Hydrologic regime and downstream movement of catfish larvae in the Madre de Dios River, southeastern Peru

MISSING IMAGE

Material Information

Title:
Hydrologic regime and downstream movement of catfish larvae in the Madre de Dios River, southeastern Peru
Physical Description:
1 online resource (84 p.)
Language:
english
Creator:
Cañas, Carlos M
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Master's ( M.S.)
Degree Grantor:
University of Florida
Degree Disciplines:
Geography
Committee Chair:
Waylen, Peter R.
Committee Members:
Mossa, Joann
Pine, William

Subjects

Subjects / Keywords:
amazon, andean, andes, catfish, de, dios, distribution, drifting, floods, forecasting, hydrologic, larva, madre, migration, model, peru, regime, reproduction, spawning, stochastic
Geography -- Dissertations, Academic -- UF
Genre:
Geography thesis, M.S.
Electronic Thesis or Dissertation
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )

Notes

Abstract:
Seasonal flow conditions and catfish larvae production were evaluated in the Madre de Dios River (southeastern Peru) in order to understand the influence of environmental conditions on the reproductive behavior of a group of large catfishes (Siluriformes: Pimelodidae), important species in commercial fisheries, that annually arrive into this headwater region for spawning. A simple stochastic model is presented to describe the influence of the natural flow regime of this river on the downstream transportation of catfish larvae. Temporal and spatial drifting patterns of these catfish larvae were also analyzed and described within the river channel, and the significant ecological role of this region was emphasized in the completeness of life cycle of this important Amazon fish resource. One year of daily river stage records were related to weekly larval catches to determine the association between floods and spawning events, and on the basis of hydro-climatologic characteristics of Andean-Amazon regions, available long-term historical rainfall records were employed to determine the inter-annual variability of floods within the Madre de Dios Basin. Larval catches obtained from five cross sectional transects placed in a 12 km-section of the channel of Madre de Dios River, and surface and deep samples, were employed to assess larvae drifting within the channel. Major larval drift occurred during the high water season, and specifically they appeared to be associated with stages of over the 5 m, a level which served as a physical indicator of the river, triggering spawning responses of these species, termed a ?Biologic Hydrological Significant Event? (BSE). Timing of these BSEs, estimated from the historical rainfall records, described a uniform distribution during the wet season, and their inter-arrival times were exponentially distributed. These observations provided the basis of the stochastic model describing the likelihood of larvae releases from this headwater region to the lowland Amazon. Within the channel, larval drift occurred mostly along the banks of the river, however no significant spatial differences in the numbers of fish larvae moving downstream were found along the sampled sections, which indicated that spawning habitats were located near this Andean foothill region. The permanent flood pulses during the periods of larval peaks and the limited availability of floodplain areas in the channel indicated that larvae do not stay in this region. Analysis of hydrologic regime and catfish larvae drifting in the Madre de Dios Basin demonstrated the strong association between fish fauna and the aquatic environment, the fundamental role of the Amazon headwater regions in fish abundance and diversity, and the function of river channels on catfish spawning and migrations. It is fundamental to maintain and protect hydrologic conditions in the Andean-Amazon region in order to keep aquatic life and connectivity of the complete basin from the headwaters to the estuary.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (M.S.)--University of Florida, 2007.
Local:
Adviser: Waylen, Peter R.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2009-08-31
Statement of Responsibility:
by Carlos M. Cañas.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2007
System ID:
UFE0021293:00001


This item has the following downloads:


Full Text

PAGE 1

HYDROLOGIC REGIME AND DOWNSTREAM M OVEMENT OF CATFISH LARVAE IN THE MADRE DE DIOS RIVER, SOUTHEASTERN PERU By CARLOS M. CAAS A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2007 1

PAGE 2

2007 Carlos Martn Caas 2

PAGE 3

To Yrma and Carlos, my parents, for their gr eat effort in supporting my life and education. 3

PAGE 4

ACKNOWLEDGMENTS I am very appreciative to Dr. Peter Waylen, Dr. Bill Pine, and Dr. Joann Mossa, members of my advisory committee, for their perman ent academic support on the completion of this research. I am especially grateful to Dr. Peter Waylen, for his dedication, energy and encouragement to develop more interesting que stions in order to understand the hydrology of Amazonian headwaters in Peru. I want to dedicate a special gratitude and appreciation to Michael Goulding and Mimi Zarate for their wa rm reception and friendship during my arrival and time living in Gainesville. I am indebted to Dr. Michael Goulding, Dr. Ronaldo B. Barthem, Dr. Bruce R. Forsberg and Dr. Rosseval G. Leite for intellectual stimulation on Amazon fish ecology and limnology. For financial support I am indebted to th e Gordon & Betty Moore Foundation and The John D. and Catherine T. MacArthur Founda tion, under grants awar ded to Dr. Michael Goulding. Several institutions in Peru also helped with data and faciliti es to execute different phases of this research. In Lima, the Servicio Nacional de Me teorologa e Hidrografa (SENAMHI) assisted with meteorological data of the region; the Ichthyo logy Department in the Museo de Historia Natural of the Universidad Nacional Mayor de San Marcos helped with fish larva identifications and the collection of bi ological samples. For field facilities and the implementation of larvae sampling in the Madre de Dios River I would also like to thank the following institutions: The Fishermens Union, the Amazon Conservation Association (ACCA), and the Marina de Guerra del Per For logistical assistance dur ing the field research I am especially indebted to the following fishermen from Puerto Maldonado: Carlos Saire, Genaro Apaza, Exaltacin Tapara, Landers Macahuachi, and Rolando Vela. Mr. Alfredo Soza kindly allowed us to use his riverside pr operty to place river level gauges. 4

PAGE 5

Finally, I want to express my special thanks to my family in Peru, to Yrma and Carlos, my parents, for their total support and encouragement on pursuing my education goals; to my sisters Mirian, Jenny and Milagros for always being ther e; to Maria Luisa, my grandmother, who waits anxiously for my return; and to my dear aunt Elsa, who was next to my mom during my first years of childhood to take care of me. 5

PAGE 6

TABLE OF CONTENTS page ACKNOWLEDGMENTS ...............................................................................................................4 LIST OF TABLES ...........................................................................................................................8 LIST OF FIGURES .......................................................................................................................10 ABSTRACT ...................................................................................................................................12 CHAPTER 1 INTRODUCTION................................................................................................................. .14 2 RELATING HYDRO-CLIMATOLOGY OF M ADRE DE DIOS BASIN TO CATFISH LARVAE DRIFTING IN SOUTHEASTERN PERU............................................................17 Introduction .............................................................................................................................17 Literature Review ...................................................................................................................18 Modeling Approach ................................................................................................................19 Study Area ..............................................................................................................................21 Materials and Methods ...........................................................................................................22 Results .....................................................................................................................................24 Larvae and Stage in Puerto Maldonado ...........................................................................24 Analysis of Available Rainfall Records ..........................................................................25 Discussion ...............................................................................................................................27 Threshold Stage ...............................................................................................................27 Independence Criterion ...................................................................................................28 Sampling Intervals ...........................................................................................................29 Nature of Relationship when Stage exceeds Threshold ..................................................29 Spawning Rates ...............................................................................................................30 Small Sample Sizes .........................................................................................................31 Potential Sources of Errors in Forecast ...........................................................................31 Inter-Annual Variability ..................................................................................................32 Conclusions .............................................................................................................................34 3 TEMPORAL AND SPATIAL PATTERNS OF DISPERSION OF CATFISH LARVAE IN MADRE DE DIOS RIVER, PERU...................................................................................50 Introduction .............................................................................................................................50 Literature Review ...................................................................................................................51 Materials and Methods ...........................................................................................................55 Study Area .......................................................................................................................55 Sampling ..........................................................................................................................56 Statistical Analysis ..........................................................................................................58 Results .....................................................................................................................................59 6

PAGE 7

Discussion ...............................................................................................................................62 Timing and Spawning ......................................................................................................62 Environmental Conditions and Spawning .......................................................................64 Spatial Distribution of Catfish Larvae .............................................................................66 Transport of Larvae to Downstream Areas .....................................................................67 Conclusions .............................................................................................................................68 4 CONCLUSIONS.................................................................................................................. ..76 LIST OF REFERENCES ...............................................................................................................79 BIOGRAPHICAL SKETCH .........................................................................................................84 7

PAGE 8

LIST OF TABLES Table page 2-1 Periods of daily rainfall record available from three me teorological stations in the Madre de Dios Basin..........................................................................................................37 2-2 Test of Fixed Effects showing statistical differences of larva densities between weeks and seasons.........................................................................................................................37 2-3 Chi-square test values of contingenc y tables of possible estimated and observed flood events in Puerto Maldonado, considering different cumulative rainfall in each station.................................................................................................................................37 2-4 Chi-square test values of contingenc y tables of possible estimated and observed flood events in Puerto Maldonado, under tota l cumulative rainfall from both stations.... 37 2-5 Percentage of agreement between forecas ted and observed BSE when there is 75 mm of rain falling in Quincemil area........................................................................................ 37 2-6 Percentage of agreement between forecasted and observed BSE when there is 125 mm of rain falling in Pilcopata area...................................................................................38 2-7 Percentage of agreement between forecasted and observed BSE when there is 225 mm of rain falling in both assessed areas.......................................................................... 38 2-8 Number of potential Biologically Signif icant Events forecasted from historical rainfall records in the Madre de Dios Basin...................................................................... 38 2-9 Chi-Square analysis confirming no statisti cal differences between estimated floods in historical records and observed floods in the 2-years period in Puerto Maldonado..........38 2-10 Number of forecasted wet season BSEs occurring above/below median, classified according to the phase of ENSO........................................................................................ 39 2-11 Number of forecasted dry season BSEs occurring above/below median, classified according to the phase of ENSO........................................................................................ 39 3-1 Levenes test for homogeneity of larval catches variances............................................... 71 3-2 Test of Fixed Effects showing statistical differences of larva densities between weeks and seasons.........................................................................................................................71 3-3 Statistical comparisons between larval densities by transects for the high water period.................................................................................................................................71 3-4 Statistical comparison between larval densities by habitats across the channel for the high water period.............................................................................................................. .71 8

PAGE 9

3-5 Statistical comparison between larval densities by depths for the high water period....... 72 3-6 Statistical comparison between egg densitie s by transects in the Madre de Dios River for the high water period.................................................................................................... 72 3-7 Statistical comparison of water parame ters by transects for high water period................ 72 9

PAGE 10

LIST OF FIGURES Figure page 2-1 The Madre de Dios Basin. Location and its major tributaries draining from the Andes and Amazon lowlands........................................................................................................ 40 2-2 Location of meteorological stati ons in the Madre de Dios Basin ......................................41 2-3 Monthly Precipitation (mm) in Pilcopata, Quincemil, and Puerto Maldonado rainfall stations in the Madr e de Dios Basin..................................................................................41 2-4 Larval density and stage records in Puerto Maldonado during the sampled period ..........42 2-5 Relationship between larval density and stage records in Puerto Maldonado, 2006.........42 2-6 Cumulative Frequency of Biologically Significant Even ts during high water period in Madre de Dios River for 2005 and 2006 .......................................................................43 2-7 Probability Distribution of inter-arrival times of Biol ogically Significant Events in Puerto Maldonado during 2005 and 2006..........................................................................43 2-8 Anticipated cumulative distribution of la rvae associated with BSE occurring at any time during the spawning/rainy season.............................................................................. 44 2-9 Correlation factors obtained for cumula tive rainfall periods in Pilcopata and Biologically Significant Even ts in Puerto Maldonado.......................................................44 2-10 Correlation factors obtained for cumula tive rainfall periods in Quincemil and Biologically Significant Even ts in Puerto Maldonado.......................................................45 2-11 Comparison between 8-days period of cumulative rainfall in Pilcopata and Biologically Significant Even ts in Puerto Maldonado.......................................................45 2-12 Comparison between 5-days period of cumulative rainfall in Quincemil and Biologically Significant Even ts in Puerto Maldonado.......................................................46 2-13 Comparison between cumulative rain fall of Quincemil and Pilcopata during Biologically Signigicant Even ts in Puerto Maldonado......................................................46 2-14 Probability and Poisson dist ributions of forecasted floods in the Madre de Dios Basin for wet and dry seasons...................................................................................................... 47 2-15 Cumulative Frequencies of Biologically Significant Events forecasted from 30 years of historical rainfall records............................................................................................... 47 2-16 Probability Distribution of inter-arrival times between forecasted Biologically Significant Events in Puerto Maldonado........................................................................... 48 10

PAGE 11

2-17 Joint Probability Distribu tion of larva catches when surpassing the threshold river level....................................................................................................................................48 2-18 Probability Distribution of estimated la rvae exported and pa ssing Puerto Maldonado..... 49 2-19 Percentage of agreement and no-ag reement between forecasted and observed Biologically Significant Even ts in Puerto Maldonado.......................................................49 3-1 Location of the study area and sampli ng transects around Puer to Maldonado, Peru........ 73 3-2 The Madre de Dios Basin, its Andean and lowland tributaries, and location of Stage in Puerto Maldonado.......................................................................................................... 73 3-3 Daily water level variation for the Madre de Dios River in Puerto Maldonado, Peru (December 2004 February 2007).................................................................................... 74 3-4 Average larval density and water leve l in Madre de Dios River during 2006................... 74 3-5 Average egg density and water level in Madre de Dios River.......................................... 75 11

PAGE 12

Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science HYDROLOGIC REGIME AND DOWNSTREAM M OVEMENT OF CATFISH LARVAE IN THE MADRE DE DIOS RIVER, SOUTHEASTERN PERU By Carlos M. Caas August 2007 Chair: Peter R. Waylen Major: Geography Seasonal flow conditions and catfish larvae production were evaluated in the Madre de Dios River (southeastern Peru) in order to understand the influence of environmental conditions on the reproductive behavior of a group of large catfishes (Silurif ormes: Pimelodidae), important species in commercial fisheries, that annually arrive into this headwater region for spawning. A simple stochastic model is presented to describe the influence of the natura l flow regime of this river on the downstream transportati on of catfish larvae. Temporal and spatial drifting patterns of these catfish larvae were also analyzed and descri bed within the river channel, and the significant ecological role of this region wa s emphasized in the completeness of life cycle of this important Amazon fish resource. One year of daily river stage r ecords were related to weekly la rval catches to determine the association between floods and spawning ev ents, and on the basis of hydro-climatologic characteristics of Andean-Amazon regions, available long-term historical rainfall records were employed to determine the inter-annual variability of floods within the Madre de Dios Basin. Larval catches obtained from fi ve cross sectional transects pl aced in a 12 km-section of the channel of Madre de Dios River, and surface an d deep samples, were employed to assess larvae drifting within the channel. 12

PAGE 13

Major larval drift occu rred during the high water season, and specifically they appeared to be associated with stages of over the 5 m, a le vel which served as a physical indicator of the river, triggering spawning responses of thes e species, termed a Biologic Hydrological Significant Event (BSE). Timing of these BSEs, es timated from the historical rainfall records, described a uniform distribution during the we t season, and their inte r-arrival times were exponentially distributed. These observations pr ovided the basis of the stochastic model describing the likelihood of larvae releases from this headwater region to the lowland Amazon. Within the channel, larval dr ift occurred mostly along the ba nks of the river, however no significant spatial differences in the numbers of fish larv ae moving downstream were found along the sampled sections, which indicated that spawning habita ts were located near this Andean foothill region. The permanent flood pulses during the periods of larval peaks and the limited availability of floodplain areas in the chan nel indicated that larvae do not stay in this region. Analysis of hydrologic regime and catfish la rvae drifting in the Madre de Dios Basin demonstrated the strong association between fi sh fauna and the aquatic environment, the fundamental role of the Amazon headwater regi ons in fish abundance and diversity, and the function of river channels on catfi sh spawning and migrations. It is fundamental to maintain and protect hydrologic conditions in the Andean-Amazon region in orde r to keep aquatic life and connectivity of the complete basin from the headwaters to the estuary. 13

PAGE 14

CHAPTER 1 INTRODUCTION The Andes represents one of the three principal ecologicalgeological zones of the Amazon System, embracing six major Amazon headwater tr ibutaries along most of its western border (Goulding et al. 2003). Hydrologic ch aracteristics of thes e tributaries are heavily influenced by local rainfall (Barthem et al. 2003; Poveda et al. 2005), which, combined with high rates of erosion, topography and land cover vegetation, create in the river channel environmental conditions to support very important ecological processes (Poff et al. 1997; Goulding et al. 2003; Junk & Wantzen 2004). Migrations for reproduction are one of the most important behavioral responses of fish to the annual natural hydrolog ic regime of this la rge basin (Lowe-McConnel 1975; Barthem & Goulding 1997; Poff et al. 1997; Junk & Wantzen 2004), and in the Madre de Dios headwater region (southeaste rn Peru), large catfishes (Siluriformes: Pimelodidae) have adapted and synchronized their reproductive beha vior and gonad development to the seasonal flood pulses (Munro 1990; Barthem & Goulding 1997; Poff et al. 1997; Caas 2000; Welcomme & Halls 2002; Barthem et al. 2003 ). After spawning, survival of catfish larvae also depends on flooding events, which enhances th e connectivity of these running ecosystems and support larval drifting from this upland region to the lowla nds (Araujo-Lima & Oliveira 1998; Oliveira & Araujo-Lima 1998; Araujo-Lima et al. 2001). The Natural Flow Regime paradigm postulate s that streamflow quantity and timing drive the distribution of aquatic fauna and the connect ivity of flowing water ecosystems (Poff et al. 1997). This concept is used to describe the rela tionship between the downstream drift of catfish larvae and hydrologic regime of the Madre de Di os Basin, and to develop a simple stochastic model of larval releases from this Amazon-Andes headwater region to the lowlands. This research shows that the temporal variability of catfish larvae subjected to the likelihood of 14

PAGE 15

occurrences of Biological Hydrol ogic Significant Events (BSE), wh ich herein are considered to be signified by critical river le vel stage associated with peaks of larval production measured in the Madre de Dios River. Partial Duration Series (PDS) analysis has been employed in the past to evaluate flood risks and related problems (Ashkar & Rouselle 1981; Waylen & Woo 1983), and this method is used to evaluate the probabi lity of frequency and timing of BSEs during the flooding period in the Madre de Dios Basin. Given the limitation of flow information for this basin and the notable interaction between precipitation in the easte rn Andes and streamflow from the headwaters of the Amazon (Poveda et al 2006), historical ra infall records of two meteorological stations of the Madre de Dios Basin are used to pred ict, or hindcast longer term BSE records and characterize the inter-annual vari ability of hydrologic conditions in the basin. It is the intention of the first part of this research to introduce a simple stochastic model that describes the strong influence of the natural flow regime on the re leases of larvae of the large migratory catfishes of Amazon. The river channel is also of critical importance for the re production of Amazonian large catfishes, as it is the principa l habitat through which fish larv ae are dispersed downstream to nursery habitats, as well as being the spaw ning ground where breeding occurs (Goulding 1980; Oliveira & Araujo-Lima 1998; Nakatani et al. 2 001). Dispersion of eggs and larvae are usually hypothesized to constitute the most fragile phase of fish natural history (Araujo-Lima & Oliveira 1998; Oliveira & Araujo-Lima 1998; Nakatani et al. 2001). Most fish species in the Amazon Basin spawn at the beginning of and/or duri ng the high water season as a response to the physical and chemical changes associated w ith large rivers (Goulding 1980, 1988; Barthem & Goulding 1997; Araujo-Lima & Oliveira 1998). Fl oods provide favorable conditions in river channels in both upstream and downstream direc tions (Poff et al. 1997; Goulding et al. 2003). 15

PAGE 16

Temporal and spatial distributions of catfish larvae from the Madre de Dios River were examined as a function of seasonal flow character istics of the principal channel. Five cross section transects were established and sampled on a weekly basis during 2006. Two years of river stage records were used to characterize the annual hydrologic regime of the Madre de Dios River, and larval catches were used to determine the influence of seasonal floods on larval drift patterns, and thus indirectly on spawning times. Th e second part of this research highlights the importance of understanding temporal and spatial pa tterns of catfish larval drift in this Amazon headwater region and provides fundamental information for conservation and fishery management of migratory sp ecies in the Amazon Basin. The construction of the Transoceanic Highway is now underway in southeastern Peru, and two dams on the Madeira River are planned n ear Porto Velho, Rondnia (Brazil). Both the highway and dams represent large economic devel opment projects with probable direct impacts on the flow regime of the Madre de Dios River and some of its tributaries. These impacts in turn could negatively impact long-distance migrator y catfishes. This resear ch not only provides scientific information to aid the understanding of the close association between fish resources and flow regime in freshwater ecosystems, but also to present a baseline for future assessments in this headwater region after these large projects are completed. 16

PAGE 17

CHAPTER 2 RELATING HYDRO-CLIMATOLOGY OF MADR E DE DIOS BASIN TO CATFISH LARVAE DRIFTING IN SOUTHEASTERN PERU Introduction Natural flow regime is the principle variable driving both the distri bution and abundance of freshwater aquatic fauna, and the integrity and c onnectivity of flowing wate r ecosystems (Poff et al. 1997). Flood events are strongly influenced by seasonal precipitation patterns within the Amazon headwater region, and in Madre de Dios Basin (Peru) it has been noted that they are related to the initiation of spawning migration behavior of a group of large catfishes (Barthem et al. 2003). The timing of flood events is critical ecologically in the life cycles of these species mature adults swimming upstream to reproduce, and larvae drifting downstream to colonize new habitats their life cycles are timed to take advantage of flood occurrences in spawning areas (Low-McConnell 1975; Barthem & Goulding 1997; Po ff et al. 1997). Currently, construction of the Transoceanic Highway is underway in southeas tern Peru, and two dams at Madeira River are being considered to be built near Porto Velho, Rondnia (Brazil). Both large development projects represent sources of si gnificant land use change and poten tial alteration of the natural flow regime of Madre de Dios Basin. This study postulates a simple model relating basin hydroclimatologic conditions to catfish larvae production. The concept of Partial Duration Series (PDS) of hydrologic events on the Madre de Dios River is employed to establish a stochastic fl ood model of catfish larvae releases from the headwaters to lowland Amazon. Two years of daily rainfall records from three meteorological stations, and daily observed stage measuremen ts at Puerto Maldonado, are available for comparison to weekly captures of fish larvae The derived relationshi p is extended to those historical rainfall records which precede the limite d period of stage records in order to evaluate the inter-annual variability of the number and timing of flood events within the basin. This 17

PAGE 18

assessment considers a Biological Hydrologic Significant Event (BSE) to be one which equals or exceeds a critical stage associated with peaks of larval production in the Madre de Dios River. This chapter shows the relationship between temporal variability of catfish larvae production and the likelihood of such BSE events. The results provide a base line agains t which potential future scenarios of flow regime altera tions and their associated cons equences on larval production at other similar regions of the Amazon System can be compared. Literature Review All river basins display a natural hydrologic signature, or regime, which exerts control over productivity of aquatic life and habitat condi tions (Poff et al. 1997; Junk & Wantzen 2004; Winemiller 2005). The regime of most rivers in the Amazon System consists of high and low water periods, with a clear regional influence of seasonal precipitation as the principal generator of floods (Welcomme & Halls 2002). Spawning migration of large catfishes along major tributaries of the Amazon System is one of the ma in responses of aquatic fauna to natural flow regime (Low-McConnell 1975; Welcomme 1979; Goulding 1980). During high flows ideal habitat conditions are created to ensure downstream transportation of new fish generation and to avoid predator presence in the main channel (Welcomme & Halls 2002). Long term flow records are available for so me areas of the Brazilian Amazon, but the headwater regions, located among the four Andean countries of Ecuador, Colombia, Bolivia, and Peru, have received little atte ntion, therefore the regions hydro -climatology is poorly understood (Goulding et al. 2003). Poveda et al. (2006) point out the unique interrelationship between the climate of the Amazon Basin, precipitation on the slopes of the eastern An des, and streamflow returning towards the lowlands, and the possibilit y of longer term persis tence of drought/flood conditions within the system because of these feedbacks. Giving the close association between 18

PAGE 19

freshwater ecology and the regional hydrologic cycle, such longer te rm trends may also manifest themselves in ecosystems. The limited available flow data are assessed statistically by means of a probability analysis of flood characteristics. The an alysis of PDS is based on th e model proposed by Todorovic & Zelenhasic in 1970, based on earlier works of La ngbein (1949), and has been accomplished to evaluate flood risks and rela ted problems (Todorovic 1978; Ashka r & Rousselle 1981; Waylen & Woo 1983; Woo & Waylen 1986; Adamowski 2000; Rmillard et al. 2004). PDS evaluates the probability characteristics of flow attributes such as frequency, duration, magnitude, and timing above some selected threshold level (Wayle n & Woo 1983). The number of events surpassing this level during the flooding period follows a Po isson distribution as the threshold becomes sufficiently distant from the mean (Wayle n & Woo 1983; Bras & R odrguez-Iturbe 1985; Rasmussen & Rosbjerg 1991; Rmillard et al. 20 04) and the inter-arrival time between such events exhibit an exponential distribution (Evans et al. 1993). In large tropical rivers fish biology is synchronized to the flood cycle. Most large catfishes of the Amazon complete long distance migrations upr iver to spawn in the ri ver channel just prior to, or during, flood phase (Goul ding 1980; Barthem & Goulding 1997) and the larval drift during the high water season (Araujo-Lima & Oliveira 1998; Welcomme & Halls 2002). Therefore, the occurrence of larvae in the main channel is associated, in a co mplex fashion, to the timing and frequency of flood events. Modeling Approach Some simple assumptions controlling the expect ed larval counts are considered within this model: 1) the timing of BSEs a nd the number of adult catfish r eady for spawning are uniformly distributed, and 2) the interarrival times between BSEs ar e exponentially distributed. If accumulated larvae waiting to be released in a next BSE are produced at a uniform rate, the 19

PAGE 20

distribution of the number of la rvae associated with each BS E should itself be exponentially distributed. These assumptions can be adjusted within this simple model based on additional information that future research will generate in both the hydro-climatological and biological environments. The rate of spawning follows a temporal distri bution, S(t), which is temporally related to optimal conditions for the transport and survival of the spawn. This distribution is related to the historic characteristics of the high flow season and the number of mature fish, M, in that year (j), M j The number of mature fish which have alrea dy spawned or are ready to spawn on day t, in year j, m(t,j), is given by: jMtSjtm )(),( Characteristics of floods occurrences are based on the selection of a truncation level of the river, which is determined by the nature of the flood-related problem (Waylen & Woo 1983). A BSE, defined as a particular river stage threshol d related to peaks of la rval production, is the basis of modeling flood characteri stics and larval numbers. The number of larvae released (n) during the k th BSE of a year is proportional to the time since the last BSE ( k -1). 1................. .......... ).........,( 1.....).........),((),(1 1 1 k jtmn kjtmjtmnk k k The number of BSE in a given high flow seas on is a random variable, k, which follows a Poisson probability distribution (Todorovic 1978 ), determined by the average number of BSE per year (), )( k e kPk and, the inter-arrival times (s) between these BSEs are exponentially distributed (Evans et al. 1993). 20

PAGE 21

)/(1)(xexsF where: = Average length of inter-arrival times x = Number of days between BSEs The hydroloclimatological characteristics of th e basin therefore control the number and timing of BSEs, which in part control expect ed larval production a nd release. If these propositions can be verified and parameterized then it is possible to model the likelihood of larval releases at Puerto Maldonado at various times during the high flow season. Study Area The Madre de Dios Basin drains an area of approximately 90,000 km 2 of the eastern flank of Cordillera de los Andes in southeastern Peru, from over 4,000 m in elevation to the lowland Amazon at 200 m (Barthem et al. 2003). It is the most western headwater basin of the Madeira River, the largest tributary of the Amazon Ba sin (Goulding et al. 2003) (Figure 2-1). Puerto Maldonado is the largest city within the Madre de Dios Basin and it is the principal urban center for commercial activity and econom ic development of southeastern Peruvian Amazon (Caas 2000, Goulding et al. 2003). Field work for this study wa s carried out near Puerto Maldonado, located in the Amazon lowlands at 256 meters elevation on the right bank of the Madre de Dios River. Upstream of Puerto Maldonado there are nine principal tributaries of the Ma dre Dios Basin. Six of these tributaries rise in the south with most of their areas located in the Andes, and three flow from northern, mostly lowland, areas (Fig ure 2-2). Discharges from thes e nine tributaries control the flow behavior of the principal ri ver. Daily water level (stage) m easured at Los Amigos Research 21

PAGE 22

Station between September 2001 and January 2003 is the known basis in this basin for the expected periods of high and low water, and flood events (Barthem et al. 2003). Monthly precipitation records at Pilcopata (32 years), Quincemil (25 years) and Puerto Maldonado (50 years) obtained from the Servicio Nacional de Meteorologa e Hidrografa (SENAMHI), indicate that the rainy season gene rally extends between October and April, and lower precipitation prevails duri ng the remainder of the year (F igure 2-3). They also indicate high geographic variability arising from elevation and location with respect to the Andes, which act as an orographic barrier to the southeast trades importing atmospheric water and exporting surface waters (Poveda et al. 2006). Hydrologic characteristics of the basin vary as a function of elevat ion and season. In the Andean region, channels gradients are steeper, wi th smaller, shallower channels, low suspended sediments, low temperatures and high levels of dissolved oxygen. The flow regime is highly sensitive to local precipitation, and no snow melt effect has been observed (Barthem et al. 2003). In lowland areas (400m and below) rivers fl ow over a flatter t opography, with large and numerous meanders, large channel widths, a nd surrounded by dense forests (Goulding et al. 2003). During field observations in 2006 at Puerto Maldonado, the channel width averaged 425 m, and the stage range was 9.21 m. Materials and Methods Two meteorological stations (Pilcopata and Qu incemil) are located on the eastern foothills of the Cordillera de los Andes, within the two most important Andean headwater tributaries, the Alto Madre de Dios and Inambari Rivers, respecti vely (Figure 2-2). The ra infall station in Puerto Maldonado is located at 256 m, Pilcopata at approximately 900 m and Quincemil at approximately 600 m. After year 2000 precipitation in the Pilc opata area was monitored from Chontachaca rainfall station, locat ed 50 km from the former ra infall station and at similar 22

PAGE 23

elevation. A total of 30 years of historical daily rainfall records were available for analysis (Table 2-1). A system of board gauges was installed at the right bank of Madre de Dios River to register water level measurements. Recordings were taken twice daily, at 8:00 am and 6:00 pm, from which a daily average river le vel was obtained between December 4 th 2004 and December 31 st 2006. Weekly larva sampling was completed throughout 2006 using a 365 micron ichthyoplankton net with a circular mouth of 47 cm and 1.50 m of length. A General Oceanics 2030R mechanical flowmeter attach ed at the open section of the net permitted calculation of the volume of water filtered during th e sampling. Samples were taken to the laboratory, where large catfish larvae were sorted out by using a Zeiss Stemi DV4 Stereo Microscope. Biological materials were stored at Museo de Historia Natural UNMSM in Lima, Peru. Larval counts and the volume of water filtered were used to ca lculate larva density values (individuals per 50m 3 ) in order to relate them to variations in stage. Larva densities and observed stage records were used to identify a threshold water level which triggered larval production in the channel. In order to consider the independence of consecutive BSEs, an arbitrary period of 3 days was established as the minimum time to separate two flood events. This criterion was based on the high frequency of short-period peaks observed in the hydrograph, on the vastness of the basin that includes several tributaries discharging independently into the main stream, and on th e requirement to obtain a reasonable number of BSEs for statistical analyses. Times of critical stage were correlated with varying periods of cumulative antecedent rainfall from each meteorological station (1 da y, 2 days, etc) to determine the volumes of cumulative rainfall in the headwaters most re lated to the generation of BSEs at Puerto 23

PAGE 24

Maldonado. Observed PDS and antecedent precip itation values from P ilcopata and Quincemil meteorological stations were re lated to establish a minimum ra infall value necessary in the Andean headwaters to generate an event. This criterion was then applied to the historical daily rainfall records for the two meteorological stations, to identify simila r rainfall conditions and postulate the annual number and timing of BSE pr ior to the limited peri od of observed floods. Flood frequency characteristics were calculated on the basis of a water year, considered in the Madre de Dios Basin from September 1 st to August 30 th Results Larvae and Stage in Puerto Maldonado Average larval densities varied significantly between the high and low water periods (Figure 2-4 and Table 2-2). After week 43 (October 21 st 2006) larval densities increased dramatically from 24 per 50 m 3 to 156, the maximum larval production observed during 2006. Meanwhile, between weeks 41 and 43 (October 7 th to October 21 st 2006), the 400 m channel of Madre de Dios River rose almost 4 m to 5.84 m. After this point, and until the end of 2006, larva densities remained high (values greater than 44 larvae per 50 m 3 ), but variable. Although the relationship between stage and la rval production was complex, the 5m mark seemed to approximate a break point between high a low values of la rval production (Figure 25). At lower stages, most values of larva densit ies were at background leve ls, with an average of 10 larvae/50m 3 ; and over this threshold, larval production was extremely variable, with both high and low values and an average of 60 larvae/50 m 3 Twenty BSEs were identified within the 2006 stage records, 18 of them occurring during the wet season. The cumulative frequency of da tes of events implied a uniform distribution throughout the high water period, commencing during the second ha lf of October and extending through the second half of Marc h (Figure 2-6). Although the inte r-arrival times of observed 24

PAGE 25

BSEs followed an exponential distribution with an average of 4 days (Figure 2-7), the assumption of Poisson distributed annual numbers of BSEs could not be te sted due to the very limited period of flood data. The av erage number of BSEs in the tw o years of observations was 8 for the high water period, and 1 for the dry seaso n. A preliminary application of the stochastic model based on information obtained from the short period of observed BSEs suggested an exponential cumulative distribution of larvae duri ng the rainy season (Figure 2-8). High values of larvae were impossible (the i nfeasible zone in Figure 2-8), during the early rainy season, as even if a BSE were to occur, an insufficient period of time has passed since the beginning of spawning to accumulate such large numbers of larvae. Analysis of Available Rainfall Records Considering the large areal extent of the basin and the lack of flow information, 30 years of daily rainfall records from two Andean meteorol ogical stations represented a major, and perhaps unique, source of hydroclimatological information from which to extend knowledge of historic BSEs in the region. To this end, a detailed analys is of the association between observed rainfall data and the observed floods -2005 and 2006was performed to determine whether sufficient evidence exists to forecast floods on the basis of precipitation. BSEs were best correlated with rainfall cumu lated over the prior 8 days at Pilcopata and 5 days at Quincemil (Figure 2-9; Figure 2-10). Rainfall totals cumulated over the appropriate period, were then related to starting dates of observed BSEs. A 75 mm-rainfall total appeared to be the minimum required from each of the selected stations to attain cr itical stage at Puerto Maldonado (Figure 2-11; Figure 2-12). An alternative approach that utilizes both stations records and entertains the possibility that rain falling in both headwater tributaries might, together, generate a BSE, indicated that BSEs in Puerto Maldonado were sustained when the sum of the two rainfalls exceed an enveloping line of about 200 mm of rainfall (Figure 2-13). 25

PAGE 26

In order to test the efficiency of these gr aphically derived rainfall thresholds, several possible values were numerically compared to the observed (2005-2006) se ries of BSEs. A chisquare statistical analysis of contingency tables between observed and forecast BSEs, suggested optimal values of 75 mm at Quincemil, 125 mm at Pilcopata, and a 225 mm combined total (Table 2-3; Table 2-4). Forecasted BSEs base d on the Quincemil criteria yield 70% agreement with the observed BSEs, 67% of agreement using Pilcopata, and 73% agreement for the combined totals (Table 2-5; Table 2-6; Tabl e 2-7). These values constituted the basis to reconstruct BSE occurrences for th e historical rainfall records. Th e combined total criterion was used when both records were available. Two possi ble errors in forecasting represented potential limitations of the model: 1) when a BSE is forecasted based on upstream rainfall and none is observed and 2) no BSE is forecast and one occurs. A total of 313 potential BSEs were identifie d from the 30 years of available historical rainfall records, 225 during the high water period and 88 during the low water period (Table 28). The average number of estimated floods wa s 8.03 during the wet season and 2.93 in the dry remarkably close top to the observed figures over the short period of stage records and not statistically different (p < 0.05) (Table 2-9). The annual number of potential events was fitted by a Poisson distribution (Figure 214), and their dates appeared to be reasonably uniform, but showed a slightly reduction in fr equencies after day 100 (December 9 th ) and again in the dry season (Figure 2-15). Inter-arrival times between potential BSEs were fitted by an exponential distribution, with an estimated mean of 9.54 days (Figure 2-16), which again was close to the observed value of 7 days, once the arbitrary minimu m separation figure of 3 days is added to the observed mean inter-arrival time of about 4 days. 26

PAGE 27

Discussion Larval presence in Madre de Dios Basin is related to the occurrence of BSEs during the high flow season, while very low values were encountered during most of the dry season, suggesting that hydrologic f actors are triggering reproductio n activity of large catfish populations and the release of larvae. The number of larvae associated with such events varies markedly, but appears to be represented by the simple threshold and stochastic model presented. While the model complements initial pr opositions about the arrival of long-distance migratory catfishes for spawning (Caas 2000; Barthem et al. 2003; Goulding et al. 2003), several key issues related to the model require further discussion. Threshold Stage The threshold of 5m is arbitrary, base d purely on observed conditions, and could potentially be redefined to a slightly lower valu e (Figure 2-4; Fi gure 2-5). However, this stage should not be viewed as the dir ect generator of observed larval densities, but merely as an indicator that both hydrologic a nd biologic variables are related to a common, unobserved cause the hydro-climatological conditions upstream in the Andean region. Therefore no stage measured at Puerto Maldonado will unambiguously define events releasing larvae completely satisfactorily Larvae in the channel are clea rly identifiable at low background levels even under the 5meter level of the stage (Figure 2-5), indicating a redu ced reproductive activity of fish fauna in the basin during the dry season and between BS Es. Larval catches during BSEs represent reproductive activity of catfishes during these events and the lengt h of time since the previous BSE. 27

PAGE 28

Independence Criterion The threshold was also selected to identify a reasonably large PDS fo r adequate parameter estimation and modeling, while ensuring, as much as possible, independent BSEs, and excluding small hydrologic events of no biological significance during the dry season. Taesombut & Yevjevich (1978) propose a minimum number of days between two independent hydrologic events of 5 + ln(basin area, km 2 ). Application of this formula to the Madre de Dios (16 days) and its major tributaries (14 days) yield only one event per year. The effect of precipitation on discharge in Andean tributaries is almost immediate, and it is common for rivers to expe rience high stage during and i mmediately following rainfall occurrences, and to cease as soon as rainfall ceases (Barthem et al. 2003; Goulding et al. 2003). Stage at Puerto Maldonado (Figure 2-4) exhibits high va riability with frequent peaks of short duration surpassing the selected threshold clear indication of the influence of Andean tributaries supporting the shor t period used to separate consecutive floods. Numerically, independent events could be deri ved by increasing the truncation level, but figure 2-4 indicates that, although hydrologically and st atistically reasonable, this has no biological basis. Nonindependent, or clusters of events can be handled statistically (R osbjerg 1985; Favre et al. 2002), however, in this model it is ultimately inter-arrival times which are of greatest significance to larval releases, and empirically at least, these still appear to follow an exponential distribution (Figure 2-7). Clustering of even ts would tend to produce a grea ter frequency of both shorter (within clusters) and longer (between clusters) inter-arrival times (such as figure 2-15), which still could be modeled by an exponential like a gamma distribution (Evans et al. 1993) or generalized Pareto (Waylen & Laporte 1999). 28

PAGE 29

Sampling Intervals The selection of a critical st age is further clouded by the logi stic restrictio n that a full sampling of larvae could only be carried out once a week. The time and resources required to properly sample at appropriate width and dept h intervals, is further compounded by the time taken to identify and count larv ae. Weekly sampling clearly misse s the peak stages (and larval production) of several potentially large BSE events. In an effort to reduce both work load and the interval between samples in the future, an effort is being made to identify sampling points (in terms of horizontal position and depth) which are most representative of cross sectional mean larval counts at various stages. In this way, future daily measurements may be made at this point in order that biological and hydrol ogic processes to be sampled at the same frequency. This may lead to some loss of accuracy in larval counts, but the relationship can be continuously updated. Nature of Relationship when Stage exceeds Threshold A notable feature of the rela tionship between larval producti on and stage (Figure 2-5) is that maximum observed densities of larvae appear to diminish with larger values of stages in excess of the threshold. This is explicable in te rms of the joint probabil ities of both peak stage levels and the distribution of larval densities, as opposed to an y physical or biolog ical process. In hydrology, the probability of the size of flood even ts above a threshold is commonly represented by an exponential-type distribution (Waylen & La porte 1999). The probab ility distribution of larval catches has also been shown to follow an exponential distribution. Both flood size above threshold, and larval densities, ar e independent under the assumption that the latter is a function of the length of the time since the last BSE (Fi gure 2-17), rather than the size of the flood. Considering both variables, a joint probability distribution results whic h exhibits a pattern similar to that observed when stages exceed 5 m (Figure 2-5; Figure 2-17). 29

PAGE 30

Spawning Rates In terms of biological processes, an assumption of uniformly distributed of rate of spawning is generally accepted (Pine, personal communication). No estimates thus far exist of this important variable in the basin, however, if the simple PDS model holds true, then counts of larvae exported [{larval density (per 50 m 3 ) x estimated discharge (m 3 s -1 )}/50] are themselves exponentially distributed with a mean of 2902 (s -1 ) (Figure 2-18). This average figure results from an average inter-arrival time of approximately 7 days (4 average days + 3 days of arbitrary minimum separation) (Figure 27), yielding an approximate ba sin-wide spawning rate of 414 larvae s -1 dy -1 (2902 s -1 /7dy), which is about half of the observed background less of larval counts when no BSE are occurring. Observed data hint at a decline in this rate over the rainy season (Figure 2-4), but current data provide insufficient evidence to reject the most basic hypothesis of uniformity. Local fisherman have noted that sexually mature catfish migr ate upstream during floods, suggesting that BSEs may have a far more complex -way-gate functi on, not only rel easing larvae from upstream, but also enabling mature adults to migrate upstr eam to the spawning areas. If the latter is the case, then the density of larvae released by a BSE will also be a function of the number of adults that successfully migrated during the previous BSE. The first BSE of a year would initiate upstream migration, but release no larvae. For instance, in Figure 2-4 the rising stage experienced in week 39 (September 23 rd ) is not associated with any larva released beyond background levels while the next hi gh stage of week 43 (October 21 st ) produced the highest larval density of the year. No previous research has related adult catf ishes ready for spawning and larvae being delivered during a same wet peri od. In the future, this information should be acquired and confirm/deny the assumption of ra te of larval production within the model. 30

PAGE 31

Small Sample Sizes The 20 observed BSEs represent a small sample size upon which to support the model. However, the larger sample provided by the es timation of potential BS Es from historical precipitation gives strong validat ory information. This approach is supported geographically, as proximity to the Andes causes the rainfall patter n to be the principal variable controlling the annual hydrologic regime in Madre de Dios Basi n (Barthem et al. 2003; Goulding et. al 2003); and from the perspective of data availability, as daily precipitation are mo re abundant than stage in the basin. The 8-days and 5-days of cumulative rain fall periods obtained for Pilcopata and Quincemil, respectively, can be in terpreted in terms of the dist ances involved and the topography (Figure 2-2). From Pilcopata (900 m), the Alto Madre de Dios tributary drains over gentler terrain and describes a wide, braided and shallo w channel that discharges into the Madre de Dios, before flowing approximately 600 km as a wider and meandering channel to Puerto Maldonado. From Quincemil (600 m), th e Araza River runs over steep terrain with rapid-flowing waters, discharging into the Inamba ri River, a braided river channe l that enters to the Madre de Dios approximately 100 km above Puerto Mald onado. Unfortunately, there are no water level data from these two sites, or any other, headwate r tributaries, which should be ideal situation to obtain PDS closer to the spawning grounds. Potential Sources of Errors in Forecast Cumulative rainfalls obtained from the two And ean tributaries, potentially responsible for each BSE, performed satisfactorily when us ed to forecast BSEs for the period 2005-2006. However approximately 30% of the forecasts (b oth BSE and no BSE) did not agree with the observed data. Potential no-agreemen ts arise when rainfall indicat es a BSE, but none occur, and when BSEs occur but none are forecast. The former indicates that, either rainfall events were 31

PAGE 32

highly localized, or that, given the non-linear nature of the hydrologic system, basin conditions, particularly dry antecedent conditions, may not ha ve converted this input into the anticipated flow. The latter reflects either the inadequacy of the spatial coverage of the meteorological network (2 stations in 90,000 km 2 the biologically signifi cant portions of which are mountainous) or, once again, the over simplification of the proposed rainfall/runoff representation (Figure 2-19). Cumulative frequencies of the dates of forecas ted floods from the historical rainfall data (Figure 2-15) displayed a reasonably uniform distribution thro ughout most of the high water period, with a slightly greater frequency of floods between Oct ober and the middle of December. This hydrologic condition indicates that as s oon as the rainy season starts, BSEs become frequent, which, in turn, could signal additional adult catfishe s to swim upstream to spawn. It can be argued that BSEs estimated in this way are a good estimate of hydrologic conditions in the basin. However the frequency of these potential ev ents did not exactly fit in a Poisson distribution (Figure 2-14). Despite this, their inter-arrival times during the rainy season, the controlling factor in the number of larvae exported, are exponentially distributed (Figure 216). Inter-Annual Variability The inter-arrival times of BSEs particularly during the wet s eason, has to this point been considered to be a simple random variable. Ho wever, BSEs can be closely linked to regional climatology, much of which in South America has been shown (Grimm et al. 2000) to be controlled at lower frequencies by the El NioSouthern Oscillation (ENSO) phenomenon (see for example Poveda et al. (2006) with regard to the Colombian Amazon, and Garreaud et al. (2003) for the Bolivian/Peruvian Altiplano). A pplication of the hyperg eometric probability distribution (Grimm et al. 2000) provides a simple means to identify any potential impact of 32

PAGE 33

ENSO in a time series. Each for ecast annual BSE is assigned to one of three mutually exclusive classifications of ENSO phase, provided, a priori by the Center for Ocean-Atmosphere Prediction Studies (COAPS) at Florida State University ( http://coaps.fsu.edu/jma.shtml ). As various lead/lag associations ex ist between ENSO phase and re gional climates throughout South America, the classifications are extended to consid er the phase in the year prior and post to the forecast BSE count, as well as the synchronous phase. The time series (Table 2-8) is sorted ( N = 28) and divided into two categories, above median and below median years (n = 14 in each ). A count is kept of the total number ( k ) of cold, neutral and warm phase year s in the series (roughly; k cold = 8, k neutral = 14, k warm = 6, although this varies slightly with the lead/lag employed), and of the number of these years, x falling above/below median (Table 2-10; Table 2-11). Assuming a null hypothesis of a random sampling scheme, the probability of x should follow a hypergeometric distribution as follows: pxkNn k x Nk nx N n(|,,) Tables 2-10 and 2-11 suggest that there is no association between rainfall conditions at the two Andean stations likely to generate BSEs, and the phase of ENSO. Although an association of regional drought with warm phases of ENSO (El Nio events) and excess rain during cold phase (La Nia) was found in the Altiplano (Tapley & Waylen 1990; Garreaud et al. 2003), the lead relationship precludes the use of ENSO indicators as forecasters of stream flow characteristics in this region. 33

PAGE 34

Conclusions This simple stochastic model is presented to explain characteristics of the complex, and previously unmonitored, crucial relationship be tween larval production sent downstream to the lower Amazon Basin, and the stage of the Madre de Dios River. Physically, it is well based in the hydroclimatology of the region as well as both the limited biological data and anecdotal observations of local fishermen. The concept and us e of PDS is well establ ished in the field of hydrology and has a sound theoretical basis in st atistics. The numbers of potential BSEs are equally likely to occur through out the wet season (timehom ogeneous Poisson process); and their inter arrival times are exponentially dist ributed. The simple assumption of uniformly distributed timings of BSEs and spawning during the wet season appears reasonable on the basis of stage and rainfall derive d records, however, a variety of non-homogeneous and seasonally varying rates of both variables could be incorp orated if necessary (Todorovic 1978; Waylen & Laporte 1999; Remillard et al. 2004). The requi rements of the model are fairly basic and appropriate for this part of th e world. They can, however, be combined in a variety of ways to produce results which appear to emul ate a complex bio-hydrologic system This collation of hydrologic and biologic data albeit for only a limited period of time is unique in the headwater regions of the southern Peruvian Amazon. Recording the flow regime of this huge basin was achieved by daily water leve l observations in Puerto Maldonado, which is well situated for flood recording purposes with re spect of the major tributaries and provides a good flooding-foot print of the complete ba sin, as supported through the calibration of the rainfall-BSE to extend the period of records. The current data not only provide the basis for the parameterization of this proto-t ype model, but may also furnis h information concerning a more efficient biological sampling scheme at a fr equency more appropriate to the hydrologic processes. 34

PAGE 35

This research also suggests the possibility of feedbacks between biological and climatological systems. The lack of flooding reco rds is a common denominator in remote areas of the Amazon headwaters. Since larva modeling analysis needed long-term flooding records, this limitation was surmounted by using rainfall records, the most abundant hydrological surrogate in these regions, and they perfor med reasonably for forecas ting the derived annual counts of BSEs. This type of association be tween weather conditions and aquatic biological responses, apparently not directly associated, should be included into experiment design for future research. External factor s such as climate change should be considered in the potential variability of these environmental associations. Larvae of long-distance migratory catfish observed in the Madre de Dios Basin drifting to lowlands, complemented with the upstream adult migrations previously observed in the basin, confirmed the connection of the whole Amazon System. Survival of these catfish species depends on two key displacements during differe nt phases of their life history, which are strongly related to flood timing and the connectivity of the comp lete system. Any changes to flow regime or disruption on th e connection of this river will not only put populations of large Amazon catfish at risk, but influence the survival of other fish species dependent upon the lateral connection between river channe l and adjacent floodplains. Consid eration of the migration of catfish and other species need to be incorporated into any envi ronmental assessment of projects in the Amazon System, especially those related w ith the use of water res ources, such as dam and reservoirs, which potentially alter the connectivit y of this large basin. Increased deforestation rates in Amazon headwaters, in association with the expansion of the agriculture frontier and road construction, will also have potential effects in catfish migration by altering the nature of 35

PAGE 36

flow regime in these upstream regions. Adults an d larvae rely on the predictability of flooding patterns, for spawning and for drifting, respectively. Finally, the simple framework of this stochastic model, well-supported by hydroclimatological and biological concepts, is able to be improved with more detailed information about fish biology and climatology of the region, and able to be applied and adapted to similar scenarios along the eastern flank of Cordillera de los Andes in other countries such as Colombia, Ecuador, Peru, and Bolivia. 36

PAGE 37

Table 2-1. Periods of daily rain fall record available from thr ee meteorological stations in the Madre de Dios Basin. Meteorological Station Peri od Records not available Pilcopata / Chontachaca Jan 1 st 1966 Dec 31 st 1967 Jan 1 st 1970 Dec 31 st 1986 Jan 1 st 2001 Dec 31 st 2006 Dec 1974 Dec 1978 Quincemil Jan 1 st 1966 Dec 31 st 1970 Jan 1 st 1973 Dec 31 st 1976 Jan 1 st 1998 Dec 31 st 2006 Apr 1973 May 1988 Puerto Maldonado Jan 1 st 2000 Dec 31 st 2006 Aug 2006 Table 2-2. Test of Fixed Effects (SAS GL IMMIX Procedure Type III, p < 0.0001) showing statistical differences of larval densities between weeks and seasons. Effect DF DF F value Pr > F Season 1 10 128.19 < .0001 Week 1 251 89.21 < .0001 Week Season 1 251 48.34 < .0001 Table 2-3. Chi-square test valu es of contingency tables of po ssible estimated and observed flood events in Puerto Maldonado, considering diffe rent cumulative rainfall in each station. Highest X 2 is shown in bold. 75 mm 100 mm 125 mm 150 mm 175 mm 200 mm Quincemil 154.98 131.53 93.82 64.56 52.36 47.91 Pilcopata 69.33 77.11 84.26 82.88 54.8 41.07 Table 2-4. Chi-square test valu es of contingency tables of po ssible estimated and observed flood events in Puerto Maldonado, considering to tal cumulative rainfall from both stations. Highest X 2 is shown in bold. 200 mm 225 mm 250 mm 275 mm 300 mm Pilcopata + Quincemil 155.33 167.81 138.28 141.30 101.39 Table 2-5. Percentage of agr eement between forecasted and observed BSE when there is 75 mm of rain falling in Quincemil area. Quincemil (75 mm) Observed No BSE Observed BSE Forecast No BSE 304 (40%) 41 (6%) Forecast BSE 181 (24%) 228 (30%) 37

PAGE 38

Table 2-6. Percentage of agreement between forecasted and observed BSE when there is 125 mm of rain falling in Pilcopata area. Pilcopata (125 mm) Observed No BSE Observed BSE Forecast No BSE 319 (42%) 82 (11%) Forecast BSE 165 (22%) 185 (25%) Table 2-7. Percentage of agreement between forecasted and observed BSE when there is 225 mm of rain falling in both assessed areas. Quincemil + Pilcopata (225 mm) Observed No BSE Observed BSE Forecast No BSE 331 (44%) 50 (7%) Forecast BSE 153 (20%) 217 (29%) Table 2-8. Number of potential Biologically Significant Events forecasted from historical rainfall records in the Madre de Dios Basin. Records for wet periods for years 1986 and 2006 (**) were incomplete, a nd not included for comparisons. Year High Water Low Water Year High Water Low Water 1966 8 4 1981 11 2 1967 10 2 1982 9 1 1968 7 6 1983 10 2 1969 10 4 1984 6 2 1970 9 4 1985 4 1 1971 8 0 1986 ** 1 1972 4 0 1998 7 4 1973 7 3 1999 8 3 1974 9 2 2000 9 5 1975 8 3 2001 9 5 1976 9 5 2002 6 5 1977 6 3 2003 10 6 1978 6 2 2004 11 4 1979 8 1 2005 9 3 1980 7 2 2006 ** 3 Table 2-9. Chi-Square analysis (p< 0.05; DF=1) confirming no statistical differences between estimated floods in historical records a nd observed floods in the 2-years period in Puerto Maldonado. High Water Low Water Estimated Floods 8.03 2.93 Observed Floods 8.00 1.00 38

PAGE 39

Table 2-10. Number of forecasted wet season BSEs occurring above/below median, classified according to the phase of ENSO (based on COAPS at FSU), and considering the phase in the prior, cont emporary and next years. ENSO Phase Prior year Contemporary year Next year Cold 3 4 4 Above median Neutral 9 7 7 Warm 2 3 3 Cold 4 3 2 Below median Neutral 5 8 7 Warm 5 3 3 Table 2-11. Number of forecasted dry season BSEs occurring above/below median, classified according to the phase of ENSO (based on COAPS at FSU), and considering the phase in the prior, cont emporary and next years. ENSO Phase Prior year Contemporary year Next year Cold 4 4 3 Above median Neutral 7 7 8 Warm 3 3 3 Cold 4 4 4 Below median Neutral 6 8 6 Warm 4 2 4 39

PAGE 40

Figure 2-1. The Madre de Dios Basin. Location and its major tributaries draining from the Andes and Amazon lowlands. 40

PAGE 41

River Stage Puerto Maldonado Pilcopata Meteorological Station Quincemil Meteorological Station Figure 2-2. Location of meteorol ogical stations in the Madre de Dios Basin. Pilcopata (900 m) and Quincemil (600 m) are located in the Andes. Puerto Mal donado (256 m), where river stage was placed, is loca ted in the Amazon lowlands. Puerto Maldonado 123456789101112 0 500 1000 1500 2000 Pilcopata 123456789101112 0 500 1000 1500 2000 123456789101112 Monthly Precipitation (mm) 0 500 1000 1500 2000 Monthly Precipitation (mm)Quincemil Figure 2-3. Monthly Precipitat ion (mm) in Pilcopata (32 years), Quincemil (25 years), and Puerto Maldonado (50 years) rainfall stations in the Madre de Dios Basin. 41

PAGE 42

0 20 40 60 80 100 120 140 160 180 147101316192225283134374043464952WeeksLarval density (50m3)0 1 2 3 4 5 6 7 8 9 10Stage (m) Larval density Stage (m) Figure 2-4. Larval density and stage records in Puerto Maldonado showing the seasonal pattern of larval production during the sampled period. 0 40 80 120 160 200 012345678910Stage (m)Larval density (50m3) Figure 2-5. Relationship between larval density and stage records in Puerto Maldonado, 2006. 42

PAGE 43

0 10 20 30 40 50 60 70 80 90 100 050100150200250Water year dayCumulative Frequency Figure 2-6. Cumulative Frequenc y of Biologically Significant Events during high water period in Madre de Dios River for 2005 and 2006. 0 0.2 0.4 0.6 0.8 1 024681012141618DaysProbability Probability Distribution Estimated Probability Distribution Figure 2-7. Probability Di stribution of inter-arrival times of Biologically Significant Events in Puerto Maldonado during 2005 and 2006 (Average = 4 days). 43

PAGE 44

Days into Spawning Season 0306090120150180210240 Probability of z larvae or more 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 Infeasible Zone 1200 2400 3600 4800 400 800 1600 2000SepOctNovDecJanFebMarApr Figure 2-8. Anticipated cumulative distribution (y) of la rvae (z) associated with BSE occurring at any time (x) during the spawning/rainy season. The Infeasibl e zone indicates values which will not be encountered due to the proximity of the date to the start of the spawning/rainy season. 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 1234567891011Days of cumulative rainfallCorrelation Factor Figure 2-9. Correlation factors obtained for cumulative rainfa ll periods in Pilcopata and Biologically Significant Even ts in Puerto Maldonado. 44

PAGE 45

0 0.04 0.08 0.12 0.16 0.2 1234567891011Days of cumulative rainfallCorrelation Factor Figure 2-10. Correlation factors obtained for cu mulative rainfall periods in Quincemil and Biologically Significant Even ts in Puerto Maldonado. 0 50 100 150 200 250 300 350 3.004.005.006.007.008.009.0010.00StageCummulative rain 8 days Figure 2-11. Comparison between 8-days period of cumulative rainfa ll in Pilcopata and Biologically Significant Even ts in Puerto Maldonado. 45

PAGE 46

0 50 100 150 200 250 300 350 400 450 3.004.005.006.007.008.009.0010.00StageCummulative rainfall 5 days Figure 2-12. Comparison between 5-days period of cumulative rainfall in Quincemil and Biologically Significant Even ts in Puerto Maldonado. 0 50 100 150 200 250 300 350 050100150200250300350400QuincemilPilcopata Figure 2-13. Comparison between cumulative rainfall of Quincemil and Pilcopata during Biologically Significant Even ts in Puerto Maldonado. 46

PAGE 47

0 0.05 0.1 0.15 0.2 0.25 123456789101112Floods per yearRelative Frequency Wet Dry Poisson Wet Poisson Dry Figure 2-14. Probability and Poisson distributio ns of forecasted floods in the Madre de Dios Basin for wet and dry seasons (Flood aver age: wet season = 8.03; dry season = 2.93). 0 20 40 60 80 100 050100150200250300350400Water year dayCumulative Frequency Total Cumulative Frequency Cumulative Frequency WET Cumulative Frequency DRY Figure 2-15. Cumulative Frequencies of Biologically Significant Events forecasted from 30 years of historical rainfall records. 47

PAGE 48

0 0.2 0.4 0.6 0.8 1 02 04 06 08 0Inter-arrival daysProbability 1 0 0 Probability Distribution Estimated Probability Distribution Figure 2-16. Probability Distri bution of inter-arrival times be tween forecasted Biologically Significant Events in Puerto Maldonado. 0 1 0 1 0 1 0 2 0 2 0.3 0 3 0.4 0 4 0.5 0 5 0 6 0 6 0 7 0 8 0 9 Stage of River above Threshold (Relative to Mean) 012345 Number of Larvae (Relative to Mean) 0 1 2 3 4 5 Probability 0 1 2 3 4 5 012345 Probability Number of Larvae (Relative to Mean)Stage of River above Threshold (Relative to Mean)Joint Probability Marginal Probability Larvae Marginal Probability BSE Peak0 0.2 0.2 0.4 0.4 0.6 0.6 0.8 0.8 1.0 1.0 Figure 2-17. Joint Probability Distribution of larval catches when surpassing the threshold river level. 48

PAGE 49

0 0.2 0.4 0.6 0.8 1 0100020003000400050006000700080009000Larvae per secondProbability Estimated Probability Distribution Probability Distribution Figure 2-18. Probability Dist ribution of estimated larvae exported and passing Puerto Maldonado (Average = 2902 larvae per second). 0 10 20 30 40 50 No Observed/No ForecastedObserved/ForecastedObserved/No ForecastedNo Observed/ForecastedCategories of ComparisonAgreement (%) Quincemil Pilcopata Pilcopata + Quincemil Figure 2-19. Percentage of agreement (in solid color) and no-agreement (in horizontal pattern color) between forecasted and observed Biologically Significant Events in Puerto Maldonado. 49

PAGE 50

CHAPTER 3 TEMPORAL AND SPATIAL PATTERNS OF DI SPERSION OF CATFISH LARVAE IN MADRE DE DIOS RIVER, PERU Introduction Fish migrations for reproduction are considered one of the main behavioral responses to the annual hydrologic regime occurring in large tropical rivers (Lowe-McConnell 1975; Munro 1990; Poff et al. 1997; Junk & Wantzen 2004). In the Amazon Basin, a group of large catfishes (Siluriformes: Pimelodidae) develop long distan ce longitudinal migrations for reproduction. These migrations are most evident during the rising water period but, depending on the species, may occur at other times of the annual cycle as well. Several species of migratory catfish swim upstream in the principal tributaries from lowla nd areas toward the headwaters in the Andean foothills (Barthem et al. 2003). In Peru, the Madr e de Dios Basin represents one of the AndeanAmazon headwaters where the commercial fisher y exploits large catfish (Caas 2000; Barthem et al. 2003). While sexually mature individuals have been found, no larval fish collections have been made to determine whether this region is a spawning ground and how fish spawning may relate to the annual hydrologic regime. Survival of Amazonian large catfishes is thought to be dependent on annual flooding events as the adult migrations and gonad deve lopment are synchronized with the seasonal flood pulses (Munro 1990; Poff et al. 1997; Caas 2000; Welcomme & Halls 2002; Barthem et al. 2003). Flooding events are also li kely important for larval development as the seasonal floods help to disperse drifting larvae from the upland region to the lowlands (Araujo-Lima & Oliveira 1998; Oliveira & Araujo-Lima 1998; Araujo-Lima et al. 2001). The pur pose of this chapter is to examine the relationship between the larval produ ction and seasonal flow characteristics of the Madre de Dios Basin. To achieve this, temporal a nd spatial distributions of catfish larvae in the Madre de Dios River collected during 2006 were analyzed. Five cross sectional transects were 50

PAGE 51

sampled on a weekly basis using an ichthyoplank ton net. At each transect a known volume of water was sampled for each catch and a suit e of physio-chemical parameters measured. Icthyoplankton samples were then sorted and catfi sh larvae were counted. Results for the Madre de Dios River showed a temporal variability wi th significant higher larval density values during the rainy/high water periods of the year. Larval de nsity within the main river in this section of Madre de Dios River was not significant differe nt by depth, but did differ laterally across the river width. River flow characteri stics provided favorable conditions for fishes to spawn and for larvae to be displaced downstream and survive. By understanding patterns of larval dispersion of these species we will be able to identify thei r spawning areas and determine their relationship with the seasonal flooding of the Madre de Dios Basin. This study represents not only the first analysis of larvae drifting in th e Southern Peruvian Amazon, but al so the first larval research developed in one of the headwaters of the Madeira River, the largest tributary in the Amazon Basin. This work also provides a basis for impor tant research on catfishes in the Amazon Basin which is fundamental for developing important co nservation plans, given the importance of these species to local peoples and the continuing threats in this region from poor land use practices and proposed construction of hydroelectric f acilities in the southwestern Amazon. Literature Review In freshwater ecosystems, fish species have ev olved strategies to regulate timing of their sexual cycles with ideal environmental conditio ns to ensure their reproductive success (Bye 1984). In temperate regions day length, temperature, and food availability predominate as environmental factors influencing timing of repr oduction (Bye 1984). Most authors have agreed that photoperiod is the pr incipal environmental factor contro lling spawning in most Salmonids (Scott 1990). Temperature is thought to be an im portant ultimate factor influencing fertilization success and survival of larvae of Cyprinids (Hontela & Stacey 1990) In tropical regions, several 51

PAGE 52

fish species have predictable a nd restricted spawning periods, whic h are related to seasonal rains, flooding of rivers, or the monsoons (Munro 1990). Spawning activity of tr opical fishes may be initiated by environmental changes such as yearly flooding events, which are associated with an increment of food availability in the aquatic ecosystem, and with a direct effect on fish larvae that ensures their displacement to area s for feeding and protection (Bye 1984). It is commonly thought that the natural flow re gime of a river is important in structuring the productivity and connectivity of riverine ecosystems (Poff et al. 1997). Hydrological conditions of tropical rivers are determined larg ely by annual cycles of rainfall and desiccation occurring at these regions (Lowe-McConnell 1975; Welcomme & Halls 2002), and regional variability of rainfall patterns, as well as geology, topography and land cover vegetation, influence in the final characteristi cs of the natural flow regime of a river (Poff et al. 1997). Five critical components including magnitude, freque ncy, duration, timing and rate of change of hydrologic conditions are said to control most ecological processes in river ecosystems (Poff et al. 1997). In many rivers, timing and frequency of fl ows are thought to be cr ucial to trigger adult fish migration to spawning areas and to tr ansport larvae downstr eam (Poff et al. 1997; Welcomme & Halls 2002; Junk & Wantzen 2004), their spawning activity is timed with the occurrence of the flooding period (Lowe-Mc Connell 1975; Bye 1984; Welcomme & Halls 2002). Flooding events also drive a change of ch emical contents and nutrient status of the parental river because of its lateral connection with the ad jacent floodplain, where nutrient cycles, primary and secondary production, and organic matter decomposition occurs in higher levels (Goulding 1980; Junk & Wantzen 2004). This s easonal change gives the river channel the ideal conditions not only for hosting spawning grounds, but also for transporting larvae downstream. 52

PAGE 53

In general, larvae drifting have received le ss attention than adult fish migrations in freshwater ecosystems. Abundance and migratory data of adult catfishes exist from monitoring commercial fishery catches, since this information represents the basis to detect impacts of overfishing, one of the major concerns about fish population management (Goulding 1979; Welcomme 1979; Caas 2000; De Resende 2004; Filho & Schulz 2004; Payne et al. 2004). Larval fish ecology represents a valuable knowle dge and a major significance for the life history of most freshwater fishes, as their requirements are often differ ent than those of adult fishes (Scheidegger & Bain 1995; Robinson et al. 1998) and their distribution often permits a good inference about spawning ground locations and dispersion processes, which are also two important conditions to consider in conservati on and management of fi sh populations (Oliveira & Araujo-Lima 1998). In temperate rivers of North America, larval studies have been conducted to understand the effects of navigation, flood contro ls, reservoirs and impoundments in the community structure of larvae and their drifting process (Harrow & Schlesinger 1980; Holland & Sylvester 1983; Brown & Coon 1994; Scheidegger & Bain 1995). In the Mi ssouri River the comple x pattern of flow between the lower portion of the river and the tributar ies drives the productivity and maintenance of larval fish diversity (Bro wn & Coon 1994). Scheidegger & Bain (1995) suggested that flow regulation in Tallapoosa River, a tr ibutary of Alabama River, app eared to lead degradation of microhabitats of fish larvae, and protection of th ese microhabitats should have priority in order to protect fish populations in this river. Flow regulation is also thought to have significantly altered the recruitment of most native fish species in Caspaspe River in Australia (Humphries et al. 2002). A variety of life histor y and recruitment styles of freshwater fishes have been identified in Murray-Darling Basin (Austra lia) based on different hydrological conditions 53

PAGE 54

occurring within a range of ge ographical zones (Humphries et al. 1999). In South America, most larval studies have been concentrated in Brazilian Amazon, and they have generated information about larval development and drifting pattern related to hydrologic conditi ons in the Central Amazon (e.g. in rivers Solimes and Amazon), wh ere the presence of large floodplains is the principal feature. In such regions, larval production is strongl y related to annual hydrological conditions occurring in the prin cipal river, which is produced by the influence of upstream tributaries (Araujo-Lima & O liveira 1998; Oliveira & Araujo -Lima 1998; Araujo-Lima et al. 2001; Nakatani et al. 2001). Egg and larval stages constitute the most fragile and important phases on fish life, since natural fish population continua nce will be based on the surviv al and success of recruitment during these dispersal phases (Araujo-Lima & Oliveira 1998; Olivei ra & Araujo-Lima 1998; Araujo-Lima et al. 2001; Nakatani et al. 2001). Seas onal increments of water flow in most rivers are critical not only to create channel conditions where spaw ning process occurs (Harrow & Schlesinger 1980; Barthem & Goulding 1997; Ar aujo-Lima et al. 2001; Welcomme & Halls 2002), but also as the main aquatic highway thro ugh which fish larvae will reach and occupy their final habitats (Scheidegge r & Bain 1995; Oliveira & Araujo-Lima 1998; Humphries et al. 2002; Humphries 2005). After spawning, downstream displacement ensures eggs and larvae drift-transportation to nursery areas, where larger amount of f ood and protection are available (Goulding 1980; Lowe-McConnell 1987; Araujo-Lim a & Oliveira 1998; Na katani et al. 2001; Humphries et al. 2002; Humphrie s 2005). Flushing condition also influences fish community structure, decreasing the likeli hood of predator presence on the wa y downstream of most of fish larvae (Junk & Wantzen 2004). Fish migration, in any of its directions, represents one of the most important factors to maintain basin producti vity and integrity in rivers (Welcome & Halls 54

PAGE 55

2002) -larval fish not only grow to become adult fish, but larval and juvenile fish are also important components of the food web as prey items for many othe r species of fish, invertebrates, birds, and mammals related to freshwater ecosystems. Hydro-climatologically, the Amazon lies in a f eed-back between the vast rainforest, that recycles almost one half of the precipitation it receives and the Andes, which acts as an orographic barrier importing atmospheric water and exporting surface waters to the Amazon System (Poveda et al. 2006). Rive r level variability in most of Amazon headwaters, such as the Madre de Dios Basin, is controlle d principally by rainfall, snow melt does not cause influence in flow regime in the Andean region (Barthem & Goulding 1997; Barthem et al. 2003). Most of the highest annual rainfall values in the large Amazon Basin are re ached in the eastern Andean foothills with records about 8000 mm (Barthem et al. 2003). Materials and Methods Study Area The study area was located around Puerto Maldona do city, the principal urban center in the Madre de Dios Basin. Puerto Maldonado is located at 480753E and 8607445N coordinates (19L UTM zone/Datum WGS84), in the confluence of the rivers Madr e de Dios and Tambopata, at 256 meters above sea level (Figure 3-1). In fr ont of Puerto Maldonado, the river channel was approximately 425 meters wide, the deepest part of the channel varied between 10 m and 13 m, and water level variation reached 9.21 mete rs between 2005 and 2006 (field observations). Barthem and his research team (2003) reporte d that 8.9 meters was the maximum annual fluctuation of the Madre de Dios River betw een 2001 and 2003 near the confluence of Los Amigos River. Physical and chemical conditions of the river vary based on these hydrologic periods, with evident increment of water level, average velocity, water discharge, and sediments in suspension during the wet season (Barthem et al. 2003). 55

PAGE 56

The Madre de Dios River is the largest river in southeastern Peru a nd one of the 10 largest rivers in the Peruvian Amazon. Fl owing generally from west to ea st, it has a drainage area of approximately 90,000 km 2 (Goulding et al. 2003). Its natural flow regime is largely influenced by the high rainfall variability caused by its proximity to the Andes (Kalliolla & Puhakka 1993; Puhakka & Kalliolla 1993; Puhakka et al. 1993; G oulding et al. 2003). Nine important tributaries discharge in the Madre de Dios River, through which water leaves the basin. Six tributaries drain from highlands areas over 4000 meters above sea le vel, and three drains in lowlands areas and drain at the northern area of the basin (Goulding et al. 2003) (Figure 3-2). Rainfall records obtained mostly from two Andean meteorological stations from the Servicio Nacional de Meteorologa e Hidrografa (SENAMHI), the Peruvian Weather Service Institution, indicated a highly variability in precipitati on pattern within the drainage ba sin. These records also suggested that in general the rainiest season in the basi n takes place between October and April and the drier season does between May and September. Sampling During year 2006 a weekly larva sampling wa s carried out along five cross channel transects located on the Madre de Dios River near the city of Puerto Maldonado. All transects crossed the channel from bank to bank, three of them (named: La Cachuela, El Balcon, and Puerto Maldonado) were located in the Madre de Dios River upstream of the confluence with the Tambopata River, and a fourth transect (named Bajo Madre de Dios) was placed downstream of the confluence. The fifth transect (named Herrera ) was in the Tambopata River, approximately 4 km upstream of the mouth (Figure 3-1). Tran sects in the Madre de Dios River were approximately 3.5 km from each other, with the most upstream transect placed 12 km from the most downstream. The Tambopata River transect was re-located 1.5 km upstream during the dry 56

PAGE 57

season due to channel shallowness. For the fina l analysis, this new location was treated as a different transect, mostly sampled during the dry season. Across each transect, sampling was carried out at five designated habitats categories: Cutting banks, referred to erosive areas usually with high velocities; filling banks, included depositional areas and with lower ve locities; the main channel, that included the deepest part of the channel; and zones in between the main chan nel to the cutting bank and main channel to the filling bank. Samples at the bank habitats were ta ken within distances no larger than 20 meters off shore. Surface samples were taken at one meter depth, and deep samples corresponded to 70% the total depth at each of th ese five habitat categories. Physic al and chemical parameters of water were obtained at surface of each sampling point. A 365-micron ichthyoplankton net with a circ ular mouth of 0.47 m diameter, 1.50 m long, conical shaped, and fitted with a removable collection bottle with two 365-micron mesh windows was used to perform the larva sampli ng. A mechanical flowmeter General Oceanics 2030R was attached at the open section of the net to calculate water filtered during each sampling. Depth measurements were obtained by using a Lowrance Fish-finding Sonar & Mapping GPS LMS-480M. Dissolved oxygen and temperature were measured with an oxymeter YSI Model 550A, conductivity was obtained w ith a conductivimeter YSI 30, and pH was measured with a Hanna Tester HI98130. A range finder Nikon ProStaff Laser440 was used to record channel width measurements and late ral distances between the sample points. Water level measurements were obtained from a gauge stage installed at right bank of Madre de Dios River in Puerto Maldonado city (Figure 3-1). Daily records were taken twice a day (8:00 am and 6:00 pm) between December 4 th 2004 and December 31 st 2006. During all mornings of 2006, channel width was also recorded with the range finder in front of the water 57

PAGE 58

level stage. Integrated water velocity was recorded at five equidistant points across the channel, also in front of the river stage. Discharges of the Madre de Dios River were estimated with the velocity-area technique, which i nvolved the calculation of discha rge segments across the channel by using the mean section method. Larva was collected at the five transects dur ing day time hours and in the same order to standardize timing of collection at each site Night time sampling was not included since no differences were found between day and night larval sampling (Araujo-Lima et al. 2001). At each habitat the net was placed at the selected de pth and left for 3 or 5 minutes, depending on the conditions of stream velocity. A concrete block helped the net to reach deep positions and the boat was kept stopped by using an anchor. After lif ting the net, all the co ntent retained in the collection bottle was rinsed, and the sample was placed into flasks of 500 ml. and preserved in 4% formalin solution. Flasks were taken to labo ratory facilities instal led in Puerto Maldonado, where larvae were sorted out. Eggs were also counted, although they were not possible to differentiate into any specific fish group. A Ze iss Stemi DV4 Stereo Microscope was used to identify and count catfish larvae. Biological material was finally stored at Museo de Historia Natural UNMSM in Lima, Peru. Statistical Analysis Larva counts were standardized to dens ity units (number of larvae per 50 m 3 ) in order to make comparisons between periods, transect, and habitat catches. Statistical analyses were performed by using SAS software and all statis tical significance was re ported at p < 0.05. A non parametric method was used, the Generalized Linear Fixed Model procedure (GLIMMIX), which allowed comparing larval densities at various temporal and spatial levels. During comparison analyses, seasons (two), transects (six ), and habitats (five) were the fixed effects within the GLIMMIX procedure. Because the variances were not proportional to the means, we 58

PAGE 59

specified within the model procedure that la rval catches followed a Negative Binomial Distribution. If larval density differed by fixed effects the Le ast Square Means procedure was used to determine which effects (seas ons, transects or habitats) differed. Results A total of 44,984 catfish larvae were captured along the five transects around Puerto Maldonado city. Larval catches presented a seasonal pattern during the sampled period. High variability characterized the samples, and Levene test indicated non homoge neity of variances in the weekly larva catches, that is the variance in each of the sa mples were not proportional to the respective means (Table 3-1). Two hydrological periods in Madre de Dios River were identified in the 2-year stage records in Puerto Maldonado: a low water period that occurred between May and September, and a high water period for the rest of months (f rom January to April, and from October to December). The river fluctuated 9.21 meters, wi th the highest peak reached for January 3 rd 2006 and the lowest for September 22 nd 2005 (Figure 3-3). During 2006 estimated larval densities presen ted significant differences between weeks and hydrological periods (Table 3-2). Lower mean values of week densities were found between April and September, and higher values occurred during the beginning and the end of the year (January to March, and October to December). Th e 400-meters channel of Madre de Dios River rose up in 3.99 meters betw een weeks 41 and 43 (October 7 th to October 21 st 2006). Higher larval densities occurred during week 44 (October 28 th 2006) until the last week of the year. Lower density values appeared between weeks 17 and 41 (April 22 nd and October 7 th 2006). Between January 3 rd and April 17 th larval densities were also high, but with a tendency to decrease as the flood season ended after week 17 (April 22 nd ). A peak of larva density similar to those found at the end of the year occu rred during the second week (January 9 th ) (Figure 3-4). 59

PAGE 60

Since presence of most larv ae occurred during the high wa ter season, all comparison was focused on catches for this hydrologic peri od. Least Square Means procedure indicated significant differences between transects El Ba lcon and Puerto Maldonado (both in Madre de Dios River), and between transects from the Tam bopata River and Puerto Maldonado tr ansect in the Madre de Dios River (Table 3-3). During the high water season larval drift was evaluated by habitats and depths identified within the cross channel transects. Statistical comparison of Least Square Means between the five arbitrary habitat categories i ndicated clear differences between larval densities obtained near the banks and those obtained in areas closer to the main channel. Larvae found in habitats related to the main channel did not present statistical differences. Larvae found at filling banks were significantly greater than those found at the cutt ing banks (Table 3-4). No statistical differences in larval catches were found between 1-meter de pth samples and those ta ken at depths greater than 1 meter (Table 3-5). Eggs were also captured during the sa mpling, and their abundances (eggs/50m 3 ) were significantly increased for weeks 42 and 43 (October 14 th and October 21 st 2006) in a better synchronization with rising water dates. During the beginning of the year, egg presence was detected during most of February, with a reducing pattern until week 10 (March 6 th 2006) (Figure 3-5). We were not able to differentiate the species whos e those eggs belonged, but their temporal presence, and mainly their large increm ent as the waters star ted rising, was a closer indicative to detect spawning periods and potentially, spawning grounds. No significant differences in egg densities were found between th e four transects placed in the Madre de Dios River (Table 3-6). 60

PAGE 61

Channel characteristics were recorded in front of the water level stage. Channel width kept fairly constant values during most of the ye ar, between 414 and 446 mete rs, the maximum width was reached in January 2006. At this same point, the highest discharge was 4125.48 m 3 /sec, and the mean surface velocity was 0.789 m/sec. Chemi cal characteristics of both rivers varied accordingly to the hydrologic periods. Conductivity kept higher values in the Madre de Dios River over 80 micro-Siemens/cm during most of the dry season, w ith the highest average value for July 2006 (93.44 micro-Siemens/cm); during th e lower season values fluctuated between 62 micro-Siemens/cm and 75 micro-Siemens/cm. In the Tambopata Rive r conductivity varied between 27.8 micro-Siemens/cm in April and 58 .7 micro-Siemens/cm in September. Dissolved oxygen in the surface of the Madre de Dios River increased gradually afte r the dry season, from a minimum value for October (5.53 mg/litter) to a maximum value in May (7.90 mg/l). Similar pattern of oxygen variability was observed for Tam bopata River, with values in the range of 5.53 mg/l in October and 7.64 mg/litter in May. Surf ace temperatures were higher during the dry season, their values increased from 22.2 C r eached for May to 27.6 C for September in the Madre de Dios River. Temperatures in the Tambop ata River varied in a similar pattern than those recorded in the Madre de Dios River. Values for pH were only obtained for the dry season and they followed a decreasing pattern, from a maxi mum 7.6 recorded in July to 6.30 in October. Madre de Dios River presented higher values of pH than the Tambopata River. Water characteristics varied be tween transects, and variati on was related to transects location with respect to the tributary entrance. In the Madre de Dios River, conductivity was not significantly different between the three transects located upstream Tambopata River, but they were different than Bajo Madre de Dios transect, located downstream this tributary; and they were also different than tran sect placed within Tambopata River. Oxygen values were not 61

PAGE 62

significantly different between the four transects of Ma dre de Dios River, but they were with those located in Tambopata River. No statistica l differences in Temper ature appeared between most of transects, except for transect Herrera that kept higher values than the rest (Table 3-7). Discussion Timing and Spawning The Madre de Dios Basin plays an important role in the recruitment of large catfishes (Family Pimelodidae) into the Amazon System. In this region, densitie s of larval catfish collected weekly at fixed transect locations de monstrated strong seasonal patterns with highest catch rates measured during times of highest flow. Significantly high er densities of catfish larvae in the river channel between Oc tober and December indicated that spawning activity of these species was synchronized with the time of wate r level rising, and it star ted in the middle of October for year 2006. High values of larvae were also encountered during the beginning of the year 2006 (January to March), which constituted part of the preceding spawning period of these species within the basin (2005-2006). Based on the temporal distribution of larval production observed during 2006, spawning activity of large catfishes in the Ma dre de Dios Basin was found to be between October and March. In other tropical regions a similar period occurs, such as in the Campaspe River (Australia) the peak time larval drift was betwee n October and December (Humphries & King 2003). In temper ate regions, spawning occurs at different period, in the Colorado River for example, most of cyprinids and cat ostomids spawn primarily during March to June (Robinson et al. 1998). Because the Colora do River is in the northern hemisphere the seasonal timing of spring floods is different than the other examples, however the response in the fish populations is likely the same. The largest numbers of larva catches were samp led for the fourth week of October, larval density values went up 5 to 10 times with respect to the previous week in all transects (maximum 62

PAGE 63

average: 180 individual per 50 m 3 in El Balcon transect). This w eek represented the on-set of the spawning activity of these catfishes in this basi n for 2006, and although we di d not register larval stage, the presence of the yolk s ac in most of the indi viduals indicates a life time of a few hours to days after hatching, which also suggests that this downstream migration of large catfish larvae represents the primary dispersi on from spawning grounds in Madre de Dios River. Frequency of flood events during the high water period in the ar ea is supporting this tran sportation and appears to be very linked with spawning behavior of large catfishes. In the Central Amazon catfish larvae have also been reported to be drifting the river in large quantities during the last months of the year (Araujo-Lima & Oliveira 1998), and othe r species such as Characiforms in the Amazon spawn also during this period (Oliveira & Araujo-Lima 1998). Although eggs captured in the nets were not able to be identified, their presence can give a more exact time-indicator of the on-set of spaw ning season and potentially location of spawning grounds in the Madre de Dios River. Time for embryo differentiation for two species of these large catfishes is known to take between 3 to 7 hours (Nakatani et al 2001), and time from fertilization to hatching lasts up to 16 hours fo r many other characiforms in the Amazon (AraujoLima 1994). The large increment in e gg densities found between October 14 th and October 21 st which matches closer in time to rising water dates, confirm with another evidence that spawning period in Madre de Dios River dur ing 2006 started with the rising of river level for the middle of October. Further, considering that both, the average surface velocity of the river during the flooding reached up to 0.789 m/sec (2.84 km/hour) and the time of embryo differentiation is less than 7 hours, then the spawning grounds could be potentially located around 20 km upstream from the sampling area. An area of rapids is loca ted within this distance in the Madre de Dios River (Locality: La Cachuela, Madre de Dios, Pe ru), where the increased turbulence caused by 63

PAGE 64

flood events give the water a better oxygenation and velocity, potentially favorable conditions for these catfishes to spawn. Environmental Conditions and Spawning It is known that environmental factors are infl uencing spawning activity of most fishes in the world, and some authors consider that enviro nmental factors are even more important in the spawning activity than biologica l factors (Lowe-McConnell 1987; Munro 1990; Humphries et al. 1999). In temperate zones photoperiod is consider ed the environmental condition best related with fish spawning, since climate and length-day vary markedly from season to season, and it can be detected by fishes (Sumpter 1990). In Ka lkaveem River (Russia) downstream migration of young coho salmon (Oncorhynchus kisutch ) was thought to be cont rolled by temperature (Pavlov & Maslova 2006); in the Nile River spawning of the cyprinid Barbus bynni coincides with the increment of water level and inundati on of shallow areas (Hontela & Stacey 1980); and in the Little Colorado River it is likely that temperature and photoperiod are important spawning cues for four native fishes (Robinson et al. 1998). The influence of precipitation from the close An dean Region in the flow regime of Madre de Dios Basin is apparent, rivers get flooded a short period after rainfall events occur and cease as soon as precipitation stops (Goulding et al 2003). During 2006 rainfall records from two Andean foothill meteorological stations within the basin indicated that rainy season started between the first and second weeks of October, a nd for those dates the wa ter level of the large Madre de Dios River rose up in almost 4 meters In Amazon headwaters regions, such the Madre de Dios Basin, hydro-climatologi cal conditions are controlling spawning activity of catfish larvae. Floods drive changes in th e physical and chemical conditions of the river that can be the signal for catfishes that a favorable environment is ready to ensure the survival of the new fish generation. Based on the classifi cation of fish reproductive st yles suggested by Balon (1984), 64

PAGE 65

these large catfishes (Family Pimelodidae) woul d be considered pelagic spawners, non-guarders and less protective, producers of eggs in large numbers, and with a long larval period terminated by metamorphosis. Then, these characteristics sugge st that successfulness of Amazonian catfish spawning depends on adults by producing and placi ng their larvae in optimal habitat conditions. Chemical characteristics are also potential signa ls for fishes to spawn, and changes of their levels in the water are a consequence of the hydrologic periods (Lowe-McConnell 1987; Barthem et al. 2003; Junk & Wantzen 2004). Significant differences were found mostly for conductivity and dissolved oxygen valu es between transects located in different tributaries, and between transects placed upstream and downstream of the tributary entrance. Differences in conductivity in the Amazon are given by geol ogy through which a basin passes, and its variability in the same river is related to the water period (Barthem et al. 2003). Between October and December, the starting of spawning period, tr ansects in Madre de Dios River presented significantly higher values of conductivity and dissolved oxygen than Tambopata River. Interestingly, the Tambopata River presented more dissolved oxygen values previous to the first significant rising of water level (October), after which values decreased for the spawning period. This can be explained by a dam effect the la rger Madre de Dios produces on the Tambopata River by this time. Regarding to larval produc tion, no significant differen ces were encountered between transects upstream and downstream Tam bopata, which could be indicating not only the low contribution of catfish larvae of this tributar y in the basin, but also that its lower oxygen and conductivity values could be the missing potential conditions that prepare the river channel for spawning. Upstream of La Cachuela transect, an area of rapids exists in the Madre de Dios River, which can be contributing with oxygen cond itions. Transport of sediments from tributaries upstream of the Madre de Dios River seems to be larger than sediment transportation in 65

PAGE 66

Tambopata River (Barthem et al. 2003), which caused the high values of conductivity. Since temperature was not significantly different between transects and ri vers, it could be said that these two chemical variables, oxygen and conducti vity, signal this group of catfishes to the spawning areas. Spatial Distribution of Catfish Larvae Larva densities found in transe cts placed in the Madre de Dios River were significantly greater than those found in the Tambopata transects. No statistical differences were found between transects placed upstream and downstr eam from the entrance of the tributary Tambopata, which indicated the poo r contribution in larval production from this tributary in the Madre de Dios Basin. It is likel y that physical and chemical char acteristics of Tambopata River did not represent the best conditions for spawning of these species. A non-homogeneous larval distribution was advised across each transect. Significant differences in larva densities were found between the five designated habitat categories. Larval densities along side banks were significantly greater than larvae found in areas related to the middle channel. A similar larval distribution acro ss the channel has been reported for larvae in other regions (Harrow & Schlesinger 1980; Holla nd & Sylvester 1983; Araujo-Lima & Oliveira 1998; Robinson et al. 1998; Hum phries & King 2003). Considering only cutting and filling bank habitats, higher larval density values were found in the depos ition zone (filling bank) where lower stream velocities occurred. In Madre de Dios River, catfis h larvae are being transported by the currents and getting clumped at these low-velocity habitats to then occupy areas downstream for feeding and hiding. Young larvae of many othe r species prefer sedimentation areas of the channel, which are in general c onsidered habitats with better conditions such as lower abrasion process of sediments and more food availabi lity (Harrow & Schlesinger 1980; Brown & Coon 1994; Scheideggeer & Bain 1995; Oliveira & Araujo-Lima 1998; Robinson et al. 1998; 66

PAGE 67

Humphries et al. 2002). Some aut hors have also found that high flow velocity habitats appear to increase larval mortality (Robinson et al. 1998). Most of caught larvae were in the early stages of development either with yolk sac or switching to initial phases of exogenous feeding (vestiges of yolk), which meant swimming was not their practice ye t. It was stream velocity that transported them to depositional areas as the flood water in creased. Some authors consider these habitats also suitable and preferred by larvae in olde r stages, which already possess swimming abilities, are able to avoid high stream velocities, and need to forage for food (Robinson et al. 1998). No significant differences were found in vertical distribution of larvae in the river channel, which indicated that a massive larvae downstr eam displacement was occurring in most of channel sections in Madre de Dios River. A similar homogeneous vertical dispersion of fish larvae was advised in the upper Colorado River (Valdez et al. 1985). In Central Amazon, larvae of two species of Characiforms were found to be distributed signi ficantly different in the water column, and this was attributed to a reduction of inter specific competence for space and food in the ample floodplain (Oliveira & Araujo-Lima 1998). Transport of Larvae to Downstream Areas Flooding also produces a lateral connection of running water ec osystems with the adjacent floodplain. The exchange of nutrients and energy from terrestrial phases to the main channel creates a suitable environment for spawning and feeding of adults, juvenile, and larvae (Bye 1984; Lowe-McConnell 1987; Junk & Wantzen 2004) Flooding also emphasizes the dynamic concept of connection and conti nuously changes of physical c onditions along the river (Vannote et al. 1980). In Madre de Dios River, since lateral inundations are short in time, but very frequent, floods are more related to give the optimal conditions to the river to host spawning grounds, and also make the river to become the highway through which larvae will be transported downstream, connecting spawning and nur sery areas of these species, geographically 67

PAGE 68

far away apart. Nursery areas of these species ar e suggested to be locate d in the lowlands, close to the discharge area of the Amazon River next to the Atlantic (B arthem & Goulding 1997). When compared with lowland areas, floodplains in Madre de Dios River are very limited, and if they occur, it will be for quite a few hours. Floodplains f acilitate better c onditions for the development of microbenthos and zooplankton, principal food sources for fish larvae when they switch from endogenous (from yolk sac) to e xogenous feeding (Humphries et al. 2002). The presence of young larvae in all th e transects during the same samp ling periods indicates they are moving downstream and not staying in this region since food offer is very limited, and they take advantage of high flows to reach more favorab le areas downstream for feeding. Predation is another consideration in this headwater region. Large floodplains have been associated with larger number of larvae because they find pl aces for shelter and protection from predators (Oliveira & Araujo-Lima 1998; Brown & Coon 1994) No many areas for shelter, such as vegetation at margins, exist in Madre de Dios banks, which do not offer conditions to larvae for shelter. Further, drifting duri ng high water diminishes the mort ality by predation (Araujo-Lima & Oliveira 1998). Floods in headwaters are necessary to ensure these individuals to reach areas with better conditions to survive and grow. Conclusions Timing of spawning of Amazon large catfishes is driven by hydrologic regime in the Madre de Dios River. Larval production was ev idenced to be significan tly higher during months of high water season than in the low water seaso n. The outstanding increment of larvae presence in the channel also corresponded with starting mo ments of the rainy season, which confirmed the strong link between starting time of spawning behavior of these species and the hydroclimatologic conditions of the en tire basin. The period of reproduc tion of these species in the 68

PAGE 69

Madre de Dios Basin extends from October to Marc h, and this research actually obtained partial samples of two consecutive periods: the end of 2005-2006 and the beginning of 2006-2007. Spatial patterns of larvae within the channe l were driven by hydrologic forces of flooding events. Morphological characteri stics of most captured larvae suggested they were not good swimmers, and given the signifi cant differences in larval pr esence found between filling and cutting banks in all the study ar ea, it is the floods that concen trate larvae in the depositional zones of the channel. The filling banks could act as potential floodplain areas for these larvae in this region, however, given the high variability in water levels recorded in the hydrograph, this condition will last for a short period of time, not being completely beneficial for larvae, in the case they were able to swim and sel ect this habitat to stay and grow. Channel conditions during high water levels featured by temporary short flood peaks, suggest that flooding events are di rectly related with the transportation of larvae to downstream areas. No adequate areas for feeding and protecti on exist in the channel of the Madre de Dios River, since floodplain areas are short in space and time. The almost permanent flood pulses in the Madre de Dios channel, caused by the discharg es of Andean headwater tributaries, will give this river the capacity of constantly transpor ting larvae downstream and no time to create large floodplain as those existing in the lowlands. Floodplains are thought to be the best habitats for survival of larvae, and in Madre de Dios Basin this condition was not occurring to support catfish larvae growth. The necessity of spatial connection along th e large Amazon Basin, between the spawning areas in the headwaters, such as Madre de Di os Basin, and lowland areas where larvae will survive successfully, is sugge sted by this research. Amazon headwater regions do not support areas for nursing larvae of these large catfishes, wh ich are conditioned to reach, in the short term, 69

PAGE 70

lowland areas to survive, and it is flood events that s upport them to achieve th is goal. This spatial connection is also necessary for adult catfishes that, during high water periods, will find the favorable conditions for swim upstream and reach the spawning areas. Two critical moments in natural life history of these catfishes depend on the longitudina l connection of the complete basin, which is controlled by the flow regime. Phys ical interruption of this route or alteration of the temporal pattern of hydrologic regime in Amazon Basin will put under serious risk the populations of these species, by stopping dispersion of larvae or altering the natural upstream migration of adults. Further, any detriment on these catfish populations will also affect important economic activity such as fisheries, and also protein availability for local stakeholders established along the Amazon rivers in the Madeira valley, principal consumers of these species. Finally, natural flow regime in the Amazon h eadwaters is very sensitive to alteration by land use practices, and it is critical not only for su rvival of catfish populations, but also of other species that depend on the floodplain in the lo wland Amazon. Large projects under development in Madre de Dios Basin, such as the Interocean ic Highway, have to c onsider their potential influence in changing flow conditions of this regi on; such alterations will have an effect far downstream. In a similar situation are considered large projects in Brazilian Amazon related to use of water resources, such as dams in the Madeira River. It is critical to include as potential effects of these projects the like lihood of disruption of critical r outes for long distance migratory catfishes. These species consider, from the Atlantic to the Andes, their living habitat, and their natural life history is defined by achieving their principal goal of reach ing both the headwaters near the Andes (in Ecuador, Colombia, Peru, and Bolivia) and the nursery areas near the estuary (in Brazil). 70

PAGE 71

Table 3-1. Levenes test fo r Homogeneity of Larval Ca tches Variances (p < .0001) Source DF Sum of Squares Mean Square F Value Pr > F Season_Week 53 5.7426E9 1.0835E8 3.57 < .0001 Error 2263 6.862E10 30320552 Table 3-2. Test of Fixed Effects (SAS GLIMMIX Procedure Type III, p < 0.0001) showing statistical differences of larva de nsities between weeks and seasons. Effect Num DF Den DF F value Pr > F Season 1 10 128.19 < .0001 Week 1 251 89.21 < .0001 Week*Season 1 251 48.34 < .0001 Table 3-3. Statistical comparisons between larval densities by transects fo r the high water period (Least Square Means. GLIMMIX Procedure, p < 0.0001). Transect name Transect name Degrees of Freedom Pr > | t | La Cachuela El Balcon 2305 0.4524 La Cachuela Puerto Maldonado 2305 0.0003 La Cachuela Bajo Madre de Dios 2305 0.7060 La Cachuela Herrera 2305 0.0535 La Cachuela Candamo Port 2305 0.0929 El Balcon Puerto Maldonado 2305 < .0001 El Balcon Bajo Madre de Dios 2305 0.2619 El Balcon Herrera 2305 0.1749 El Balcon Candamo Port 2305 0.3090 Puerto Maldonado Bajo Madre de Dios 2305 0.0012 Puerto Maldonado Herrera 2305 < .0001 Puerto Maldonado Candamo Port 2305 < .0001 Bajo Madre de Dios Herrera 2305 0.0272 Bajo Madre de Dios Candamo Port 2305 0.0453 Herrera Candamo Port 2035 0.6612 Table 3-4. Statistical comparison be tween larval densities by habitats across the channel for the high water period (Least Square Mean s. GLIMMIX Procedure, p < 0.0001). Habitat Habitat Degrees of Freedom Pr < | t | Cutting bank Filling bank 2307 < .0001 Cutting bank Middle channel 2307 < .0001 Cutting bank Cutting Middle 2307 < .0001 Cutting bank Filling Middle 2307 < .0001 Filling bank Middle channel 2307 < .0001 Filling bank Cutting Middle 2307 < .0001 Filling bank Filling Middle 2307 < .0001 Middle channel Cutting Middle 2307 0.0809 Middle channel Filling Middle 2307 0.2527 Cutting Middle Filling Middle 2307 0.5458 71

PAGE 72

Table 3-5. Statistical comparison between larval densit ies by depths for the high water period (Least Square Means. GLIMMIX Procedure, p < 0.0001). Sample depth Sample depth Degrees of Freedom Pr < | t | 1-meter Surface > 1-m surface 2313 0.3539 Table 3-6. Statistical comparison between egg dens ities by transects in th e Madre de Dios River for the high water period (Least Squares Means. GLIMMIX Procedure, p < 0.0001). Transect Transect Degrees of Freedom P > | t | La Cachuela El Balcon 1868 0.7574 La Cachuela Puerto Maldonado 1868 0.4834 La Cachuela Bajo Madre de Dios 1868 0.0026 El Balcon Puerto Maldonado 1868 0.6998 El Balcon Bajo Madre de Dios 1868 0.0074 Puerto Maldonado Bajo Madre de Dios 1868 0.0208 Table 3-7. Statistical comparison of water parame ters by transects for high water period (Least Squares Means. GLIMMIX Procedure, p < 0.0001). Transect name Transect name Conductivity (microSiemens/cm) Oxygen (mg/L) Temperature (C) La Cachuela El Balcon 0.8774 0.3156 0.7260 La Cachuela Puerto Maldonado 0.9277 0.0623 0.3628 La Cachuela Bajo Madre de Dios < .0001 0.0483 0.0501 La Cachuela Herrera < .0001 < .0001 < 0001 La Cachuela Candamo Port < .0001 0.0003 0.0968 El Balcon Puerto Maldonado 0.9494 0.388 0.5757 El Balcon Bajo Madre de Dios < .0001 0.3292 0.1072 El Balcon Herrera < .0001 < .0001 < 0.0001 El Balcon Candamo Port < .0001 < .0001 0.1751 Puerto Maldonado Bajo Madre de Dios < .0001 0.9068 0.2920 Puerto Maldonado Herrera < .0001 < .0001 < .0001 Puerto Maldonado Candamo Port < .0001 < .0001 0.3843 Bajo Madre de Dios Herrera < .0001 < .0001 0.0009 Bajo Madre de Dios Candamo Port < .0001 < .0001 0.9606 Herrera Candamo Port 0.0050 < .0001 0.0021 72

PAGE 73

Figure 3-1. Location of the st udy area and sampling transects around Puerto Maldonado, Peru. River Stage Puerto Maldonado Figure 3-2. The Madre de Dios Basin, its Andean (south) and lo wland (north) tributaries, and location of Stage in Puerto Maldonad o, where field work was carried out. 73

PAGE 74

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 3-Dec-043-Mar-053-Jun-053-Sep-053-Dec-053-Mar-063-Jun-063-Sep-063-Dec-06 DaysWater level (m) Figure 3-3. Daily water level vari ation for the Madre de Dios Ri ver in Puerto Maldonado, Peru (December 2004 to February 2007). 0 20 40 60 80 100 120 140 160 180 147101316192225283134374043464952WeeksLarval density (50m3)0 1 2 3 4 5 6 7 8 9 10Stage (m) Larval density Stage (m) Figure 3-4. Average larv al density (larvae/50m 3 ) and water level in Madre de Dios River during 2006. 74

PAGE 75

0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 1357911131517192123252729313335373941434547495153WeeksEgg density (50m3)0 1 2 3 4 5 6 7 8 9 10Stage (m) Egg density Stage (m) Figure 3-5. Average egg density (eggs/50m 3 ) and water level in Madre de Dios River. 75

PAGE 76

CHAPTER 4 CONCLUSIONS River dynamics and the reproductive behavior of fish are profoundly related in the Madre de Dios headwater region of the southwestern Amazon Basin in southeastern Peru. Several large catfish species and probably many smaller spec ies of various fish groups have evolved to take advantage of, and now depend on, hydrologic conditions associated with the Andes-Amazon interface zone. The downstream drift of catfish larv ae in the Madre de Dios River, as well as the presence of ripe adults, strongl y indicates that these species spawn in this headwater region during the high water period. The nurse ry habitats for seve ral of the large catfish that spawn in the Madre de Dios headwater region are found only in the Amazon River estuary, more than 3,500 km downstream. Longitudinal migrations of both adults moving upstream and newly born fish downstream, demonstrate that these species ha ve evolved to use nearly the entire east-west extension of the Amazon Basi n for their life histories. A simple stochastic model was proposed to characterize fish larvae recruitment from the Madre de Dios headwaters to the Amazon lowland rivers. Biological a nd hydrological variables were employed to strengthen the model. Simple assumptions about the uniform temporal distribution of flooding events and larvae were tested with the support of the long term hydrologic data derived from historical precipitation records. Th is approach, of generating long term hydrologic conditions from precipitation info rmation, is reasonable and also supported by the nature of the hydro-climatologic conditions characterizing the Andes-Amazon headwater region of the Madre de Dios Basin, where rainfa ll patterns are strongly as sociated with river dynamics. Nevertheless, considering the great spatial variation within this headwater basin, and especially the high Andes and lo wland Amazon, it is necessary to also consider additional potentially important variables, other than just water level, th at could produce environmental 76

PAGE 77

signals that determine when large migratory cat fish spawn. Sediment types, pH and conductivity are among the variables considered. The concept of a two-way gate is introduced to help explain the fundamental role of flooding events in the life cycles of large migratory catfishes in the Amazon Basin. Flow characteristics of the Madre de Dios head waters during the high water period provide environmental signals that induce large catfishes to migrate upstr eam to seek spawning habitats in Andean foothill tributaries. These open-ga te signals are considered biologically significant events, and they are considered as being sent randomly during the rainy season, which in turn will also determine the intensity of the next spawning based on the number of ripe adults present. In this sense the flow regime of the Madre de Dios River heavily determines when and if ripe adults migrate to the headwater region and the nu mber of fish larvae that will be released to migrate downstream. The model is also designed to incorporate othe r non-uniform variables that are expected to be derived from future research. These addi tional variables should im prove forecasting of spawning events and also provide more detailed information on the interaction of variables and hydrologic regimes. Although there is only limited hydr o-climatic information available for most of the Madre de Dios Basin and only short-term fish larvae sampling, the model presented here should nevertheless provide a basis for which to consider how headwater hydrologic regimes need to be maintained for migratory catfish life cycles. Larva production that is released from the Madre de Dios headwater region demonstrates the importance of the Andes-Amazon interface zone in the maintenance of fish abundance and diversity over a much broader area within th e Amazon Basin. Observed drifting patterns of catfish larvae indicate the importance of river channels as the principal habitat where spawning 77

PAGE 78

and downstream/upstream transportation take pla ce. No significant spatial differences were found in the numbers of fish larvae moving downs tream in the Madre de Dios River. This indicates that spawning habitats are near the Andes or in the Andean foothills, and probably mostly the latter. Since many of the species are important commercially the maintenance of hydrologic conditions in the Amazon-Andes headwa ters has important implications for the maintenance of commercial fisheries from the headwaters to the estuary. 78

PAGE 79

LIST OF REFERENCES Araujo-Lima, C.A.R.M. 1994. Egg size and larval development in Central Amazon fish. Journal of Fish Biology 44: 371-389. Araujo-Lima, C.A.R.M. & Oliveira, E.C. 1998. Tran sport of larval fish in the Amazon. Journal of Fish Biology 53: 297-306. Araujo-Lima, C.A.R.M., da Silva, V.V., Petry, P., Oliveira, E.C. & Moura S.M.L. 2001. Diel Variation of larval fish abundance in th e Amazon and Rio Negro. Brazilian Journal of Biology 61(3): 357-362. Adamowski, K. 2000. Regional analysis of annua l maximum and partial duration flood data by nonparametric and L-moment periods. Journal of Hydrology 229: 219-231. Ashkar, F. & Rousselle, J. 1981. Design Discharg e as a Random Variable: A Risk Study. Water Resources Research 17(3): 577-591. Balon E.K. 1984. Patterns in the Evolution of Repr oductive Styles in Fishes. In: Potts, G.W. & Wootton, R.J., eds. Fish Reproduction. Strate gies and Tactics. Academic Press Inc. (London) Ltd. London, pp. 35-53. Barthem, R. & Goulding, M. 1997. The Ca tfish Connection: ecology, migration, and conservation of Amazon predators. Colu mbia University Press. New York. Barthem, R., Goulding M., Forsberg B., Caas, C. & Ortega, H. 2003. Aquatic Ecology of the Rio Madre de Dios. Scientific Bases fo r Andes-Amazon Headwaters Conservation. Amazon Conservation Association. Grfica Biblos S.A. Lima. Bras, R.L. & Rodrguez-Iturbe, I. 1985. Random Functions in Hydrology. Addison Wesley, Raeding, MA. Brown, D.J. & Coon, T.G. 1994. Abundance and Asse mblage Structure of Fish Larvae in the Lower Missouri River and its Tributaries. Transactions of the American Fisheries Society 123: 718-132. Bye, V. 1984. The Role of Environmental Fact ors in the Timing of Reproductive Cycles. In: Potts, G.W. & Wootton, R.J., eds. Fish Re production. Strategies and Tactics. Academic Press Inc. (London) Ltd. London, pp. 187-205. Caas, C. 2000. Evaluacin de los Recursos Pesqueros en la Provincia de Tambopata, Madre de Dios. Conservation International-Peru. Serie Tecnica 1. Lima. De Resende, E.K. 2004. Migratory Fishes of th e Paraguay-Paran Basin excluding the Upper Paran Basin. In: Carolsfeld, J., Harvey, B., Ross, C. & Baer, A., eds. Migratory Fishes of South America. Biology, Fisheries, and Conser vation Status. World Fi sheries Trust. World Bank. International Development Research Center, pp. 103-159. 79

PAGE 80

Evans, M., Hastings, N. & Peacock, B. 1993. Statistical Distributions. Wiley, New York. Favre, A-C., Musy, A. & Morganthaler, S. 2002. Two-site modeling of rainfall based on the Neyman-Scott process. Water Re sources Research 38(12): 1307-1343. Filho, E.Z. & Schulz, U.H. 2004. Migratory Fishes of the Uruguay River. In: Carolsfeld, J., Harvey, B., Ross, C. & Baer, A., eds. Mi gratory Fishes of South America. Biology, Fisheries, and Conservation Status. World Fisheries Trust. World Bank. International Development Research Center, pp. 161-197. Garreaud, R.D., Vuille, M., Clement, A.C. 2003. Th e climate of the Altiplano: observed current conditions and mechanisms of past cha nges. Palaeogeography, Palaeoclimatology, Palaeoecology 194: 5-22. Goulding, M. 1979. Ecologia da Pesca do Rio Made ira. Conselho Nacional de Desenvolvimento Cientfico e Tecnolgico. Instituto Naciona l de Pesquisas da Amaznia. Manaus. Amazonas Goulding, M. 1980. The Fishes and the Forest: e xplorations in Amazonian Natural History. University of California Pr ess. Berkeley, California. Goulding, M. 1988. Ecology and management of migr atory food fishes of the Amazon Basin. In: Almeda, F. & Pringle, C.M., eds. Tropical Rainforests: Diversity and Conservation. San Francisco: California Academy of Sciences, pp. 71-85. Goulding, M., Barthem, R. & Ferreira, E. 2003. The Smithsonian Atlas of the Amazon. Smithsonian Books. Washington and London. Goulding, M., Caas, C., Barthem, R., Forsberg, B., & Ortega, H. 2003. Amazon Headwaters: River, Wildlife and Conservation in Southeas tern Peru. Amazon Cons ervation Association. Grfica Biblos S.A. Lima. Grimm, A.M., Barros, V.R. & Doyle, M.E. 2000. Climate variability in southern South America associated with El Nio and La Nia events. Journal of Climate 13(1): 35-58. Harrow, L.G. & Schlesinger, A.B. 1980. The Larval Fish Recruitment Study. Omaha Public Power District. 92pp. Holland L.E. & Sylvester, J.R. 1983. Distributio n of Larval Fishes Related to Potential Navigation Impacts on the Upper Mississippi River, Pool 7. Tr ansactions of the American Fisheries Society 112: 293-301. Hontela, A. & Stacey, N.R. 1990. Cyprinidae. In : Munro, A.D., Scott, P. & Lam T.J., eds. Reproductive Seasonality in Teleosts: Envir onmental Influences. CRC Press, Inc. Boca Raton. Florida, pp. 53-77. 80

PAGE 81

Humphries, P., King, A.J. & Koehn, J.D. 1999. Fish, flows and flood plains: links between freshwater fishes and their environment in the Murray-Darling River system, Australia. Environmental Biology of fishes 56: 129-151. Humphries, P., Serafini, L.G. & King, A.J. 2002. River regulation and fi sh larvae: variation through space and time. Fr eshwater Biology 47: 1307-1331. Humphries, P. & King, A.J. 2003. Drifting fish larvae in Murray-Da rling Basin rivers: composition, spatial and temporal patterns and distance drifted. Paper presented at Downstream, Movement of Fish in the Murray-Darling Basin, Canberra Workshop 3-4 June 2003. Humphries, P. 2005. Spawning time and early life history of Murray cod, Maccullochella peellii peelii (Mitchell) in an Australia n river. Environmental Bi ology of Fishes 72: 393-407. Junk, W.J. & Wantzen, K.M. 2004. The Flood Pulse Concept: New Aspects, Approaches, and Applications an Update. In: Welcomme, R.L. & Petr, T., eds. Proceeding of the Second International Symposium on the Management of Large Rivers for Fisheries. Volume 2. Food and Agriculture Organization & Me kong River Commission. FAO Regional Office for Asia and the Pacific, Bankgok, pp. 117-149. Kalliolla, R. & Puhakka, M. 1993. Geografa de la Selva Baja Peruana. In: Kalliolla, R.P., Puhakka, M. & Danjoy, W., eds. AMAZONI A PERUANA. Vegetaci n Hmeda Tropical en el Llano Subandino. Proyecto Amazonia Universidad de Turku (PAUT). Oficina Nacional de Recursos Naturales (ONERN), Jyvskyl, pp. 9-21. Langbein, W. 1949. Annual floods and the partial duration flood series. American Geophysical Union Transactions 30: 879-881. Lowe-McConnell, R.H. 1975. Fish Communities in Tropical Freshwaters: Their Distribution, Ecology, and Evolution. London. Longman. Lowe-McConnell, R.H. 1987. Ecological Studies in Tropical Fish Communities. Cambridge University Press. Cambridge. Munro, A.D. 1990. Tropical Freshwater Fish. In: Munro, A.D., Scott, P. & Lam T.J., eds. Reproductive Seasonality in Teleosts: Envir onmental Influences. CRC Press, Inc. Boca Raton. Florida, pp. 145-239. Nakatani, K., Agostinho, A.A., Baumgartner, G., Bi aletzki, A., Sanches, P.V., Makrakis, M.C. & Pavanelli, C.S. 2001. Ovos e Larvas de Peix es de gua Doce: Desenvolvimento e Manual de Identificao. Ictioplncton. EDUEM, Maring. 378 pp. OBrien, J. 2007. ENSO Index according to JMA SSTA (1868-present). Center for OceanAtmospheric Prediction Studies (COAPS). Florid a State University. Tallahassee. Florida. < http://coaps.fsu.edu/jma.shtml >. June 2007. 81

PAGE 82

Oliveira, E.C. & Araujo-Lima, C.A.R.M. 1998. Distribuco das larvas de Mylossoma aureum e M. duriventre (Pisces: Serrasalmidae) nas margens do rio Solimes, AM. Revista Brasileira de Biologia 58(3): 349-358. Pavlov, D.S. & Moslova, E.A. 2006. Downstr eam Migration and F eeding of Young Coho Salmon Oncorhynchus kisutch in the Northern Part of the Range in Kamchatka. Biology Bulletin 3(33): 248-259. Payne, A.I., Sinha, R., Singh, H.R. & Hug S. 2004. A Review of the Ganges Basin: Its Fish and Fisheries. In: Welcomme, R.L. & Petr, T., eds. Proceedings of the Second International Symposium on the Management of Large Rivers for Fisheries. Sust aining Livelihoods and Biodiversity in the New Millennium. Volume 1. Poff, N.L., Allan, J.D., Bain, M.B., Karr, J.R., Prestegaard, K.L., Ritcher, B.D., Sparks, R.E. & Stromberg, J.C. 1997. The Natural Flow Regime A Paradigm for River Conservation and Restoration. BioScience 47: 769-784. Poveda, G., Waylen, P., Pulwarty, R. 2006. Annual and inter-annual variab ility of the present climate in northern South America and southern Mesoamerica. Palaeogeography, Palaeoclimatology, Palaeoecology 234(1): 3-27. Puhakka, M. & Kalliolla, R.P. 1993. La Vegetaci n en Areas de Inundacin en la Selva Baja de la Amazona Peruana. In: Kalliolla, R.P., Puhakka, M. & Danjoy, W., eds. AMAZONIA PERUANA. Vegetacin Hmeda Tropical en el Llano S ubandino. Proyecto Amazonia Universidad de Turku (PAUT). Oficina N acional de Recursos Naturales (ONERN), Jyvskyl, pp. 113-138. Puhakka, M., Kalliolla, R.P., Salo, J. & Rajasilta, M. 1993. La Sucesin Forestal que sigue a la Migracin de Ros en la Selva Baja Peruana. In: Kalliolla, R.P., Puhakka, M. & Danjoy, W., eds. AMAZONIA PERUANA. Vegetacin Hmeda Tropical en el Llano Subandino. Proyecto Amazonia Universidad de Turku (PAUT). Oficina Nacional de Recursos Naturales (ONERN), Jyvskyl, pp. 167-201. Rasmussen, P.F. & Rosbjerg, D. 1991. Evaluation of risk concepts in partial duration series. Stochastic Hydrology and Hydraulics 5: 1-16. Rmillard, L., Rousselle, J., Ashkar, F. & Sparks, D. 2004. Analysis of the Seasonal Nature of Extreme Floods across Canada. Journal of Hydrologic Engineering 9(5): 392-401. Rosbjerg, D. 1985. Estimation in partial duration series with independent and dependent peak values. Journal of Hydrology 76: 183-195. Robinson, A.T., Clarkson, R.W. & Forrest, R.E. 1998. Dispersal of Larval Fishes in a Regulated River Tributary. Transactions for the American Fisheries Society 127: 772-786. Scheidegger, K.J. & Bain, M.B. 1995. Larval Fi sh Distribution and Microhabitat use in FreeFlowing and Regulated Ri vers. Copeia 1: 125-155. 82

PAGE 83

Scott, A.P. 1990. Salmonids. In: Munro, A.D., Scott, P. & Lam T.J., eds. Reproductive Seasonality in Teleosts: Environmental Influe nces. CRC Press, Inc. Boca Raton. Florida, pp. 33-51. Sumpter, J.P. 1990. General Concepts of Seas onal Reproduction. In: Munro, A.D., Scott, P. & Lam T.J., eds. Reproductive Seasonality in Teleosts: Environmental Influences. CRC Press, Inc. Boca Raton. Florida, pp. 13-31. Tapley, T. & Waylen, P. 1990. Spatial variabilit y of annual precipitation and ENSO events in western Peru. Hydrological Sciences Journal 35(4): 429-446. Taesombut, V. & Yevjevich, V. 1978. Use of partia l flood series for estimating distribution of maximum annual flood peak. Colorado St ate University. Hydrology paper. 71pp. Todorovic, P. & Zelanhasic, E. 1970. A stochas tic model for flood analysis. Water Resources Research 6(6): 1641-1648. Todorovic, P. 1978. Stochastic models of fl oods. Water Resources Research 14(2): 345-356. Valdez, R.A., Carter, J.G. & Ryel, R.J. 1985. Drift of larval fishes in the upper Colorado River. Proceedings of the Western Association of Fish and Wildlife Agencies 65:171-185. Vannote, R.L., Minshall, G.W., Cummins, K.W., Sedell, J.R. & Cushing, C.E. 1980. The River Continuum Concept. Canadian Journal of Fisheries and Aquatic Sciences 37: 130-137. Waylen, P. & Woo, M. 1983. Stochastic analysis of high flows in some Central Bristish Columbia rivers. Canadian Journal of Civil Engineering 10(2): 205-213. Waylen P. & Laporte, M.S. 1999. Flooding and th e El Nio-Southern Oscillation phenomenon along the Pacific coast of Costa Rica. Hydrological Processes 13(16): 2623-2638. Welcomme, R.L. 1979. Fisheries Ecology of Floodplain Rivers. Longman Inc. New York. Welcomme, R. & Halls, A. 2002. Dependence of tropi cal river fisheries on flow. Paper presented at the International Confer ence on Environmental Flows for River Systems. 3-8 March 2002, Cate Town, South Africa. Winemiller, K.O. 2005. Floodplain river food we bs: generalizations and implications for fisheries management. In: Welcomme, R.L. & Petr, T., eds. Proceedings of the Second International Symposium on the Management of Large Rivers for Fisheries Volume 2. Mekong River Commission, Phnom Penh, Cambodia. Woo M. & Waylen, P. 1986. Probability stud ies of floods. Applied Geography 6: 185-195. 83

PAGE 84

BIOGRAPHICAL SKETCH Carlos Canas was born in 1973 in Chimbote, Ancash, Peru, where he grew up and got his elementary and high school education. In 1991 he moved to Lima, and in 1997 he obtained his bachelors degree in biology at the Universidad Nacional Mayor de San Marcos Lima, Peru, with specialization in hydrobiology and fisheries. During his year s as an undergraduate student, he worked extensively on freshwater fishes in the Department of Ichthyology of the Museo de Historia Natural Javier Prado Lima, Peru. In the early 1990s he also began participating in scientific expeditions to various Peruvian rivers. As a fish ecologist he has 7 years of field experience in the Amazon Basin and Andes of Pe ru, and has conducted aq uatic research with Conservation International (C I) and the Amazon Conservation Association (ACA). In 2005 he enrolled in the Masters program of the Geograp hy Department of the Univ ersity of Florida to pursue a specialization in hydrol ogy and freshwater ecosystems, and to gain knowledge in GIS and remote sensing. At present he is active in sc ientific research and c onservation efforts in the Madre de Dios Region of southeastern Peru. 84