bot 51 115 120. Qual l s RG, Richardson CJ. 2003. Factors controlling concentration, export, and decomposition of dissolved organic nutrients in the Everglades of Florida. Biogeochemistry 62: 197 229. Quinn JF, Dunham AE. 1983. On hypothesis testing in ecology and evolution. Am Nat 122:602 617. Raich JW, Schlesinger WH. 1992. The global carbon dioxide flux in soil respiration and its relationship to vegetation and climate. Tellus B 44:81 99. Rietkerk M, Dekker SC, Wassen MJ, Verkroost AW, Bierkens MF. 2004. A putative mechan ism for bog patterning. Am Nat 163: 699 708. Rietkerk M, van de Koppel J. 2008. Regular pattern formatio n in real ecosystems. Trends Ecol Evol 23:169 175. Ross MS, Mitchell Bruker S, Sah JP, Stothoff S, Ruiz PL, Reed DL, Jayachandran K, Coultas CL. 2006. Interaction of hydrology and nutrient limitation in the ridge and slough landscape of the southern Everglades. Hydrobiologia 569 :37 59. Roughgarden J. 1983. Competition and t heory in community ecology. Am Nat 122:583 601. Scheffer M. 1999. Searching explanations of nature in the mirror world of math. Conserv Ecol 3:11.
166 Schipper LA, McLeod M. 2006. Subsidence rates and carbon loss in peat soils following conversion to pasture i n the Waikato R egion, New Zealand. Soil Use Manage 18:91 93. Schlesinger WH. 1977. Carbon balanc e in terrestrial detritus Annu Rev Ecol Syst 8:51 81. Schortemeyer JL. 1980. An evaluation of water management practices for optimum wildlife benefits in Conser vation Area 3A. Ft. Lauderdale (FL): Florida Game and Fresh Water Fish Commission. 172 p. Sience Coordination Team (SCT). 2003. The role of flow in the Everglades ridge and slough landscape Miami (FL): Science Coordinating Team, South Florida Ecosystem Re storati on Working Group 62 p. Sinden Hempstead M, Killingbeck KT. 1996. Influences of water depth and substrate nitrogen on leaf surface area and maximum bed extension in Nymphaea odorata Aquat Bot 53:151 162. Smil V. 1985. C arbon, nitrogen, sulfur: Huma n I nterference in Grand Biospheric Cycles. New York (NY): Plenum Press. 459 p. Snyder JM, Richards JH. 2005. Floral phenology and compatibility of sawgrass, Cladium jamaicense (Cyperaceae). Am J Bot 92:736 743. Sophocleous M. 1985. The role of specific yi eld in ground water recharge estimations: a numerical study. Ground Water 23 :52 58. Sorensen KW. 1993. Indonesian peat swamp forests and their role as a carbon sink. Chemosphere 27:1065 1082. Stern J, Wang Y, Gu B, Newman J. 2007. Distribution and turnover of carbon in natural and constructed wetlands in the Florida Everglades. Appl Geochem 22:1936 1948. Strack M, Waddington JM, Tuittila ES. 2004. Effect of water table drawdown on northern peatland methane dynamics: Implications for clima te change. Global B iogeochem Cy 18:GB4003. Suding KN, Gross KL, Houseman GR. Alternative states and positive feedbacks in restoration ecology. Trends Ecol Evol 19:46 53. Sulman BN, Desai AR, Schroeder NM, Ricciuto D, Barr A, Richardson AD, Flanagan LB, Lafleur PM, Tian H, Chen G, Grant RF, Poulter B, Verbeeck H, Ciais P, Ringeval B, Baker IT, Schaefer K, Luo Y, Weng E. 2012. Impact of hydrological variations on modeling of peatland CO 2 fluxes: Results from the North American Carbon Program site synthesis. J Geophys Res 117: G01031.
167 Sumner DM. 2007. Effects of capillarity and microtopography on wetland specific yield. Wetlands 27 693 701. Suzuki S, Ishida T, Nagano T, Waijaroen S. 1999. Influences of deforestation on carbon balance in a natural tropical peat swamp fo rest in T hailand. Environ C Biol 37:115:128. Thornley MN, Johnson IR. 1990. Plant and crop modeling; a mathematical approach to plant and crop physiology. Oxford: Clarendon. 669 p. Toth LA. 1987. Effects of hydrologic regimes on lifetime production and nutrient dyn amics of sawgrass. South Florida Water Management District Tech Pub 87 6. West Palm Beach: FL 38 p. Troxler PR, van der Valk AG, Newman S, Coronado CA, Troxler Gann TG, Childers DL, Orem WH, Sklar FH. 2009. Heterogeneity of phosphorus distribution in a pa tterned landscape, the F lorida Everglades. Plant Ecol 200:83 90. Tuittila E, Komulainen VM, Vasander H, Laine J. 1999. Restored cut away peatlands as a sink for atmospheric CO 2 Oecologia 120:563 574. Updegraff K, Bridgham SD, Pastor J, Wieshampel P, Harth C. 2001. Response of CO 2 and CH 4 emissions from peatlands to warming and wate r table manipulation. Ecol Appl 5:311 326. Urban NH, Davis SM, Aumen NG.1993. Fluctuations in sawgrass and cattail densities in Everglades Water Conservation Area 2A under varyin g nutrient, hydrologic and fire regimes. Aquat Bot 46:203 223. Vaithiyanathan P, Richardson CJ. 1998. Biogeochemical characteristics of the Everglades sloughs. J Environ Qual 27:1439 1450. Vasander H, Jauhiainen J. 2001. Measuring of CO 2 emissions in tropi cal peatlands. Trop Peatlands 1:16 19. Wagner SW, Reicosky DC. 1992 Closed chamber effects of leaf temperature, canopy photosynthesis, and evapotranspiration. Agron J 84: 731 738. Watters JR, Stanley EH. 2006. Stream channels in peatlands: the roll of biol ogical process in controlling channel form. Geomorphology 89 :97 110. Wang C, Jang J, Zhang Q. 2006. Soil respiration in six te mperate forests in China. Glob Change Biol 12:2103 2114. Watts DL, Cohen MJ, Heffernan JB, Osborne TZ. 2010. Hydrologic modification and the loss of self organized feedbacks in the ridge slough mosaic of the Everglades. Ecosystems 13:813 827.
168 Weaver HA, Speir WH. 1960. Applying basic soil water data to water control problems on Everglades peaty muck. USDA Agr icultural Research Service Washington (DC): Government Printing Office. 15 p. Weisner SEB, Miao SL. 2004. Use of morphological variability in Cladium jamaicense and Typha domingensis to understand vegetation changes in an Everglades marsh. Aquat Bot 78:319 335. Wetzel PR, van der Valk AG, Newman S Gawlik DE, Gann TGT, Coronado Molina CA, Childers DL, Sklar FH. 2005. Maintaining tree islands in the Florida Eveglades: nutrient redistri bution is the key. Front Ecol Env 3:370 376. Wetzel PR, Sk lar FH, Coronado CA, Troxler TG, Krupa SL, Sullivan PL, Ewe S, Price RM, Newman S, Orem WH. 2011. Biogeochemical processes on tree islands in the Greater Everglades: initiating a new paradigm. Crit Rev Environ Sci Technol 41:670 701. White WN. 1932. A meth od of estimating ground water supplies based on discharge by plants and evaporation from soil. US Geological Survey Water Supply Paper 659 A. Washington (DC): United States Government Printing Office.105 p. Wigley TML, Schimel DS. 2000. The Carbon Cycle. C ambridge University Press. 292 p. Wilsnack MM, Welter DE, Montoya AM, Restrep JI, Obeysekera J. 2001. Simulating flow in regional wetlands with the MODFLOW wetlands package. J Am Water Res As 37:655 674. Wsten JHM, Ismail AB, van Wijk ALM. 1997. Peat subs idence and its practical implications: a case study in Malaysia. Geoderma 78:25 36. Wood JM, Tanner GW. 1990. Graminoid community composition and structure within four Everglades management areas. Wetlands 10:127 149 Wu Y, Sklar FH, Rutchey K. 1997. Analy sis and simulations of fragmentation patter ns in the Everglades. Ecol Appl 7: 268 276. Wu Y, Wang N, Rutchey K. 2006. An analysis of the spatial complexity of ridge and slough patterns in the Everglades ecosystem. Ecol Compl 3:183 192. Xu M, Qi Y. 2001. So il surface CO 2 efflux and its spatial and temporal variations in a young ponderosa pine plantation in northern California. Glo b Change Biol 7:667 677. Yuan J, Delesantro J, Casey S, Paz S, Watts D, Cohen J. 2012. Analysis of patch geometry characteristics in the ridge slough patterned landscape in the Everglades. Proceedings of the 9 th INTECOL International Wetlands Conference Wetlands in a Complex World June 3 8, 2012 Orlando Florida. Publication available online at:
169 http://www.conference.ifas.ufl.edu/intecol/presentations/Posters/Intercol_patch_g eometry_final.pdf Zaffke M. 1983. Plant communities of Water Conservation Area 3A: base line documentation prior to the operation of S 339 and S 340. Technical Memorandum DRE 164. West Palm Beach (FL): South Florida Water Management District. 61 p. Zweig CL, Kitchens WM. 2008. Effects of landscape gradients on wetland vegetation communities: Information for large scale restoration. Wetlands 28:1086 1096. Zweig CL, Kitchens WM. 2009. Multi state succession in wetlands: a novel use of state and transition models. Ecology 90:1900 1909.
170 BIOGRAPHICAL SKETCH Danielle Watts developed a passion for natural systems while growing up as a Keys. As an undergraduate at the University of Florida, Watts majored in wildlife ecology and conservation while learning that not all plants should be petted and the unobserved parts of an ecosystem are sometimes the most interesting. Watts accomplished her dream of visiting Africa and living abroad by serving as a Peace Corps Volunteer in Guinea, West Africa. While sitting on top of a plateau during the sub Saharan Harmattan season, Watts realized she had a passion for the low, wet areas of the world. Ch asing after a desire to study these systems, she completed a M.S. degree in interdisciplinary ecology at the University of Florida investigating landscape pattern dynamics in the Everglades. Those studies led to the research accomplished in this dissertati on.