1 EFFECT OF SAMPLING METHOD AND TIME INTERVAL ON AN INDEX OF WETLAND CONDITION By ERICA CATHERINE HERNANDEZ A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2014
2 Â© 2014 Erica Catherine Hernandez
3 To my precious loved ones, Severine, Leonardo, Alison and Parker
4 ACKNOWLEDGMENTS I would like to thank my advisor, Dr. Mark Brown, and my committee members Dr. Kelly Chinners Reiss and Dr. Mark Clark for their thoughtful encouragement, insight and time in helping me on this journey. This research was supported by a grant from the Florida Department of Envi ronmental Protection (FDEP). Additionally I would like to acknowledge and thank my colleagues at the H.T. Odum Center for Wetlands for their unwavering dedication in helping to collect, classify and enter in data for this study Carrie Boyd, Valerie Burke tt, Robert Compton, Anthony Davanzo, Jenet Dooley, Alison Hernandez, Torren Hoyord, Kelly Reiss, and Sean Sharp. Finally thank you to the landowners and ma nagers whom granted permission to study their wetlands.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ .......... 8 LIST OF ABBREVIATIONS ................................ ................................ ............................. 9 ABSTRACT ................................ ................................ ................................ ................... 10 CHAPTER 1 REVIEW OF LITERATURE ................................ ................................ .................... 12 Introduction ................................ ................................ ................................ ............. 12 Statement of the Problem ................................ ................................ ....................... 12 Indices of Biotic Integrity (IBIs) ................................ ................................ ............... 13 IBI Response to Natural and Anthropogenic Variation ................................ ............ 14 Fish Based IBI ................................ ................................ ................................ .. 15 Vegetation Based IBI ................................ ................................ ........................ 17 Macrophytes as Indicators of Wetland Condition ................................ .................... 19 Sampling Methodology Effect on Interpretation of Ecological Phenomena ............. 21 Vegetation Patterns in Wetland s ................................ ................................ ............. 23 Study Objectives ................................ ................................ ................................ ..... 25 2 EFFECT OF TIME ON CONSISTENT AND REPEATABLE MACROPHYTE INDEX FOR WETLAND CONDITION ................................ ................................ ..... 27 Introduction ................................ ................................ ................................ ............. 27 Statement of the Problem ................................ ................................ ................. 28 Review of Literature ................................ ................................ ......................... 29 Study Objectives ................................ ................................ .............................. 30 Methods ................................ ................................ ................................ .................. 31 Site Description ................................ ................................ ................................ 31 Macrophyte FWCI ................................ ................................ ............................ 33 Sampling Description ................................ ................................ ....................... 35 Statistical Analyses ................................ ................................ .......................... 36 Results ................................ ................................ ................................ .................... 36 Effect of Time on Macrophyte FWCI ................................ ................................ 37 Community Similarity ................................ ................................ ........................ 37 Discussion ................................ ................................ ................................ .............. 38 Natural Variation in Wetland Communities ................................ ....................... 39 Differences in Spec ies Composition ................................ ................................ . 40 Differences in FWCI Score ................................ ................................ ............... 41
6 Study Limitations ................................ ................................ .............................. 43 Conclusion ................................ ................................ ................................ .............. 43 3 SAMPLING DESIGN EFFECT ON INTERPRETATION OF WETLAND CONDITION ................................ ................................ ................................ ............ 55 Introduction ................................ ................................ ................................ ............. 55 Statement of the Problem ................................ ................................ ................. 55 Review of Literature ................................ ................................ ......................... 56 Sample design and scale ................................ ................................ ........... 56 Temporal and spatial patterns in wetland vegetation ................................ . 57 Interpreting wetland condition ................................ ................................ .... 58 Study Objectives ................................ ................................ .............................. 59 Methods ................................ ................................ ................................ .................. 59 Site Description ................................ ................................ ................................ 59 Sampling Methodology ................................ ................................ ..................... 60 Community Structure ................................ ................................ ........................ 62 Wetland Condition ................................ ................................ ............................ 63 Results ................................ ................................ ................................ .................... 64 Wetland Condition Comparison ................................ ................................ ........ 65 Proportion of Indicator Species for Calculating Metrics ................................ .... 65 Community Structure ................................ ................................ ........................ 66 Diversity statistics ................................ ................................ ...................... 66 Proportion of species by growth habit ................................ ........................ 66 Shared species ................................ ................................ .......................... 67 Discussion ................................ ................................ ................................ .............. 68 Examples of Potential Variabi lity ................................ ................................ ...... 69 Study Limitations ................................ ................................ .............................. 71 Conclusion ................................ ................................ ................................ .............. 71 4 SUMMARY AND CONCLUS IONS ................................ ................................ .......... 84 LIST OF REFERENCES ................................ ................................ ............................... 90 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 97
7 LIST OF TABLES Table page 2 1 Metrics included in final Florida Wetland Condition Index calculation by wetland type. ................................ ................................ ................................ ...... 45 2 2 Individual metric valu es and final Florida Wetland Condition Index (FWCI) scores for herbaceous depression (HD) in 2011 and 2012. ............................... 46 2 3 Individual metric values and final Florida Wetland Condition Index (FWCI) sc ores for forested depression (FD) in 2011 and 2012. ................................ ...... 46 2 4 Individual metric values and final Florida Wetland Condition Index (FWCI) s cores for forested strand or floodplain (FSF) in 2011 and 2012. ....................... 47 2 5 Pearson coefficient (r ) for individual metrics and community diversity measures. ................................ ................................ ................................ ........... 47 2 6 Proportion of vegetation b y growth habit for each wetland type each year. ........ 48 3 1 Metrics included in final Florida Wetland Condition Index calculation by wetland type. ................................ ................................ ................................ ...... 73 3 2 with data from different sampling methods. ................................ ........................ 73 3 3 Individual metric values and final Florida W etland Condition Index (FWCI) herbaceous depression wetlands. ................................ ................................ ...... 74 3 4 Individual metric values and final Florida Wetland Condition Index (FWCI) forested depression wetlands. ................................ ................................ ............ 75 3 5 Individual metric values and final Florida Wetland Condition Index (FWCI) forested strand or floodplain wetlands. ................................ ............................... 76 3 6 Regression and Stud for individual metrics ................................ ........ 77 3 7 measures ................................ ................................ ................................ ............ 77
8 LIST OF FIGURES Figure page 2 1 Geographic location of wetlands, herbaceous depression (HD), forested depression (FD) and forested strand or floodplain (FSF). ................................ .. 49 2 2 Assessment area (0.5 ha) and sampling layout. ................................ ................. 50 2 3 Macrophyte FWCI scores in 2011 vs. 2012 linear regression ............................. 51 2 4 Percent cover by height class for three wetland types ................................ ........ 52 2 5 Percent shared species at a wetland between years by wetland type ................ 53 2 6 Percent of species co occurring at a cumulative number of wetlands in 2011 and 2012 ................................ ................................ ................................ ............ 54 3 1 Locations of wetlands in study herbaceous depression (HD) , forested depression (FD), for ested strand floodplain (FSF) ................................ .............. 78 3 2 Box plot of wetland size for three wetland types ................................ ................. 79 3 3 Layout of sampling area within th e same Assessment Area (AA). ..................... 80 3 4 Condition scores compared by method and year ................................ ............... 81 3 5 Proportion of species by growth habitat fo r each wetland type and year. ......... 82 3 6 Percent shared species detected by both methods within the same year, or within method . ................................ ................................ ................................ .... 83
9 LIST OF ABBREVIATIONS AA As sessment Area CC Coefficients of Conservatism FD Forested depression wetland FQAI Floristic Quality Assessment Index FSF Forested strand or floodplain wetland FWCI Florida Wetland Condition Index (transect) h a Hectare HD Herbaceous depression wetland IB I Index of Biological Integrity m meter NWCA National Wetland Condition Assessment 2011 (quadrat)
10 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the D egree of Master of Science EFFECT OF SAMPLING METHOD AND TIME INTERVAL ON AN INDEX OF WETLAND CONDITION By Erica Catherine Hernandez August 2014 Chair: Mark Brown Major: Interdisciplinary E cology The vegetation portion of the Florida Wetland Conditio n Index (FWCI) provided consistent and repeatable measures of condition at eighteen wetlands sampled in two consecutive growing seasons. The sample wetlands reflected a gradient of adjacent land use from non impacted reference areas to wetlands imbedded w ithin silviculture, cattle pasture and residential areas. Wetlands were described as herbaceous depression (n=6), forested depression (n=5) and forested strand or floodplain wetlands (n=7) and represented different states of succession. Even though the wet lands were unique from one another and occurred across a large geographic area in Florida the FWCI results calculated for all the wetlands were representative of adjacent land use impacts and not sensitive to natural variation. During the duration of thi s study changes in weather from drought to tropical storm conditions as well as management activities such as fire and herbivory impacted wetlands. These effects were apparent in the change of species composition between sampling periods; 23% 56% of spec ies were different when resampled . Even though composition changed, the proportion of indicators remained consistent; the resulting condition scores suggested a one to one
11 relationship between sampling periods. Different sampling methods were also applied to determine if area and arrangement of sample units could alter the calculation of wetland condition score. This study did not detect a difference in condition calculations between sampling methods but recommended caution in configuring sample design due to topographic complexity and natural gradients present in wetlands.
12 CHAPTER 1 REVIEW OF LITERATURE Introduction Developing watersheds and land use impacts have downstream consequences that can be measured in changes to the biological assemblages represe nted in wetland communities (Karr et al. 1986). Even though the United States has a policy of no net loss, measuring degradation where a wetland no longer resembles historic condition can be difficult to quantify. Plants are the foundation of trophic inter actions in wetlands and provide indirect connections to other organisms that rely on wetlands (Mitsch and Gosselink 2000); monitoring plants can be a good proxy for determining wetland condition (Stapanian 2013). In trying to determine if a wetland is impa ired and no longer reflects an unimpaired system, methods to determine condition must collect a representative sample and be repeatable and consistent over time. Statement of the Problem In 1981, Karr et al. introduced an index of biologic integrity (IBI) , a method to identify species that indicated impacts to aquatic ecosystems or that were indicative of system integrity by comparisons to a reference standard . The authors encouraged the scientific community to develop metrics based on biologic response to ecosystem impacts and land use change. Since then hundreds of indexes and metrics have been developed around the world to assess ecosystem integrity (Ruaro and Gubiani 2013). Karr et al. (1986) proposed implementing recommendations made by H erricks and S chaeffer (1985) to evaluate the strength of an index for a monitoring program . M any ind ic es abide by the first two recommendations (Ruaro and Gubiani 2013 ) , measurements must be biological and representative of troph ic levels not directly
13 measured (i.e. di atoms, macroinvertebrates, macrophytes, etc.). Many indices also meet the third recommendation, measurements must be sensitive to conditions being monitored, indices demonstrate this by responding to land use change ( U.S. EPA 2002) . The last three recomme ndations are relat ed to confidence in the measurements themselves . T he measured responses must be within a range suitable for intended application, measurements should be reproducible within acceptable limits of data collected over space and time and measu rements should have low variability. Karr et al. (1986) also encouraged practitioners to test indices by expanding sampling areas to sites not used for indices development and expand application to include different gradients of disturbance. In sampling d ata for application of biological indices, coarseness of scale, size of sample area and questioning if the sample data is adequate for describing the system is also important for determining the reliability of measurements applied to calculating an IBI. Th comparatively few examples of validation in the literature. Methods developed for determining wetland condition have application for tracking net loss of wetland s (Scozzafava et al. 2009) . I n order to continue to have confidence that these methods have utility they must be tested outside of initial development to ensure that this type of wetland accounting to track net loss, is accurate. Indices of Biotic Integrity (IBIs) Biological integrity was defined by Karr a nd Dudley (1981) as the capacity of an ecosystem having species composition, diversity, and functional organization comparable to that of This definition h as been used as a basis for developing what have been termed indexes of biological integrity (IBIs), which provide a
14 community assemblage (e.g., diatoms, plants, macroinvert ebrates, fish). In theory, an ecological system either has integrity or lacks integrity. In contrast, an IBI focuses on condition, where the biological community is measured in terms of a gradient from reference standard condition (i.e., ecological system surrounded by natural landscapes with no apparent anthropogenic alterations ) to severely degraded. IBIs are composed of several individual metrics (e.g., species sensitive to disturbance, functional group members, exotic species, etc.), with individual met ric scores typically added together to reflect condition. In developing an IBI , biological data are support ed by physical and chemical parameters along a gradient of human disturbance (U.S. EPA 2002) thereby identifying biological indicators of environment al stress. In practice, IBIs have been widely developed for wetland ecosystems ( Reiss and Brown 2007; Ruaro and Gubiani 2013 ) , and Scozzafava et al. (2009) suggest that identifying wetland condition with these methods informs trends of net loss of biologi cal integrity. Wetland IBIs have shown strong correlation with adjacent land use; as land use intensity increases around a wetland sensitive species are replaced by species tolerant of dist urbance ( Lopez and Fennessy 2002 ; Miller et al. 2006; Reiss et al 2 009; Kutcher and Bried 2014 ). Even though some species may experience a time lag in response to change ( Reiss et al. 2009; Pasquaud et al. 2013 ) , IBI interpretation of condition is sensitive to changes in adjacent land use ( Reiss 2006; Mack 2007 ). IBI Re sponse to Natural and Anthropogenic Variation Natural variation due to season, weather, hydrologic flu ctuation and stochastic events a ffect s populations within wetlands ( Jeffries 2008 ; Ramberg et al. 2010; Scarsoglio et al. 2012 ; Chapin and Paige 2 013 ). Ka rr et al. (1987) suggested long term
15 datasets should be used in the development of IBIs to reflect the scope of biological response to natural variation and be distinguish able from anthropogenic change. Practitioners w ould have confidence in i ndices that demonstrated consistent measurements of condition over tim e where no anthropogenic change had occurred (Karr et al. 1987). More recently, Mack (2007), Mazor et al. (2009), and Yodder and Barbour (2009 ) have echoed this sentiment calling for IBI development based on long term datasets to capture the variability of biological response to natural and anthropogenic environmental stressors along spatial and temporal scales. In practice, long term data sets are rarely available to support IBI development of wetla nds. In light of biological response to natural and anthropogenic variation , it remains unclear if wetland IBIs identify condition consistently in different seasons and in different years. Mazor et al. (2009) examined 20 years of macroinvertebrate data w ith IBI methods and individual metrics in a watershed where management conditions were constant and disturbance was mostly absent. While the study found little variability in long term data, on a shorter time frame condition evaluations where not consisten t. Total variance in condition explained by time ranged from 5 35%. The authors concluded that benthic communities experienced high year to year variability and longer time frames and multi metric indicators were necessary for identifying accurate trends. Fish B ased IBI Early IBI development was in fish assemblages in streams and lakes. Most of the studies that explored IBI effectiveness at detecting long term trends of natural variation and biological response to land use change were in fish based IBIs. Angermeir and Karr (1986) in a multi year and seasonal study concluded the optimal sampling time for consistent evaluation with a fish IBI in small streams was in early summer to
16 compensate for seasonal differences in richness and species distribution. Ka rr et al. (1987) also found seasonal effects on species richness, even though the fish IBI being applied in small streams was sensitive to water quality and habitat changes. The IBI stability was variable based on quality of upstream habitat and stream con dition; scores were more stable in areas of higher condition than those of lower condition. Fore et al. (1994) also found greater stability in fish IBI condition at sites that maintained higher condition than degraded sites when comparing late summer and early fall sampling periods. The study concluded high variability in condition scores could be early indicators of increased anthropogenic degradation in the watershed because impacted systems no longer have the elasticity to cope with natural change. Con trary to Fore et al. (1994) and Karr et al. (1987) findings, Pyron et al. (2007) did not find higher variability in condition scores at sites with higher condition than degraded sites when comparing month to month differences in the summer for fish IBI co ndition in a 25 year data set for a large river. Pyron et al. attributed this different finding to distribution of fish in the smaller streams . Within year variation of condition defined by a multi metric index was not significantly different but sampling at the most representative period t o capture variation, such as in the summer months, was recommended. Moncayo Estrada et al. (2012) applied a fish community IBI to a 40 year data set for a shallow lake ecosystem in Mexico that experienced large natural v ariation in lake volume due to normal drought cycles . I mpacts to fish populations were not detected by the IBI during normal drought until a lack of rainfall was coupled with later human disturbance in the form of increased water use and habitat loss (Monc ayo Estrada et al. 2012) , this means the IBI was reliable at deciphering impacts from natural and anthropogenic variation .
17 Vegetation B ased IBI In an attempt to develop an IBI for coasta l wetlands of the Great Lakes in North America, Wilcox et al. (2002) found the high variability of lake levels caused extreme natural variation in vegetation communities. A valid repeatable IBI would only be possible if condition indices were developed for multiple water level historie s and time since extreme events. G iven the long term complex hydrologic cycles for the Great Lakes , the authors found the application of a plant based IBI for coastal wetlands limiting ; they were unable to separate biologic response to natural or anthropogenic caused change. Mack et al. (2008) applied the Ohio inland vascular plant based IBI to coas tal wetlands in the Great Lakes. E ven though some hydrologic characteristics were different, interior Ohio wetlands also experienced extreme hydrologic events and had significant floristic similariti es to the coastal wetlands on Lake Erie (Mack et al. 2008) . The results showed correlations between least impacted coastal marshes and high quality inland wetlands. The authors believed the inland index could be modified with success to correlate a coastal marsh index for the Great Lakes validated by a land use intensity gradient. Complexities relating to fluctuating hydrology in the Great Lakes could be addressed with IBIs developed for different successional phases of wetland plants and by developing samp ling guidelines to address plant community migration (Mack et al. 2008). Like the Great Lakes wetlands, prairie pothole s in North Dakota have dynamic hydrologic conditions with long term studies documenting large natural shifts in plant community com posi tion (Euliss and Mushet 2011 ). Evaluating the consistency of IBI scores over a four year period, Euliss and Mushet (2011 ) found IBI scores increased with increased inundation and decreased with natural dry periods. Part of
18 ) expl anation for the decrease in IBI was related to the replacement of floating or submerged native perennial species with native annual species colonized from exposed seed banks during natural drawdowns. Wilson et al. (2013) tested the consistency of an IBI de veloped for wet meadow zones in shallow open water marshes in Canadian northern prairies. Unlike the prairie pothole wetlands st udied by Euliss and Mushet (2011 ), the wetlands in Wilson et al. (2013) had less seasonal variation in hydrology, tended to have standing water in most growing seasons, and were naturally dominated by native perennial species. Even though the wet meadow zones potentially shifted in response to a natural moisture gradient, condition scores were consistent in both wet and dry years o f sampling (Wilson et al. 2013). Both Mack et al. (2008) and Wilson et al (2013) recommended not sampling if an extreme natural flooding or drought event occurred temporarily precluding the representative vegetation from that wetland. There has been some research on variation in wetland condition over time related to the floristic quality assessment index (FQA I ) , which is often used as a metric in wetland macrophyte IBIs. In development of an FQAI individual plant species are assigned a numerical value der ived from expert botanist opinion relating to habitat affinity and tolerance to disturbance ; this value is called the coefficient of conservatism (CC) after Wilhelm and Ladd (1988). To calculate the FQAI, the sum of CC scores is divided by the square root of native species richness or sometimes total species richness resulting in mean CC value for the assessment area. In wetlands, FQAI and mean CC scores have been correlated with adjacent land use and disturbance gradients ( Lopez and Fennessy 2002; Cohen et al. 2004 ). However, FQAI results have been criticized for having sensitivity to wetland area and total richness ( Francis et al.
19 2000; Matthews et al. 2005). Using a mean CC score reduced bias from total richness on the FQAI condition score ( Cohen et al. 2 004; Matthews et al. 2005), making mean CC less sensitive to sampling season effect on total richness (Francis et al. 2000). For example, in a study examining seasonal and annual FQAI scores, Lopez and Fennessy (2003) found consistency in FQAI scores in be tween years and in summer and fall comparisons in depression wetlands in Ohio. Yet in other studies (e.g. , Francis et al. 2000; Matthews et al. 2005; Bried et al. 2013) FQAI was less consistent and mean CC was recommended as a better measurement to withsta nd natural variability in wetlands related to season and time. Bried et al. (2013) found that even though there was high species turnover between seasons and some evidence of sampling error, scores were consistent between sampling periods since recruited s pecies had similar CC scores as those they replaced. In a study on Florida wetlands, Deimeke et al. (2013) applied the vegetation portion of the Florida Wetland Condition Index (FWCI), a multi metric IBI created for geographically isolated forested wetlan ds, to mature wetlands 7 8 years after the development of the FWCI. Land use around the wetlands remained constant between sampling periods. While paired wetlands shared on average just 30% of species between sampling periods, FWCI scores were significantl y correlated. Only the FQAI metric had a small but significant shift between sampling periods, with lower mean CC scores in the later sampling period. Macrophytes as Indicators of Wetland Condition Vascular plants are ubiquitous , generally immobile, resp ond to environmental changes , and relatively easy to identify to species level (U.S. EPA 2002). Many wetland IBIs are based on wetland vegetation because they are good indicators of wetland
20 condition ( Miller et al. 2006 ; Mack 2007; Stapanian 2013). As prim ary producers, wetland vegetation are part of the foundation of trophic interactions in wetlands and support many life history requirements of fauna species including structural cover and support for foraging and reproduction (Mitsch and Gosselink 2000). V egetation can indicate anthropogenic effects in a watershed by responding to and being indicators of hydroperiod, water chemistry, substrate type, landscape connectivity to seed sources, edge effects and climat e change ( Doherty et al. 1999 , Reiss and Brown 2005a ). Still, interpretation of macrophyte presence should be applied with caution due to variation in response times to an environmental change, identification and detection without an inflorescence ( U.S. EPA 2002 ) and the influence a sampling method ma y have on data interpretation ( Levin 1992; L egendre et al. 2002 ). Unique attributes specific to wetlands could result in high species turnover. Wetland vegetation composition and structure are dynamic and can fluctuate greatly along temporal and spatial gr adients (Chapin and Paige 2013). Wetland species are adapted to stress from fluctuation in hydrology that strictly aquatic or upland species could not tolerate (Mitsch and Gosselink 2000). Sometimes s tochastic events that alter water levels could lead to c onditions that preclude wetland species from staying viable in the seed bank causing un vegetated areas ( Ridolfi et al. 2006; Palanisamy and Chui 2013 ) or shifts in spatial arrangement of species ( Euliss and Mushet 2011 ) . Microtopography plays a large rol e in how hydrology drives seedling survival and heterogeneity in vegetation composition and structure (Titus 1990). The Florida Wetland Condition Index (FWCI), aforementioned in the Deimeke et al. (2013) study above, is a multi metric index of biotic inte grity developed for three
21 wetland types, both forested (Reiss and Brown 2005a ) and herbaceous geographically isolated depressions ( Lane et al. 2003) and forested strand and floodplain wetlands (Reiss and Brown 2005b). Th ree separate indices were developed for diatom , vegetation , and macroinvertebrate assemblages (Reiss et al. 2009). The FWCIs were significantly correlated with adjacent land use intensity and physical and chemical soil and water parameters (Reiss et al. 2009). Drought conditions persisted du ring the develo pment of the FWCI (Lane et al. 2003; Reiss 2005a ) and in general wetlands were only visited once for data collection (Reiss 2006). The authors recommended further validation of the FWCI by testing yearly variation in wetlands outside of thos e examined for the development of the index. Sampling Methodology Effect on Interpretation of Ecological Phenomena The concept of scale is a human construct, necessary in ecology to define an area of interest both temporally and spatially in an attempt to describe ecological phenomena (Hobbs 2003). The application of a multi scaled perspective is an integral part of studying pattern and process in any system. Due to physical and monetary constraints prohibiting complete coverage of a system being studied, ecologist can only sample a smaller area or point in time and make inferences about the larger system ( Stohlgren et al. 1997; Hobbs 2003). At issue is the reality that there is no set standard for scale and scientists become the lens from which the system in interpreted (Levin 1992). At the resolution of one scale complexity and heterogeneity may be overwhelming but at another, the pattern becomes homogenous or vice versa (Dutilleul 1993). The mechanisms that interact and force the patterns in a landscape are typically larger than the study area itself (Levin 1992); in defining a study area and applying a
22 sampling protocol it helps to understand something about those mechanisms in designing a sampling plan (Legendre et al. 2002). The size and shape of a sa mple unit could allow for inference of a spatial pattern that is an artifact of sampling and not due to ecological phenomena ( Fo rtin 1999 ; Dungan et al. 2002 ; Stohlgren et al. 2003 ). In designing a sampling protocol the main objective should be to detect p atterns in the response variable. Optimizing a design to capture real spatial patterns such as gradients and patches can help scientists reduce type I errors (i.e., detecting a response that is not actually there) (Legendre et al. 2002). Many studies have manipulated sample layouts to examine how the scale of observation could alter interpretation of a process. For example, Reed et al. (1993) found by applying multiple scales of observation, correlations between vegetation and environmental variables could change dramatically. By varying size and shape of sample layouts within the same area of a floodplain forest, Dungan et al. ( 2002 ) concluded interpretations of an ecosystem process differed with sample unit because some methods did not intercept directiona l patterns occurring in the landscape. In trying to design an appropriate sampling methodology for vegetation, understanding something about the mechanism that influence species distribution and cover can be helpful in capturing the best information neede d to study species assemblages. One of the most widely accepted predictions in ecology is that an increase in sample unit size will result in an increase in species richness (He and Legendre 2002), and to a lesser extent, the distance between sample units and the number of sample areas can also influence the relationship between area and total richness (Palmer and White 1994 ). Long thin plots tend to encounter greater
23 heterogeneity and large square or round plots will encounter greater homogeneity (Stohlgre n et al. 1995). The shape of a sample unit or more specifically the area to perimeter ratio is also integral to how effective the sample is at intercepting species richness especially if directional gradients are present (Dungan et al. 2002; Murray Hudson et al. 2012) . Vegetation in general exhibits high spatial autocorrelation; sample units closest together tend to be more similar (Dormann 2007). In pl anning a sample design scientists must acknowledge the probability of spatial autocorrelation that could b e amplified in the sample layout making environmental drivers and species interactions more difficult to separate (Peet et al. 1998 ). In the presence of spatial gradients t he starting location of a plot layout can bias the effectiveness of collecting a rep resentative sample (Stohlgren et al. 1995) . Dutilleul (1993) recommended that a completely random sampling design was not appropriate when encountering heterogeneity in the form of patches and environmental gradients, unless the sample area is highly homog enous, blocking treatments such as grouped sample units was a better design. Legendre et al. (2004) recommended adjusting sample design so it appropriately intercepted heterogeneity when spatial gradients were present. Vegetation Patterns in Wetlands In w etlands, hydrology and more specifically hydroperiod is a major force in determining wetland vegetation community structure and composition ( van der Valk 1981; Chapin and Paige 2013). Small scale variations in hydrology can have strong impacts on the spati al and temporal heterogeneity of wetland plants (Jeffries 2008). Wetland plant species do not occur randomly but in zones and patches along spatial gradients related to moisture; this is the result of interactions between slope, microtopography, water dept h, and duration and timing of inundation (Mitsch and
24 Gosselink 2000). Hydrologic fluctuations help maintain wetland species presence by precluding upland species intolerant of inundation (FNAI 2010), concentric zones of vegetation may expand and contract a long directional gradients related to depth and duration of flooding (Wilson et al. 2013). In depression wetlands, species richness decreases directionally with the highest richness in ecotone edges decreasing towards wetland interiors where longer hydrop eriods occur ( Kirkman et al. 1998; Murray Hudson et al. 2012 ). The dynamic nature of most wetlands in regard to periods of drought and flooding is reflected in the plasticity of wetland plants especially in ecotone edges that can be tolerant of both period s of inundation and dry conditions (Kirkman et al. 1998). The index of biologic integrity, FWCI is a richness based index. Metrics are calculated by counting the presence of species indicative of reference condition or associated with land use impacts as a proportion of total species richness in the sample area. As vegetation occurs along spatial and temporal gradients in wetlands it is important that sample units intercept representative areas. Murray Hudson et al. (2012) conducted a study to determine t he optimal layout and area needed to collect enough representative data in wetland depressions to analyze with the vegetation portion of the FWCI. Belted transects were split into similarly sized zones reflecting duration of inundation. The authors found t he outermost zone contained on average 87% of total species richness for the wetlands sampled. However , when condition was calculated separately for each vegetation zone no statistical difference was found in condition scores. Murray Hudson et al. (2012) c oncluded that area and effort in sampling could be minimized by only sampling the outer third of the wetland and still accurately characterize wetland condition.
25 Study Objectives The main objective of this study involved applying two different sample desi gns to make repeat field measurements on the vegetative community structure in 18 wetlands in Florida over two consecutive growing seasons. Assuming land use and management remained the same for the wetlands during the intervening year, it was hypothesized that the macrophyte community structure and composition would remain relatively constant, within the expected range of variation due to normal fluctuations in climatic regime, and therefore macro phyte FWCI scores would not be a ffected. The main question d riving this research was related to the consistency and repeatability of wetland condition scores derived from analysis of macrophyte communities, as follows: Assuming no changes in management regimes and surrounding land use intensity, what effect does ti me (and natural climatic variation) have on FWCI macrophyte scores? Will the macrophyte FWCI remain consistent form year to year? Or will it change to such an extent that condition scores will be significantly different? In other words how reliable are FW CI scores? Sampling method design and scale are important considerations in determining if a field sample is representative and adequate for interpreting ecological phenomena, in this case wetland condition. Wetland plants do not randomly distribute but ra ther occur along gradients related to microtopography and moisture (Mitsch and Gosselink 2000) . If two different sampling methods are expected to have high spatial autocorrelation but differ by size and shape, how similar will the plants species they inter cept be? Will differences to the sample layouts result in the methods intercepting distinctly different areas of a wetland and therefore not capturing the same species information? Would these potential differences in species detection impact total specie s richness and
26 therefor result in different proportion of indicator species by method? Will the differences in sample design result in data that gives a different interpretation of wetland condition by method?
27 CHAPTER 2 EFFECT OF TIME ON CONSISTENT AND R EPEATABLE MACROPHYTE INDEX FOR WETLAND CONDITION Introduction Biological integrity was introduced as a concept by Karr and Dudley (1981) as a balanced, integrated, adaptive community of organisms having species com position, diversity, and functional organization In theory, an ecological system either has integrity or does not. In practice, an index of biological integrity (IBI) provides a quantitative means of as community assemblage (e.g., diatoms, plants, macroinvertebrates, fish). The assessment of biological condition with an IBI consists of measuring the biological community in terms of a gradient from reference standard condition (i.e., ecological system surrounded by natural landscapes with no apparent anthropogenic alterations ) to severely degraded. In developing an IBI , biological data are support ed by physical and chemical parameters along a gradie nt of human disturbance (U.S. EPA 2002) thereby identifying biological indicators of environmental stress. IBIs have been widely developed for wetland ecosystems ( e.g., Reiss and Brown 2007; Ruaro and Gubiani 2013 ) . They have strong correlation with chang es to adjacent land use; as land use intensity increases around a wetland sensitive species are replaced by species tolerant of dist urbance ( Lopez and Fennessy 2002 ; Miller et al. 2006; Reiss et al 2009; Kutcher and Bried 2014 ). This makes IBIs effective t ools for monitoring if anthropogenic impacts are altering condition in wetland ecosystems (Karr and Chu 1997) or how a wetland condition responds to restoration (Matthews et al.
28 2009). Wetland IBIs can also be applied to inform trends of net loss of biolog ical integrity (Scozzafava et al.2009). Herricks and Schaeffer (1985) recommended effective biological indicators should be interpretable at several trophic levels and connected to organisms not directly monitored. As primary producers, vegetation is part of the foundation of trophic interactions in wetlands and support s many life history requirements of fauna species ( Mitsch and Gosselink 2000). Vascular plants are ubiquitous , generally immobile, respond to environmental changes , and relatively easy to id entify to species level (U.S. EPA 2002). Many wetland IBIs are based on vegetation because they are good indicators of wetland condition (Miller et al. 2006 ; Mack 2007; Stapanian 2013 ) and are responsive to hydro period, water chemistry, substrate type, la ndscape connectivity to seed sources, edge effects and climate change (Reiss and Brown 2005a, Doherty et al. 1999). Statement of the Problem Practitioners should have confidence in an IBI that has shown consistency and repeatability when land use surround ings are constant ( Karr et al. 1987 ). Further, these methods should be effective at determining the difference in biological response to natural variation and human caused impacts (Karr et al. 1987). A robust IBI methodology should be validated by assessin g wetlands not included in IBI development ( Karr et al. 1986; Reiss 2006). While many wetland IBIs have been developed since these suggestions by Karr et al. (1986 and 1987) there are not many examples of studies validating IBIs in this way. Where further research is needed is in determining if IBI measures are consistent and repeatable in light of natural variation due to effects of time and climatic variability (Wilson 2013).
29 Review of Literature Unique physical and chemical attributes of wetlands make vegetation composition and structure dynamic, sometimes fluctuating greatly along temporal and spatial gradients (Chapin and Paige 2013). Fluctuations due to season, weather impacts on hydrology and stochastic events can a ffect biological populations withi n wetlands ( Jeffries 2008 ; Ramberg et al. 2010; Scarsoglio et al. 2012 ; Chapin and Paige 2 013 ). This type of natural change could confound IBI condition results if metrics are too sensitive to normal variation. Some studies testing condition variability us ing more mobile indicators such as macro invertebrates (Mazor et al. 2009) and fish ( Angermeir and Karr 1986; Fore et al. 1994) in stream based IBIs, found high year to year and seasonal variability in population richness and distribution suggesting longer time frames were needed to develop consistent IBIs capable of reflecting representative condition. Some communities such as coastal wetlands in the Great Lakes are also highly mobile due to dramatic changes in lake levels. Other wetlands with ephemeral hy drologic regimes such as prairie potholes may experience high turnover of species or community shifts along hydrologic gradients. Even in Florida where the growing season is longer than most parts of North America, identifying plant species without an infl orescence due to sampling season can challenge species detection in a wetland. Wilcox et al. (2002) concluded that high variability in lake levels in the North American Great Lakes made the development of an IBI for coastal vegetation too complex to separ ate biological response from anthropogenic caused change. In 2008 Mack et al. addressed the Wilcox et al. (2002) concerns by applying an inland vascular plant based IBI created for Ohio to the Great Lakes coastal wetlands concluding the interior wetlands a lso experienced extreme hydrologic events and had significant
30 floristic similarities. After correlating the two wetland types by grouping land use gradients, the authors determined with some modification an IBI could be developed for the coastal marshes of Lake Erie and other communities that experience plant community migration and dynamic hydrologic conditions. North Dakota prairie potholes also have large natural shifts in plant communities due to high variance in hydrologic regimes. In a four ye ar stud y Euliss and Mushet (2011 ) determined the prairie pothole IBI scores were not reliably consistent in these wetlands due to variable water levels and natural dry down impacts to plant composition. In open water marshes in northern Canada prairies, wetlands had less seasonal variation in hydrology in a study by Wilson et al. (2013); standing water was typically present and vegetation was dominated by native perennial species. The wet meadow zone being sampled did shift along a hydrologic gradient between year s; by sampling in the most representative central area of the zone IBI condition scores were consistent in both wet and dry years (Wilson et al. 2013). In Florida, Deimeke et al. (2013) resampled forested wetlands where surrounding land use had not changed in the several years since they were first assessed for the development of an IBI. Authors found condition scores to be significantly correlated between years. Both Mack et al. (2008) and Wilson et al (2013) recommend not sampling a community if an extrem e weather event such as drought or flooding had occurred because the event could preclude the collection of a representative sample. Study Objectives It remains unclear if wetland IBIs identify condition consistently in different seasons and over time in light of potential biological response to natural and anthropogenic changes. This question is best addressed by each regionally developed
31 and tested IBI method. The main objective of this study was to make repeat field measurements on vegetative communitie s in 18 wetlands in Florida over two consecutive growing seasons. Community data were applied to the vegetation portion of the Florida Wetland Condition Index (FWCI), a multi metric index of biotic integrity developed for geographically isolated herbaceous and forested wetlands as well as forested strand and floodplain wetlands. It was hypothesized that macrophyte community composition would remain relatively constant, within the expected range of variation due to normal fluctuations in climatic regime, and therefore FWCI scores would remain constant. This study tested the consistency and repeatability of FWCI scores between sampling periods. Methods Employing standardized field sampling method s and several sta ti sti cal analysis techniques this study e xamine d the impact of time (one year) between sampling events on the consistency of wetland condition scores derived from analysis of the macrophyte communit y. In the following section, descriptions of the wetland sites and the macrophyte FWCI are given first, an d then field sampling methods and finally statistical methods that were employed are given. Site D escription Eighteen wetlands in Florida were selected for this study from a larger population of 67 wetlands that were sampled as part of the U.S. Environmen tal Protection Agency 2011 National Wetland Condition Assessment (NWCA). The 2011 NWCA study consisted of 67 wetlands in Florida selected from the U.S. Fish and Wildlife status and trends database. The NWCA site selection method employed the General Rand om
32 Tessellation Stratified survey design ( Stevens and Olsen, 2004) which generated a random point in each wetland as the center for the sample area. Wetland classes chosen from the NWCA study for FWCI application complied with wetland type under which the FWCI was developed. These were, herbaceous depression (HD , n=6), forested depression (FD , n=5) and forested strand or floodplain wetlands (FSF , proposed classification of freshwater wetlan d regions south (n=2), central (n=5), north (n=7), and panhandle (n =4) (Figure 2 1). The selected sites were also stratified along a prior i categories interpreted from remotely sensed aerial imagery representing a gradient of adjacent land use intensities. Land use for wetlands included agriculture (cattle grazing) (n=3), urban residential (n=2), reclaimed phosphate mine (n=1), silviculture and managed for est (n=6), and reference (n=6). Six of the study sites were under private ownership and twelve were own ed by public entities. Wetland sampling in 2011 occurred between May and mid October. The 2012 sampling season was shorter with site visits occurring between May through early August. The longer 2011 sampling season was due to additional data collection at wetlands not included in this study for the greater NWCA project in 2011. While all sampling occurred within the Florida growing season (USDA 1985), differences in sampling periods meant that some wetlands were resampled in earlier or later parts of the season the following year. The 2012 sampling season had more site visits earlier in the growing season, occurring on average, 60 ordinal days prior to sampling in 2011 which consisted of greater late season sampling.
33 Parts of Florida were characterized as being in moderate to severe drought in 2011 which continued into the middle of 2012 (NOAA 2011, 2012). The drought in 2012 was broken by tropical storm Debby which hit Florida on June 25 th . The rainfall from the storm was more than 20 inches in parts of the Florida panhandle and much of the northern peninsula had at least 5 10 inches of rain in one day (National Hurricane Center, 2013). Half of the wetlands in this study had standing water during one sampling period but were not inundated for both sampli ng periods. In addition to natural variation in terms of flooding and drought, some management activities occurred between sampling that impacted species presence . The managers of an herbaceous marsh in south Florida removed the cattle that had been g razing the land. In 2011, cattle trails, cow patties and evidence of grazing were present during sampling, but in 2012 the grazing pressure and trampling had been removed. Another herbaceous marsh had a prescribed fire conducted between the 2011 and 2012 s ampling events. The site had a visible reduction in vegetation biomass and an increase in bare ground at the 2012 site visit . Macrophyte FWCI Vegetation based FWCI consisted of five metrics for herbaceous depressions (HD) and forested strand or floodplain wetlands (FSF) wetlands and six for forested depression (FD) wetlands. Table 2 1 lists metrics applied to calculate vegetative FWCI scores by wetland type. Metrics included in the vegetation assemblage for the three wetland types (HD, FD, FSF) were proport ion of species indicative of tolerance to anthropogenic impacts, proportion of species indicative of sensitivity to anthropogenic impacts, proportion exotic species and mean coefficient of conservatism (CC) (Reiss et al. 2009). Tolerant metrics were based on indicator species analysis that identified
34 species persisting in wetlands impacted by intensifying land use and tolerant of physical and chemical parameters associated with increased land use intensity ( Lane et al. 2003; Reiss and Brown 2005b; Reiss 200 6) . Sensitive metrics identified indicator species that did not persist in an intensifying landscape and were instead associated with physical and chemical parameters indicative of a reference wetland (Lane et al. 2003; Reiss and Brown 2005b; Reiss 2006) . The exotic metric was based on a native or exotic status per species as defined by the United States Department of Agriculture (USDA) PLANTS Database ( http://plants.usda.gov/java/) . The coefficient of conservat ism (CC) is a numerical value derived from expert botanist opinion relating to habitat affinity and tolerance to disturbance after Wilhelm and Ladd (1988). Coefficient of conservatism values were based on Florida Department of Environmental Protection (DEP ) standard operating procedures (SOP) document DEP SOP 002/01 LT 7000 Determination of Biological Indices ( DRAFT 2012 ). If a species was not included in the DEP SOP, Reiss and Brown. (2005a, 2005b) and Lane et al. (2003) were utilized for CC values. Herba ceous wetlands had an additional metric measuring the ratio of annual to perennial species (Lane et al. 2003). Forested wetlands both FD and FSF had a metric for proportion of native an d perennial species (Reiss et al. 2005b; Reiss 2006 ). Only FD wetlands had a metric defining percent of species with a wetland indicator status of obligate or facultative wet based on Florida rule 62 340.450 , F.A.C. (Reiss 2006). Species taxonomy, duration (annual or perennial), native status and wetland indicator status wer e first based on DEP SOP or if not listed then United States Department of Agriculture PLANTS Database (2012) ( http://plants.usda.gov/java/ ).
35 Individual metric scores are based on the presence of indicator species. When the metrics were developed, specie s that correlated the best with physical and chemical parameters and land use gradients for reference wetlands were converted to a proportion based on total richness. The top 95 th and bottom 5 th percentiles were normalized and scaled from zero to ten, with ten being the optimal value for highest condition in that metric (Reiss 2006). In this study metrics were summed and scaled to 100 so that condition scores for the three wetland types could be compared where 100 represented highest condition in a wetland. Sampling D escription M acrophyte communities were sampled using 40 meter belted transects laid out in four cardinal directions from the NWCA cent er point within a 0.5 ha assessment area ( AA) (Figure 2 2). Each belted transect was divided into four , 5 meter long 1 meter wide quadrats and every rooted species within each quadrat identified to species level . FWCI was based on presence ab sence data, a species wa s present if it was rooted in one of the belted transect quadrats, regardless of number of quadrats it was recorded in. Unknown species were recorded as a pseudonym and were collected for lab identification . Data sheets were entered into a Microsoft Access Â® database and updated when unknown species were subsequently identified. Unknowns that could not b e identified were recorded but not included in FWCI analysis. The field sheets and access database were cross referenced and checked for quality assurance. Additionally all species were identified by growth habit (grass including rush and sedge, forb, shru b, tree or vine). Each growth habit category was then totaled to determine the proportion each growth habit category contributed to total richness for each wetland type. In addition to the belted transect presence absence data, five 100m Â² quadrats were als o
36 sampled within the AA in order to describe vegetation structure via p ercent cover estimates by vegetation height class groups . Statistical Analyses The comparison of scores between sampling periods for FWCI condition were expected to be close to one to one and were thus examined with linear regression. T otal richness at a wetland, total evenness, Shannon diversity (H), and Simpson diversity (S) were calculated in Microsoft Excel for both sampling periods and compared between sampling periods with Pears on correlation coefficient . In order to understand if individual wetlands experienced change in composition, v egetation data collected each year were compared for unique and shared species at the same wetland . A Jaccard distance measure was applied to calc ulate species differences between years by subtracting species in common to both sampling periods from total species richness and dividing that result by total species richness both years. Ja ccard distance = |A U B| |A B| / |A U B| Jac card ranged from zero to one, zero meaning a site shared the exact same species and one meaning no species were in common. Percent shared species was calculated as the inverse of Jaccard by subtracting the dissimilarity value from the number one and multip lying the result by 100. All calculations except for Jaccard distance were completed using Microsoft Excel v. 14 and Minitab v.15. Jaccard distance was calculated in PCORD v.6. Results The results from the field campaigns and data analysis are presented by evaluating the impact of time between sampling and the resulting differences in (2 1)
37 community composition and structure on vegetative metrics and condition scores of the FWCI. Effect of T ime on M acrophyte FWCI There were very minor changes in the componen ts of, and overall macrophyte FWCI scores between the two sampling periods . Tables 2 2 through 2 4 contain all calculated scores and individual metrics for wetlands from both sampling periods. A linear regression analysis demonstrated macrophyte FWCI score s in 2011 vs. 2012 were strongly significant (p<0.000) (Figure 2 3). The fitted slope, 0.92 (95% CI = 0.8 1.04) was close to one making FWCI 2011 scores a good predictor of 2012 scores. 4) between years (two tailed p = 0.57). FWCI condition scores ranged from 12% 94% condition (mean = 58) in 2011 and 12% 88% (mean= 59) in 2012. With the exception of the wetland indicator status metric, only applicable to the small sample size of five F D wetlands, individual metrics that compose FWCI condition had statistically significant correlation values between years (Table 2 5 ). There was no significant difference in proportion of indicator species (tolerant or sensitive) or exotic species (mean = 2.4, = 3.5) between years (two tailed p = 0.39). Community Similarity Pearson coefficients for diversity measures of community composition are presented in Table 2 5 index and the Shannon diversi ty (H) index were significantly correlated between the two sampling periods. Table 2 6 reports values for proportion of species by growth habit for each year by wetland type. For all three wetland types, growth habit data showed very little variation and demonstrated stability between years in wetland structure. Figure 2 4
38 also demonstrated stability in wetland vegetation structure by describing percent cover of species by height categories. While all three wetland types show a modest decrease in the <0.5 m height class and a more moderate increase in the 0.5m to 2m height class for 2012, most ranges for percent cover over lap within wetland type between years. Forested wetlands had the most consistency in median cover values. While FWCI condition scores showed very little variation between the sampling years and diversity statistics and vegetation structure reflected community stability; the percent of species detected at the same wetland both years demonstrated a different and more dynamic scenario. The box plot in Figure 2 5 demonstrates the range of percent shared species for the three wetland types. Mean percent species detected for all wetlands both years was 55% ( = 9). HD wetlands (n=6) had the lowest median shared species at 47% and the widest ran ge of species in common at a wetland, 44% 77%. FD wetlands (n=5) had a median shared species of 52% and demonstrated less variance in percent shared species with a range of 46% 61%. FSF wetlands (n=7) had the highest median shared species 60%, with a r ange of 46% 67% shared species detected at the same wetland between years. In general most species encountered either year were rare; 50% in 2011 and 53% of species in 2012 were only encountered at one wetland (Figure 2 6). Approximately 90% of all speci es occurred at less than a quarter of wetlands in either year indicating wetlands in this study were unique from one another. Discussion The high fidelity of FWCI scores suggests the FWCI has robust vegetation indicator species, indicators of wetland cond ition were retained even though there was close to a 50% average species difference from one year to the next. There were minor
39 change s in macrophyte FWCI scores and metrics between sampling periods despite relatively wide variation in species composition . Results demonstrated stability in vegetation FWCI scores in the two year comparison despite natural variation and changes in land management . Natural Variation in Wetland C ommunities Natural seasonal variation in weather and stochastic events can affec t hydrology in wetlands shifting spatial arrangement of species ( Scarsoglio et al. 2012; Chapin and Paige 2013; Palanisamy and Chui 2013). The changes from drought conditions to tropical storm flooding and timing differences in sampling in this study would lead to some expectation of change in community composition between years. Even so, the mean percent species not detected the second year (4 5% ) and range of species detected both years (44% 77%) was greater than was expected for the time frame of this stu dy. Condition scores were maintained between years because condition indicators persisted between years and the proportion of the indicators remained constant even though community composition varied. In a Rhode Island study testing the reliability of mea n CC scores in different seasons, Bried et al. (2013) also did not find strong correlation in species composition between sample periods in non foreste d wetlands. The authors did however find mean CC scores were maintained and determined this result was fr om species being replaced by plants with similar indicator status. Deimeke et al. (2013) noted a 70% change in species when mature forested wetlands in Florida were re sampled several years later. Deimeke et al. (2013) determined high FWCI score fidelity w as maintained between years because the proportions of indicator species stayed the same.
40 Unlike the Deimeke et al. (2013) study which sampled mature, forested, geographically clustered wetlands selected from the initial development of the FWCI; this stud y represented a wider scope of wetland types (HD, FD, FSF), state of succession and geographic area in Florida. Most species are rare in abundance and cover (Barbour et al. 1987); in this study approximately 50% of all species encountered either year occur red at only one wetland emphasizing that wetlands sampled in this study were unique from each other. The fact that these wetlands consistently maintained score fidelity between years even though the wetlands represented a wide range of wetland types and un ique compositions further illustrated the strength and consistency of the vegetation FWCI method. Differences in Species Composition Even though this study covered a short time frame, natural weather patterns, potential differences in detection and change s in land management coupled with the dynamic nature of wetlands could explain the variation in species composition in one year between sampling. The difference s in season (an average of 3 months) could have affected species detection or identification due to variation in plant phenology. Drought also appeared to affect t he wetland vegetation before tropical storm Debby in 2012. Observations of wilting, defoliation and death of vegetation were made in some of the wetlands during the periods of drought in 20 11 and early 2012. Another potential example of drought was the abundance of the species dog fennel Eupatorium capillifolium which is adapted to growing quickly in early spring before rains saturate wetlands in Florida (Sellers et al. 2009) . E. capillifol ium was observed at 44% of wetlands in 2011 and 50% of wetlands in 2012 , a large proportion if you recall from Figure 2 6 that most species encountered in this study were relatively rare. In 2011 E.
41 capillifolium was mostly observed in the <0.5 meter heigh t class, in 2012 it was often observed in the 0.5m 2m height class. This effect was visible in the box plot figures of percent cover by height class (Figure 2 4). For all wetland types, the < 0.5m height class reduced in cover from 2011 to 2012, while th e 0.5m 2m height class increased in cover from 2011 to 2012. The increase in height of this weedy species potentially shaded out or out competed species of smaller stature or emerging from the seed bank. Standing water one year but not both years was anot her factor in natural variation that could have changed species presence and composition between sampling periods. At some wetlands , field notes indicated standing water was tannic and difficult to see through. Half of the sites had standing water at least one of the sampling years ; four wetlands had standing water both sampling periods . Submerged aquatic vegetation may have been difficult to see in these wetlands. Additionally vegetation that may have emerged under drought conditions would not have survive d inundation after tropical storm Debby passed through in late June of 2012 . Differences in FWCI Score Wetland species are adapted to dynamic hydrologic conditions that strictly upland or aquatic species could not tolerate (Mitsch and Gosselink 2000). Dur ing periods of drought or fire suppression, woody species, exotics and upland species could creep into wetland edges (Myers and Ewel 1990 ; FNAI 2010 ). In the Rhode Island study, Bried et al. (2013) concluded that areas with greater climate extremes such as drought might be more vulnerable to seasonal and inter annual variation in CC scores. In North Dakota, Euliss and Mushet (2011 ) found IBI scores increased with inundation and decreased with natural dry periods in prairie po tholes. Euliss and Mushet (2011 ) attributed this result to native annual species emerging from the seed bank during
42 natural drawdowns replacing floating aquatic native perennial species; a potential bias in CC scores. In this study, variation was not detected in condition scores as a r esult of changes to species presence from drought, flooding or prescribed fire. Again, replacement species between years maintained the same indicator status. For example, as result of a prescribed fire between sampling periods, one HD wetland had 45% shar ed species detected. Total richness however, was the same both years. Even though more than half the species were different the metric values and vegetation FWCI score was not significantly different . Another HD wetland surrounded by actively grazed pastur e was the only wetland with greater than 10% gap in condition score between years; it increased by 14% in 2012. In 2011 there were pockets of standing water but most of the site was dry dog fennel ( E. capillifolium ) grew to over 2 meters in the time betwee n sampling periods but sampling in 2012 occurred after tropical storm Debby . The cover and density of E. capillifolium in addition to the standing water might have contributed to the decrease in species richness to almost half of what it was in 2011 . Even so, only the exotic metric was different between years with two species being documented in 2011 but not 2012. Competition and inundation might have prevented the persistence of these species in the wetland but adjacent pasture and inevitably of future nat ural dry downs makes it likely that these exotic species will become established again. Still the difference in vegetative FWCI scores from 12% condition to 26% the following year does not suggest that condition at the site had dramatically improved. Rathe r, this impacted site is far from reference condition but evaluating over time might detect trends or changes if the adjacent land use improved.
43 Study Limitations FWCI sampling design was applied in a modified format to meet additional study parameters re lated to NWCA. Wetland hydrology is a major driving force in determining the influence a wetland has on the life history of macrophytes (van der Valk 1981). Plant zonation in wetlands is not a random mixing but rather a response to moisture gradients, espe cially flooding depth, duration and timing (Mitsch and Gosselink 2000). The FWCI sampling methodology was designed to maximize species heterogeneity to get the most complete species area relationship to characterize a wetland by intercepting spatial gradie nts related to hydrology (Murray Hudson et al. 2012). The random placement of the AA based on the NWCA design could mean that sample areas did not capture representative samples where species composition was associated with zones along a hydrologic gradien t. Wetland type selected to meet the requirements for applying FWCI was defined more loosely in this study than the descriptions for the original wetlands used to develop the method. HD and FD wetlands in this study were much larger than those studied in t he development of FWCI ( ) , while the median size for FD and HD wetlands in this study was larger at 47 ha and 43 ha , respectively. Unlike the FD and HD wetlands, t he FSF index was similar in size to the originally studied wetlands for development . However, the FS F development was a pilot study and the g eographic location and number of FSF examples in the pilot study were the limiting factor in applying FWCI to FSF wetlands in this study. Conclusion The high correlation of condition scores between years even with large difference s in species composition between years suggested the FWCI method was
44 robust in detecting wetland condition over two consecutive growing seasons . Even though species composition varied between sampling years (mean = 4 5% were different ) new s pecies still represented the same indicators of condition , suggesting the method is not sensitive to potential natural variation in wetlands .
45 Table 2 1. Metrics included in final Florida Wetland Condition Index calculation by wetland type. Metric Herbaceous Depression Forested Depression Forested Strand or Floodplain Proportion Tolerant Species X X X Proportion Sensitive Species X X X Proportion Exotic Species X X X Modified Floristic Quality Assessment Index(FQAI) X X Mean Coefficient of Co nservatism (CC) X Annual Perennial Ratio X Proportion Native Perennial X X Percent Wetland Indicator Species X
46 Table 2 2 . Individual metric values and final Florida Wetland Condition Index (FWCI) scores calculated for herbaceous depress ion (HD) in 2011 and 2012. The coefficient of conservatism metric is repre sented by CC. Metric scores range from 0 to 10, FWCI scores range from 0 to 100, 100 being highest condition. Table 2 3. Individual metric values and final Florida Wetland Con dition Index (FWCI) scores calculated for forested depression (F D) in 2011 and 2012. The coefficient of conservatism metric is represented by CC. Metric scores range from 0 to 10, FWCI scores range from 0 to 100, 100 being highest condition.
47 Table 2 4. Individual metric values and final Florida Wetland Condition Index (FWCI) scores calculated for forested strand or floodplain (FSF ) in 2011 and 2012. The coefficient of conservatism metric is represented by CC. Metric scores range from 0 to 10, FWCI scores range from 0 to 100, 100 being highest condition. Table 2 5. Pearson coefficient (r ) for individual metrics and community diversity measures. FWCI metrics r % Tolerant 0.72 % Sensitive 0.91 % Exotic 0.54 Mean CC 0.91 Annual perennial ratio 0.93 % Native perennial 0.92 % Wetland indicator * 0.33 Community diversity measures Richness 0.87 Evenness 0.7 0 Shannon diversity 0.87 Simpson diversity 0.92 Bold r indicates p < 0.05 * metric only applicable for fi ve forested depression wetlands
48 Table 2 6. Proportion of vegetation by growth habit for each wetla nd type each year. 2011 2012 Herbaceous Depression (n=6) grass 28% 27% forb 56% 55% shrub 9% 8% tree 1% 3% vine 6% 7% Forested Strand Floodplain (n=7) grass 24% 23% forb 24% 27% shrubs 12% 13% trees 26% 26% vines 13% 10% Forested Depression (n= 5) grass 19% 14% forb 40% 38% shrub 13% 15% tree 18% 19% vines 10% 14%
49 Figure 2 1. Geographic location of wetlands, herbaceous depression (HD), forested depression (FD) and forested strand or floodplain (FSF). HD FD FSF
50 Figure 2 2 . Assessment area (0.5 ha) and sampling layout of belted transects in four cardinal directions. Transects were 40m long with eight individual quadrats 1 m wide by 5m long. All rooted vegetation in the quadrats was recorded and analyzed with the FWCI metho d for wetland condition .
51 Figure 2 3 . Macrophyte FWCI scores in 201 1 vs. 2012 linear regression were strongly significant (p<0. 000). The fitted slope, 0.92 (95% CI = 0.8 1.04) was close to one (dashed line) making FWCI 2011 scores a good predictor of 2012 scores. Herbaceous depressions are open circles, forested depressions ar e represented by diamonds, and forested strand floodplain wetlands are triangles. 1:1
52 Figure 2 4 . Percent cover by height class for three wetland types, h erbaceous depressions (HD n=6), forested depressions (FD n=5), Forested strand floodplain wetlands (FSF n=7) in 2011 (solid) and 2012 (hatched) . Interquartile box shows the 25 th 75 th quartile intercepted by the interior line for median ri chness. Box whiskers represent the entire range of richness values and the asterisks are outliers. Herbaceous depression Forested depression Forested strand or floodplain
53 Figure 2 5. Percent shared species at a wetland between years by wetland type herbaceous depression (HD), forested depression (FD) and forested strand o r floodplain wetland (FSF) . Interquartile box shows the 25 th 75 th quartile intercepted by the interior line for median shared species. Box whiskers represent the entire range of percent shared species. HD (n=6) FD (n=5) FSF (n=7)
54 Figure 2 6 . Percent of species co occurring at a cumulative number of wetlands in 2011 (solid bar) and 2012 (hatched bar). In both years, half of species occurred in only one wetland. Less than 1% of species occurred at more than nine (or half) of the wetlands. No species occurred at more than elev en wetlands.
55 CHAPTER 3 SAMPLING DESIGN EFFECT ON INTERPRETATION OF WETLAND CONDITION Introduction Scale has always been an important aspect of interpreting ecological phenomena ( Levin 1992 ; Legendre et al. 2002). Whether considering corridors for wildli fe, biotic and abiotic flows across a landscape or genetic fitness of a metapopulation, how one defined an area of interest could greatly affect the interpretation of an ecological process. There is no set standard for size, shape or frequency of sampling areas, and scientists frequently question if their data is representative of the studied community. Sampling strategies are often limited by time, monetary constraints, and site access difficulties. In maximizing area but minimizing effort during sampli ng design and data collection, it is important that a representative sample is collected to ensure results reflect an accurate portrayal of the system. Statement of the Problem Spatial gradients are an important consideration in designing the sampling met hod for studying a natural system. Species distribution can have complicated drivers; the magnitude of scale from fine to coarse and the detection of richness at different scales can lead to different impressions of homo or heterogeneity in the landscape a t different scales. In wetlands; plants are not randomly distributed but occur in zones most strongly reflective of moisture gradients (Mitsch and Gosselink 2000). In studying wetlands, capturing heterogeneity along spatial gradients related to moisture wo uld maximize the sampling effort for plant species richness representation (Murray Hudson et al. 2013).
56 Indices of biological integrity use biological species assemblage as proxies for defining wetland condition (Karr et al. 1986). Determining wetland co ndition has practical application for detecting net loss (Scozzafaza et al. 2009) evaluating wetland degradation (Stainbrook et al. 2006; Moncayo Estrada et al. 2012) or restoration success (Matthews et al. 2009). Detection of richness is strongly correlat ed with sample design (He and Legendre 2002) and s ome individual metrics and IBI methods have been criticized for being too sensitive to overall richness ( Mazor et al. 2009 ) . Aside from richness , manipulation of the sample design could also intercept diffe rent species assemblages and may not detect certain indicators if representative areas or gradients are missed. Constraints to extensive sampling are often present and practitioners will look for ways to maximize information gathered but minimize effort. I f sample design has a large influence on which species are detected in a wetland, sample design could result in an incorrect characterization of wetland condition. An inaccurate characterization of wetland condition c ould result in a net loss of wetlands. The amount of area included in analysis and how sample units intercept the study area are very important in accurately portraying wetland condition. Review of Literature Sample design and s cale Monetary resources and time constrains scientists to define a limited spatial and temporal boundary to the system they study; inference from this sample is then applied to interpret the greater ecological phenomena ( Stohlgren et al. 1997; Hobbs 2003). Typically the mechanism behind the patterns and processes being interpreted is much greater in scope than the defined sample area (Levin 1992). For this reason, the sampling design, how a sample unit is applied on the landscape and how much data is
57 collected is imperative for collecting a representative sample. Unfort unately there exists no set standard for the scale at which a system is studied (Levin 1992). The resolution of scale can affect the interpretation of field data; complexity in a landscape could appear homogenous at too small a resolution and large spatial patterns could be missed entirely (Dutilleul 1993). While not always possible, if information exists to identify spatial patterns in a landscape such as patches or gradients; this can be used to design a more effective sampling area to appropriately inter sect heterogeneity in the system (Legendre et al. 2002). The physical properties of a sample unit, size, shape, area and where it is applied in the landscape could force a pattern in the data collected that is an artifact of the sample design and not the result of an ecological process ( Fortin 1999 ; Dungan et al. 2002 ; Stohlgren et al. 2003 ). As the size of a sample unit increases greater species richness is expected to be captured (He and Legendre 2002). Plots with larger area to perimeter ratios such a s long thin plots will encounter greater heterogeneity while large square or round plots with lower area to perimeter ratios will encounter greater homogeneity (Stohlgren et al. 1995; Dungan et al. 2002 ). In general when sampling vegetation, scientists sho uld expect sample units that are close together to have greater similarity (Dormann 2007); this vegetation property can make interpretation of species assemblage more challenging in trying to identify the biotic and abiotic drivers of community composition (Peet et al. 1998). Temporal and s patial patterns in wetland v egetation Spatial and temporal patterns in wetland vegetation can be driven by small changes in hydrology (Jeffries 2008). Wetland vegetation can exists in concentric zones that expand and cont ract along directional gradients related to moisture (Wilson et al.
58 2013) or exist in pockets of heterogeneity due microtopographic effects on depth and duration of inundation (Mitsch and Gosselink 2000). In depression wetlands species richness also exists on a directional gradient, ecotone edges tend to have the highest species diversity with richness decreasing towards wetland interiors where hydroperiods tend to be lo nger (Kirkman et al. 1998; Murray Hudson et al. 2012 ). Interpreting wetland c ondition T he Florida Wetland Condition Index (FWCI) is a multi metric index of biologic integrity developed for three wetland systems, geographically isolated forested and herbaceous wetlands and forested strand and floodplain forests (Reiss et al. 2009). FWCI can b e applied to evaluate wetland condition by looking at species assemblages for diatoms, macrophytes and macroinvertebrates and comparing species presence to that of a reference standard condition. The result is a condition score which represents a percenta ge of the r eference standard (i.e. 100 %). Condition scores have strong correlations with adjacent land use gradients; indicator species reflect the biological response to complex interactions between land use impacts and physical and chemical characteristi cs of a wetland (Reiss et al. 2009). The FWCI is a richness based index; all species present contribute an equal weight in condition score calculation. Individual metrics such as species tolerant of disturbance, sensitive to disturbance, native or exotic s tatus etc. are calculated as a proportion of total species richness and summed for final condition score (Reiss et al. 2009). Murray Hudson et al. (2012) applied the vegetation portion of FWCI to geographically isolated herbaceous depressions. The authors calculated the FWCI separately for each vegetation zone in the marsh. The study found that most species richness was accounted for in the outer third portion of
59 the marsh despite this, no statistical difference was found in the resulting condition score ca lculations. Study Objectives The main objective of this study involved comparing two distinct sampling methods differing in properties of area, perimeter, and placement in a wetland and comparing the species overlap in 18 Florida wetlands during two growi ng seasons. The questions driving this research are related to the interpretation of species data in regards to accurately representing community structure and wetland condition. These questions included: If wetland plants are not distributed randomly but are instead in zones associated with moisture, would two different sampling methods differing in size and shape still intercept similar species richness and composition? Should autocorrelation be expected in species detection because of the close proximit y of the sample areas or do the sample designs intercept distinct patches of heterogeneity? Would the data collected using different sampling methods result in the same characterization of community structure? Lastly, if there were differences in species detection due to sampling method, could this result in a different interpretation of the condition of the wetland? Methods Site Description Eighteen wetlands in Florida were selected for this study from a larger population of 67 wetlands sampled as part o f the U.S. Environmental Protection Agency 2011 National Wetland Condition Assessment (NWCA). The 18 wetlands were visited twice, once in the growing season of 2011 and once in the growing season in 2012. The 2011
60 season was slightly different from 2012 in that additional field work was conducted for the NWCA. The eighteen wetlands were chosen based on wetland types that accommodated requirements for application of the macrophyte FWCI. The sites were stratified across classification of freshwater wetland regions south (n=2), central (n=5), north (n=7), and panhandle (n=4) (Figure 3 1). Figure 3 2 demonstrates the range in wetland size for the three wetland types. These were, herbaceous depression (HD) (n=5) with a medi an size of forested depression (FD) (n= 6) median size floodplain or strand forest ( FSF ) (n=7) w gradient of reference to impacted surrounding land uses a nd were each unique from one another by wetland type, geography, size and land use type. Sampling Methodology Two different sampling methods, one transect based, the other quadrat based were applied in the field for comparison of vegetation community data and to compare results from wetland condition analysis. The assessment area (AA) center was defined by the greater federal NWCA project using a method employed the General Random Tessellation Stratified survey design (Stevens and Olsen, 2004) which genera ted a random point in a wetland as the center of the sample area. This design was meant to create an unbiased survey of wetland condition over a national distribution from a relatively small sample size of wetlands. In contrast, the original macrophyte FWC I sampling method was not random but systematic; in its initial design, belted transects were placed from the wetland ecotone edge and across the hydrologic gradients within the wetland . The FWCI belted transect sampling methodology was modified to
61 accommo date the design developed for the NWCA . Data collected in quadrats for the NWCA methodology will herein be referred to as quadrat data. Data collected in the belted transects for the FWCI methodology will herein be referred to as transect data. A figure i llustrating the layout of the assessment area is given in Figure 3 3. The AA had a total area of 0.5 ha and a radius of 40 meters. The belted transect design was comprised of 40 meter belted transects laid out in four cardinal directions from the AA cent er point. The quadrat design layout was also oriented along the same cardinal direction layout with the AA. Each transect was divided into four , 5 meter long 1 meter wide units for a total area of 160 m 2 . The five 100m 2 quadrats equaled a total area of 500 m 2 . Each quadrat shared a boundary with one of four transects radiating from the AA center . Due to differences in area and shape the quadrat sampling method had a perimeter to area ratio of 0.4 and the transect sampling method had a perimeter to area ratio of 2. Twenty five square meters of sampling area overlapped between the two methods. The belted transects length exceeded t he boundary of any quadrat edge by 8 15 meters depending on the cardinal direction. Eight transects intercepted the ecotone edge of a wetland but no quadrat crossed an ecotone. For the belted transect, only rooted species were identified to species level, while for quadrat data, v egetation that was either rooted or overhanging was included and identified. The transect method took less time to ap ply in the field, about 2 hours, compared to the quadrat method , which covered 340 m 2 more area and took at least 4 hours to complete. The US Department of Agriculture PLANTS Database (USDA 2012) (http://plants.usda.gov/java/) was used as the nam ing authority. Unknown species were recorded as a pseudonym and were collected for lab identification. Data sheets were
62 entered into a Microsoft Access Â® database and updated when unknown species were subsequently identified. Unknowns that could not be ide ntified were recorded but not included in analysis. Community S tructure Basic descriptive statistics were calculated in Microsoft Excel (2010) to compare vegetation sampled by each method including total richness at a wetland, e venness (E) , Shannon divers ity (H), and Simpson diversity (S). The richness, evenness and diversity measures were then compared using the Pearson correlation coefficient using Minitab 18.104.22.168 (Â©2006 Minitab Inc.) . Each species identified per wetland type was partitioned into groups based on growth habit (e.g. grass, forb, shrub, tree, vine). For each wetland the proportion of species that made up each growth habit category was compared between the two sampling method data sets (transect and quadrat). Species data sets collected by each sampling method were compared for unique and shared species at the same wetland in order to detect spatial autocorrelation between sampling methods. A Jaccard distance measure was used to calculate percent of species that were different between metho ds each year by subtracting species in common to both sampling methods from total species richness and dividing that result by total species richness for both methods. Jaccard distance = |A U B| |A B| / |A U B| (3 1) The Jaccard values ranged from zero to one, zero meaning a site shared the exact same species and one meaning no species were in common. Percent shared species was calculated as the inverse of Jaccard by subtracting the distance value from one and multiplying the result by 1 00.
63 Wetland Condition In order to determine if sampling design intercepted different community structure and resulted in different calculated wetland condition scores; macrophyte FWCI (Lane et al. 2003; Reiss and Brown 2005b; Reiss 2006) was applied for e ach vegetation data set collected from each sampling method (transect and quadrat). Table 3 1 lists individual metrics applied to calculate FWCI condition for each wetland type HD, FD and FSF . FWCI metrics tolerant, sensitive, exotic and mean coefficient of conservatism (CC) are applicable for the three wetland types but have indicator species and CC scores specifically identified per wetland type (Reiss et al. 2009) . Tolerant metrics were based on indicator species analysis that identified species that pe rsisted in wetlands impacted by intensifying land use and were tolerant of physical and chemical parameters associated with increased land use intensity (Lane et al. 2003; Reiss and Brown 2005b; Reiss 2006). Sensitive metrics identified indicator species t hat did not persist in an intensifying landscape and were instead associated with physical and chemical parameters indicative of a reference wetland (Lane et al. 2003; Reiss and Brown 2005b; Reiss 2006). The exotic metric was based on a native or exotic st atus per species as defined by the United States Department of Agriculture PLANTS Database ( http://plants.usda.gov ) (USDA 2012). The coefficient of conservatism score is derived from the Floristic Quality Assessment I ndex (FQAI) after Wilhem and Ladd (1988) where species were assigned values on a scale of 0 10 based on expert botanist opinion of species tolerance to disturbance (Cohen et al. 2005) . Species taxonomy, duration, native status and wetland indicator status were based first on the Florida Department of Environmental Protection standard operating procedures (SOP)
64 document DEP SOP 002/01 LT 7000 Determination of Biological Indices (DRAFT 2012 ), if a species was not listed then USDA (2012) was used . The DEP SOP was also the source for CC values if a species was not listed Reiss and Brown. (2005a, 2005b) and Lane et al. (2003) were utilized , if a species did not have a CC value it was not included in scoring that metric. Herbaceous wetla nds had an additional metri c measuring the ratio of annual to perennial species (Lane et al. 2003) . Forested wetlan ds both FD and FSF wetlands had a metric for proportion of native and perennial species (Reiss and Brown 2005b, Reiss 2006). Only FD wetlands had a metric defining perc ent of species with a wetland indicator status of obligate or facultative wet (Reiss 2006) . Metric calculations were based on the proportion of qualifying indicators out of total species richness detecting in the sampling area. The outer 5 th and 95 th perce ntiles were normalized so each metric represented a zero to ten range, ten being optimal condition for that metric (Reiss 2006). Wetland condition scores were derived by combining individual metric scores and scaling values to 100% for cross wetland type c omparisons. Reference wetlands representing optimal condition would equal 100%. The proportion of indicator species (used to calculate metrics) and wetland condition score determined by each sampling method were compared using linear regression and tested Microsoft Excel v. 14 . Results The results from in situ field collection and data analysis are presented in terms of did wetland condition scores calculated from the different data sets based on samplin g methods, reflect the same interpretation of condition. Then, whether community structure and composition reflected differences in the sampling method and whether this affected condition scores will be addressed.
65 Wetland Condition C omparison The compari son of condition scores calculated from each sampling method demonstrated a strong linear relationship. Condition score results calculated for the transect method were an unbiased predictor of the quadrat score both years (Figure 3 4) (R 2 = 0.79 in 2011, R 2 = 0.94 in 2012) with a slope of 0.96 and 0.99 when the intercept was set to zero. Table 3 2 reports results from the two test suggesting there is a 1:1 relationship between condition scores calculated by either quadrat or tran sect in 2011. Figure 3 4 also examines the linear relationship within method between years (e.g. transect scores 2011 v. 2012 or quadrat scores 2011 v. 2012). The results are highly linear for both within method comparisons resulting in high R 2 values (tr ansect R 2 = 0.99, quadrat R 2 test (Table 3 2) also suggests a 1:1 relationship between years for within method calculated condition scores. Tables 3 3 through 3 5 , lists condition scores and individual metric score results by wetland , wetland type, sampling method and year. Proportion of Indicator Species for Calculating M etrics Condition score is calculated by combining several metrics, the tolerant, sensitive, exotic and mean CC metrics are all applicable to quantifying condition fo r the three wetland types. Fluctuation in the proportion of indicator species per method could be possible even if the calculated condition scores appeared 1:1. For example, if the quadrat method scored a 10 in tolerant but a 5 for sensitive and the trans ect method scored 5 for tolerant and 10 for sensitive, the proportion of indicator species would be different per method for those metrics but the final condition score would be the same, this however was not the case. Table 3 6 demonstrates that the propo rtion of indicator species for each metric had a strong linear relationship between transect and quadrat
66 methods for both years of data collection (R 2 values ranged from 0.88 1. 00). A test also suggested a 1:1 relationship for each comparison of proportion of indicators . Community S tructure Diversity s tatistics Diversity measures describing species data collected by either method did not suggest the methods sampled different community structure. Pearson coefficients for diversity measures of community composition are presented in Table 3 7 . Comparisons were significantly correlated between the two sampling methods for both years. Only Evenness (E), was not statistica lly correlated, but this was not an unexpected result due to the large size difference between transect and quadrat areas. Proportion of species by growth habit The differences in size and shape of transect and quadrat sampling areas might result in dete ction differences based on species growth habitat. For example, a large quadrat might be a better sampling tool for large woody tree species and small area more appropriate for detecting grass and forbs. Each species identified per wetland type was partiti oned into groups based on growth habit (e.g. grass, forb, shrub, tree, vine). For each wetland the proportion of species that made up each growth habit category was compared between the two sampling method data sets (transect and quadrat) (Figure 3 5). Eac h method was relatively consistent both years in detecting specific growth habits. In HD wetlands where most species were grasses or forbs, both methods appear to detect similar proportions. Where growth habit becomes more diversified among grasses and for bs, shrubs, trees and vines, the two methods began to vary
67 slightly for FD and FSF wetlands. In both wetland types the transect method detected a higher proportion of grass and forb species while the quadrat method detected a higher proportion of shrub, tr ee and vine species suggesting some variation in detection of community structure by method. Shared species Even though there was high correlation in wetland condition scores and proportion of indicator species, p ercent of species detected by both methods within a given year was lower than expected. This could indicate a difference in detection of community structure by method despite previous results comparing diversity statistics for the two sampling methods. A box plot demonstrating the range and variati on of shared species between methods and years is given in Figure 3 6. Median shared from 36% to 71%. In 2012 median shared species detected by both transects and quadrats Figure 3 6 demonstrates , median s hared species detected by the two sampling methods in the same year (e.g. transect v. quadrat 2011 or transect v. quadrat 2012) was higher than shared species detected by the same sampling method compared to the following year (e.g. transect 2011 v. 2012 or quadrat 2011 v. 2012) . The median shared species by transect only method between years was 55% ( 9) and ranged from 44% to 77% . The quadrat method had a median o f 57% ( 11) shared species detected for both years of sampling and ranged from 40% to 79% . Sampling within year had a higher median shared species percent than inter annual variation within method.
68 Discussion The wetland condition scores calculated by e ither method were highly correlated as expected, what was unexpected was that this relationship was not predicted by the percent shared species detected by the methods. Although shared species varied , the proportion of indicator species did not; this sugge sted indicators of condition persisted even if individual species presence changed. This is strong evidence for confidence in the wetland condition method. This dataset only represented 18 wetlands; the sample size was not large enough to describe the dis tribution of variance between the calculated wetland scores. Even though the condition scores were highly correlated, outlier sites had gaps in condition scores as large as 10% (four wetlands in 2011 and two in 2012). While this may not be an ecologically important difference or a statistically significant result, a larger study could find more examples of score gap that would warrant further exploration . The wide range and variability in detected shared species hints that the methods intercepted different areas at the same wetland. This was further amplified by the variation in proportion of species by growth habit detected by either method. If adjacent sampling units intercepted such variation in vegetative species, the resulting data could be interpreted as representing different community structure. This would in fact be incorrect and could mean that heterogeneity in the wetland was not well represented by a single sampling method. Physical environmental parameters could have increase d heterogeneity in a wetland and explain why some wetlands had higher or lower shared species detected by method. The following discussion explores why the methods may not be interchangeable and why caution should be applied when designing sampling methods to collect a repres entative sample.
69 Examples of Potential V ariability Plots with high perimeter to area ratios are expected to intercept greater heterogeneity on a landscape; long thin plots will intercept greater combinations of plant species interactions ( Stohlgren et al. 1995; Peet et al. 1998 ). The transect method had a perimeter to area ratio four times (transect perimeter area ratio 2) that of the quadrat method (perimeter area ratio 0.4). In wetlands where vegetation zonation occurs across spatial gradients driven by h ydrology (van der Valk 1981), a method such as the belted transects, that encounters heterogeneity may be desired as long as the transect stays within community type. The starting location can bias the effectiveness of the plot layout intercepting a repres entative sample (Stohlgren et al. 1995). A sampling regime that intercepted zones associated with the hydrologic gradient in multiple places would capture the heterogeneity of the wetland and collect a more representative sample. Crossing an ecotone incre ases potential for intercepting heterogeneity in a wetland. The transition area between a wetland and upland is dynamic. Wetland edges can be invaded by upland species, including woody and exotic vegetation during periods of drought or fire suppression ( My ers and Ewel 1990 ; FNAI 2010; Bried et al. 2013). Likewise, prescribed fire or inundation could push ecotone edges outw ard from the wetland ( Myers and Ewel 1990 ; FNAI 2010 ). By layout design, the quadrats were more interior in the AA and had the potential to capture less variation along a hydrologic gradient. While no quadrats intercepted ecotones, transects did in eight wetlands. Microtopography increases heterogeneity of species in wetlands ( Huenneke 1986 ; Titus 1990 ; Moser et al. 2007). Sampling areas th at had dynamic patches of microtopgraphy could have increased species associations. The presence of standing water, occurrence of drought and differenc es in land management activity could have
70 amplified the effect of microtopography on species composition. Notes taken in the field on the topographic complexity of some wetland AAs characterized them as having multiple horizontal plains at different elevations, animal spoil piles from burrows, plant hummocks and tussocks, and patches of bare ground. This topo graphic complexity could explain the low percentage of shared species; m ost likely transects and quadrats sampled slightly different areas as small patches of heterogeneity were intercepted. A sampling regime that intercepted the zones associated with the hydrologic gradient in multiple places is more likely to capture heterogeneity of the wetland (Murray Hudson et al. 2012). In the field, standing tanic water was sometimes noted as physically affecting detection of species by making it difficult to see spe cies presence. This would be more likely in the deeper interior zones intercepted by quadrats. Along hydrologic gradients, shallow water edges might have greater species detection due to improved visibility. Along the hydrologic gradient wetland edges may also have different species presence due to saturated but not inundated soils emerging from the seed bank. The wetland community structure within a given AA was not different but interpreting vegetation data collected for either method could have led to th at conclusion if they were not considered within the greater context of the physical gradients within the wetland. The differences in size and area of transect and quadrat sample units meant data was collected at potentially different scales, intercepting the vegetation in such a way that the interpretation of ecological phenomena, in this case community structure, could be different. The large range of shared species identified by the two sampling methods in the same wetlands indicates the AAs were not hom ogenous areas. It does
71 appear in some cases the methods sampled distinctly heterogeneous pockets within the wetland. With this small data set it was difficult to say if one method was more effective at collecting a representative sample. This was further c onfounded by the large differences in total wetland area in which the relatively small AA was embedded. However, physical attributes of any study area should be considered when planning a sampling design. Spatial gradients and how they interact with dynami c circumstances such as the presence of surface water or recent management activities such as fire must be considered when planning for a site assessment. Study Limitations Due to the random placement of the AA and the size of wetlands in this study, samp le areas may not have captured representative species composition associated with zones along a hydrologic gradient. The FWCI sampling method was developed to intercept the hydrologic and microtopographic gradients of depressional or linear wetlands to max imize species area relationships ( Reiss and Brown 2005b ; Murray Hudson et al. 2012). The random assignment of AAs in this study, based on NWCA 2011 site selection, may not have been the best application of the FWCI method to capture representative spatial gradients, especially for large wetlands. Conclusion Comparing wetland condition scores calculated from two different sampling methods demonstrated very strong indicators of condition even when the methods intercepted different vegetation in the same wet land. This emphasized the strength of the vegetation Florida Wetland Condition Index as an IBI method. Even though percent shared species varied, species indicative of condition were redundant and persisted resulting in the 1:1 condition scores. Whether a specific sampling method collected the
72 more representative sample in this study could not be determined from the small sample pool. Other studies applying the vegetation portion of the FWCI did recommend belted transects should be oriented across hydrologi c gradients and intercept ecotones to capture heterogeneity in wetlands ( Reiss and Brown 2005b ; Murray Hudson et al. 2012). The variation in growth habit and species encountered, as well as field notes characterizing the physical condition of the wetlands pointed out heterogeneity in the sampling area. It stands to reason that caution should be applied in designing a sampling method layout to ensure a representative sample is collected.
73 Table 3 1 . Metrics included in final Florida Wetland Condition Index c alculation by wetland type. Metric Herbaceous Depression Forested Depression Forested Strand or Floodplain Proportion Tolerant Species X X X Proportion Sensitive Species X X X Proportion Exotic Species X X X Modified Floristic Quality Assessm ent Index(FQAI) X X Mean Coefficient of Conservatism (CC) X Annual Perennial Ratio X Proportion Native Perennial X X Percent Wetland Indicator Species X Table 3 between wetland condition scores, calculated with data from different sampling methods. RÂ² Slope Standard error of the slope Estimated t Critical t 2011 transect v. quadrat 0.97 0.96 0.04 1.06 2.11 2012 transect v. quadr at 0.99 0.99 0.02 0.54 2.11 transect 2011 v. 2012 0.99 0.99 0.02 0.33 2.11 quadrat 2011 v. 2012 0.98 1 0.04 0.16 2.11
74 T able 3 3. Individual metric values and final Florida Wetland Condition Index (FWCI) scores calculated for transect (T) and quadrat (Q) in 2011 and 2012 i n herbaceous depression wetlands . The coefficient of conservatism metric is represented by CC. Sampled wetlands are listed by site id. Individual metric scores range from 0 to 10. FWCI score ranges 0 to 100 (100 equaling reference condition).
75 T able 3 4. Individual metric values and final Florida Wetland Condition Index (FWCI) scores calculated for transect (T) and quadrat (Q) in 2011 and 2012 in forested depression wetlands. Forested depression 1314 1325 1336 1346 3065 2011 T/Q T/Q T/Q T/Q T/Q Tol erant 8/8 6/5 6/7 5/6 10/8 Sensitive 1/1 0/0 0/1 0/0 0/0 Exotic 10/9 10/7 10/10 10/10 10/10 CC 4/5 8/7 6/7 6/7 10/10 Native Perennial 4/5 10/9 9/8 8/10 10/10 Wetland Status 10/10 5/5 8/5 7/6 7/9 FWCI 62/63 65/55 65/63 60/65 78/78 2012 T/Q T/Q T/Q T/Q T/Q Tolerant 10/10 10/10 10/10 10/10 10/10 Sensitive 0/0 0/0 0/0 0/0 0/0 Exotic 4/5 6/6 7/7 7/8 10/10 CC 4/5 6/6 7/7 7/8 10/10 Native Perennial 4/6 10/7 8/9 9/9 10/10 Wetland Status 9/7 5/2 3/4 3/5 10/9 FWCI 52/55 62/52 58/62 60/67 83/82 T he coefficient of conservatism metric is represented by CC. Sampled wetlands are listed by site id. Individual metric scores range from 0 to 10. FWCI score ranges 0 to 100 (100 equaling reference condition).
76 T able 3 5. Individual metric values and final F lorida Wetland Condition Index (FWCI) scores calculated for transect (T) and quadrat (Q) in 2011 and 2012 in forested strand or floodplain wetlands. Forested strand or floodplain 1238 1280 1324 1365 3052 3073 3077 2011 T/Q T/Q T/Q T/Q T/Q T/Q T/Q Toleran t 6/6 2/0 10/10 7/7 6/3 8/9 6/2 Sensitive 5/7 4/3 3/5 10/10 6/7 10/9 8/10 Exotic 7/8 8/7 5/8 10/10 10/10 5/10 10/10 CC 3/5 8/7 10/10 10/9 8/8 7/7 10/9 Native Perennial 5/5 9/8 6/9 10/7 8/9 10/10 10/10 FWCI 52/62 62/50 68/84 94/86 76/74 80/90 88/82 2012 T/Q T/Q T/Q T/Q T/Q T/Q T/Q Tolerant 4/4 4/0 10/10 7/7 4/2 7/10 7/6 Sensitive 3/3 3/2 3/2 10/10 9/9 10/8 7/8 Exotic 8/8 6/8 8/10 8/10 10/10 10/10 10/10 CC 5/6 9/8 10/10 9/10 8/9 7/10 10/10 Native Perennial 5/5 7/9 6/8 9/8 9/9 10/10 10/9 F WCI 50/52 58/54 74/80 86/88 80/78 88/96 88/86 The coefficient of conservatism metric is represented by CC. Sampled wetlands are listed by site id. Individual metric scores range from 0 to 10. FWCI score ranges 0 to 100 (100 equaling reference condition).
77 T able 3 proportion of indicator species detected by different sampling methods for metrics Tolerant, Sensitive, Exotic and mean coefficient of conservatism (CC). Transect v . Quadrat proportion of indicators RÂ² Slope Standard error of the slope Estimated t Critical t 2011 Tolerant 0.88 0.94 0.08 0.769 2.11 2012 Tolerant 0.92 0.94 0.07 0.82 2.11 2011 Sensitive 0.98 0.91 0.05 2.03 2.11 2012 Sensitive 0.94 0.97 0.06 0.44 2.11 2011 Exotic 0.87 0.93 0.09 0.82 2.11 2012 Exotic 0.89 0.94 0.08 0.73 2.11 2011 mean CC 0.99 0.98 0.02 1.21 2.11 2012 mean CC 1.00 0.99 0.01 0.53 2.11 Table 3 measures 2011 2012 Richness ( R) 0.87 0.94 Evenness ( E) 0.28 0.26 Shannon (H) 0.92 0.9 Simpson (S) 0.91 0.93 Note: values in bold indicate p < 0.05
78 Figure 3 1 . Locations of wetlands in study herbaceous depression (HD) , forested depression (FD), f orested strand floodplain (FSF) HD FD FSF
79 Figure 3 2. B ox plot of wetland size broken out by wetland type ; herbaceous depression (HD), forested depression (FD ), for ested strand floodplain (FSF ) . Interquartile box shows the 25 th 75 th quartile intercepted by the interior line for median richness. Box whiskers represent the entire range of richness values and the asterisks are outliers. Outlier of 4,578 ha in FSF not shown. HD (n=6) FD (n=5) FSF (n= 7)
80 Figure 3 3 . Layout of belted transect s ampling method ( left) and quadrats (right ) for the within the same 0.5 ha Assessment Area (AA).
81 Figure 3 4. Condition scores compared by method and year, blue circles herbaceous depression (HD), green triangles forested depression (FD) wetlands, blue triangles forested strand or floodplain (FSF) wetlands.
82 Figure 3 5. Proportion of species by growth habitat for each wetland type and year. transect (dotted polygon) and quadrat (solid red line)
83 Figure 3 6. Percent shared species detected by both methods within the same ye ar, Transect v. Quadrat (for 2011 and 2012) or within method, Transect 2011 v. 2012 or Quadrat 2011 v. 2012. The comparisons between methods had the higher shared species within the same year.
84 C HAPTER 4 SUMMARY AND CONCLUSIONS Wetlands provide valuable s ervices to people and wildlife (Euliss et al. 2008) both immediately adjacent to and in the greater ecosystem landscape; policies of no net loss of wetlands are in place because of this recognition. The physical qualities that wetlands possess to provide e cosystem services also make wetlands vulnerable to cumulative downstream effects of land use change and watershed impacts (Dahl 2011). In order to determine if wetlands are being impacted by anthropogenic changes, wetlands are monitored for degradation. In dexes of Biotic Integrity (IBI) evaluate changes to biological assemblages in response to wetland degradation (Karr and Dudley 1981). This is done by identifying indicator species that correlate with physical and chemical parameters of a wetland that are i ndicative of reference or impaired condition (Reiss et al. 2009). Many IBIs have been developed since the concept was first introduced ( Ruaro and Gubiani 2013 ) for many different biological assemblages (e.g., fish, macroinvertebrates, diatoms etc.) species presence are compared to a reference standard indicative of system integrity (Karr and Dudley 1981). These assemblages should reflect complex trophic interactions not directly measured and be sensitive to the conditions being monitored (Karr et al. 1986). The data used to develop an IBI should be indicative of the wetland, have reproducible results with low variability over space a nd time by not have over sensitivity to natural variation (Karr et al. 1986). Although IBI methods have been developed for many wetland systems ( Ruaro and Gubiani 2013 ) there were comparatively few examples of validation of these methods in the literature. These methods must meet the afore mentioned criteria if practitioners and
85 policy makers are to have confidence that IBI method s are important monitoring tools for protecting wetland integrity and determining if policies of no net loss are working. The Florida Wetland Condition Index (FWCI) (Reiss et al. 2009) is a multi metric index of biotic integrity developed for herbaceous ( Lane et al. 2003) and forested depressions (Reiss and Brown 2005a) and forested strand or floodplain wetlands (Reiss and Brown 2005b). Th ree separate indices were developed for diatom , vegetation , and macroinvertebrate assemblages (Reiss et al. 2009). Many IBIs are based on wetland vegetation. Macrophytes make good indicators of condition ( Miller et al. 2006 ; Mack 2007; Stapanian 2013 ); they respond to environmental change ( Doherty et al. 1999 , Reiss and Brown 2005a), are pervasive in many different types o f wetlands and as primary producers are the foundation of trophic interactions. While all three FWCI indices have statistically significant relationships with adjacent land use intensity (Reiss et al. 2009), the vegetation FWCI requires less laboratory tim e for condition calculation because plants are relatively easy to ide ntify to species in the field. In 2011 and again in 2012, eighteen wetlands were visited in the growing season and evaluated with the vegetation portion of the FWCI. These wetlands repres ented a range of geographic area across the Florida peninsula and represented three different wetland types, herbaceous depression (HD) (n=6), forested depression (FD) (n=5) and forested floodplain or strand forest (FSF) (n=7). The wetlands represented dif ferent stages of succession and di fferent adjacent land use type indicative of a gradient of intensified land use. These wetlands were never evaluated under the development of the FWCI method. By applying the FWCI at different types of wetlands with differ ent environmental variables affecting condition and visiting these wetlands in two
86 consecutive growing seasons this study aimed to identify if the FWCI would be consistent at defining condition within the expected range of variation due to normal fluctuati ons in climatic regime. Additionally, during each site visit, two sampling methods which differed in perimeter, area and orientation were applied to study how the sample unit could affect data collection. The study asked if sampling methods collected diff erent data due to physical attributes of the sample units , would the FWCI still have a consistent identification of wetland condition. By evaluating the consistency and repeatability of the FWCI and studying how condition results might be manipulated by s ample design this study contributed to building confidence in FWCI as a monitoring tool for tracking impacts to and net loss of wetlands. FWCI condition scores had statistically significant correlation between years. Indicator species either persisted bet ween years and/or the proportion of indicators remained constant. There was no indication of change to community structure between years but a difference in species composition was greater than expected. On average as many as 45% of species were different the following year. The changes to species composition indicated there was natural variation which may have been a response to climatic variables such as drought and flooding that occurred between sampling periods. FWCI condition scores had high fidelity b etween years despite shifts in species composition indicating the FWCI continued to reflect adjacent land use but was not sensitive to potential natural variation in wetlands. In comparing the two sampling methods used to collect vegetation data each samp ling period there was again a statistically significant correlation between FWCI scores. The proportion of indicator species was also significantly correlated between the
87 methods. Evidence from community structure and species composition data indicated tha t the sample units intercepted different spatial gradients and heterogeneity in the wetlands; this result was supported by Murray Hudson et al. (2012). Murray Hudson et al. (2012) sampled different vegetation zones in depression marshes; even though speci es composition varied, FWCI scores were not statistically different indicating that the proportion of indicators remained constant. In this study there was no statistical difference in condition scores calculated the different sampling methods, these resul ts did not indicate one method was preferable over the other. However, because wetlands had directional gradients related to topography and hydrology and had areas of heterogeneity the best sample unit layout should intercept these gradients so that a repr esentative sample would be collected for future studies. Both the comparison of FWCI scores between years and the comparison of scores calculated from different sampling methods demonstrated that vegetation indicators in the FWCI were robust and either per sisted or were replaced by species that indicated the same condition thereby retaining the proportion of indicators and maintaining consistent scores. Wetland condition scores in this study were representative of what would be expected for adjacent land u se classes but determining whether the scores were accurate was outside the scope of this study. What is unknowable from this study is if condition results describe the wetland or if results only describe the study area; this is a common problem in ecolog y ( Fo rtin 1999 ; Dungan et al. 2002 ; Stohlgren et al. 2003). Context is important and knowing something about the adjacent land use is helpful. Comparing the vegetation condition results to other indices such as diatoms and macroinvertebrate s could strengt hen the interpretation of wetland condition by providing
88 a more complete picture. In Reiss et al. (2009) three wetlands surrounded by an intensified land use had FWCI scores that were not in agreement for the multiple FWCI metrics. These wetlands had FWCI scores under 12% for diatom and macrophyte indices but macroinvertebrate FWCI had ranges greater than 50%. In another example from Reiss et al. (2009) a wetland embedded in a developed landscape had a high macrophyte condition score but low diatom and mac roinvertebrate scores. In fish IBI studies Karr et al. (1987) and Fore et al. (1994) suggested that systems experiencing greater anthropogenic impacts could reflect greater variability in condition scores. S ites with ecological integrity experienced little condition variability this was thought to be due to elastic properties that allowed wetlands with integrity to cope with natural variation (Fore et al. 1994) . Although Reiss et al. (2009) found some disagreement in the three different IBIs, there was agr eement in most wetlands thereby increasing confidence in the methods for interpreting condition. There is potential that indicators representing different trophic levels could respond more quickly to impacts or recover more slowly from impacts or have an e ntirely different type of spatial distribution related to physical and temporal variation. Indicators that are highly mobile and have cyclical life histories (e.g., macroinvertebrates) might experience greater spatial, seasonal and inter annual variation a ffecting species presence and distribution (Mazor et al. 2009). Mazor et al. (2009) recommended incorporating multiple metrics in indices to smooth out effects of short term variation but concluded that the most accurate indices must incorporate long term datasets into calibration. Fish IBIs are also based on highly mobile species, Fore et al. (1994) and Karr et al. (1987) found sample design had a strong effect on capturing
89 representative species that had patchy distribution due to seasonal effects; timing and location were integral to collecting a representative sample (Pyron et al. 2007) and accurately identifying condition . In Florida there are longer growing season where plants are still identifiable, plants are also typically less mobile . These physica l traits could make plants less sensitive to annual and seasonal variation; and thereby result in less variability in the FWCI. Caution should be applied when relying on vegetation as proxy for condition because of potential time lag in vegetation response to impact s (Reiss et al. 2009). This study demonstrated there was potential for high species turnover in a period of a year affecting specie composition. Deimeke et al. (2013) also demonstrated high species turnover of macrophytes on a longer time scale o f seven to eight years with little variability of FWCI condition scores. In both studies land use remained relatively constant which emphasized FWCI is sensitive to land use impacts (i.e., scores reflected adjacent land use) but not natural variation. Vali dation from other indices would help strengthen this conclusion. Some developers of IBI methods have advocated for harnessing long term data sets from institutions to help in the calibration of IBIs to identifying long term cycles and variability. Even tho ugh this study did not find variation in FWCI scores, and identified indicator species that either persisted or were replaced by species with the same status; long term data of plant distribution and climate could help identify patterns in high species tur nover. This type of data may not be practically acquired but with the importance of spatial and temporal scale to IBI consistency; IBIs that are not validated and that were developed with short time frames may not be as beneficial for tracking net loss of wetlands.
90 LIST OF REFERENCES Angermeier PL, Karr JR (1986) Applying an Index of Biotic Integrity Based on Stream American Journal of Fisheries Management 6:418 429. doi: 10.1577/154 8 8659(1986)6<418 Barbour, MG , Burk JH , and Pitts WD (1987) Terrestrial Plant Ec ology. Second Edition. Benjamin Cummings Publishing Company, Menlo Park, CA . Bried JT, Jog SK, Matthews JW (2013) Floristic quality assessment signals human disturbance over na tural variability in a wetland system. Ecological Indicators 34:260 267. doi: 10.1016/j.ecolind.2013.05.012 Chapin DM, Paige DK (2013) Response of Delta Vegetation to Water Level Changes in a Regulated Mountain Lake, Washington State, USA. Wetlands 33:431 444. doi: 10.1007/s13157 013 0401 5 Cohen MJ, Carstenn S, Lane CR (2004) Floristic quality indices for biotic assessment of depressional marsh condition in Florida. Ecological Applications 14:784 794. Dahl TE (2011) Status and Trends of Wetlands in the Con terminous United States 2004 to 2009. U.S. Department of the Interior; Fish and Wildlife Service, Washington, D.C. 108. Deimeke E, Cohen MJ, Reiss KC (2013) Temporal stability of vegetation indicators of wetland condition. Ecological Indicators 34:69 75. d oi: 10.1016/j.ecolind.2013.04.022 Review of Technical & Scienti fic Literature ( 1990 1999 ) A report to United States Environmental Protection Agency Assessment of Wetlands Wo rkgroup Dormann CF (2007) Effects of incorporating spatial autocorrelation into the analysis of species distribution data. Global Ecology and Biogeography 16:129 138. doi: 10.1111/j.1466 8238.2006.00279.x Dungan, JL, Perry, JN, Dale, MRT, Legendre, P, Cit ron Pousty, S, Fortin, MJ, Jakomulska, A, Miriti, M, Rosenberg MS (2002) A Balanced View of Scale in Spatial Statistical Analysis. Ecography 25:626 640. Dutilleul P (1993) Spatial heterogeneity and the design of ecological field experiments. Ecology 74:164 6 1658. Euliss NH, Mushet DM (2011) A multi year comparison of IPCI scores for prairie pothole wetlands: implications of temporal and spatial variation. Wetlands 31:713 723. doi: 10.1007/s13157 011 0187 2
91 Florida Departm ent of Environmental Protection (201 2) DEP SOP 002 01 LT 7000 Determination of Biological Indices Draft http://www.dep.state.fl.us/water/sas/qa/docs/62 160/lt 7000 biological indices.pdf , Acces sed 1 June 2011 Florida Natural Areas Inventory. (2010) Guide to Natural Communities. http://www.fnai.org/natcom_accounts.cfm/freshwater , Accessed 28 April 2014 Fore LS, Karr JW, Conquest LL (1994) Statistical properties of an index of biological integrity used to evaluate water resources. Canadian Journal of Fisheries and Aquatic Sciences 51:1077 1087. Fortin M J (1999) Effects of quadrat size and data measurement on the detection of bound aries. Journal of Vegetation Science 10:43 50. Francis CM, Austen MJW, Bowles JM, Draper WB (2000) Assessing Floristic Quality in Southern Ontario Woodlands. Natural Areas Journal 20:66 77. He F, Legendre P (2002) Species Diversity Patterns Derived from Sp ecies Area Models. Ecology 83:1185 1198. Herricks EE, Schaeffer DJ (1985) Can we optimize biomonitoring? Environmental Management 9:487 492. doi: 10.1007/BF01867323 Hobbs NT (2003) Challenges and opportunities in integrating ecological knowledge across sca les. Forest Ecology and Management 181:223 238. doi: 10.1016/S0378 1127(03)00135 X Huenneke LF, Sharitz RR (1986) Microsite abundance and distribution of woody seedlings in a South Carolina cypress tupelo swamp. American Midland Naturalist 115:328 335. Jef fries M (2008) The spatial and temporal heterogeneity of macrophyte communities in thirty small, temporary ponds over a period of ten years. Ecography 31:765 775. doi: 10.1111/j.0906 7590.2008.05487.x Karr J, Dudley D (1981) Ecological Perspective on Water Quality Goals. Environmental Management 5:55 68. Karr JR, Fausch KD, Angermeier PL, et al. (1986) Assessing Biological Integrity in Running Waters A Method and Its Rationale. Karr JR, Yant PR, Fausch KD, Isaac J (1987) Spatial and Temporal Variability of the Index of Biotic Integrity in Three Midwestern Streams. Transactions of the American Fisheries Society 116:1 11. doi: 10.1577/1548 8659(1987)116<1 Karr, JR., Chu EW (1997) Biolog ical monitoring and assessment: Using multimetric indexes e ffectively. Sea ttle, WA: University of Washington, EPA 235 R 97 001
92 Kirkman KL, Drew MB, West LT, Blood ER (1998) Ecotone Characterization between upland Longleaf Pine/wiregrass stands and seasonally ponded isolated wetlands. Wetlands 18:346 364. Kutcher TE, Bried JT (20 14) Adult Odonata conservatism as an indicator of freshwater wetland condition. Ecological Indicators 38:31 39. doi: 10.1016/j.ecolind.2013.10.028 Lane, C.R. 2000. Proposed ecological regions for freshwater wetlands of Florida, Masters Thesis, University o f Florida Lane CR, Brown MT, Murray hudson M, Vivas MB (2003) The wetland condition index indicators of wetland condition for isolated depressional herbaceous wetlands in Florida Report Submitted to the Florida Department of Environmental Protection Contra ct #WM 683. Development 1 165. Legendre P, Dale MRT, Fortin M, et al. (2002) The consequences of spatial structure for the design and analysis of ecological field surveys . Ecography 25:601 615. Legendre P, Dale MRT, Fortin M, et al. (2004) Effects of spati al structures on the results of field experiments . Ecology 85:3202 3214. Award lecture. Ecology 73:1943 1967. Lopez RD, Fennessy SM (2002) Testing the floristic quality a ssessment index as an indicator of wetland condition. Ecological Applications 12:487 497. Mack JJ (2007) Developing a wetland IBI with statewide application after multiple testing iterations. Ecological Indicators 7:864 881. doi: 10.1016/j.ecolind.2006.11. 002 Mack JJ, Avdis NH, Braig EC, Johnson DL (2008) Application of a vegetation based index of biotic integrity for Lake Erie coastal marshes in Ohio. Aquatic Ecosystem Health & Management 11:91 104. doi: 10.1080/14634980701880823 Matthews JW, Tessene P a., Wiesbrook SM, Zercher BW (2005) Effect of area and isolation on species richness and indices of Floristic Quality in Illinois, USA wetlands. Wetlands 25:607 615. doi: 10.1672/0277 5212(2005)025[0607:EOAAIO]2.0.CO;2 Matthews JW, Spyreas G, Endress AG (2009 ) Trajectories of vegetation based indicators used to assess wetland restoration progress. Ecological Applications 19:2093 2107. Mazor RD, Purcell AH, Resh VH (2009) Long term variability in bioassessments: a twenty year study from two northern California streams. Environmental management 43:1269 86. doi: 10.1007/s00267 009 9294 8
93 Miller S, Wardrop D, Mahaney W, Brooks R (2006) A plant based index of biological integrity (IBI) for headwater wetlands in central Pennsylvania. Ecological Indicators 6:290 312. doi: 10.1016/j.ecolind.2005.03.011 Mitsch, W J , Gosselink, JG ( 2000 ) Wetlands, 3rd edition. John Wiley and Sons, Inc. New York, New York Moncayo Estrada R, Lyons J, Escalera Gallardo C, Lind OT (2012) Long term change in the biotic integrity of a shallow t ropical lake: A decadal analysis of the Lake Chapala fish community. Lake and Reservoir Management 28:92 104. doi: 10.1080/07438141.2012.661029 Moser K, Ahn C, Noe G, Survey USG (2007) Characterization of microtopography and its influence on vegetation pat terns in created wetlands. Wetlands 27:1081 1097. Myers, R, Ewel, J (1990) Ecosystems of Florida. University of Central Florida Press, Gainesville, FL Murray Hudson M, Lane CR, North S, Brown MT (2012) Macrophyte Species Distribution, Indices of Biotic Int egrity, and Sampling Intensity in Isolated Florida Marshes. Wetlands. doi: 10.1007/s13157 012 0278 8 National Hurricane Center. (2013) Tropical Cyclone Report Tropical Storm Debby (ALO42012) 23 27 2012 http://www.nhc.noaa.gov/data/tcr/AL042012_Debby.pdf Ac cessed 1 December 201 3 . NOAA National Climatic Data Center, State of the Climate: Drought for August 2011, http://www.ncdc.noaa.gov/sotc/drought/2011/8 Accessed 16 February 2014 NOAA National C limatic Data Center, State of the Climate: Drought for June 2012, http://www.ncdc.noaa.gov/sotc/drought/2012/6 Accessed 16 February 2014 Palanisamy B, Chui TFM (2013) Understanding wetland plant dynamics in response to water table changes through ecohydrological modelling. Ecohydrology 6:287 296. doi: 10.1002/eco.1268 Palmer MW, White PS (1994) Scale dependence and the Species Area relationship. The American Naturalist 144:717 740. Pasquaud S, Cou rrat A, Fonseca VF, et al. (2013) Strength and time lag of relationships between human pressures and fish based metrics used to assess ecological quality of estuarine systems. Estuarine, Coastal and Shelf Science 134:119 127. doi: 10.1016/j.ecss.2013.02.00 2 Peet RK, Wentworth TR, White PS (1998) A flexible, multipurpose method for recording vegetation composition and structure. Castanea 63:262 274.
94 Pyron M, Lauer TE, LeBlanc D, et al. (2007) Temporal and spatial variation in an index of biological integrity for the middle Wabash River, Indiana. Hydrobiologia 600:205 214. doi: 10.1007/s10750 007 9232 9 Ramberg L, Lindholm M, Hessen DO, et al. (2010) Aquatic ecosystem responses to fire and flood size in the Okavango Delta: observations from the seasonal floodp lains. Wetlands Ecology and Management 18:587 595. doi: 10.1007/s11273 010 9195 x Reed, R.A., Peet, R.K., Palmer, M.W., White PS (1993) Scale dependence of vegetation woodland. Journal of Vegetation Science 4:329 340. Reiss KC, Brown MT (2005)a The Florida Wetland Condition Index (FWCI) Developing biological ind icators for isolated depressional forested wetlands. Final report to the Florida Department of Environmental Protection http://www.dep.state.fl.us/labs/library/index.htm Accessed 1 February 2011 Re iss, KC, Brown MT (2005)b Pilot study The Florida Wetland Condition Index (FWCI): preliminary development of biological indicators for forested strand and floodplain wetlands. F inal report to the Florida Departm ent of Environmental Protection http://www.dep.state.fl.us/labs/library/index.htm Accessed 1 February 2011 Reiss K (2006) Florida Wetland Condition Index for depressional forested wetlands. Ecological Indicators 6:337 352. doi : 10.1016/j.ecolind.2005.03.013 Reiss KC, Brown MT (2007) Evaluation of Florida palustrine wetlands: Application of USEPA levels 1, 2, and 3 assessment methods. EcoHealth 4:206 218. doi: 10.1007/s10393 007 0107 3 Reiss KC, Brown MT, Lane CR (2009) Characte subtropical wetlands: the Florida wetland condition index for depressional marshes, depressional forested, and flowing water forested wetlands. Wetlands Ecology and Management 18:543 556. doi: 10.1007/s11273 009 9132 z water table feedbacks on the stability and resilience of plant ecosystems. Water Resources Research 42:n/a n/a. doi: 10.1029/2005WR004444 Ruaro R, Gubiani Ã‰A (2013) A scientometric assessment of 30 years of the Index of Biotic Integrity in aquatic ecosystems: Applications and main flaws. Ecological Indicators 29:105 110. doi: 10.1016/j.ecolind.2012.12.016 temporal stochastic resonance ind uces patterns in wetland vegetation dynamics. Ecological Complexity 10:93 101. doi: 10.1016/j.ecocom.2011.11.003
95 Scozzafava , M et al. ( 2009 A powerful supplement to status and trends http://www.fws.gov/Wetlands/Status And Trends 2009/index.html Accessed 11 June 2014 Sellers et al. ( 2009 ) Dogfennel (Eupatorium capillifolium) size at application affects herbicide efficacy. We ed technology 23.2: 247 250 Stainbrook KM, Limburg KE, Daniels R a., Schmidt RE (2006) Long term changes in ecosystem health of two Hudson Valley watersheds, New York, USA, 1936 2001. Hydrobiologia 571:313 327. doi: 10.1007/s10750 006 0254 5 Stapanian M a. , Adams J V., Gara B (2013) Presence of indicator plant species as a predictor of wetland vegetation integrity: a statistical approach. Plant Ecology 214:291 302. doi: 10.1007/s11258 013 0168 z Stevens DL, Olsen AR (2004) Spatially balanced sampling of nat ural resources. Journal of the American Statistical Association 99:262 278. doi: 10.1198/016214504000000250 Stohlgren ATJ, Falkner MB, Schell LD (1995) A Modified Whittaker Nested Vegetation Sampling Method Reviewed work ( s ): A Modified Whittaker nested vegetation sampling method. Vegetatio 117:113 121. Stohlgren TJ, Chong GW, Kalkhan MA, et al. (1997) Multiscale sampling of plant 1074. Stohlgren TJ, Barnett DT, Kartesz JT (2 003) The rich get richer: patterns of plant invasions in the United States. Frontiers in Ecology and the Environment 1:11. doi: 10.2307/3867959 Titus J (1990) Microtopography and Woody Plant Regeneration in a Hardwood Floodplain Swamp in Florida. Bulletin of the Torrey Botanical Club 117:429 437. U.S. Environmental Protection Agency ( 2002 ) methods for evaluating wetland condition: using vegetation to assess environmental conditions in wetlands. Office of Water, U.S. Environmental Protection Agency, Washingt on, DC. EPA 822 R 02 020. United States Department of Agriculture (1985) Hydric soils of the s tate of Florida. Washington DC: Soil Conservation Service United States Department of Agriculture PLANTS Database (2012) http:// http://plants.usda.gov Accessed 12 April 2014 62:688 696.
96 Wilcox DA, Meeker JE, Hudson PL, et al. (2002) Hydrologic variability and the evaluation hydrologic variability and the application of index of biotic. Wetlands 22:588 615. Wilhelm, G, Ladd D (1988) Natural area assessment in the Chicago region. Transactions of the 53 rd North American Wildlife and Natural Rescources Conference p361 375 Louisv ille, Kentucky. Wildlife Management Institute, Washington, D.C. Wilson MJ, Bayley SE, Rooney RC (2013) A plant based index of biological integrity in permanent marsh wetlands yields consistent scores in dry and wet years. Aquatic Conservation: Marine and Freshwater Ecosystems 709:n/a n/a. doi: 10.1002/aqc.2354 Yoder CO, Barbour MT (2009) Critical technical elements of state bioassessment programs: a process to evaluate program rigor and comparability. Environmental monitoring and assessment 150:31 42. doi: 10.1007/s10661 008 0671 1
97 BIOGRAPHICAL SKETCH Erica C. Hernandez was born and raised in Miami, Florida. Erica attended Florida State University , where she graduated with a Bachelor of Science in environmental s tudies. For the ne xt ten years Erica was very fortunate to work in grant funded research through state and university partnerships studying rare ecosystems throughout the state of Florida. She has been able to work and live both in remote pristine wilderness and bustling ci ties while learning about the unique flora and fauna of Florida. In 2011, Erica was accepted and enrolled in the School of Natural Resources and Environment to pursue a Master of Science degree in interdisciplinary e cology under the guidance of Mark Brown . Upon completion of her M.S. degree, Erica looks forward to continuing her career of spending time outside in the field to understand and protect the beautiful state of Florida.