UUUUpper PPPParaguay RRRRiver BBBBasin GGGGIS DDDDatabase - Expanding the Pilot Project -

Upper Paraguay River Basin GIS DatabaseUpper Paraguay River Basin GIS DatabaseUpper Paraguay River Basin GIS DatabaseUpper Paraguay River Basin GIS Database ---- Expanding the Pilot Project Expanding the Pilot Project Expanding the Pilot Project Expanding the Pilot Project ----

Edited by

Mario Cardozo Dawn Browne

Montserrat Carbonell

Prepared by the “Upper Paraguay River Basin GIS Database Consortium”

BoliviaBoliviaBoliviaBolivia BrazilBrazilBrazilBrazil ParaguayParaguayParaguayParaguay Pamela Rebolledo Fábio Ayres Aníbal Aguayo Heidi Resnikowski Luiz Benatti Rob Clay Julia Boock Claudia Mercolli Lindalva Cavalcanti Laura Rodríguez Mário Dantas Oscar Rodas Gislaine Disconzi Wolf Eberhardt Eliani Fachim Nelson Laturner Bill Liu Humberto Maciel Carlos Padovani Sylvia Torrecilha Ayr Trevisanelli U.S.AU.S.AU.S.AU.S.A. Dawn Browne Mario Cardozo Kristine Kuhlman Montserrat Carbonell Dick Kempka Nancy Thompson

December 2004

Printed by: Ducks Unlimited, Inc., Memphis, TN, U.S.A. Prepared with financial assistance of Ducks Unlimited, Inc., USDA Forest Service and US Fish and Wildlife Service

Copyright: The organizations responsible for this publication have waived copyright. ISBN: 1 932052 17 8 Suggested citation: Cardozo, M., Browne, D. & Carbonell, M. (Eds.) 2004. Upper Paraguay River Basin GIS Database, Expanding the Pilot Project. Ducks Unlimited, Inc., Memphis, TN, USA. Available from: Ducks Unlimited, Inc., One Waterfowl Way, Memphis, TN 38120-2351, U.S.A. http://www.ducks.org/conservation/latinamerica_projects.asp The presentation of material in this book and the geographical designation employed do not imply the expression of any opinion whatsoever on the part of Ducks Unlimited, Inc. concerning the legal status of any country, area or territory, or concerning the delimitation of its boundaries or frontiers. The information contained in this book and accompanying maps are unsuited for, and shall not be used for any regulatory purpose of action, nor shall the report or accompanying maps be the basis for any determination relating to impact assessment or mitigation. 2004

Contents

1. INTRODUCTION/ INTRODUCCIÓN /INTRODUÇÃO ............................................ 1

2. EXPANDING THE PILOT PROJECT ......................................................................... 7

2A. RELATÓRIO TÉCNICO DE ATIVIDADES DA ECOTRÓPICA NA ESTRADA PARQUE TRANSPANTANEIRA ............................................................................................ 7 2B. LAND COVER CLASSIFICATION FOR THE UPPER PARAGUAY RIVER BASIN PILOT PROJECT AREA ................................................................................................ 40 2C. UPPER PARAGUAY RIVER INUNDATION PREDICTION USING RAINFALL AND NDVI: BASIN AND SIX SUB BASIN MODELS ...................................................................... 69

3. FURTHER CONSERVATION EFFORTS IN THE UPRB........................................ 87

3A. BOLIVIA: RESULTS OF THE PANTANAL PROGRAM IN BOLIVIA (WWF BOLIVIA)......... 87 3B. BRAZIL: PLANNING AND MANAGEMENT OF THE “NASCENTES DO RIO TAQUARI” STATE PARK 89 3C. BRAZIL: FLOOD MONITORING IN THE PANTANAL WETLAND............................. 95 3D. BRAZIL: RE-DRAWING THE PANTANAL WETLAND DELINEATION IN BRAZIL, BOLIVIA AND PARAGUAY............................................................................................... 98 3E. BRAZIL: FIRE MONITORING AND ANALYSIS FOR THE BRAZILIAN PANTANAL ......... 103 3F. PARAGUAY: USE OF REMOTE SENSING TO PLAN WATERBIRD CONSERVATION IN THE CENTRAL PARAGUAYAN CHACO ........................................................................ 110

4. ACKNOWLEDGEMENTS....................................................................................... 120

This page left blank intentionally

1

1. INTRODUCTION/ INTRODUCCIÓN /INTRODUÇÃO The results of the work conducted for the advancement of the “Upper Paraguay River Basin GIS Database Project” (UPRB Project) are compiled in the present report. During the Pilot Project (2000-2003) remote sensing and geographic information system (GIS) techniques were applied to a trinational pilot area. These methods were tested and adjusted, and environmental data of regional significance for resource management of the UPRB were obtained. The results of the Pilot Project were published in 2003, including descriptions of the methods employed and the GIS data produced1. The report also emphasized the challenge presented when working with sixteen partner organizations in five countries, Bolivia, Brazil, Paraguay, Canada and the U.S.A. Because the main goal of the project was to build technical capacity, promote transboundary cooperation and standardization of methods for easy sharing of information, the results of the Pilot Project should not be measured only by its products, but also by the network of professionals formed and the GIS capacity built in each of the partner organizations. The wider application of methods, data and results led to the successful implementation of conservation projects in the UPRB. This second report is a compilation of the Pilot Project results and ancillary information generated through conservation projects implemented by the different partner organizations. Since the completion of the Pilot Project in 2003, the consortium of partner organizations has completed a number of additional studies including an impact analysis along the Transpantaneria highway in Mato Grosso, Brazil, mapping vegetation and land use (Section 2a); a land cover classification system for the UPRB, for use with remotely sensed data (Section 2b); and a model for river water level prediction using Normalized Difference Vegetation Index (NDVI), precipitation and river water level data for several sub-basins in the UPRB (Section 2c)2. GIS and remote sensing are technologies broadly employed as resource-management tools, and the methods used, data gathered and results obtained through the Pilot Project and ancilliary activities have already been applied to conservation projects by different partners in South America. Only a few examples have been included in this CD-atlas to provide an idea of the potential application of a GIS database for the conservation, management and sustainable use of natural resources in the UPRB. These include the delineation of buffer zones and environmental zoning for the management plan for the Taquari State Park in Mato Grosso do Sul, Brazil (Section 3b); an analysis of maximum and minimum inundation areas for four sub-basins of the Brazilian UPRB

1 Browne, D., Carbonell, M. & Kempka,D. (Eds.) 2003. Upper Paraguay River Basin GIS Database, Pilot Project I. Ducks Unlimited, Inc., Memphis, TN, USA. 2 The editors of this report have made only minor changes to the papers and are not responsible for their contents; for full versions or for more information about each study please contact the corresponding authors.

2

(Section 3c); preliminary results of wetland delineation (Section 3d) a study of fire incidence (Section 3e); and a classification of wetlands in the central Chaco region of Paraguay (Section 3f). As a complement to this report, a CD-atlas was prepared including an interactive map with selected data generated through the Pilot Project and ancilliary activities in which these datasets can be viewed within a simple ArcGIS desktop application. To demonstrate the usefulness of analyzing environmental data in GIS format, this application allows the user to select and display different datasets of relevance for conservation in the UPRB. The complete datasets from the UPRB Project have not been included in this CD-atlas due to size limitations, but are available through an FTP server accessible via the internet.3

3 Please contact mcarbonell@ducks.org or dbrowne@ducks.org for FTP access information.

3

Introducción Los resultados del trabajo realizado para expandir el proyecto “Upper Paraguay River Basin GIS Project” (Proyecto UPRB) han sido compilados en el presente informe. Durante el Proyecto Piloto (2000-2003) técnicas de sensoramiento remoto y sistemas de información geográfica (SIG) fueron aplicadas a un área piloto trinacional. En el proceso, estos métodos se evaluaron y ajustaron, obteniéndose datos ambientales de importancia regional para el manejo de los recursos de la Cuenca Alta del Rio Paraguay (CARP). Los resultados del Proyecto Piloto fueron publicados en 2003 en un informe describiendo los métodos empleados y datos de SIG generados4. En ese informe también se enfatizó el desafío que representó trabajar con 16 organizaciones en cinco países: Bolivia, Brasil, Paraguay, Canadá y los Estados Unidos. El propósito primario del proyecto ha sido la capacitación técnica, la promoción de cooperación transfronteriza y la estandarización de métodos para facilitar el intercambio de información, por lo tanto los resultados del Proyecto Piloto no han de medirse sólo por sus productos, sino también por la red de profesionales formada, la capacitación en SIG lograda, y la más amplia aplicación de estos métodos, datos y resultados a la exitosa implementación de proyectos de conservación en la CARP. Por tanto, este informe es una compilación de los resultados del Proyecto Piloto y los proyectos de conservación desarrollados por las distintas organizaciones involucradas. Desde la finalización del Proyecto Piloto en 2003, el consorcio de organizaciones del Proyecto UPRB ha completado varios estudios adicionales tales como: un análisis de impacto ambiental a lo largo de la carretera “Transpantaneira” en Mato Grosso, Brasil, incluyendo mapeo de la vegetación y uso de la tierra (Sección 2a); un sistema de clasificación de cobertura terrestre para la CARP y uso con datos de sensores remotos (Sección 2b); y un modelo de predicción de nivel de río para algunas sub-cuencas de la CARP, usando “Normalized Vegetation Difference Index” (NDVI), precipitación y datos historicos de nivel de río (Sección 2c)5. Las tecnologías de SIG y sensoramiento son ampliamente utilizadas como herramientas para el manejo de recursos, y los métodos empleados, datos recolectados y resultados obtenidos en el Proyecto Piloto y actividades adicionales ya han sido aplicados a proyectos de conservación por las distintas organizaciones participantes en América del Sur. Sólo una selección de estos proyectos ha sido incluida en este informe para mostrar las aplicaciones potenciales de una base de datos en SIG para conservación,

4 Browne, D., Carbonell, M. & Kempka,D. (Eds.) 2003. Upper Paraguay River Basin GIS Database, Pilot Project I. Ducks Unlimited, Inc., Memphis, TN, USA. 5Los editores de este informe sólo han hecho cambios mínimos a los trabajos incluidos y no son responsables por sus contenidos; para versiones completas o más información sobre estos trabajos, por favor contacte a los autores correspondientes.

4

manejo, y uso sustentable de recursos naturales en la CARP. Estos proyectos incluyen la delimitación de zonas de amortiguamiento y zonificación ambiental para el plan de manejo del Parque Estadual Taquari en Mato Grosso do Sul, Brasil (Sección 3b); un análisis de área de inundación máxima y mínima para cuatro sub-cuencas de la sección brasileña de la CARP (Sección 3c); resultados preliminares de estudios sobre delimitación de humedales (Sección 3d); y distribución de incendios (Sección 3e); y una clasificación de humedales en la región central del Chaco paraguayo (Sección 3f). Como complemento a este informe, se ha preparado un CD-atlas incluyendo un mapa interactivo con parte de los datos generados en el Proyecto Piloto y varias actividades relacionadas. En este mapa los datos pueden verse en una aplicación simple de ArcGIS desktop. Para demostrar la utilidad de analizar datos ambientales en formato SIG, esta aplicación permite seleccionar y visualizar diferentes datos de relevancia para conservación en la CARP. Los datos completos del Proyecto UPRB no han sido incluidos en el CD-atlas por limitaciones de espacio, pero están disponibles a través de un servidor FTP por internet6.

6 Contacte a dbrowne@ducks.org o a mcarbonell@ducks.org para información sobre accesso al servidor de FTP.

5

Introdução Os resultados do trabalho realizado para a expansão do projeto “Upper Paraguay River Basin GIS Database Project” (Projeto UPRB) foram compilados no presente relatório. Durante o Projeto Piloto (2000-2003) técnicas de sensoreamento remoto e sistemas de informação geográfica (SIG) foram aplicadas a uma área piloto trinacional. No processo, estes métodos foram avaliados e ajustados, obtendo-se dados ambientais de importância regional para o manejo dos recursos da Bacia do Alto Paraguai (BAP). Os resultados do Projeto Piloto foram publicados num relatório sobre os métodos usados e os dados de SIG gerados7. Nesse relatório também se enfatizou o desafio que representou trabalhar com 16 organizaçõoes em cinco países: Bolivia, Brasil, Paraguai, Canadá e Estados Unidos. Como o propósito primário do projeto tem sido de capacitação técnica, promoção da cooperação transfronteriza e estandardização de métodos para facilitar o intercâmbio de informação, os resultados do Projeto Piloto não devem medir-se só pelos seus produtos, mas também pela rede de profissionais formada, a capacitação em SIG obtida, e uma aplicação mais ampla dos métodos, dados e resultados à bem sucedida implementação de projetos de conservação na BAP. Por tanto, este relatório é uma compilação dos resultados do Projeto Piloto e projetos de conservação desenvolvidos pelas organizações envolvidas. Desde a finalização do Projeto Piloto em 2003, o consórcio de organizações do Projeto UPRB completou vários estudos adicionais como: uma análise de impacto ambiental ao longo da estrada parque “Transpantaneira” no estado de Mato Grosso, Brasil, incluindo o mapeamento da vegetação e uso do solo (Secção 2a); um sistema de classificação da cobertura terrestre para a BAP e uso com dados de sensores remotos (Secção 2b); e um modelo de de nível de rio para algumas sub-bacias da BAP, usando “Normalized Vegetation Difference Index” (NDVI), precipitação e dados anteriores de nível de rio (Secção 2c) 8. As tecnologias de SIG y sensoreamento remoto são amplamente utilizadas como ferramentas para o manejo de recursos, e os métodos usados, dados coletados e resultados obtidos no Projeto Piloto e atividades adicionais já foram aplicados a projetos de conservação na América do Sul pelas organizações participantes. Apenas uma seleção destes projetos foram incluídos neste relatório para mostrar as aplicações potenciais de uma base de dados em SIG para conservação, manejo, e uso sustentável dos recursos naturais na BAP. Estes projetos incluem a delimitação de zonas de amortecimento y zonificação ambiental para o plano de manejo do Parque Estadual 7 Browne, D., Carbonell, M. & Kempka,D. (Eds.) 2003. Upper Paraguay River Basin GIS Database, Pilot Project I. Ducks Unlimited, Inc., Memphis, TN, USA. 8 Os editores deste relatório só fizeram mudanças mínimas aos trabalhos incluídos e não são responsáveis pelos seus conteúdos; para versões completas ou mais informação sobre estes trabalhos, por favor contacte aos autores correspondentes..

6

Taquari em Mato Grosso do Sul, Brasil (Secção 3b); uma análise da máxima e mínima área de inundação para quatro sub-bacias da seção brasileira da BAP (Secção 3c); resultados preliminares de estudos sobre delimitação de zonas húmidas (Secção 3d) e distribução de incêndios (Secção 3e) na BAP; e uma classificação de zonas húmedas na região central do Chaco paraguaio (Secção 3f).

Como complemento a este relatório, um CD-atlas foi preparado incluindo um mapa interativo com parte dos dados gerados no Projecto Piloto e actividades relacionadas. Neste mapa os datos podem ser vistos numa simples aplicação de ArcGIS desktop. Para demostrar a utilidade de analizar dados ambientales em formato SIG, esta aplicação permite seleccionar e visualizar diferentes dados de relevância para conservação na BAP. Os dados completos do Projeto UPRB não foram incluídos no CD-atlas por limitações de espaço, mas estão disponíveis através de um servidor FTP na internet9.

9 Por favor, contacte mcarbonell@ducks.org ou dbrowne@ducks.org para informação sobre acesso ao servidor FTP .

7

2. EXPANDING THE PILOT PROJECT

2A. RRRRELATÓRIO TÉCNICO DE ELATÓRIO TÉCNICO DE ELATÓRIO TÉCNICO DE ELATÓRIO TÉCNICO DE ATIVIDADES DA ATIVIDADES DA ATIVIDADES DA ATIVIDADES DA EEEECOTRÓPICA NA COTRÓPICA NA COTRÓPICA NA COTRÓPICA NA EEEESTRADA STRADA STRADA STRADA PPPPARQUE ARQUE ARQUE ARQUE TTTTRANSPANTANEIRARANSPANTANEIRARANSPANTANEIRARANSPANTANEIRA

WOLF EBERHARD10

SSSSUMMARYUMMARYUMMARYUMMARY This document reports the results of the project conducted by Ecotrópica (Fundação de Apoio à Vida nos Trópicos) with the support of Ducks Unlimited, Inc. (DU) to gather data related to the current state of land use and vegetation along the “Estrada Parque Transpantaneira” (Highway-Park Transpantaneira) (EPT), for a Geographic Information System (GIS) database. In a first phase, project activities focused on the: − Upgrade of hardware and software for the GIS laboratory, − Compilation of ancillary data, and their metadata, related to the cartography of the

Upper Paraguay River Basin section in the state of Mato Grosso, and − Creation of an index of aerial photography for the same area. During the second phase of the project, anthropogenic and/or natural impacts on the vegetation along the EPT were assessed, for which land cover was mapped in detail and properties were located in the study area. RRRRESUMOESUMOESUMOESUMO O presente relatório refere-se às atividades da Ecotrópica (Fundação de Apoio à Vida nos Trópicos) desenvolvidas em parceria com a Ducks Unlimited, Inc. (DU) com o objetivo de realizar levantamento de dados e desenvolver estudos sobre o estado atual da cobertura vegetal, ocupação e uso do solo da Estrada Parque Transpantaneira (EPT) para colaborar na criação do SIG Pantanal. Numa primeira fase os trabalhos realizados focalizaram os seguintes aspectos: − Otimização da sala de geoprocessamento com a colocação de alguns itens de

infraestrutura,

10 Fundação Ecotrópica, Cuibá, Mato Grosso, Brazil; dewolf@terra.com.br. Este relatório contém treixos de relatórios parciais internos realizados para a Ecotrópica durante 2002-2003 pelo profissional E. A. Silveira.

8

− Atualização de softwares e dos equipamentos de informática para as atividades técnicas, administrativas e de gestão dos recursos, e

− Coleta de dados secundários abordando a compilação de metadados de referências cartográficas e confecção de um índice de aerofotografias, ambos referentes à região da Bacia do Alto Paraguai, no Estado de Mato Grosso.

Numa segunda fase, as atividades do projeto enfocaram a avaliação dos impactos antrópicos e/ou naturais sobre a vegetação ao longo da EPT, bem como o mapeamento da vegetação e a localização dos limites de propriedades ao longo da Unidade de Conservação com os nomes dos respectivos donos e, ainda, a digitalização da EPT e estradas secundárias. IIIINTRODUÇÃONTRODUÇÃONTRODUÇÃONTRODUÇÃO A existência de um número reduzido de informações sobre o Pantanal levou a Ducks Unlimited Inc. (DU) a estimular a criação de um Sistema de Informação Geográfica (SIG) com dados sobre a região. A idéia foi contemplar a região do Pantanal do Brasil, da Bolívia e do Paraguai, numa proposta de trabalho conjunto. Foram, então, realizadas reuniões de estudo e vários treinamentos para nivelamento de conhecimentos sobre o uso de programas específicos de computador, para capacitação do pessoal dos três países para dar início à criação do SIG Pantanal. Com a finalidade de adequar técnicas e procedimentos relacionados com trabalhos de campo e tratamento de imagens de satélite, foi realizado o Projeto Piloto I que contemplou uma região comum ao Brasil, Bolívia e Paraguai. O Projeto Piloto I foi concluído com sucesso e os resultados já foram divulgados através de trabalhos impressos. Com base nos resultados obtidos com a experiência do Projeto Piloto I foram iniciadas atividades com vistas à realização deste estudo ao longo da transpantaneira. Este projeto visa contemplar o Estado de Mato Grosso e teve suas atividades voltadas para uma Unidade de Conservação: a Estrada Parque Transpantaneira que está localizada no município de Poconé, sudoeste do Estado de Mato Grosso e tem início no quilómetro (km) 17 da estrada MT 060, no posto de fiscalização da FEMA e IBAMA. No início da década de 70 teve início a construção da parte da MT 060, Poconé-Porto Joffre, que passou a se chamar Transpantaneira. Ela faz a ligação da cidade de Poconé ao complexo hoteleiro Porto Joffre na margem direita do Rio Cuiabá, com uma extensão de 146 km com 116 pontes. A Transpantaneira é um dique construído em pleno Pantanal, na direção geral norte-sul, que represa a água durante o período da cheia que acontece na direção geral leste-oeste. Com a finalidade de permitir a passagem da água sem destruir o aterro, foram deixadas 116 passagens, onde algumas são naturais como

9

a do Rio Bento Gomes, Rio Pixaim, Rio Cassange e inúmeros corixos. A construção deste aterro, de alto impacto à zona inundável, permite, aos criadores de gado da região, escoar a sua produção durante quase o ano todo, bem como, facilita o acesso de turistas até o Rio Cuiabá, no final da Transpantaneira. O material necessário à construção do aterro foi retirado de caixas de empréstimo localizadas à esquerda e à direita da Transpantaneira. As cavidades resultantes da retirada de terra, alguns com mais de 300 metros cúbicos (m3) cada um, resultaram em reservatórios de água que, ao longo do tempo, se transformaram em sistemas de criação de vários tipos de pequenos animais (crustáceos, moluscos, peixes, pequenos répteis) e numerosas plantas que participam da dieta preferencial de muitos pássaros, répteis e mamíferos. Em conseqüência da fartura de alimento, durante o período da vazante, ocorre uma concentração natural da fauna ao longo da Transpantaneira criando assim um atrativo para visitação e turismo. Houve um momento em que a concentração de jacarés (Caiman yacare) se transformou numa tragédia, pois muitas centenas deles foram atropelados pelos veículos que transitavam pela Transpantaneira. Em 1996 a Estrada Transpantaneira foi transformada, pelo Decreto Estadual nº 1028, de 26 de julho, em uma Unidade de Conservação Estadual: a Estrada Parque Transpantaneira. No presente estudo sobre esta Unidade de Conservação, os impactos foram avaliados apenas qualitativamente, sendo considerada a presença de espécies exóticas (invasoras), áreas queimadas, placas de publicidade, erosão do aterro, abertura de drenos, construção de diques e desmatamentos. MMMMETODOLOGIAETODOLOGIAETODOLOGIAETODOLOGIA

Metadatos

Depois de escolhido e separado o material cartográfico de interesse para a Bacia do Alto Paraguai (BAP), foi contratado um técnico para fazer o lançamento de 171 registros de metadados em “software” apropriado, o MetaLite. Após a conferência de todos os registros, foi gerada uma cópia do banco de dados MetaLite.dbf. Este arquivo poderá ser consultado e atualizado através do MS Access 2000 ou então poderá ser colocado no Programa de Metadados (MetaLite) substituindo o MetaLite.dbf original, previamente renomeado para MetaLite.old. Esta opção permite ao Coordenador do SIG Pantanal escolher os registros de maior interesse e gerar apenas os arquivos de exportação desejados.

Pluviometria Em virtude da indisponibilidade de dados atuais sobre estações e níveis pluviométricos na região da EPT, foi utilizado um mapa (Figura 1) de Precipitação Média – Mês de Junho de 1979 do trabalho Estudo de Desenvolvimento Integrado da Bacia do Alto Paraguai

10

(EDIBAP), que apresenta características semelhantes às do semi-árido, com precipitações anuais em torno de 1100 milimetros (mm).

Figura 1. Mapa de precipitação na BAP, região de Mato Grosso

Fotoíndices A região do Pantanal do Brasil e a BAP estão contidas no interior do retângulo formado pelas seguintes coordenadas geográficas: W60°, S14° e W52°30´, S19°, o que significa um mosaico de 25 folhas na escala de 1:250 000 da base cartográfica oficial. Com base nestas coordenadas e no contorno da BAP para a região de Mato Grosso, foi feita uma solicitação à 5ª Divisão de Levantamento do Exército para confecção dos fotoíndices necessários para a cobertura da área. Foram gerados 113 fotoíndices com uma cobertura de 00°30´x 00°30´, cada um, equivalente a uma folha na escala de 1:100 000 da base cartográfica oficial. Os fotoíndices, depois de identificados, conforme a grade da Figura 2, foram escanerizados em scanner de rolo marca Ocê, modelo 9400, com resolução de 300 dpi, em preto e branco e armazenados em CD-ROM com arquivos no formato tif.

11

Figura 2. Grade de localização e distribuição dos fotoíndices

Avaliação dos Impactos Ambientais na EPT Os impactos ambientais foram avaliados qualitativamente, portanto, não foi realizada uma avaliação da sua grandeza ou extensão. Foi anotada somente a presença do impacto e descritas algumas considerações sobre a sua importância. Entre os impactos ambientais observados na EPT, destacam-se a presença de espécies exóticas (invasoras), queimadas, placas de publicidade, erosão do aterro, abertura de drenos, construção de diques e desmatamento.

Mapeamento Fitofisionômico A caracterização da vegetação ao longo da Estrada Parque Transpantaneira, foi realizada verificando-se os tipos vegetacionais que ocorrem na margem da Estrada e em incursões nas estradas vicinais, principalmente naquelas que dão acesso às propriedades. Assim, o efeito de borda e as áreas degradadas às margens da Estrada, que impedem a visualização dos tipos fitofisionômicos, foram contornados.

12

A coleta de dados teve início com viagens de campo realizadas no período de dezembro de 2002 a janeiro de 2003 onde, partindo-se de Porto Joffre, sistematicamente o veículo foi parado sobre cada ponte com a finalidade de colher suas coordenadas métricas (Universal Transverse Mercator - UTM) e anotar informações com fotografias sobre a vegetação. As coordenadas foram registradas em um Garmin GPSMAP 195 configurado para UTM com Datum WGS84. As fotografias digitais foram obtidas através de uma câmera Kodak DC 290 Zoom. Com a finalidade de melhorar a autonomia em horas de trabalho da câmera Kodak DC 290 Zoom, foi realizada uma adaptação num inversor de corrente, marca Trancham, com entrada de 12 VCC e saída de 110 VCA com 280 W de potência, para que pudesse ser plugado no sistema elétrico (acendedor de cigarros) de qualquer veículo. Foi realizada, também, uma pequena modificação para que o cabo do Digital Câmera AC Adapter pudesse receber uma extensão de 10 m para permitir ao operador sair do veículo e subir em cima da capota para obter melhor ângulo de visualização. Esta modificação permitiu que a câmera fosse utilizada indefinidamente e sem necessidade da utilização das suas baterias originais (Figuras 3 e 4). O sistema também recebeu uma tomada para plugar o Garmin GPSMAP 195 que tem um grande consumo de energia. Seis baterias AA permitem uma autonomia de apenas 6 (seis) horas de operação contínua. O sistema foi testado em automóvel, barco e avião monomotor.

Inversor de correnteAdaptador AC (127V)para câmera

Plug DC (12V) paraGPSMAP 195

Plug para sistemaelétrico veicular

Garmin GPSMAP 195Antena externaGPSMAP 195

Conector DC (6V)para câmera

Extensão com 10,0 mde comprimento

Sistema de inversão do corrente

Figura 3. Fonte de energia para funcionamento simultâneo de Garmin GPSMAP 195 e da câmera digital Kodak DC 290 Zoom

13

Adapador AC paracâmera DC 290 Zoom

Plug DC paraGPSMAP 195

Inversor de corrente

Painel trazeiro do inversor de corrente

Figura 4. Conexão do Adaptador AC para a câmera Kodak DC 290 Zoom Na seqüência, foram utilizadas as folhas Poconé, MIR 404 e Ilha Camargo, MIR 418, do Plano de Conservação da Bacia do Alto Paraguai (PCBAP) (PNMA, 1997), na escala de 1:250 000, para suporte inicial e orientação no estudo da vegetação existente ao longo da EPT (Figura 5). Para a digitalização do curso da EPT e estradas secundárias, foram utilizadas as mesmas folhas, porém, não do PCBAP e sim da base cartográfica oficial (Diretoria do Serviço Geográfico do Exército, ou DSG). Estes mapas foram escanerizados num scanner de rolo marca Ocê, modelo 9400, com 300 dpi de resolução e posteriormente foram georreferenciadas, no IDRISI (SIG da ClarkLabs), usando quatro interseções do grid UTM como pontos de controle. Estes bitmaps foram importados para o CartaLinx (Software para digitalização de vetores da ClarkLabs) e usados como “backdrop image” para obtenção, por digitalização “on screen”, dos vetores correspondentes à EPT e estradas secundárias. Num segundo momento foram utilizadas duas imagens de satélite Landsat Enhanced Thematic Mapper (ETM)+7, bandas 5, 4 e 3 (Red, Green, Blue – RGB), WRS 227/071, de julho de 2000, correspondente à região de Poconé e WRS 227/072, de maio de 2001, correspondente à região de Ilha Camargo. As duas imagens foram unificadas num mosaico. Os “layers” origindados pelo CartaLinx foram carregados no ArcView 3.2.a, onde a nova “view” foi projetada para UTM com Datum WGS84. A imagem (mosaicada) também foi importada para o ArcView e georreferenciada com auxílio do “align tool”. Com a ferramenta apropriada foram gerados dois “buffers”: o primeiro de 1 km para cada lado da EPT, que tinha sido decidido de comum acordo entre DU, ECOTRÓPICA e FEMA, como área de trabalho e tem uma superfície de 29 463 ha; o segundo “buffer”, de 10 km, surgiu de uma necessidade estética para melhorar a apresentação, porque o “buffer” de 1 km é muito estreito e longo e, em função da escala do mapa, quase não se torna visível.

14

Figura 5. Mapa preliminar da vegetação ao longo da EPT (PCBAP).

15

Inicialmente foi realizada uma classificação não supervisionada das imagens com a ferramenta “categorize”. O resultado não foi satisfatório considerando o detalhe de mapeamento requerido, motivo pelo que se procedeu a realizar uma interpretação visual apoiada nos dados de observação no campo e fotografias. Utilizando-se as ferramentas de desenho do ArcView foram digitalizadas todas as unidades vegetacionais, bem como as alterações antrópicas, ex.: área urbana, garimpo, depósito de lixo, etc. com o respectivo registro na tabela do banco de dados (Figura 6).

Figura 6. Resultado da digitalização das unidades vegetacionais dentro do limite de 10

km para cada lado da EPT. Para facilitar a análise da composição das unidades vegetacionais foram utilizadas algumas fotografias justapostas, obtidas em seqüência, que ilustram a vegetação com maior amplitude (Figuras 7, 8, 9 e10).

16

Figura 7. Fotografia central de um grupo de cinco fotos.

Figura 8. Unidade vegetacional composta por 5 fotografias. Vista de um “capão”.

Figura 9. Unidade vegetacional composta por 7 fotografias num ângulo de 180°. A EPT é visível no estremo esquerdo, bem como, no estremo direito da composição. Exemplo

de um cerradão (savana florestada).

Figura 10. Exemplo de campo inundável (savana gramíneo-lenhosa) com “capões” ao fundo. Nesta composição de 7 fotografias pode-se observar o represamento das águas

causado pelo aterro da EPT.

17

Na “view” correspondente ao mapeamento da fitofisionomia da EPT foram colocadas, somente como ilustração, três fotografias daquelas que foram utilizadas na identificação das unidades vegetacionais. Elas foram relacionadas com as classes correspondentes no mapa de fitofisionomias para mostrar o tipo de vegetação predominante naquela região. Através da ferramenta “Hot Link”, as fotografias foram associadas a um símbolo imitando um disparo de flash de máquina fotográfica e correspondem àquela classe vegetacional onde está localizado o símbolo (Figura 11).

Figura 11. Exemplo de associação de fotografia a uma unidade vegetacional. Abaixo encontra-se uma relação numérica das fotografias utilizadas, com a descrição da paisagem correspondente. − Foto 01. Floresta estacional semidecidual aluvial (mata ciliar do Rio Cuiabá, Porto

Joffre). − Foto 02. Comunidade pioneira em primeiro plano e ao fundo Cambarazal (Fazenda

Rodeio da Campina Grande, antiga Fazenda Joffre).

18

− Foto 03. Comunidade pioneira (espinheiral) (Fazenda Rodeio da Campina Grande). − Foto 04. Vista parcial de savana gramíneo lenhosa (campo inundável), ao fundo

Cambarazal (Região dos Campos de Joffre). − Foto 06. Savana gramíneo lenhosa (campo inundável). (Região dos Campos de Joffre,

próximo da divisa com o Leirinho) − Foto 08. Savana gramíneo lenhosa (campo inundável); em primeiro plano a

Brachiaria subquadripara, espécie de origem africana que foi introduzida no Pantanal e é considerada uma invasora agressiva.

− Foto 09. Savana gramíneo lenhosa (campo inundável); margem da Transpantaneira com duas espécies invasoras Panicum maximum (campim-colonião) e Brachiaria subquadripara (capim-braquiaria).

− Foto 10. “Murundu”, pequeno monte de terra com vegetação mais alta, cercado pelo campo inundável.

− Foto 11. Vista da Estrada Parque Transpantaneira e ao fundo uma “cordilheira de mata”, extensa faixa de floresta.

− Foto 12. Cambarazal, formação arbórea com dominância de Cambará (Vochysia divergens).

− Foto 13. Vista parcial de um cambarazal, com espécies pioneiras de embaúba (Cecropia pachystachya).

− Foto 14. Cambarazal invadindo um campo inundável. − Foto 15. Cambarazal adulto, apresentando dois estratos, o arbóreo formado por

cambará (Vochysia divergens) e o arbustivo com outras espécies. − Foto 16. Vista geral de pastagem manejada, fazenda Campo Largo. − Foto 17. Pastagem manejada, ao fundo a Savana Florestada (Cerradão); fazenda

Campo Largo − Foto 18. Pastagem manejada. − Foto 19. Pastagem manejada. Fazenda Campo Largo. − Foto 20. Contato entre Floresta Estacional Semidecidual e Comunidades Pioneiras

(próximo ao Leirinho).

19

−

Foto 01 Foto 02

Foto 03

Foto 06

Foto 08

Foto 04

Foto 07

Foto 09

20

Foto 10 Foto 11

Foto 12

Foto 14

Foto 16

Foto 13

Foto 15

Foto 17

21

Foto 18 Foto 19

Foto 20

Limites e Propriedades A falta de demarcação das propriedades na região do Pantanal envolve questões relacionadas à redistribuição de terras entre herdeiros. Outro fator que contribuiu para a atual situação da questão fundiária, ocorreu após a abertura da Transpantaneira, onde os proprietários que ficaram sem acesso à estrada, acabaram acordando ou trocando lotes de terras para obter o acesso. O custo da demarcação, as dificuldades para a construção e manutenção das cercas, aliados a outros fatores, levou os proprietários a acordos verbais. Tais acordos foram possíveis pelo fato de que a maioria dos proprietários têm parentescos entre si. A região que apresentou os maiores problemas com demarcação de terras, foi a que compreende o trecho do km 85 (Fazenda Nova Berlim) até o km 107 (Pousada Recanto do Jaguar), onde cinco proprietários foram identificados até o momento. Para o levantamento de dados foi organizada uma viagem de campo que permitiu visitar todas as propriedades que tinham limites com a EPT. Este trabalho relacionou a existência de 46 propriedades que encostam na EPT, cujos nomes e proprietários foram cadastrados no banco de dados. A posição dos limites das propriedades junto da EPT foi registrada no GPSMAP 195, juntamente com todas as

22

pontes. Posteriormente estes dados foram descarregados do GPS para o CartaLinx, a projeção UTM foi convertida para coordenadas geográficas com Datum WGS84 e exportados para o ArcView 3.2.a, pelo processo anteriormente descrito. A utilização destes dados permitiu a elaboração do Mapa de Limites e Propriedades. RRRRESULTADOSESULTADOSESULTADOSESULTADOS

Metadados

Foram compilados 171 títulos de referências cartográficas que estão armazenados em um arquivo *.dbf. Este arquivo poderá ser aberto para consulta e edição dos dados através do MS Access 2000, bem como poderá ser introduzido no MetaLite para consultas, alterações e exportação de dados.

Fotoíndices Foram confeccionados 113 fotoíndices de aerofotos realizadas no período de 1964 a 1967, através do Projeto AF-63-32, AST-10 (Brasil–USA). Uma cópia digital de todos os índices foi encaminhada em CD_ROM à DU e os originais em papel encontram-se na Ecotrópica.

Impactos ambientais sobre a EPT Espécies Exóticas Espécies Exóticas Espécies Exóticas Espécies Exóticas A presença de espécies exóticas, refere-se à introdução de forrageiras para plantio de pastagens, que na região são basicamente duas espécies: a Brachiaria subquadripara e a B. humidicola. A primeira tem ampla dispersão sendo encontrada desde os ambientes mais alagados, sobre macrófitas aquáticas, até áreas mais secas; enquanto que a segunda teve sua distribuição restrita apenas às áreas em que foi introduzida. O controle da B. subquadripara é difícil, uma vez que apresenta crescimento rápido competindo com as espécies de gramíneas nativas e inclusive com as macrófitas aquáticas (aguapés). A espécie também é favorecida por proprietários que a plantam como forma de combater outras espécies invasoras, como o algodão-bravo (Ipomea carnea). QueimadasQueimadasQueimadasQueimadas As queimadas são outra ameaça às áreas da região, principalmente onde não há gado bovino, pois esses locais acumulam biomassa rapidamente e o fogo pode alastrar-se ali com rapidez dificultando o seu controle. Os locais onde foram observados vestígios de fogo estavam em áreas de savana arborizada e parque. Estes tipos de vegetação são utilizados como áreas de pastagens naturais, sendo que durante o período seco o capim

23

nativo torna-se duro e é recusado pelo gado. A queima provoca o brotamento das espécies de gramíneas proporcionando a reabilitação da área para o gado. PublicidadePublicidadePublicidadePublicidade Placas de publicidade foram colocadas sem nenhum controle dos órgãos competentes e ao longo da Estrada elas estão penduradas em árvores e até mesmo fincadas nas áreas de campo, impedindo a visão da paisagem. Estas placas variam em tamanho, e anunciam principalmente as pousadas e também empresas de telefonia. Algumas trazem informações sobre a quilometragem, sendo a única informação disponível para os usuários porque a Estrada não é sinalizada. A maioria das placas foi observada nos primeiros 50 km. ErosãoErosãoErosãoErosão Os processos erosivos sobre o aterro da Estrada são pontuais, sendo observados somente sulcos, não configurando problema grave. Estas alterações foram observadas principalmente nos desvios, construídos para contornar as pontes quebradas. AterrosAterrosAterrosAterros Os aterros alteram toda a dinâmica hídrica da paisagem dificultando o escoamento das águas. A EPT é a maior expressão desse impacto, mas também pode ser observado nas estradas que acessam as sedes das propriedades. Muitos aterros são construções antigas, o que não chega a ser uma preocupação, pois os processos de adaptação das espécies às mudanças do meio físico já ocorreram. A divisão das propriedades e a abertura de pousadas provocam uma nova demanda por estradas, o que poderá aumentar este tipo de impacto a médio e longo prazo. Drenos e diquesDrenos e diquesDrenos e diquesDrenos e diques A abertura de drenos e construção de diques também causam alterações na dinâmica hídrica. Este impacto ambiental foi observado na planície do Pantanal antes do início da EPT (no km 10), em área de uso agropecuário. O barramento das águas pode causar sérios prejuízos econômicos e ecológicos, pois áreas próximas aos diques permanecem mais tempo inundadas. DesmatamentosDesmatamentosDesmatamentosDesmatamentos Os desmatamentos, embora pouco observados, merecem uma maior atenção, pois estão ocorrendo nas áreas mais altas, recobertas com vegetação de floresta ou savana florestada. Estas áreas constituem abrigo para a fauna e para o gado bovino durante a época de cheia no Pantanal. As áreas desmatadas estão sendo destinadas à pastagem, com plantio de braquiária, o que reduz a biodiversidade.

24

AssoreamentoAssoreamentoAssoreamentoAssoreamento Embora não citado entre os impactos ambientais, o assoreamento é um tema muito discutido entre os proprietários da região. Muitos argumentam que os leitos de rios, corixos e vazantes estão assoreados e que deveriam ser drenados como forma de aumentar o escoamento das águas e reduzir o avanço das espécies invasoras (pombeiros, algodão-bravo, pimenteira, etc.). No entanto, o Pantanal é uma planície de sedimentação, onde os processos estão em pleno andamento. Ao longo da EPT não há grandes indícios de assoreamentos, estes são visíveis, principalmente, nas calhas dos grandes rios que nascem nos Planaltos e que cortam a Planície do Pantanal, como o Rio Cuiabá e o Rio Paraguai. Há, entretanto, que se considerar o barramento das águas provocado pelo aterro da EPT (Oliveira, 1998; Campos Filho, 2002) pois este sim provoca o processo de colmatação dos canais que drenam as águas. Grandes blocos de biomassa (batumes) ficam presos aos pilares das pontes e impedem a passagem da água e no período seco a biomassa morre, deposita-se nas baías, nas caixas de empréstimo e nos canais sob as pontes. As alterações no meio físico estimulam a colonização por novas espécies nos ambientes em processo de mudança. Intervenções antrópicas para restabelecer a dinâmica hídrica dos corixos e vazantes desencadearão novos processos ecológicos, mas não garantem o retorno das comunidades vegetais que existiram no passado. Restos de manutençãoRestos de manutençãoRestos de manutençãoRestos de manutenção Outro impacto, de caráter estético, é o dos restos de madeira, encontrados sob as pontes, que são constituídos por pedaços de tábuas e vigas que foram abandonados depois de trabalhos de manutenção. Em todas as pontes foi observado este impacto. Embora a madeira possa se decompor ao longo do tempo, os restos de madeira causam uma forte impressão de abandono e falta de consciência ecológica. A Figura 12 mostra a distribuição percentual dos vários tipos de impactos ambientais observados na EPT.

Placas15%

Espécie exotica25%

Queimada15%

Dreno10%

Aterro10%

Erosão10%

Desmatamento10%

Dique5%

Figura 12. Impactos ambientais observados na Estrada Parque Transpantaneira.

25

Mapeamento Fitofisionômico Foi produzido um mapa de vegetação, uso e ocupação da região da EPT, no município de Poconé, MT, Brasil (Figura 13). A nomenclatura das unidades vegetacionais obedeceu às normas do Manual Técnico de Vegetação Brasileira do Instituto Brasileiro de Geografia e Estadística (IBGE) (IBGE, 1992) e a paleta de cores foi especialmente montada para este estudo. A legenda mostra, além dos símbolos e a descrição de cada unidade, o número total de unidades de cada classe (Σ) com a correspondente área em hectares (ha) (Figuras 14 e 15).

26

Figura 13. Mapa de vegetação, ocupação e uso do solo.

27

Figura 14: Legenda dos tipos de vegetação

28

Figura 15: Legenda dos tipos de vegetação (continuação da Figura 14).

29

Observações da Flora da Estrada Parque Transpantaneira A flora caracterizada ao longo da EPT, pertence em sua grande parte à Savana, sendo os elementos florestais situados em encraves (cordilheiras e capões). As margens da EPT geralmente têm influência antrópica, sendo a flora caracterizada como ruderal, portanto a caracterização foi realizada sempre em unidades menos impactadas e distantes do efeito de borda da Estrada, com exceção para as comunidades pioneiras. No total foram observadas 76 espécies, pertencentes a 44 famílias, sendo as mais importantes Poaceae (6), Bignoniaceae (5), Fabaceae (4) e Mimosaceae (4) (Tabela 1).

Tabela 1. Espécies observadas nas unidades de mapeamento ao longo da Estrada Parque Transpantaneira, município de Poconé, MT. Fa: Floresta Estacional Semidecidual Aluvial; Fb: Floresta Estacional Semidecidual da s Terras Baixas; Cb: Floresta Estacional Decidual das Terras Baixas; Sd: Savana Florestada; Sa: Savana Arborizada; Sp: Savana Parque; Sg: Savana Gramineo-Lenhosa; Pa: Comunidades Pioneiras; Ag: Agropecuária.

Família/Família/Família/Família/Espécie Nome regional Fa Fb Cb Sd Sa Sp Sg Pa Ag

AmaryllidaceaeAmaryllidaceaeAmaryllidaceaeAmaryllidaceae Agave americana L. x AnacardiaceaeAnacardiaceaeAnacardiaceaeAnacardiaceae Astronium fraxinifolium Schott Gonçaleiro x x Myracroduon urundeuva (Enbgl.) Fr.All. Aroeira x AnnonaceaeAnnonaceaeAnnonaceaeAnnonaceae Anona dioica St. Hil. Arixicum x ApocynaceaeApocynaceaeApocynaceaeApocynaceae Aspidosperma australe M. Arg. Peroba x Aspidosperma cylindrocarpon M. Arg. Peroba Rhabidadenia pohlii M. Arg. x Thevetia bicornuta M. Arg. Lingua-de-vaca x x ArecaceaeArecaceaeArecaceaeArecaceae Acrocomia aculeata (Jacq.) Lodd. Bocaiuva x x Bactris sp. Tucum x Copernicia alba Morong Caranda x Scheelea pharelata (Mart) Burret Acuri x x x AsteraceaeAsteraceaeAsteraceaeAsteraceae Aspilia latissima Malme Mirassolzinho x x x BignoniaceaeBignoniaceaeBignoniaceaeBignoniaceae Jacaranda cuspidifolia Mart. Caroba x Tabebuia aurea (Manso) B. et H. Paratudo x Tabebuia heptaphylla (Vell.) Tol. Piúva x x x Tabebuia sp. Ipê amarelo x Thevertia bicornuta Muell. Arg. Leiterinho x BombacaceaeBombacaceaeBombacaceaeBombacaceae Pseudobombax longiflorum (Mart. et Zucc.)

Embiruçu x

30

Família/Família/Família/Família/Espécie Nome regional Fa Fb Cb Sd Sa Sp Sg Pa AgBombacaceae Bombacaceae Bombacaceae Bombacaceae (cont.) Pseudobombax marginatum (St. Hil.) Rob. Embiruçu-da-

mata x

BoraginaceaeBoraginaceaeBoraginaceaeBoraginaceae Cordia glabrata (Mart.) DC. Louro x BromeliaceaeBromeliaceaeBromeliaceaeBromeliaceae Bromelia balansae Mez Gravatá x CactaceaeCactaceaeCactaceaeCactaceae Cereus peruvianus Mill. Urumbeba x CaesaCaesaCaesaCaesalpinaceaelpinaceaelpinaceaelpinaceae Bauhinia forficata Link. Pé-de-boi

(branca) x

Bauhinia sp. Pata-de-vaca x Cassia grandis L. f. Canafístula x x CecropiaceaeCecropiaceaeCecropiaceaeCecropiaceae Cecropia pachystachya Trec. Embaúba x x ChrysobalanaceaeChrysobalanaceaeChrysobalanaceaeChrysobalanaceae Couepia uiti (Mart. et Zucc.) Bth. Pateiro x Licania minutiflora (Sag.) Fritsch Cedro-d'água x Licania parvifolia Huber Pimenteira x ClusiaceaeClusiaceaeClusiaceaeClusiaceae Rheedia brasiliensis (Mart.) Pl. et Tr. Cupari x CombretaceaeCombretaceaeCombretaceaeCombretaceae Buchenavia tomentosa Eichl. Tarumarana x Combretum laxum Jacq. Pombeiro-branco x ConvolvulaceaeConvolvulaceaeConvolvulaceaeConvolvulaceae x x Ipomoea carnea Jacq. Algodão-bravo CyperaceaeCyperaceaeCyperaceaeCyperaceae Cyperus giganteus Vahl Pirizeiro x Rhynchospora trispicata (Ness) Steud. Capim-navalha x DilleniaceaeDilleniaceaeDilleniaceaeDilleniaceae Curatella americana L. Lixeira x ErythroxylaceaeErythroxylaceaeErythroxylaceaeErythroxylaceae x Erythroxylum anguifugum Mart. Pimenteirinha x x Sapium obovatum Kl. Sarã-de-leite x EuphorEuphorEuphorEuphorbiaceaebiaceaebiaceaebiaceae Alchornea castaneifolia (Willd.) A. Juss. Sarã x FabaceaeFabaceaeFabaceaeFabaceae Albizia samam (Jacq.) F. v. M. Farinha-seca x Dipteryx alata Vog. Cumbaru x Machaerium aculeatum Raddi Espinheiro x Machaerium hirtum (Vell.) Stellf. Barreirinho x x SmilacaceaeSmilacaceaeSmilacaceaeSmilacaceae Smilax fluminensis Steud. Japecanga x MalphiguiaceaeMalphiguiaceaeMalphiguiaceaeMalphiguiaceae Byrsonima orbignyana A. Juss. Canjiqueira x MarantaceaeMarantaceaeMarantaceaeMarantaceae Thalia geniculata L. Caeté x x

31

FamFamFamFamília/ília/ília/ília/Espécie Nome regional Fa Fb Cb Sd Sa Sp Sg Pa AgMelastomataceaeMelastomataceaeMelastomataceaeMelastomataceae Mouriri guianensis Aubl. Roncador x MimosaceaeMimosaceaeMimosaceaeMimosaceae Anadenathera colubrina (Vell.) Brenan Angico x Inga vera ssp. affinis (DC) TD. Penn. Ingá x Mimosa pellita H. et B. Espinheiro x MonimiaceaeMonimiaceaeMonimiaceaeMonimiaceae Siparuna guianensis Aubl. Negramina x MoraceaeMoraceaeMoraceaeMoraceae Ficus gardneriana (Miq.) Miq. Figueira-mata-

pau x x

MyrtaceaeMyrtaceaeMyrtaceaeMyrtaceae Eugenia pitanga (Berg) Nied. Pitanga x Psidium kennedyanum Morong Araçazinho x NinphaeaceaeNinphaeaceaeNinphaeaceaeNinphaeaceae Ninphaeae prolifera Wiersema Lagartixa x OchnaceaeOchnaceaeOchnaceaeOchnaceae Ouratea sp. Curte-seco x OnagraceaeOnagraceaeOnagraceaeOnagraceae Ludwigia tomentosa (Cambess.) Hara Florzeiro x PoaceaPoaceaPoaceaPoacea Brachiaria humidicola (Rend.) Schweich Braquiária x Paspalum Hydrophilum Henr. Felpudo x Andropogon bicornis L. Capim-rabo-de-

burro x

Axonopus sp. Capim-mimoso x Brachiaria subquadripara (Trin.) Hitchc. Tanner-grass x x Paspalum repens Berg. Capim-fofo x x PolygonaceaePolygonaceaePolygonaceaePolygonaceae Coccoloba molis Casar Belém x Coccoloba ochreolata Wedd. Uveira-do-mato x Triplaris americana L. Novateiro x x PontederiacePontederiacePontederiacePontederiaceaeaeaeae Eichhornia azurea (Sw.) Kunth. Camalote x Eichornia crassipes (Mart.) Solms Camalotinho x PontederiaceaePontederiaceaePontederiaceaePontederiaceae Pontederia parviflora Alexander Guapé x RubiaceaeRubiaceaeRubiaceaeRubiaceae Alibertia edulis (L. l. Rich.) A. C. Rich. Marmelada-bola x Genipa americana L. Jenipapo x Fagara hassleriana Chod. Mamica-de-

porca x x

SapindaceaeSapindaceaeSapindaceaeSapindaceae Dilodendron bipinnatum Radlk. Mulher-pobre x SterculiaceaeSterculiaceaeSterculiaceaeSterculiaceae Helicteres guazumaefolia H.B.K. Rosquinha x TyphaceaeTyphaceaeTyphaceaeTyphaceae Typha angustifolia L. Tabôa x

32

Família/Família/Família/Família/Espécie Nome regional Fa Fb Cb Sd Sa Sp Sg Pa AgVitaceaeVitaceaeVitaceaeVitaceae Cissus erosa L. C. Rich. Cipó-de-arraia-

liso x x x x

VochysiaceaeVochysiaceaeVochysiaceaeVochysiaceae Vochysia divergens Pohl Cambará x x x ZingiberaceaeZingiberaceaeZingiberaceaeZingiberaceae Costus cf. arabicus L. Cana-do-brejo x

Observações da Fauna da Estrada Parque Transpantaneira Foram realizadas observações sobre a fauna ao longo da EPT e nas estradas vicinais que dão acesso às propriedades. Os locais de maior avistamento de animais foram na estrada de acesso à Fazenda Pouso Alegre (km 33) e na região dos Campos de Joffre. Os dados foram separados de acordo com os grupos, sendo que os jacarés são avistados ao longo de toda a Transpantaneira, principalmente próximo às pontes (Tabela 2).

Tabela 2. Lista de espécies de répteis observados na Estrada Parque Transpantaneira,

município de Poconé, MT.

Ordem Ordem Ordem Ordem / Subordem / Família Espécie Nome regional CrocodiliaCrocodiliaCrocodiliaCrocodilia Alligatoridae Caiman yacare Jacaré SquamataSquamataSquamataSquamata Serpentes Boidae Eunectes murinus Sucuri

As aves representam o grupo de maior facilidade para observação e são facilmente avistadas nos locais abertos como a Savana Gramíneo-lenhosa e Savana Parque. Ninhos de Tuiuiu também foram avistados (Tabela 3).

Tabela 3. Lista de espécies de aves observadas na Estrada Parque Transpantaneira, Município de Poconé, MT.

Ordem/Família Espécie Nome regional AnseriformesAnseriformesAnseriformesAnseriformes Anatidae Dendrocygna sp. Marreca Anhimidae Cairina moschata Pato-do-mato Chauna torquata Tachã Anhima cornuta Anhuma CaprimulgiformesCaprimulgiformesCaprimulgiformesCaprimulgiformes Caprimulgidae Nytidromus sp. Curiango CharadriiformesCharadriiformesCharadriiformesCharadriiformes Jacanidae Jacana jacana Cafezinho

33

Ordem/Família Espécie Nome regional Caprimulgiformes Caprimulgiformes Caprimulgiformes Caprimulgiformes (cont.) Ardeidae Egretta thula Garcinha-branca Ardeidae Tigrisoma lineatum Socó-boi Ciconiidae Jabiru mycteria Tuiuiu Ardeidae Casmerodius albus Garça-branca-grande FalconiformesFalconiformesFalconiformesFalconiformes Falconidae Herpetotheres cachinnans Acauã Caracara plancus Gavião caracará Accipitridae Rostrhamus sociabilis Gavião caramujeiro Busarellus nigricollis Gavião-belo GalliformesGalliformesGalliformesGalliformes Cracidae Ortalis canicollis Arancuã Gruiformes Aramidae Aramus guarauna Carão Eurypygidae Eurypyga helias Pavãozinho-do-pará OiciconiforOiciconiforOiciconiforOiciconiformesmesmesmes Ciconiidae Ciconia maguari Maguari Passeriformes Passeriformes Passeriformes Passeriformes Emberezidade Pardaria capitata Cardeal PsittaciformesPsittaciformesPsittaciformesPsittaciformes Psittacidae Anodorhynchus hyacinthinus Arara-azul-grande Psittacidae Aratinga aurea Periquitos RheiformesRheiformesRheiformesRheiformes Rheidae Rhea americana Ema TinamiformesTinamiformesTinamiformesTinamiformes Tinamidae Crypturellus undulatus Jaó TrogoniformesTrogoniformesTrogoniformesTrogoniformes Trogonidae Ceryle torquata Martim pescador

Os mamíferos mais comuns são as capivaras, sendo observado, também com facilidade, o cervo-do-pantanal. A Fazenda Pouso Alegre foi o local de maior avistamento de grupos de animais; no restante da EPT os animais geralmente foram avistados isolados ou em grupos pequenos. Informações sobre onças foram obtidas na região da floresta semidecidual das terras baixas, na Pousada do Jaguar, onde o proprietário informou sobre o avistamento de 4 animais, e na Fazenda Esperança também foi relatado o avistamento. As duas propriedades estão situadas na mesma região (Tabela 4).

Tabela 4. Lista de espécies de mamíferos observados na Estrada Parque Transpantaneira, município de Poconé, MT.

Ordem/família Espécie Nome regional ArtiodactylaArtiodactylaArtiodactylaArtiodactyla Tayassuidae Tayassu tajacu * Caititu Cervidae Blastocerus dichotomus Cervo-do-pantanal Ozotoceros bezoarticus Veado campeiro

34

Ordem/família Espécie Nome regional CarnívoraCarnívoraCarnívoraCarnívora Felidae Panthera onca ** Onça-pintada PrimatesPrimatesPrimatesPrimates Callithricidae Alouatta caraya Bugio RodentiaRodentiaRodentiaRodentia Hydrochaeridae Hydrochaeris hydrocaeris Capivara * Criado em sede de pousada. ** Não houve observação direta; as informações sobre avistamentos são provenientes de proprietários, peões e usuários da EPT.

Limites e Propriedades Foi produzido um mapa apresentando a localização das pontes, estradas secundárias da EPT, os limites entre as propriedades que chegam à beira da Estrada e o nome dos respectivos proprietários (Figura 15). RECOMENDAÇÕES

Mapeamento Fitofisionômico

Com a finalidade de melhorar a ilustração do Mapa de Vegetação, Uso e Ocupação da EPT, sugere-se fotografar todas as diferentes unidades vegetacionais ao longo da EPT existentes no mapa, mostrando os diferentes tipos de vegetação e/ou qualquer outra alteração da paisagem. A seguir estas fotos poderão ser ligadas às classes de vegetação constantes da legenda, através da ferramenta “Hot Link” do ArcView 3.2.a.

Ameaças É indispensável a permanente vigilância sobre as atividades de impacto ambiental envolvendo a região da EPT e entorno. Degradação das pastagens naturais Muitas áreas de pastagem natural no Pantanal foram degradadas e estão sendo convertidas em pastagens cultivadas, sendo que a principal espécie utilizada para o plantio é a braquiária. Aumentar a produtividade das pastagens é uma das iniciativas tomadas pelos produtores para fazer frente à competitividade das pastagens cultivadas nas áreas fora do Pantanal. A viabilidade econômica das fazendas de gado no Pantanal deve estar contida no plano de manejo da EPT, como forma de garantir a conservação das paisagens no Pantanal.

35

Figura 15. Mapa das divisas das propiedades que encostam na EPT

36

Embora seja um tema polêmico, a forma de ocupação que os antigos fazendeiros realizaram no passado e que propiciou a conservação da paisagem pantaneira, já não é suficiente para gerar renda para manter as propriedades (Oliveira, 1998; Campos Filho, 2002). Embora muitas propriedades estejam incrementando a renda através da abertura de pousadas; uma maioria carece de estrutura e informação sobre o turismo ecológico, confundindo-o com o turismo rural. Como nem todas as propriedades terão vocação e capital para investir no turismo, será necessário discutir formas de incremento na renda dos produtores. Técnicas de manejo e recuperação de pastagens naturais e certificação de carne orgânica, seriam as principais alternativas no momento. Abertura de novas pousadas e construções de estradas e sedes de fazenda As medidas supracitadas são de fundamental importância para evitar o abandono das atividades agropecuárias e o parcelamento do solo. Cada pedaço de terra desmembrado, significa pelo menos uma nova sede e uma nova estrada. As estradas são altamente impactantes no Pantanal, pois quando construídas sobre o campo são aterradas, o que altera o ciclo hidrológico e, quando construídas sobre cordilheiras (Florestas e Cerradão), causam desmatamento e fragmentação de hábitats. Abertura de trilhas ecológicas ao longo da EPT Na faixa de 300m que compreende a EPT, deve-se ter cuidado ao planejar trilhas ecológicas, pois nesta área predominam as comunidades pioneiras de macrófitas aquáticas e pombeiros. Este tipo de comunidade se estabeleceu, principalmente após a abertura da Transpantaneira, colonizando as caixas de empréstimo. Novas intervenções poderão desestruturar a estrutura das comunidades. Recuperação das caixas de empréstimo As caixas de empréstimo ao longo da EPT, são um dos impactos ambientais causados ao Pantanal e ao mesmo tempo um dos atrativos turísticos, pois armazenam água para o período seco, atraindo os animais e criando nichos para as plantas aquáticas, que adaptaram-se rapidamente a este ambiente (Campos Filho). Este novo nicho, criado após a abertura da estrada, encontra-se em pleno processo de sucessão ecológica. As comunidades aquáticas estão sendo substituídas por comunidades arbustivas e arbóreas (pombeiros) à medida em que aumenta o processo de colmatação e dos períodos de seca. As mudanças nos recursos para a fauna, aparentemente tem causado, segundo relatos da população local, uma redução no número de animais. A recuperação destas áreas deve ser amplamente discutida e questionada, devendo ter como referência a conservação e não a restauração dos ambientes já impactados e em pleno processo de sucessão ecológica. Assim podem-se selecionar áreas, com relevante interesse cênico, como as áreas de Savana Gramíneo-lenhosa (campo), para que sirvam como áreas piloto onde se possam ser implementadas práticas de manejo e monitoramento da biodiversidade.

37

Questão fundiária Sobre a questão fundiária há de se pensar em um amplo acordo com os proprietários para que se efetue a regularização das propriedades em conjunto, de forma a reduzir os custos. Tal demarcação feita por GPS e plotada em imagem de satélite, solucionará a questão, facilitando futuros parcelamentos e controle dos nomes dos proprietários por parte do órgão gestor da UC EPT. A regularização fundiária é de vital importância para que não ocorra o parcelamento do solo paralelo a UC EPT. Embora a escrituração de terras no Pantanal deva ser superior a 100 ha, existem casos de propriedades com área menor do que o permitido. Esse tipo de “micro-parcelamento” deve ser evitado para que se possa ter um mínimo de garantia de conservação das paisagens da UC EPT. Outra medida preventiva deve ser no sentido de promover a conscientização dos grandes proprietários que confrontam com a EPT, destacando a importância da conservação ambiental da paisagem pantaneira para a manutenção das atividades econômicas relacionadas ao ecoturismo. Esta medida de caráter preventivo, poderá evitar o parcelamento das propriedades paralelas à EPT. Para as pequenas propriedades já estabelecidas podem ser proporcionadas alternativas econômicas de manejo sustentável de modo a integrá-las às atividades relacionadas ao ecoturismo.

Resíduos Sobre o destino dos resíduos dos estabelecimentos ao longo da EPT, não há problemas graves com tratamento que recebem, até porque, são poucos estabelecimentos e a maioria tem como clientela turistas, o que acaba reforçando a necessidade da manutenção de uma postura ecologicamente correta e preocupação com o destino correto dos resíduos. Devem ser realizadas campanhas de orientação dos proprietários de lanchonetes e pousadas para o manuseio correto do lixo; os proprietários devem, também, ser orientados no sentido de permitir a introdução de técnicas e tecnologia acessíveis ao nível dos empreendimentos, o que poderá garantir melhores resultados.

Atividades impactantes Instalação de Placas de publicidade em locais indevidos, queimadas não autorizadas, desmatamentos, construções de aterros e abertura de drenos, refletem a fraca atuação dos órgãos públicos estaduais responsáveis pela fiscalização. A ausência do poder público é uma das reclamações dos produtores e donos de pousadas. Tais práticas devem ser combatidas com esclarecimentos à população da região sobre a importância da UC EPT. Esta atividade deve ser realizada através de cursos de educação ambiental envolvendo os sindicatos de produtores rurais e entidades de classe do setor, como o Núcleo de Criadores de Cavalo Pantaneiro. É importante ressaltar que no decorrer do

38

trabalho a maior parte dos produtores entrevistados desconheciam a finalidade da Unidade de Conservação.

Fiscalização Reforço da atividade de fiscalização, no sentido de estruturação dos meios de fiscalização, também serão necessários. Ressaltando que a Fundação Ecotrópica já promoveu um curso de treinamento em fiscalização e educação ambiental para a abordagem de turistas para a Polícia Militar Ambiental e técnicos da FEMA. Sobre queimadas, o IBAMA já treinou uma brigada contra incêndio na região.

Espécies exóticas Sobre as espécies exóticas de braquiária, serão necessários estudos mais aprofundados sobre Dinâmica de População e monitoramento para a espécie Brachiaria subquadrípara. Estudos sobre a possibilidade de plantio de espécies forrageiras nativas podem ser encontrados na EMBRAPA Pantanal e, finalmente, estimular os produtores para o cultivo das espécie nativas e evitar a espécie exótica.

Erosão Os processos erosivos sobre o aterro da EPT são pouco freqüentes e, aparentemente, não apresentam maiores problemas com a preservação ambiental. Os trechos com maior incidência desses processos, que são as cabeceiras de pontes, sofrerão intervenções que estão previstas no projeto BID-Pantanal. Informações técnicas adicionais sobre os processos erosivos foram tratadas pelo EIA/RIMA da Rodovia MT – 060, Estrada Parque Transpantaneira. GGGGLOSSÁRIOLOSSÁRIOLOSSÁRIOLOSSÁRIO − Assoreamento: Diminuição da profundidade de cursos de água por deposição de

sedimentos − Batume: Ilha flutuante composta por vários tipos de vegetação. − BID: Banco Interamericano de Desenvolvimento. − Caixa de empréstimo: Cavidade retangular, às margens da EPT, resultante da

retirada de terra utilizada para formar o aterro da Estrada. − Capão: Vegetação florestal situada em relevo mais alto e não inundável. − Cênico: Referente à cena, paisagem. − Colmatação: É o aumento de material arrastado e depositado em determinado lugar

ao ponto de dificultar, ou mesmo, impedir a passagem da água. − DU – Ducks Unlimited, Inc, www.ducks.org − ECOTRÓPICA: Fundação de Apoio à Vida nos Trópicos, www.ecotropica.org.br

39

− EIA/RIMA: Estudo e Relatório de Impacto Ambiental. − EPT: Estrada Parque Transpantaneira. − FEMA: Fundação Estadual do Meio Ambiente, www.fema.mt.gov.br − IBAMA: Instituto Brasileiro do Meio Ambiente e dos Recursos Renováveis da

Amazônia, www.ibama.gov.br − Inversor de corrente: Dispositivo eletrônico que transforma a corrente elétrica

contínua (DC 12V) em corrente alternada (AC 120 V). − Murundu: Pequeno monte de terra com vegetação mais alta, cercado pelo campo

inundável. − Ruderal: Vegetação que cresce, preferencialmente, em torno das habitações

humanas e estradas. − TCBR: Tecnologia e Consultoria Brasileira S. A., Cuiabá, MT. − TRANSPANTANEIRA: Estrada Estadual de Mato Grosso, MT 060, que liga a Rodovia

Federal BR 070 a Porto Joffre, município de Poconé, MT.

RRRREFERÊNCIAS BIBLIOGRÁEFERÊNCIAS BIBLIOGRÁEFERÊNCIAS BIBLIOGRÁEFERÊNCIAS BIBLIOGRÁFIFIFIFICASCASCASCAS Campos, Filho L.V.S. 2002. Tradição e ruptura. Cultura e ambientes pantaneiros. Entrelinhas, Cuiabá, MT. 184p. Oliveira, B.A.C. 1998. Valor recreativo da Rodovia Transpantaneira: uso turístico e conservação no Pantanal Mato-Grossense. Dissertação, Universidade Federal de Mato Grosso, Instituto de Biociências, Programa de Pós-graduação em Ecologia e Conservação da Biodiversidade. Cuiabá, MT. 102p. Instituto Brasileiro de Geografia e Estatística (IBGE). 1992. Manual técnico da vegetação brasileira. Rio de Janeiro: IBGE. Programa Nacional de Meio Ambiente (PNMA), ed. 1997. Plano de Conservação da Bacia do Alto Paraguai – PCBAP/Projeto Pantanal, Programa Nacional do Meio Ambiente. Vol. 3. Brasília: PNMA.

40

2222BBBB.... LLLLAND AND AND AND CCCCOVER OVER OVER OVER CCCCLASSIFICATION FOR THLASSIFICATION FOR THLASSIFICATION FOR THLASSIFICATION FOR THE E E E UUUUPPER PPER PPER PPER PPPPAAAARAGUAY RAGUAY RAGUAY RAGUAY RRRRIVER IVER IVER IVER BBBBASIN ASIN ASIN ASIN PPPPILOT ILOT ILOT ILOT PPPPROJECT ROJECT ROJECT ROJECT AAAAREAREAREAREA MARIO CARDOZO11, W. THEODORE MEALOR, JR.12 AND DAWN BROWNE13

AAAABSTRACTBSTRACTBSTRACTBSTRACT The Upper Paraguay River Basin (UPRB) is a large watershed in Bolivia, Brazil and Paraguay that contains the Pantanal wetland complex. Both the basin and the Pantanal have vegetation types characteristic of several ecoregions, including Chaco, Cerrado, and Chiquitano Dry Forest. This study developed a hierarchical land cover classification system based on those found in the literature and modified for local conditions. One- and two-date variants of an unsupervised classification method using Landsat data were employed to test the applicability of the proposed system. The results show that the applied techniques are useful in discriminating land cover classes based on vegetation physiognomy (forest, savanna, and grassland) and flooding seasonality. Accuracy assessment proved the two-date approach more accurate for land cover classification indicating the importance of explicitly incorporating seasonality when mapping dynamic landscapes such as the Pantanal. IIIINTRODUCTIONNTRODUCTIONNTRODUCTIONNTRODUCTION This study addresses a section of one of the largest interior freshwater wetlands in the world, the Pantanal, and the drainage basin that encloses it, the Upper Paraguay River Basin (UPRB), located in southern South America (Figure 1). The purpose of this study is to develop a prototype land cover classification system of the UPRB that can be used in the classification of remotely sensed data. Because the existing land cover classification systems for the UPRB have been developed in different countries, utilizing different methodologies based on different criteria, and at different scales and levels of complexity, there is a need for their standardization when the various politically and naturally diverse territories of the region are regarded as a unit. Although global and regional vegetation classification systems that can be applied to the UPRB exist (Zeilhofer and Schessl, 1999; Paranhos, 2000), studies that integrate locally developed classification systems are lacking. The mapping of land cover is generally performed with the aid of remote sensing technology, which constitutes a powerful tool for the identification and quantification of land cover types in wetlands (Lyon, 2001). Remotely sensed data can provide a multitemporal description of land cover of large and/or inaccessible areas that vary seasonally and throughout the years. Experiments have demonstrated that seasonal 11 The University of Texas at Austin, TX, U.S.A; mcardozo@mail.utexas.edu 12 The University of Memphis, TN, U.S.A; wmealor@memphis.edu 13 Ducks Unlimited, Inc., Memphis, TN, U.S.A; dbrowne@ducks.org

41

variability must be considered for the proper classification of certain types of land cover (Lunetta and Balogh, 1999).

Figure 1. Location of UPRB, the Pantanal, and Pilot Area.

In the Pantanal area, the characterization of landscape has been performed mainly through mapping techniques rather than the use of multispectral classification (MME, 1982a; PNMA, 1997). Although Landsat images are mentioned as source data for many studies and maps of the area, the classification of land cover in these studies usually does not involve spectral pattern recognition and is generally performed through manual digitization guided by visual interpretation and supported by ground reference data (MME, 1982a; PNMA, 1997). With this study, a pioneering effort was made to develop and test a land cover classification system for the UPRB, for use with multispectral classification of remotely sensed data. The logistical impossibility of testing the applicability of the scheme in the entire UPRB led to the selection of a specific study site (pilot area) to evaluate single-

42

date versus two-date approaches of a remote sensing technique for land cover classification. This study provides insights into the efficiency of Landsat image classification for the spectral differentiation of land cover at a broad scale in the pilot area, having limited ancillary data and fieldwork to assist the process. It was designed to evaluate the importance of utilizing multitemporal data for the proper characterization of the land cover types of the UPRB where seasonal flooding has a profound effect on vegetation and land use patterns. SSSSTUDY TUDY TUDY TUDY AAAAREAREAREAREA The results of this study are presented at two different spatial extents. The entire UPRB is addressed in the development of a land cover classification system for the region, while a pilot area in the UPRB is the object of testing this system and automated image classification procedures for use of Landsat data (Figure 1). The UPRB, located in Bolivia, Brazil, and Paraguay, is at the junction of three South American ecoregions, Cerrado, Chaco, and Chiquitano Dry Forest, which present different vegetation types in wetland and upland environments. There is a lack of agreement on the spatial extent and boundaries of the UPRB. Some authorities believe that the UPRB extends westward into the Paraguayan and Bolivian Dry Chaco (Browne et al., 2003; TNC, 2002). Liu (2003) states that the addition of the Dry Chaco sub-basin would represent an increase of 50 percent in the total area of the UPRB. The contribution of the Dry Chaco sub-basin in the hydrological balance of the UPRB is uncertain (Liu, 2003). The Pantanal’s areal extent represents close to 28 percent of the UPRB and occupies approximately 137,000 km2 (Liu, 2003) of areas below 200 m of altitude in Bolivia, Brazil and Paraguay (EMBRAPA, 2003). The Pantanal is characterized as an extensive wetland having vegetation cover conditioned to standing water, and seasonal and periodic flooding. This dynamic landscape is diverse temporally and spatially and is regionally and internationally valuable because of its large extent, scenic beauty, relative pristine state, and high biological richness (Swarts, 2000). The pilot area consists of a trinational section of the central-southern Pantanal and UPRB, included in the Landsat scene corresponding to the World Reference System 2 (WRS-2) path 227 row 74, and encompassing 30,674 km2. The pilot area is in the Chaco-influenced Pantanal portion of the UPRB.

43

MMMMETHODSETHODSETHODSETHODS

Land Cover Classification SystemLand Cover Classification SystemLand Cover Classification SystemLand Cover Classification System Because different sources classify the vegetation of the Pantanal and the UPRB differently, there is a need to establish a standard system that is comprehensive for the region. The proposed land cover classification system is based on the integration of land uses and vegetation types described in the literature for the UPRB, but it also takes into consideration factors that facilitate the classification of remotely sensed data in automated procedures. In the UPRB region, the classification system with the most detailed hierarchical structure is the official Brazilian vegetation classification system (IBGE, 1992); as a result, this system is used as the basis for the land cover classification scheme developed in this study. Other classification systems and vegetation studies were reviewed and integrated to complete the scheme and conform it to local conditions. Most of these studies have in common the use of physiognomy (forest, savanna, and grassland) as one of the basic parameters of classification (Table 1). Another parameter commonly used is general appearance. In forests general appearance refers to deciduousness (phenology) or openness and in savannas to density and spatial arrangement of arboreal stratum.

Table 1. Major vegetation studies reviewed

Authors of studies Authors of studies Authors of studies Authors of studies related to vegetation related to vegetation related to vegetation related to vegetation

classificationclassificationclassificationclassificationSpatial extentSpatial extentSpatial extentSpatial extent DescriptionDescriptionDescriptionDescription

Parameter(s) of classif ication (in Parameter(s) of classif ication (in Parameter(s) of classif ication (in Parameter(s) of classif ication (in hierarchical order when hierarchical order when hierarchical order when hierarchical order when

applicable)applicable)applicable)applicable)

IBGE, 1992 Brazil Technical handbook of vegetation

Natural/anthropogenic, vegetation physiognomy, climate, physiology,

appearance, elevation

PNMA, 1997 Brazilian UPRBVegetation study and maps for conservation

plan

Natural/anthropogenic, vegetation physiognomy, climate, physiology,

appearence, elevation

Machado and Ribeiro, 1994 Brazilian Cerrado Vegetation study Vegetation physiognomy

UNA and GTZ, 1991 Paraguayan Western Region

Vegetation and land use study and map

Natural/anthropogenic, vegetation physiognomy, appearance, flora

MAG and BGR, 1999 Paraguayan Western Region

Vegetation study and map

Vegetation physiognomy, physiology, flora

Mereles, 2000 Paraguayan Pantanal Ecological study for conservation plan Flooding seasonality, environment

WWF, 2002 Bolivian Pantanal Ecological study for conservation plan

Vegetation physiognomy, appearance, physiology, flora

This study UPRB Land cover classification system

Natural/anthropogenic, vegetation physiognomy, appearance, flora, flooding seasonality (as attribute)

44

Other criteria employed to classify vegetation in the existing systems include: climate, which indicates subjection to dry seasons; physiology, which refers to availability/use of water (xerophytic, mesophytic, hydrophytic); elevation (montane, submontane, lowland, and alluvial); flooding seasonality; and flora, which alludes to predominance of vegetation species. Only parameters that could be standardized for the majority of the vegetation classes found in the literature were used as classification criteria in the system developed in this study.

Image Data and PreprocessingImage Data and PreprocessingImage Data and PreprocessingImage Data and Preprocessing Procedures Procedures Procedures Procedures Landsat 5 TM and Landsat 7 ETM+ satellite data were used in this study (). Analysis of Landsat data was performed with the software Erdas Imagine 8.5. The dates of imagery were selected based on precipitation and river gauge data of stations located in the pilot area or its proximity (Figures 2, 3, 4, and 5; Table 2). Three dates of imagery were selected to depict different seasons and flood extents. The 06/09/1997 image corresponds to an extraordinary flood event that occurred during the high-flood season of 1997. The other two dates correspond to comparable relatively typical water years: one image is from the low-flood season (11/14/1999) and the other from the high-flood season (07/30/1998).

Figure 2. Hydrological stations in or close to Pilot Area.

45

River Gauge - Ladário

1

2

3

4

5

6

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Month

Riv

er G

auge

(m)

199719981999Average (1984-2000)

Year

Date of selected images

Figure 3. River gauge at Ladário

Precipitation - Corumbá

0

50

100

150

200

250

300

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Month

Pre

cipi

tatio

n (m

m)

199719981999Average (1984-2000)

Year

Date of selected images

Figure 4. Precipitation at Corumbá

46

Precipitation - Bahía Negra

0

50

100

150

200

250

300

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Month

Prec

ipita

tion

(mm

)

1997

1998

1999

Average(1986-2000)

Year

Date of selected images

Figure 5. Precipitation at Bahía Negra

Table 2. Selected images

Bahía Bahía Bahía Bahía NegraNegraNegraNegra CorumbáCorumbáCorumbáCorumbá

6/9/1997 TM 97.7 87.9 5.53 cheia (high flood)

extraordinary

7/30/1998 TM 12 0.1 4.43 cheia (high flood)

common

11/14/1999 ETM+ 102 82.1 1.19 seca (low flood)

common

Flood Flood Flood Flood season season season season

in in in in PantanalPantanalPantanalPantanal

Year flood Year flood Year flood Year flood eventeventeventevent

Landsat Landsat Landsat Landsat image dateimage dateimage dateimage date SensorSensorSensorSensor

Total monthly Total monthly Total monthly Total monthly precipitation (mm)precipitation (mm)precipitation (mm)precipitation (mm)

Monthly Monthly Monthly Monthly average river average river average river average river

gauge at gauge at gauge at gauge at Ladário (m)Ladário (m)Ladário (m)Ladário (m)

The three selected images were co-registered and georeferenced by Image Links, Inc., a value added vendor, to Universal Transverse Mercator (UTM) World Geodetic System (WGS) 1984 Zone 21, using reference points from a one-kilometer Digital Elevation Model (DEM) and available topographic maps, obtaining a Root Mean Square Error (RMSE) of less than one pixel. A radiometric correction technique, histogram matching, was applied to the 07/30/1998 image in order to compensate for its different pixel value range in comparison to the 11/14/1999 image. The 07/30/1998 image was resampled to match

47

the pixel value range of the 11/14/1999 image because the latter originated from a sensor with improved signal to noise ratio (SNR), Landsat 7 ETM+. Principal Component Analysis (PCA) was applied to minimize redundancies in the dataset and prepare it for optimal performance in the unsupervised classification. Studies show that preprocessing the data with PCA may improve the separability of land cover types in the classification (Li and Yeh 1998; Lunetta 1998; Yuan, Elvidge, and Lunetta 1998). Additionally, PCA was utilized in this study in order to evaluate an automated approach generalizable and more practical for users with varied technological resources in the countries that share the Pantanal and UPRB. The selected components amounted to more than 96 percent of the variance of the original dataset (Tables 3 and 4).

Table 3. Eigenvalues of One-date PCA

ComponentComponentComponentComponent EigenvalueEigenvalueEigenvalueEigenvalue % Variance% Variance% Variance% Variance % % % % AccumulativeAccumulativeAccumulativeAccumulative

1 1360.28 65.48 65.482 553.30 26.64 92.123 122.22 5.88 98.00

4* 41.46 2.00 100.00* discarded component

Table 4. Eigenvalues of Two-date PCA

ComponentComponentComponentComponent EigenvalueEigenvalueEigenvalueEigenvalue % Variance% Variance% Variance% Variance % % % % AccumulativeAccumulativeAccumulativeAccumulative

1 1388.025 58.89 58.892 575.338 24.41 83.313 192.500 8.17 91.474 110.859 4.70 96.18

5* 42.183 1.79 97.97 6* 37.629 1.60 99.56 7* 7.055 0.30 99.86 8* 3.224 0.14 100.00

*discarded components

Image Classification ProcedureImage Classification ProcedureImage Classification ProcedureImage Classification Procedure Because of the lack of extensive field-based data and the remoteness of the pilot area, an unsupervised method, iterative self-organizing data analysis (ISODATA), was selected

48

for the classification of land cover. Two approaches of this method were compared. In the first, the first three components derived from bands 3, 4, 5, and 7 of the 07/30/1998 image (Table 3) were classified into 120 classes using the ISODATA method. This number of classes was selected to maximize the spatial separability of classes.

In the second approach, the first four components derived from bands 3, 4, 5, and 7 of each of the 07/30/1998 and 11/14/1999 stacked images were used as input data for the unsupervised classification (Table 4). The same number of classes, 120, was utilized for the classification process. Because phenological differences in the vegetation were considered factors affecting seasonal variation in land cover, a second image corresponding to a contrasting season was included to account for intra-annual variation in vegetation. To finalize the imagery classification process of both approaches, the 120 classes were recoded to pixel values of one to eight to reassign the spectral classes to common land cover categories using the classes of a land cover classification scheme proposed in this study for the UPRB. The reassignment of the classes was performed through visual interpretation of the original images with the help of vegetation maps.

Open Water Delineation and Classification FinalizationOpen Water Delineation and Classification FinalizationOpen Water Delineation and Classification FinalizationOpen Water Delineation and Classification Finalization Single band density slicing was used to classify water pixels in the three images selected. This procedure can provide results that are comparable in accuracy to more elaborate and time-consuming techniques (Frazier and Page, 2000). Band 5 (mid infrared) was selected for use because it allows better water pixel separability than other infrared and visible bands. To add the multitemporal data of flood extent to the classified images, a new image was created encompassing the water-pixel data from the three dates. The pixels of the three flood-extent images were recoded to a value of 100 and then the images were overlaid to develop a composite of flood extent. The overlay process added the values of pixels that coincided spatially so the resulting image presented pixel values of 100, 200 and 300. Pixels with values of 300 were considered permanent water, while pixels with values of 100 and 200 were considered floodable and seasonally flooded areas, respectively. The multitemporal flood extent image was then merged with both classified images (one- and two-date approaches). The pixels of the two new images were recoded to classes conforming to flood seasonality with the exception of the forest classes, which were not reclassified because the remotely sensed data employed do not permit accurate delineation of flooded forests. A majority statistical filter with a moving

49

window of 5x5 pixels was used to enhance the continuity of dominant land cover classes.

Field Data for Evaluation of Classification ApproachesField Data for Evaluation of Classification ApproachesField Data for Evaluation of Classification ApproachesField Data for Evaluation of Classification Approaches Overall, user’s and producer’s accuracies, and the KHAT statistic were calculated to evaluate and compare one versus two-date approaches of classification. A total of 199 oblique photographs (snapshots) taken in overflights in Bolivia, Brazil, and Paraguay were used for classification accuracy assessment (Figure 6). Global Positioning System (GPS) coordinates were assigned to each photograph during the overflights.

Figure 6. Reference data for accuracy assessment.

Only photographs showing a predominant or single land cover were used. Because the reference photographs correspond to dates different than those of the Landsat images used in the classification procedures (December 2000 and March 2002), only the

50

physiognomy of vegetation was considered in the identification of land cover types in the reference photographs. LLLLANANANAND D D D CCCCOVER OVER OVER OVER CCCCLASSIFICATION LASSIFICATION LASSIFICATION LASSIFICATION SSSSYSTEM YSTEM YSTEM YSTEM RRRRESULTS ESULTS ESULTS ESULTS (UPRB)(UPRB)(UPRB)(UPRB) The proposed integrative hierarchical classification system for the UPRB (Table 5) combines five schemes and several vegetation studies identified in the literature (Table 6). Because criteria and scale of application differ among the schemes, the levels of the proposed classification system are based upon physiognomy, phenology, and flora. Criteria such as elevation (submontane, alluvial) and physiological factors related to use/availability of water (mesoxerophytic, xerophytic) are not consistently used to create fourth and fifth levels because sufficient data are not available to standardize the various forests types described in the literature. Still, tentative fourth and fifth classification levels are included in Table 5 to incorporate the most significant vegetation types found in the literature. The first level of the proposed system is the most general and divides land cover into natural and anthropogenic classes. The natural classes refer to natural communities, while anthropogenic classes represent artificially impounded open water (reservoirs, canals) and lands altered from a natural state by human occupancy (urban land, agricultural land, reforestation, bare land). Natural classes may include areas occupied and used by humans, such as cattle ranching in savannas, only if these areas maintain the basic natural structure. The placement of natural/anthropogenic classes in the first level is based on the Brazilian vegetation classification system (IBGE, 1992). In the natural category, the second level subdivides vegetation into three physiognomies based on the study by Ribeiro and Machado (1998): forests (more than 50 percent tree coverage), savannas (less than 50 percent tree coverage), and grasslands (predominantly herbaceous). Other classes in the natural category include open water (streams and lakes) and bare land (floodable areas in succession). The third level characterizes forests in regard to their appearance/phenology in semideciduous, deciduous, and evergreen categories. Savannas and grasslands are floristically classified in this level; they correspond to either Chaco or Cerrado ecoregions. Attributes employed for the characterization of wetland/upland habitats in the region relate frequency of flooding to vegetation types. The “permanently flooded lands” are palustrine areas normally characterized by aquatic and/or hydrophytic vegetation (Anderson et al., 1976), but they can also include non-vegetated inundated areas. Permanently flooded lands include swamplands, lagoons, and plains and depressions that are always inundated. The “seasonally flooded lands” are those conditioned by recurring inundation, and are considered part of the seasonal wetland complex. “Floodable lands” are potentially floodable natural lands, which are disturbed by

51

inundation irregularly or at lower frequencies. Finally, the “non-floodable lands” do not experience periodical or sporadic flooding.

Table 5. Proposed UPRB classification system

1. Natural land 1.1. Open water 1.2. Forest

1.2.1. Evergreen forest 1.2.1.1. Riparian evergreen forest b 1.2.1.2. Xerophytic evergreen forest – cerradão (Cerrado) d 1.2.1.3. Submontane open rainforest d

1.2.2. Semideciduous forest 1.2.2.1. Alluvial semideciduous forest c 1.2.2.2. Lowland semideciduous forest d 1.2.2.3. Submontane semideciduous forest (Chiquitano Dry Forest) d 1.2.2.4. Xerophytic semideciduous forest (Chaco) b, c, d 1.2.2.5. Mesoxerophytic semideciduous forest (Chaco) b, c

1.2.3. Deciduous forest 1.2.3.1. Lowland deciduous forest d 1.2.3.2. Submontane deciduous forest d 1.2.3.3. Deciduous forest (Chiquitano Dry Forest) c, d

1.3. Savanna 1.3.1. Cerrado savanna

1.3.1.1. Treed Cerrado savanna c, d 1.3.1.2. Parkland Cerrado savanna b, c, d

1.3.2. Chaco steppe savanna 1.3.2.1. Treed Chaco steppe savanna c, d 1.3.2.2. Parkland Chaco steppe savanna b, c, d

1.4. Grassland 1.4.1. Cerrado grassland a, b, c, d 1.4.2. Chaco grassland a, b, c, d

1.5. Bare land b, c, d 2. Anthropogenic Land

2.1. Open water 2.2. Urban land/Built up area 2.3. Agricultural land 2.4. Reforestation 2.5. Bare land

Attributes of flooding a: permanently flooded land b: seasonally flooded land c: floodable land d: non floodable land

52

Table 6. Proposed and reviewed vegetation classes

Vegetation Vegetation Vegetation Vegetation formformformform Proposed classesProposed classesProposed classesProposed classes Reviewed classes included in proposed Reviewed classes included in proposed Reviewed classes included in proposed Reviewed classes included in proposed

classesclassesclassesclasses

Riparian evergreen forest (1.2.1.1) Gallery forest (Ribeiro and Machado, 1998); evergreen forest (Zeilhofer and Schessl, 1999)

Xerophytic evergreen forest-cerradão (1.2.1.2)

Cerradão (Ribeiro and Machado, 1998); forested savanna (IBGE, 1992; PNMA, 1997); dense arboreal

savana (MME, 1982a)

Submontane open rainforest (1.2.1.3) Submontane open rainforest (IBGE, 1992; PNMA, 1997)

Alluvial semideciduous forest (1.2.2.1)

Alluvial semideciduous seasonal forest (IBGE, 1992; PNMA, 1997)

Lowland semideciduous forest (1.2.2.2)

Lowland semideciduous seasonal forest (IBGE, 1992; PNMA, 1997)

Submontane semideciduous forest (Chiquitano Dry Forest) (1.2.2.3)

Submontane semideciduous seasonal forest (IBGE, 1992; PNMA, 1997); Chiquitano semideciduous

forest (WWF, 2002)

Xerophytic semideciduous forest (Chaco) (1.2.2.4)

Forested steppe savanna (IBGE, 1992; PNMA, 1997); quebracho blanco forest (UNA, and GTZ, 1999);

xerophytic forest (MAG, and BGR, 1999); floodable forest (MAG, and BGR, 1999)

Mesoxerophytic semideciduous forest (Chaco) (1.2.2.5)

Forested steppe savanna (IBGE, 1992; PNMA, 1997); quebracho colorado forest (UNA, and GTZ, 1991;

MAG and BGR, 1999)

Lowland deciduous forest (1.2.3.1) Lowland deciduous seasonal forest (IBGE, 1992; PNMA, 1997)

Submontane deciduous forest (1.2.3.2)

Submontane deciduous seasonal forest (IBGE, 1992; PNMA, 1997)

Deciduous forest (Chiquitano Dry Forest) (1.2.3.3) Chiquitano deciduous forest (WWF, 2002)

Treed Cerrado savanna (1.3.1.1) Treed savanna (IBGE, 1992; PNMA, 1997)

Parkland Cerrado savanna (1.3.1.2) Parkland savanna (IBGE, 1992; PNMA, 1997)

Treed Chaco steppe savanna (1.3.2.1) Treed steppe savanna (IBGE, 1992; PNMA, 1997); xeric savanna (Frey, 1995)

Parkland Chaco steppe savanna (1.3.2.2)

Parkland steppe savanna (IBGE, 1992; PNMA, 1997); hydromorphic savannas (MAG and BGR, 1999)

Cerrado grassland (1.3.1.1) Gramineae-woody savanna (IBGE, 1992; PNMA, 1997)

Chaco grassland (1.3.2.1) Gramineae-woody steppe savanna (IBGE, 1992; PNMA, 1997)

Forest

Savanna

Grassland

53

The land cover classification system proposed in this study does not present an explicit category for wetlands because the existing regional literature is neither specific nor detailed in the characterization of wetlands. This fact reflects the complex dynamics of vegetation successional patterns caused by the spatially and temporally variable flooding patterns of the study area (Ponce and da Cunha, 1993). Discussion of the proposed land cover classes is presented in the following sub-sections. Each class of the proposed system is accompanied by its respective order number found in Table 5.

Open Water (1.1)Open Water (1.1)Open Water (1.1)Open Water (1.1) The open water category includes natural linear lotic (flowing) water bodies such as streams, and enclosed lentic (non-flowing) water bodies such as lakes (Anderson et al., 1976). In the Brazilian section of the Pantanal, seasonal water bodies, such as river corridors temporarily connecting depressions and seasonal lagoons, are described (Allem and Valls, 1987). However, in this study, open water refers only to permanent features and seasonal bodies of water are considered floodable lands.

Evergreen Forest (1.2.1)Evergreen Forest (1.2.1)Evergreen Forest (1.2.1)Evergreen Forest (1.2.1) Evergreen forests are found only in seasonally flooded and non-floodable lands, in the Brazilian section of the UPRB. Some authors mention the presence of seasonally flooded evergreen forest (1.2.1.1) associated with streams (Zeilhofer and Schessl, 1999; Frey, 1995). Zeilhofer and Schessl (1999) observed this forest type in sites located in the northern Brazilian Pantanal, on fluvial deposits along lower terraces of riverbanks. This forest type presents a low arboreal stratum of 8–11 m and predominance of Vochysia divergens can occur (Zeilhofer and Schessl, 1999). Another evergreen forest in the Cerrado, generally located on non-floodable lands, is the “xerophytic evergreen forest- cerradão” (1.2.1.2) (Ribeiro and Machado, 1998). In classification schemes directly related to the Pantanal and the UPRB, cerradão appears as a subtype of cerrado (PNMA, 1997). Its flora reflects that of the Cerrado, but in this study, due to its physiognomy, cerradão is included within the forests. Its arboreal stratum is low (8–16 m) (PNMA, 1997) and predominantly continuous, presenting 50 to 90 percent of coverage (Ribeiro and Machado, 1998). Species present in the cerradão of the UPRB include Carycar brasiliense, Salvertia convallariodora, Dimorphandra mollis, Astronium graveolens, Qualea grandifolia, Anadenanthera pelegrina, Hymenaea stigonocarpa, Qualea parviflora, Kielmeyera coriacea, among others (PNMA, 1997). Another type of non-floodable evergreen forest included in the UPRB is submontane open rainforest (“floresta ombrófila aberta submontana”) (1.2.1.3), which is of Amazonian influence (IBGE, 1992). Open rain forest presents four typical

54

physiognomies: (1) forest with palm (floresta-de-palmeira, cocal), (2) forest with woody vines (floresta-de-cipó, cipoal), (3) forest with bamboo (floresta-de-bambu, bambuzal), and (4) forest with Phenakospermum quianense (floresta-de-sororoca, sororocal) (IBGE, 1992). All open rain forests contain gaps or openings; subtypes are differentiated by altitudinal gradient: (1) lowland, (2) submontane, and (3) montane (IBGE, 1992). In the UPRB, only the subtype submontane has been identified (PNMA 1997), however, it is not specified to which of the four typical open rainforest physiognomies it corresponds. The highest submontane open rain forest identified in the region presents an arboreal stratum of 30 m (PNMA, 1997).

SeSeSeSemideciduous Forest (1.2.2)mideciduous Forest (1.2.2)mideciduous Forest (1.2.2)mideciduous Forest (1.2.2) In the UPRB, semideciduous forest can be found in seasonally flooded, floodable, and non-floodable lands. The forests classified as semideciduous in Brazil are comprised of 20–50 percent deciduous trees (PNMA, 1997). They are found in areas where a rainy summer is followed by a marked dry season (IBGE, 1992). Amazonian genera predominate, such as Parapiptadenia, Peltophorum, Carianiana, Lecythis, and Tabebuia, among others (IBGE, 1992). According to altitude, the following subtypes are recognized in Brazil: (1) alluvial, (2) lowland, (3) submontane, and (4) montane (IBGE, 1992). In the UPRB, all subtypes are identified except the montane (PNMA, 1997). The alluvial semideciduous forest (1.2.2.1) is floodable and associated with streams and water bodies. In the UPRB, physiognomically it resembles other riparian forests found in Brazil, but it is floristically distinct because of the presence of western Amazonian species (PNMA, 1997). A highest canopy of 30 m was observed for this forest type in the Brazilian UPRB (PNMA, 1997). Non-floodable lowland semideciduous forest (1.3.2.2) in the Brazilian section of the UPRB is characterized by an evergreen middle stratum formed predominantly by acuri palms (Scheelea phalerata) (PNMA, 1997). A highest canopy of 25 m was observed for this forest type (PNMA, 1997). The submontane subtype (1.2.2.3) is present in non-floodable terrains of higher elevation (more than 150 m) (PNMA, 1997). Certain species of the genera Cedrela, Tabebuia, and Aspidosperma are frequent. Typically, a submontane semideciduous forest comprises a stratum of shrubs and tree seedlings. A highest canopy of 30 m was observed for this forest type in the UPRB (PNMA, 1997). Landivar (2001) states that the submontane semideciduous forests described in the Brazilian section of the UPRB are part of the Chiquitano Dry Forest ecoregion, which is more conspicuous in Bolivia. The seasonally flooded and floodable Chaco forests of the region are described as semideciduous in Paraguayan literature (UNA and GTZ, 1991; MAG and BGR, 1999). The Chaco section of the UPRB presents xerophytic (quebracho blanco forest) (1.2.2.4) and mesoxerophytic (quebracho colorado forest) (1.2.2.5) semideciduous forests (UNA and

55

GTZ, 1991). These forest types overlap with the “forested steppe savanna” registered in the Brazilian vegetation classification system. Physiognomically a forest, this vegetation type is classified as a subtype of steppe and presumably is the single descriptor of several Chaco forest formations in the Brazilian literature (PNMA, 1997). Formations designated as “xerophytic forest” (1.2.2.4) differ from one location to the other in the Paraguayan Chaco (MAG and BGR, 1999). The type present in the UPRB thrives on sandy soils of hydric origin (MAG and BGR, 1999). This forest is short in stature, relatively dense, and usually comprised of three strata (canopy, understory, herbaceous cover) (MAG and BGR, 1999). Emerging specimens of the highest stratum may give it an open forest physiognomy. Some of the most common species mentioned for the xerophytic forest are: Pisionia sapallo, Anadenantera colubrina, A. peregrina, Aspidosperma quebracho-blanco, Schinopsis heterophylla, Capparis retusa, Ruprechtia triflora, Ximena americana, Acacia praecox, Mimosa velloziana, Capparis speciosa, Bauhinia sp., Zizipus mistol, Croton sp., Oxalis erosa, Dickya sp., Bromelia hyernimi, Rivina humilis, Eupatorium squarrosoramosum, among others (MAG and BGR, 1999). It is present in seasonally flooded and floodable areas (MAG and BGR, 1999). The “quebracho colorado” forest, called in the present study “mesoxerophytic semideciduous forest” (1.2.2.5) is described as belonging to the mesoxerophytic unit of the Chaco. This forest type is seasonally flooded (MAG and BGR, 1999). The predominant species is Schinopsis balansae; also common in the highest stratum are: Astronium urundeuva, Caesalpina paraguariensis, Phyllostylon rhamnoides, Calycophyllum multiflorum, Arecastrum romanzoffianum (MAG and BGR, 1999). Another source (UNA and GTZ, 1991) refers to the same formation as “quebrachal de quebracho colorado." This forest type is characterized as being the tallest (15 m) in the Chaco and localized in the lowest elevations (UNA and GTZ, 1991). However, “quebrachal de quebracho colorado" includes larger areas and does not exactly coincide in spatial distribution in comparison to the “quebracho colorado” forest.

Deciduous Forest (1.2.3)Deciduous Forest (1.2.3)Deciduous Forest (1.2.3)Deciduous Forest (1.2.3) Deciduous forests are subjected to two contrasting seasons: a rainy period followed by a long dry season. More than 50 percent of the canopy of this forest has deciduous tree specimens that lose their leaves during the dry season (IBGE, 1992). In Bolivia, types of deciduous forests are described as part of the section of the Chiquitano Dry Forest ecoregion in the UPRB (WWF, 2002). The literature suggests that these forests can be located in either floodable or non-floodable terrains (WWF, 2002). In the Brazilian classification system, subtypes of deciduous forest include: (1) alluvial, (2) lowland, (3) submontane, and (4) montane (IBGE, 1992). In the UPRB, only the lowland (1.2.2.2) and submontane (1.2.2.3) subtypes are described and both are in non-

56

floodable areas of the state of Mato Grosso do Sul. These subtypes present Afro-amazonian and Argentinean-andean species. Anadenanthera, Phyllostylon, Hymenaea, and Aspidosperma are frequent genera in these forests (PNMA, 1997). In the UPRB, the highest canopies observed present 20 m (PNMA, 1997). Regionally, these forests are known as “dry forests” (“matas secas”) and prefer calcareous substrates (PNMA, 1997).

Cerrado Savanna (1.3.1)Cerrado Savanna (1.3.1)Cerrado Savanna (1.3.1)Cerrado Savanna (1.3.1) In the proposed UPRB classification system, Cerrado savanna includes all arboreal-herbaceous formations that belong to or are influenced by the Cerrado floristic province. In the Brazilian official classification system (IBGE, 1992), Cerrado savannas are subdivided into four subtypes: (1) forested savanna (“cerradão”), (2) treed savanna (“cerrado”), (3) parkland (“savana parque – parque de cerrado”), and (4) Gramineae-woody savanna (“savana gramíneo-lenhosa”). These four subtypes can be found in the UPRB. However, in the present study, the forested and the Gramineae-woody savannas are excluded from the savanna class and placed in the forest and grassland classes, respectively, because of physiognomic differences (Ribeiro and Machado, 1998). The treed Cerrado savanna (1.3.1.1), known regionally as “cerrado,” and the parkland (1.3.1.2) subtype are considered of natural and/or anthropogenic origin (PNMA, 1997). In certain areas, especially those employed for cattle ranching, the Cerrado savannas are subjected to periodical fires, usually of annual frequency (PNMA, 1997). Sparsely distributed trees averaging 5 m of height with thin and crooked trunks compose the arboreal stratum of the treed Cerrado savanna subtype (PNMA, 1997). The herbaceous stratum is continuous and dominated by species from the Gramineae, Cyperaceae, and Eriocaulaceae families (PNMA, 1997). The most frequent species include: Curatella americana, Terminalia argentea, Luehea paniculata, Dimorphandra mollis, Annona crassifolia, Stryphnodendron sp. (PNMA, 1997). A sparse shrubby stratum is sometimes present, in which short palms predominate; the most common of them are from the genera Astrocaryum and Diplothemium (PNMA, 1997). The RADAM Brazil project (MME, 1982b), which refers to vegetation regions rather than vegetation types, presents two subcategories for this subtype: (1) without gallery forest, and (2) with gallery forest. The parkland subtype is characterized by a predominant herbaceous stratum comprised of hemicryptophytes and geocryptophytes, where species of Gramineae and Cyperaceae are the most common (PNMA, 1997). Isolated nano and microphanerophytes (3–5 m) form the highly sparse arboreal stratum (PNMA, 1997). In certain areas, parklands present formations composed exclusively or predominantly by specimens of one tree species, such as Tabebuia aurea or Curatella Americana (PNMA, 1997). In the RADAM project (MME, 1982a), parkland is classified an herbaceous formation regionally known as campo sujo (shrubby rangeland); however, Ribeiro and Machado (1998) classify it as a savanna.

57

Chaco Steppe Savanna (1.3.2)Chaco Steppe Savanna (1.3.2)Chaco Steppe Savanna (1.3.2)Chaco Steppe Savanna (1.3.2)

Chaco steppe savanna includes all arboreal-herbaceous formations that belong to or are influenced by the Chaco floristic province: treed (1.3.2.1) and parkland (1.3.2.2) steppe savanna subtypes. These vegetation types are present in the Chaco of Brazil, Paraguay, and Bolivia. Steppe savanna is the term used in the Brazilian classification system to denote vegetation types from the Chaco and Caatinga floristic provinces. This descriptor was originally employed in Africa to designate a tropical vegetation type with characteristics of steppe, which is subjected to a double seasonality marked by two annual dry seasons (IBGE, 1992). Chaco savannas have also been regarded as xeric savannas (Frey, 1995). The Chaco penetrates Brazil near the confluence of the Apa and Paraguay rivers and expands northward, bordering the Paraguay River, to the State of Mato Grosso (PNMA, 1997). Chaco steppe savanna is usually associated with saline soils (IBGE, 1992). In Brazil, the steppe savanna is subdivided into: (1) forested steppe savanna, (2) treed steppe savanna, (3) parkland steppe savanna, and (4) Gramineae-woody steppe savanna (IBGE, 1992). All of these subtypes are present in UPRB (IBGE, 1992). The forested subtype has been described in the forest section, and the Gramineae-woody steppe savanna is described among the herbaceous formations. The treed Chaco steppe savanna (1.3.2.1) presents the same floristic characteristics as the forested steppe savanna, but it consists of shorter trees that are less densely grouped, forming an open arboreal stratum coexisting with a continuous but seasonal herbaceous stratum (PNMA, 1997). The predominant species of phanerophytes belong to the genera Prosopis, Acacia, Ziziphus, and Celtis. It can be of anthropic origin, when trees of forests are extracted favoring the extension of the herbaceous stratum (IBGE, 1992; PNMA, 1997). More sparsely dispersed trees characterize the arboreal stratum of the parkland Chaco steppe savanna (1.3.2.2). This subtype usually occurs in depressions that become inundated during the rainy season because of their poor drainage (PNMA, 1997). Species of the genera Acacia and Machaerium are common; in the dry season, Leptochloa sp. spreads through formerly flooded areas (PNMA, 1997). This steppe savanna subtype also includes the regionally named formation “carandazal” located in floodable areas where the palm Copernicia alba forms almost pure groups (PNMA, 1997). In the Paraguayan literature, “carandazal” is mentioned as “hydromorphic savanna.” It occurs on structured soils rich in clay, with predominance of gleycosol and vertisol, seasonally flooded during three to six months per year due to precipitation and river flooding (MAG and BGR, 1999). It lacks an intermediate stratum; the ground-level herbaceous stratum is formed by annual marshy vegetation (MAG and BGR, 1999).

58

Cerrado Grassland (1.4.1)Cerrado Grassland (1.4.1)Cerrado Grassland (1.4.1)Cerrado Grassland (1.4.1)

Cerrado grasslands are found in seasonally flooded, floodable, and non-floodable areas and is designated in Brazilian classification systems as Gramineae-woody savanna. Cerrado grassland can be of natural and/or anthropogenic origin (PNMA, 1997). In the natural state, it is constituted by grasses and scattered shrubs, small isolated trees and stem-less palms (PNMA, 1997). Fire favors the establishment of geophytes over hemicryptophytes (PNMA, 1997). The most common woody species are from the genera Andira, Cassia, Byrsomina, and Bauhinia (PNMA, 1997). Frequent palms are Attalea sp., Orbignya eichleri, Allagoptera campestris; the herbarceous Gramineae stratum is mostly comprised of species of Axonopus, Andropogon, Aristida, Tristachya, Paspalum, Hemarthria, Digitaria, Panicum, and Brachiaria (PNMA, 1997).

Chaco Grassland (1.4.2)Chaco Grassland (1.4.2)Chaco Grassland (1.4.2)Chaco Grassland (1.4.2) Chaco grasslands are found in seasonally flooded, floodable, and non-floodable areas. It is a predominantly herbaceous formation of Chaco origin in which Paratheria prostrata, Aristida sp., and Elyonurus sp. are common (PNMA, 1997). Scattered thorny shrubs, such as Celtis sp., also can be found (PNMA, 1997).

Land Cover in Pilot AreaLand Cover in Pilot AreaLand Cover in Pilot AreaLand Cover in Pilot Area Vegetation in the pilot area is dominated by Chaco formations, which include semideciduous and deciduous forest, steppe savanna and Chaco grassland. These vegetation types, particularly the savannas and grasslands, are subjected to periodic inundation, occurring on seasonally flooded and floodable lands. Cerrado formations and submontane open rain forest have not been mapped in the pilot area (IBGE, 1992; PNMA, 1997; MME, 1982b). Chiquitano Dry Forest formations are present in the northwestern corner of the pilot area. Anthropogenic categories (agriculture land, urban land) are also observed. IIIIMAGE MAGE MAGE MAGE PPPPROCESSING ROCESSING ROCESSING ROCESSING RRRRESULTS ESULTS ESULTS ESULTS (P(P(P(PILOT ILOT ILOT ILOT AAAAREAREAREAREA)))) A total of ten land cover classes were identified in the two classified images: (1) open water, (2) deciduous forest, (3) semideciduous forest, (4) savanna, (5) floodable savanna, (6) seasonally flooded savanna, (7) grassland, (8) floodable grassland, (9) seasonally flooded grassland, and (10) bare land (Figure 7). The extent of open water is the same in the two classified images because it results from the multi-flood extent image. All pixels previously classified as open water that did not match the permanent water class of the multi-flood extent image were reclassified to floodable or seasonally flooded grasslands.

59

Figu

re 7

. La

nd c

over

cla

ssifi

catio

n re

sults

.

60

It was not possible to discriminate the anthropogenic versus natural classes through the unsupervised classification approaches performed in this study. Natural land cover classes were distinguished to the third level of the proposed classification system.

Land Cover ClassesLand Cover ClassesLand Cover ClassesLand Cover Classes Bare Land Areas without vegetation, excluding open water, but including parts of urban areas and roads, were classified as “bare land.” Based on visual assessment of the unprocessed images, the majority of the areas classified as bare land, in both approaches, correspond to natural areas, which generally appear as some type of Chaco grassland or savanna on Brazilian and Paraguayan vegetation maps (UNA and GTZ, 1991; PNMA, 1997). In the process of assigning classes in the two-date approach, it was observed that a large extent of the bare lands present on the 11/14/1999 image appeared as grasslands or savannas on the 7/30/1998 image. Therefore, the two-date classified image presents a lower extent of bare lands and more grasslands and savannas. The difference of 16% corresponds mainly to grasslands and savannas that are located in non-floodable areas (as determined by the three-date flood extent image), since the bare lands classified in floodable and seasonally flooded areas were reclassified, in both approaches, as categories of grasslands (Table 7). The seasonality or change in the land cover of these areas may be caused by flooding that was not detected in this study or it may be caused by vegetation succession promoted by other natural causes or anthropogenic intervention.

Table 7. Coverage of land cover types in the two classified images.

Area (ha)Area (ha)Area (ha)Area (ha) % (A)% (A)% (A)% (A) Area (ha)Area (ha)Area (ha)Area (ha) % (B)% (B)% (B)% (B)

Open Water 24,532.92 0.80 24,532.92 0.80 0

Deciduous Forest 113,223.24 3.69 72,177.30 2.35 1.34

Semideciduous Forest 1,192,097.52 38.86 1,368,078.48 44.60 -5.74

Savanna 255,674.43 8.34 381,838.59 12.45 -4.11

Floodable Savanna 87,663.78 2.86 182,256.84 5.94 -3.08

Seasonally Flooded Savanna 14,999.31 0.49 17,749.17 0.58 -0.09

Grassland 76,386.42 2.49 332,717.04 10.85 -8.36

Floodable Grassland 599,774.40 19.55 483,843.87 15.77 3.78

Seasonally Flooded Grassland 53,833.77 1.76 48,570.66 1.58 0.17

Bare Land 649,259.46 21.17 155,680.38 5.08 16.09

Total 3,067,445.25 100.00 3,067,445.25 100.00 -

One-date approachOne-date approachOne-date approachOne-date approach Two-date approachTwo-date approachTwo-date approachTwo-date approachClassClassClassClass Difference Difference Difference Difference

in % (A - B)in % (A - B)in % (A - B)in % (A - B)

61

Grasslands and Savannas Classes of grasslands and savannas were considered seasonally flooded, floodable, or non-floodable. The floodable classes correspond to areas of potential flood, while the extent of the seasonally flooded classes is considered the minimum area periodically flooded. The seasonally flooded areas in the pilot area correspond mainly to grasslands and palm savannas (parkland Chaco steppe savanna) from the Chaco ecoregion (MME, 1982b; UNA and GTZ, 1991; IBGE, 1992; PNMA, 1997). Savannas and grasslands of uncertain floristic origin were encountered in the Bolivian section of the pilot area. Anthropogenic areas with pastures could not be spectrally differentiated and were classified as grasslands. Grasslands (as well as other vegetation types) may also include lagoons covered by aquatic vegetation. Semideciduous Forest Semideciduous forests in the study area are predominantly comprised of non-floodable forests but also include riparian forest and forest in low areas that might be seasonally flooded or in floodable areas. The forests of this class are considered semideciduous based on Brazilian and Paraguayan vegetation and land cover maps of the pilot area (MME, 1982b; UNA and GTZ, 1991; IBGE, 1992; PNMA, 1997). It was not possible to divide this class into the various semideciduous forest types occurring in the area. Deciduous forest types could be mistakenly included in this class, particularly in the Brazilian section, where deciduous forest is found in the east-southern corner of the scene. The two-date approach presents a greater extent of semideciduous forest, close to 6 percent more than that of the one-date approach. This discrepancy is mainly due to the differences in the classification of deciduous forest. Deciduous Forest The decision to classify certain forested areas as deciduous, mainly in Bolivia, was based on the literature (WWF, 2002). In both classification approaches, this class of deciduous forests is spectrally distinct. Particularly, in the two-date approach, it was observed that in 7/30/1998 the vegetation cover is scarce in comparison to the evident forest cover present in 11/14/1999. This reaffirms the strong seasonality of these forests. Areas of deciduous forest found on vegetation maps of the Brazilian section (MME, 1982b) were not differentiated from the semideciduous forest class and do not present the same seasonality as the identified deciduous forest when comparing the two dates of imagery (07/30/1998 and 11/14/1999).

Accuracy Assessment Accuracy Assessment Accuracy Assessment Accuracy Assessment It was not possible to evaluate all ten classes of the classified images because the reference data do not include deciduous forest. Additionally, flooding aspects had to be

62

disregarded because the photographs used as reference data were taken at different times of the year (December 2000 and March 2002). Therefore, in the comparison between the two approaches of land cover classification, only five combined classes were considered: (1) open water, (2) forest, (3) savanna, (4) grassland, and (5) bare land. The results show the two-date approach to be superior to the one-date approach (Tables 8, 9, and 10). The greatest difference between the two approaches is the classification of bare land. In the one-date classification, the user’s accuracy of the bare land class is extremely low, approximately 8%. In the two-date classification, the user’s accuracy of the bare land class is greater, around 45%, but it is still the lowest among the classes. In both cases, the main source of error lies in the confusion of bare land with savanna. The two-date approach also shows better results in the classification of savanna, but this class still presents confusion with forest, bare land, and grassland. Regarding the latter class, the two-date approach presents greater user’s accuracy (75% versus 70%), but lower producer’s accuracy (69% versus 82%), which indicates that the two-date approach produced a grassland class with less error, but that it also is less inclusive of correctly classified grasslands than that of the one-date approach. In general, grassland shows less confusion with bare land than does savanna. The classes presenting the greatest values of accuracy, in both approaches, are open water and forest. As expected, the open water class presents the same accuracy for both approaches. Forest yielded similar accuracy results in both approaches.

Table 8. Accuracy assessment results for one-date approach.

Open Forest Savanna Grassland Bareland TotalOpen Water 4 0 0 1 0 5

Forest 1 67 10 3 0 81Savanna 0 1 7 4 0 12

Grassland 1 0 16 45 2 64Bareland 0 2 30 2 3 37

Total 6 70 63 55 5 199199199199

ClassClassClassClassUser's User's User's User's

Accuracy Accuracy Accuracy Accuracy (%)(%)(%)(%)

Producer's Producer's Producer's Producer's Accuracy Accuracy Accuracy Accuracy

(%)(%)(%)(%)Open Water 80.00 66.67Forest 82.72 95.71Savanna 58.33 11.11Grassland 70.31 81.82Bareland 8.11 60.00

Overall AccuracyOverall AccuracyOverall AccuracyOverall Accuracy: 63.32%KHAT StatisticKHAT StatisticKHAT StatisticKHAT Statistic: 0.5066

Classiffied Classiffied Classiffied Classiffied DataDataDataData

Reference DataReference DataReference DataReference Data

63

Table 9. Accuracy assessment results for two-date approach.

Open Water Forest Savanna Grassland Bareland TotalOpen Water 4 0 0 1 0 5

Forest 1 68 11 6 0 86Savanna 0 0 36 10 0 46

Grassland 1 2 10 38 0 51Bareland 0 0 6 0 5 11

Total 6 70 63 55 5 199199199199

ClassClassClassClassUser's User's User's User's

Accuracy Accuracy Accuracy Accuracy (%)(%)(%)(%)

Producer's Producer's Producer's Producer's Accuracy Accuracy Accuracy Accuracy

(%)(%)(%)(%)Open Water 80.00 66.67

Forest 79.07 97.14Savanna 78.26 57.14

Grassland 74.51 69.09Bareland 45.45 100.00

KHAT StatisticKHAT StatisticKHAT StatisticKHAT Statistic: 0.6563

Reference DataReference DataReference DataReference Data

Classified Classified Classified Classified DataDataDataData

Overall AccuracyOverall AccuracyOverall AccuracyOverall Accuracy: 75.88%

Table 10. Comparison of classification approaches.

One-date approach

Two-date approach

Accuracy 63.32% 75.88%KHAT 0.5066 0.6563

Variance 0.0016 0.0018Z statistic 12.67 15.51

Z statistic: 2.5722

Error matrixParameter

One-date versus two-date approach

Finally, the two-date approach has a greater KHAT statistic (0.66 versus 0.51) and overall accuracy (76 versus 63%), indicating not only that this classification approach is more accurate when compared to the one-date method, but also in regard to random classification (Lillesand and Kiefer 2000). The Z test (Congalton and Green 1999)

64

comparing the KHAT statistic values of the error matrices confirms that the accuracy results of the two classification approaches differ significantly at the 95% confidence level (Z = 2.57) (Table 10). CCCCONCLUSIONSONCLUSIONSONCLUSIONSONCLUSIONS This study standardizes the language of disparate land cover/vegetation classification schemes into one integrated system for the entire UPRB. The proposed system is open ended, has a clear hierarchical structure of three levels, and presents classification criteria for flooding that permits the addition of other land cover and vegetation types and levels of finer scale. In the classification of vegetation, hierarchical priority was given to physiognomy, which is present in the structure of the majority of the classification schemes reviewed, and has proven to be, in this study, one parameter that allows clear separability of vegetation classes in the visual interpretation and automated classification of Landsat imagery. The image processing results, limited to the pilot area, clearly indicate that the seasonality of vegetation and flooding must be taken into account; multitemporal data are particularly important for the identification and differentiation of savannas and grasslands, which are highly variable in the region, presumably a result of the alternation of contrasting seasons. The two-date unsupervised classification approach has a significantly greater overall accuracy and KHAT statistic in comparison to the one-date approach, which indicates that a single scene cannot provide data to properly classify the land cover in the pilot area. Of all classes, the most problematic is “bare land,” which designates areas recuperating from natural or anthropogenic disturbances (e.g., fire, flood). The two-date classification shows greater accuracy mainly because of the better classification of bare land and savanna. However, the two-date classification still presents an overall accuracy below the 85% acceptable for mapping (Anderson et al. 1976). The addition of images from intermediate seasons could provide a better characterization of the seasonality of vegetation and improve the classification results, especially in the grassland and savanna areas. The methods suggested consider the characteristics of the pilot area and the results achieved are specific to this area. It cannot be generalized that the land cover classification system and/or the multispectral classification method are appropriate for other areas of the UPRB; it can only be suggested that these methods will work in areas presenting similar environmental conditions and land cover types. In areas such as the uplands, where flooding and vegetation seasonality may not be an issue, the temporal variability might not be as important. In areas subjected to seasonal changes where, for example, bare land can be associated specifically with only one vegetation type, a multitemporal classification will not be as critical.

65

In the pilot area, the multitemporal aspect proved to be fundamental for land cover classification. Although the differentiation of the savanna and grassland physiognomies was not supported by strong values of accuracy, the comparison of the two classification approaches demonstrated that large extents of the savanna and grassland classes were not recognized when only one date of imagery was used. Also noted is that the overestimated bare land class generated in the one-date classification approach cannot be indisputably associated with a single land cover type, which reinforces the importance of using multiple dates to depict intra-annual variability of land cover even in a single classification. Further research should experiment with the addition of more dates of imagery and other classification procedures to enhance discrimination between savannas and grasslands. AAAACKNOWLEDGEMENTSCKNOWLEDGEMENTSCKNOWLEDGEMENTSCKNOWLEDGEMENTS For revisions and important suggestions, thanks to Richard Kempka (Ducks Unlimited), Thad Wasklewicz (The University of Memphis), Montserrat Carbonell (Ducks Unlimited), Kelley Crews-Meyer (The University of Texas at Austin) and Kenneth Young (The University of Texas at Austin). This study was conducted with the financial and logistic support of the U.S. Forest Service, the U.S. Fish and Wildlife Service, Ducks Unlimited and the University of Memphis. RRRREFERENCESEFERENCESEFERENCESEFERENCES Anderson, J.R., Hardy, E.E, Roach, J.T., and Witmer, R.E. 1976. A land use and land cover classification system for use with remote sensor data. Geological Survey Professional Paper 964. Washington, D.C: U.S. Geological Survey. Browne, D., Carbonell, M. and Kempka, D., eds. 2003. Upper Paraguay River Basin GIS Database Project, Pilot Project I. Memphis, TN: Ducks Unlimited, Inc. Center for Disease Control and Prevention (CDC). 2000. Geographic Boundaries. http://www.cdc.gov/epiinfo/shape.htm (last accessed 21 July 2003) Congalton, R. G. and Green, K. 1999. Assessing the Accuracy of Remotely Sensed Data: Principles and Practices. Boca Raton, FL: Lewis Publishers. Frazier, P. and Page, K. 2000. Water body detection and delineation with Landsat TM data. Photogrammetric Engineering & Remote Sensing 66: 1461–1467. Frey, R. 1995. Flora and vegetation of “Las Piedritas” and the margin of Laguna Caceres, Puerto Suarez, Bolivian Patanal. Bulletin of the Torrey Botanical Club 122: 314–319.

66

Instituto Brasileiro de Geografia e Estatística (IBGE). 1992. Manual técnico da vegetação brasileira (technical handbook of the Brazilian vegetation). Rio de Janeiro: IBGE. Landivar, R. 2001. Chiquitano Dry Forests. Ecoregions. World WildLife Fund. http://www.worldwildlife.org/wildworld/profiles/terrestrial_nt.html. (last accessed 17 July 2003) Li, X. and Yeh, A.G. 1998. Principle component analysis of stacked multi-temporal images for the monitoring of rapid urban expansion in the Pearl River Delta. International Journal of Remote Sensing 19: 1501–1518. Lillesand, T. M. and Kiefer, R.W. 2000. Remote sensing and image interpretation. New York: John Wiley & Sons, Inc. Liu, W. T. 2003. Upper Paraguay River Basin hydrology analysis II – sub basin hydrology. Forthcoming. Lyon, J. 2001. Wetland Landscape characterization: GIS, remote sensing, and image analysis. Chelsea, MI: Ann Arbor Press. Lunetta, R. S. 1998. Applications, project formulation, and analytical approach. In Remote sensing change detection, environmental monitoring methods and applications, ed. Lunetta, R. S. and C. D. Elvidge, 1–19. Chelsea: Ann Harbor Press. Lunetta, R. S. and Balogh, M. 1999. Application of multi-temporal Landsat 5 TM imagery for wetland classification. Photogrammetric Engineering & Remote Sensing 65: 1303–1310. Mereles, F., ed. 2000. Iniciativas transfronterizas para el Pantanal (Paraguay): Delimitación, áreas de conservación, plan de conservación. Paraguay: DeSdelChaco, The Nature Concervancy, U.S. Agency for International Development. Ministerio de Agricultura y Ganadería (MAG) and Bundesanstalt für Geowissenshaften und Rohstoffe (BGR). 1999. Proyecto de Sistema Ambiental del Chaco (Project of the Environmental System of Chaco). Asunción: MAG and BGR. Ministério das Minas e Energia (MME). 1982a. Levantamentos dos Recursos Naturais (inventory of natural resources). Vol. 26. Projeto RADAMBRASIL. Rio de Janeiro: MME. Ministério das Minas e Energia (MME). 1982b. Levantamentos dos Recursos Naturais (inventory of natural resources). Vol. 27. Projeto RADAMBRASIL. Rio de Janeiro: MME.

67

Olson, D. M., Dinerstein, E., Wikramanayake, E.D., Burgess, N.D., Powell, GVN., Underwood, C., D’Amico, J.A., Strand, H.E., Morrison, J.C., Loucks, C.J., Allnutt, T.F., Lamoreux, J.F., Ricketts, T.H., Itoua, I., Wettengel, W.W., Kura, Y., Hedao, P., and Kassem, K. 2001. Terrestrial ecoregions of the world: a new map of life on Earth. BioScience 51: 933–938. Ponce, V. M., and da Cunha, C.N. 1993. Vegetated earthmounds in tropical savannas of central Brazil: a synthesis. Journal of Biogeography 20: 219–225. Programa Nacional de Meio Ambiente (PNMA), ed. 1997. Plano de Conservação da Bacia do Alto Paraguai – PCBAP/Projeto Pantanal, Programa Nacional do Meio Ambiente. Vol. 3. Brasília: PNMA. Ribeiro, J., and Machado, B. 1998. Fitofisionomias dos bioma cerrado. In Cerrado: ambiente e flora, 89–166. Planaltina: Empresa Brasileira de Pesquisa Agropocuária (EMBRAPA). Swarts, F. 2000. The Pantanal in the 21st Century: for the planet’s largest wetland, uncertain future. In The Pantanal of Brazil, Bolivia and Paraguay: selected discourses on the world’s largest remaining wetland system, ed. F. A. Swarts, 1–22. Gouldsboro, PA: Hudson MacArthur Publishers. The Nature Conservancy (TNC). 2002. Pantanal 2002, compilado de datos ambientales y socioeconómicos. (CD) TNC. Universidad Nacional de Asunción (UNA) and Gesellschaft für Technische Zusammenarbeit (GTZ). 1991. Vegetación y uso de la tierra de la Región Occidental. San Lorenzo: UNA y GTZ. U.S. Geological Survey (USGS). 1999. Drainage Basins. http://edcdaac.usgs.gov/gtopo30 /hydro/sa_basins.asp (last accessed 25 July 2004) World Wildlife Fund (WWF). 2002. Priorización de cuencas para la conservación del Pantanal boliviano. Santa Cruz: WWF. Yuan, D., Elvidge, C.D. and Luneta, R.S. 1998. Survey of multispectral methods for land cover change analysis. In Remote sensing change detaction, environmental monitoring methods and applications, ed. Lunetta, R. S. and Elvidge, C.D., 21–39. Chelsea: Ann Harbor Press.

68

Zeilhofer, P., and Schessl, M. 1999. Relationship between vegetation and environmental conditions in the Northern Pantanal of Mato Grosso. Journal of Biogeography, 27: 159–168.

69

2C. UUUUPPER PPER PPER PPER PPPPARAGUAY ARAGUAY ARAGUAY ARAGUAY RRRRIVER IVER IVER IVER IIIINUNDATION NUNDATION NUNDATION NUNDATION PPPPREDICTION REDICTION REDICTION REDICTION UUUUSING SING SING SING RRRRAINFALL AND AINFALL AND AINFALL AND AINFALL AND NDVI:NDVI:NDVI:NDVI: BBBBASIN AND ASIN AND ASIN AND ASIN AND SSSSIX IX IX IX SSSSUB UB UB UB BBBBASIN ASIN ASIN ASIN MMMMODELSODELSODELSODELS

WILLIAM T. LIU14 AND FÁBIO M. AYRES15

AAAABSTRACTBSTRACTBSTRACTBSTRACT Inter annual variation of monthly River Water Level (RWL) at Ladário Hydrological Station (LHS) was used to study the flood conditions of the entire Upper Paraguay River Basin (UPRB). Six sub basin models were also constructed to investigate the contribution of water collected in the upper sub basins on individual sub basin RWL estimations. The responses of RWL to Precipitation (PCP) and to Normalized Difference Vegetation Index (NDVI) were analyzed. A combined use of PCP and NDVI data is proposed to predict RWL at LHS, which monitors the water collected from the upper UPRB watershed. Inundation area estimation as a function of RWL proposed by Hamilton et al. (1996) was applied to predict Pantanal inundation area using both recorded and predicted RWL data of 1981 to 2000. Our technique demonstrated that by applying RWL prediction model and the inundation area estimation model proposed by Hamilton et al. (1996), the Pantanal inundation area extent could be predicted one month in advance with reasonable success. Therefore, the statistical approach presented herewith may provide a useful tool to predict RWL and, hence, prevent flood damage in high RWL periods as well as control river transportation traffic in order to prevent riverbank erosion during low RWL periods. For future studies, an adequate hydrological simulation model based on a high accuracy digital elevation model and a rainfall forecasting system, such as a radar system, are needed to fulfill a real time flood advancing prediction and mitigation tasks. IIIINTRODUCTIONNTRODUCTIONNTRODUCTIONNTRODUCTION The Pantanal wetland is one of the world’s largest continental wetlands, and collects the drainage water from the Upper Paraguay River Basin (UPRB). The Pantanal holds one of the most diversified human patrimonies, housing innumerous flora and fauna. The distinct annual cycle of inundation provides rich natural resources for economically important fish and wildlife species, and more recently, ecological tourism. Fish yields may vary from the alteration of their life cycle caused by seasonal and annual variation of inundation patterns (Welcomme, 1985). Animals may have difficulties finding dry refuges during unpredicted high flood years (Mittermeier et al., 1990). Therefore the occasional unpredicted high floods and severe droughts affect fish and wildlife as well as important cattle production areas. During the past 30 years, agricultural activities and 14 Universidade Católica Dom Bosco, Campo Grande, MS, Brazil. E-mail address: will@ucdb.br 15 OIKOS (Cooperativa de Trabalho Sócio-ambiental), Campo Grande, MS, Brazil; fabio_oikos@pop.com.br

70

cattle production in the catchment have increased considerably. Improper land management on predominantly sandy soils in the upper basin continues to cause serious erosion and acceleration of sediment load in the flood plain. In addition to the gradual loss of ranchlands, unpredicted floods often result in serious cattle loss. Also, due to recent increase of river transportation, riverbanks erosion in low water level periods becomes a more serious issue every year. Therefore, the purpose of this study is to develop an alternative river water level prediction model in order to predict the occurrences of high flood for suitable mitigation planning and low river water level for river traffic controls. Several researchers have investigated Pantanal inundation behaviors. Giddlings and Choudhury (1989) have observed that the Pantanal normal maximum water surface occurs in April, which was detected by the microwave polarization difference temperature images collected from Nimbus-7 37 GHz Scanning Multifrequency Microwave Radiometer (SMMR) data from 1979 to 1985. Galdino and Clarke (1997) predicted the possible peak RWL occurrence month based on the probability analysis of 6 RWL classes obtained from the RWL data for the period of 1990 to 1996. They observed that the annual occurrence of maximum monthly RWL at LHS shifted gradually from April to July while the monthly RWL decreased from over 6 meters to lower than 4 meters. Hamilton et al. (1996) presented a Pantanal inundation area estimation approach using passive microwave remote sensing data obtained from Nimbus-7 SMMS. They constructed a mixing model that calculated typical fractions of water, inundation area and no-inundation area from representative cells and then extended to the whole region using Landsat TM images. They observed that the maximum inundation occurred as early as February in the north and as late as June in the south, reflecting the gradually delayed drainage time in responding to rainfall from the upper to the lower part of the UPRB. They also presented an inundation area estimation equation that calculates the inundation area as a function of the previous two-month averaged RWL monitored at LHS. The equation is described as follows: Equation 1 INUNDAt = 18.52 (RWL t+1 + RWL t+2 )/2 – 17.309 (1) Where: INUNDAt = Inundation area of month t (km); RWL t+1 and RWL t+2 = river water level (m) of second and third month after month t, respectively. Although the inundation area estimation equation proposed by Hamilton et al. (1996) is very useful for long-term hydrology and climate diagnosis, it is not for prediction

71

purposes, since the estimation is based on the RWL data two months after the inundation occurrence. Normalized Difference Vegetation Index (NDVI) is an index that provides a standardized method of comparing vegetation greenness between satellite images, also provides a measure of surface soil moisture conditions (Liu et al., 1994; Liu and Kogan, 1996; Liu and Juarez, 2001). NDVI reaches its maximum value at a threshold rainfall amount beyond which further increase of rainfall will not increase surface greenness or the NDVI value. Therefore, NDVI fails to indicate high RWL caused by excesses of rainfall. In order to overcome this, a combination of both rainfall and NDVI data are suggested for developing a RWL prediction model. A combined use of the inundation area estimation method proposed by Hamilton et al. (1996) and the RWL prediction model constructed in this study will be demonstrated and may provide a useful tool for climatic statistical modeling for flood extent prediction. MMMMETHODSETHODSETHODSETHODS

Study AreaStudy AreaStudy AreaStudy Area

The UPRB has a catchment area of 485,000 km2 and is located in central South America, covering parts of western Brazil, eastern Bolivia and northeastern Paraguay. Annual mean temperature stays relatively constant ranging from 24.5°C to 25.6°C. Annual precipitation (PCP) ranges from 1100 mm in the west to 1650 mm in the east with a distinct wet (October to April) and dry season (June to August). Monthly rainfall in the dry season is generally less than 50 mm/month and in the wet season is usually higher than 100mm/month. Monthly rainfall in both transit months (May and September) varies between 50 to 100 mm/month. Table 1 shows the climate data of several meteorological observation stations located in the UPRB (DNMET, 1992).

Table 1 – Climatic data of the Upper Paraguay River Basin (Data source: DNMET, 1992) ________________________________________________________________________________

Station Name Latitude Longitude Mean Total Rainfall temperature °C/mm/year

________________________________________________________________________________ Corumbá 19.05°S 57.30°W 25.0 1118 Coxim 18.30°S 54.46°W 24.5 1501 Cáceres 16.03°S 57.41°W 25.2 1460 Cuiabá 15.33°S 56.07°W 25.6 1315 Diamantino 14.24°S 56.27°W 25.2 1619

________________________________________________________________________________ Figure 1 illustrates the UPRB boundary that extends from latitude 14°S to 23°S and longitude 52°W to 61°W. The Pantanal wetland depression (Figure 1) occupies an area of

72

137,000 km2 and is formed by a mosaic of alluvial fans of Pleistonncene origin (Klammer, 1982), collecting the drainage water from the UPRB watershed. The Pantanal flat plains have a gentle slope of 2.5 cm/km with a gradual increase of elevation from west (80 m) to east and north (250 m). The upland drainage basin consists of flat plains to the west and elevated plateaus and low mountains to the north and east with elevation ranging from 250 m to 1200 m. The UPRB has very distinct hydrological features. First, some of its tributaries may change their course from year to year, which is caused by the high velocity of river water acting on a fragile and curved sandy riverbank.

Figure 1. Location of the Upper Paraguay River Basin (UPRB) with its upper and lower divisions and Ladário Hydrological Station and six rainfall stations, including Arenapólis, Quebo, Porto Estrela, Ponte Cabaçal, N. S. Livramento and Barão de Melgaço. Figure 2 shows numerous intersecting tributaries. Second, due to the Pantanal depression formed by a mosaic of alluvial fans, inundated sub basins often invade adjacent sub basin flood plains instead of discharging into the river. Third, there is a distinct flood wave delay of one to five months from north to south due to slow ground water drainage in the flat floodplains (Hamilton et al. 1996).

73

Figure 2. Aerial photo of the Upper Paraguay River tributaries in the Pantanal Nabileque River region.

Data UsedData UsedData UsedData Used River Water Level Data RWL data recorded at the LHS (Latitude: 19° 05´S; Longitude: 57° 30´W) for the period of January 1981 to December 2000 provided by the Brazilian Marine Corps at Corumbá, Mato Grosso do Sul State, were used in this study. The RWL measured at the LHS monitors the total water drained from the upper part of the UPRB. For the same period, monthly PCP data of six rainfall stations, including Arenápolis (14.51°S, 56.1°W), Quebo (14.65°S, 56.11°W), Porto Estrela (15.31°S, 56.23°W), Ponte Cabaçal (15.47°S, 57.9°W), N. S. Livramento (15.77°S, 56.35°W) and Barão de Melgaço (16.19°S, 55.95°W), provided by the Brazilian National Water Agency (Agencia Nacional de Agua, ANA) were used. All six rainfall stations are located outside the Pantanal wetland area and are concentrated in the northern part of the sub basin. Averaged values of six station rainfall data were used to represent the rainfall amount received in the upper part of the UPRB. NDVI data NDVI monthly maximum value composite data (Holben, 1986) with a resolution of 8 km by 8 km for the period of August 1981 to December 2000, provided by the Distributed Active Archive Center at Goddard Space Flight Center (DAAC/GSFC) of NASA were used in this study (AVHRR-Advanced Very High Resolution Radiometer). According to Eidenshink et al. (1997), these data have already been processed with radiometric calibration, using the NOAA standard method (Kidwell, 1995; Rao and Chen, 1995), and

74

atmospheric corrections including Raleigh scattering using method of Gordon et al. (1988) and ozone absorption using the method of Fleig et al. (1983). NDVI is a ratio of the reflectance values of AVHRR channels 2 (Ch2: 0.725 -1.10 µm ) and 1 (Ch1: 0.58-0.68 µm Ch1: 0.58-0.68 µm), and is expressed by the following equation: Equation 2. NDVI=(Ch2-Ch1)/(Ch2+Ch1). (2) The watershed drainage area monitored by LHS was used to obtain average monthly NDVI. In the present study, only the upper part of the basin was used to construct the RWL prediction model. The heavy black line in the middle of Figure 1 shows the division of the UPRB. Only the area north of this line was under consideration since the RWL at LHS collects all water drained from this catchment area. The interflow effect between sub basins outside this area was not considered

BAP Model Construction and ValidationBAP Model Construction and ValidationBAP Model Construction and ValidationBAP Model Construction and Validation According to Hamilton et al. (1996), there is a delay of 2 to 6 months for the peak monthly RWL at LHS in response to peak rainfall in the Pantanal. The progressive flood wave along regional hydrological flow paths starts in the northern and eastern parts of the Pantanal and progressively moves towards the southern regions. Therefore, correlation analysis between RWL and PCP was carried out to estimate the delay of RWL responding to PCP. Correlation analysis between RWL and NDVI was also carried out to estimate the delay in variation of RWL responding to NDVI. The months with the higher correlation coefficients between RWL and PCP and between RWL and NDVI were used as independent variables for model construction. A stepwise, multiple linear regression technique was applied to construct the model. Due to the lack of NDVI data for 1994, the available data were divided into in two periods: August 1981 to December 1993 and January 1995 to December 2000. The NDVI data for August 1981 to December of 1993 were used for model construction. Usually, the model-predicted RWL values obtained from the data set used for model construction are called simulated values. These simulated values were not real predicted values but were adjusted values obtained from the model construction process using a statistical best-fit curve approach. Therefore, for the purpose of validating the model, an independent NDVI data set of January 1995 to December 2000 was used to run the model and to compare the predicted and observed RWL values. The inundation area estimation method presented by Hamilton et al. (1996) was used to estimate the inundation area for the period of 1981 to 2000 (Equation 1). Further, Equation 1 and the RWL prediction model were used to predict the inundation area. The results were compared to the inundation area estimation using observed RWL data.

75

DDDDISCUSSION AND ISCUSSION AND ISCUSSION AND ISCUSSION AND RRRRESULTS ESULTS ESULTS ESULTS

River Water Level AnalysisRiver Water Level AnalysisRiver Water Level AnalysisRiver Water Level Analysis

Table 2 summarizes the occurrence of both maximum and minimum RWL for the period of 1981 to 2000. From the comparison of 20 annual cycles of RWL monitored at the LHS, it was observed that the occurrence of maximum RWL month shifted from April in higher flood years to July in lower flood years. This means that in high rainfall years, a higher rainfall amount in the wet season causes higher surface runoff water. On the other hand, the minimum RWL occurred mostly in December but in some years it occurred in either November or January. It was found that in the wet season, the highest RWL of 6.57 m occurred in April of 1988 and the lowest RWL of 4.2 m occurred in July of 1986. In the dry season, the highest RWL (3.30 m) occurred in December of 1992 and the lowest RWL (1.19 m) occurred in November of 1999.

Table 2. Maximum and minimum River Water Level (RWLmax and RWLmin) at Ladário Hydrological Station for the period of January 1981 to 2000. (Data Sources: Brazilian

Navy at Corumbá, MS)

Year RWLmax (m) Month RWLmin (m) Month 1981 5.43 5 1.95 12 1982 6.41 4 2.52 12 1983 5.35 5 2.39 12 1984 5.04 6 2.47 11 1985 6.04 4 1.56 13 1986 4.22 7 1.32 12 1987 4.94 6 1.40 11 1988 6.57 4 1.41 12 1989 6.10 5 1.87 13 1990 4.45 6 1.95 12 1991 5.40 5 2.37 12 1992 5.35 6 3.25 13 1993 5.13 5 1.42 12 1994 3.91 7 1.42 11 1995 6.50 4 2.02 12 1996 5.05 6 1.97 12 1997 5.66 5 2.13 13 1998 4.61 6 2.17 11 1999 4.58 6 1.19 11 2000 4.62 6 1.21 11

River Water Level Responses to PCP and NDVIRiver Water Level Responses to PCP and NDVIRiver Water Level Responses to PCP and NDVIRiver Water Level Responses to PCP and NDVI

76

Inter-annual and seasonal hydrological variability were analyzed by comparing the time series plots of RWL, PCP and NDVI from 1981 to 1993. Figures 3 to 6 show the time series plots of four years representing four extreme RWL cases: − A maximum RWL of 6.57m in the wet season occurred in April 1988 (Figure 3); − A minimum RWL of 1.19m in the dry season occurred in November 1999 (Figure 4); − A maximum RWL of 3.30m in the dry season occurred in December 1992 (Figure 5); − A minimum RWL of 4.20 in the wet season occurred in July 1986 (Figure 6).

0

0 , 1

0 , 2

0 , 3

0 , 4

0 , 5

0 , 6

0 , 7

1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 1 4

T im e P e r i o d (m o n th )

Wat

er L

evel

(10m

), ND

VI a

nd P

CP

(m)

W a te r L e ve l (1 0 )N D V I P C P (m m )

Figure 3. Time series plot of monthly water level at Ladário Hydrological Station regional averaged monthly NDVI and precipitation for the UPRB from 01/1988 to

02/1989

0

0 ,1

0 ,2

0 ,3

0 ,4

0 ,5

0 ,6

0 ,7

1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 1 4

T im e P e rio d (m o n th )

Wat

er L

evel

(10m

), ND

VI a

nd P

CP

(m)

W a te r L e ve l (1 0 )N D V I P C P (m m )

Figure 4.... Time series plot of monthly water level at Ladário Hydrological Station and

regional averaged monthly NDVI and precipitation of UPRB for the period of 01/1999 to 02/2000

77

0

0 , 1

0 , 2

0 , 3

0 , 4

0 , 5

0 , 6

0 , 7

1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 1 4

T i m e P e r i o d (m o n th )

Wat

er L

evel

(10m

), N

DVI

and

PCP

(m)

W a t e r L e ve l (1 0 )N D V I P C P (m m )

Figure 5. Time series plot of monthly water level at Ladário Hydrological Station and

regional averaged monthly NDVI and precipitation of UPRB for the period of 01/1992 to 02/1993

0

0 ,1

0 ,2

0 ,3

0 ,4

0 ,5

0 ,6

0 ,7

1 2 3 4 5 6 7 8 9 10 11 12 13 14

T im e P e rio d (m o n th )

Wat

er L

evel

(10m

), ND

VI a

nd P

CP (m

)

W a ter Leve l (10 )N D V I P C P (m m )

Figure 6. Time series plot of monthly water level at Ladário Hydrological Station and

regional averaged monthly NDVI and precipitation of UPRB from 01/1986 to 02/1987. It was observed that NDVI reached a value of around 0.6 mostly in November and remained above this value for a period as long as 8 to 9 months (November to June or July). This indicates that the vegetation surface reaches its maximum greenness right after the beginning of the wet season. Further increase of rainfall amount in the following months does not increase NDVI. NDVI maintains its maximum value for 2 to 3 months after the end of the wet season (May to July) and drops to its minimum at the end of dry season (mostly in September). In the Pantanal, with a smooth slope of

78

2.5cm/km, a higher soil moisture condition indicated by vegetation greenness generally persists for a longer period of time. The slow discharge of subsurface groundwater caused the RWL at LHS to decrease to 1.19 m in 1999, the driest year. Therefore, NDVI during the drying phase behaves similarly to RWL at its decreasing phase. Because free surface water absorb both visible and near infrared bands, such as AVHRR bands 1 and 2, values of NDVI close to zero (shallow water) or negative NDVI values (deep water) occur in the wet season. In the Pantanal, NDVI saturates at a low value of around 0.6, due to the larger flooded area and innumerous small shallow ponds. This limits the application of NDVI to infer a high RWL in the wet season. However, in peak rainfall seasons, monthly PCP data from December to March were mostly higher than 150 mm. Considering an evapotranspiration rate of 150 mm/month in the region, most of the excess water (PCP over 150 mm) contributes to surface runoff, which results in a higher than anticipated RWL. Therefore, in the wet season, PCP provides better high RWL estimates. In contrast, the dry season monthly PCP was lower than 50 mm/month or close to zero for several months, which does not provide a fair amount of low RWL values maintained by the ground water. Due to the fact that PCP provides better estimates in the wet season and NDVI provides better estimates in the dry season, both PCP and NDVI were used to develop the RWL prediction model. The six rainfall stations used for calculating PCP data are located in the upper part of the basin, approximately 400 km north of the RWL station. Therefore, a correlation process was carried out to find the RWL delay responding to the rainfall received in the upper part of the basin. The correlation process was also performed for NDVI to RWL and PCP to NDVI. Table 3 shows the correlation coefficient (r) of RWL to PCP and RWL to NDVI with a time lag of RWL from zero to the particular month showing the highest r-value. The highest r-value of 0.809 was obtained for RWL responding to 5 months of PCP and a highest r-value of 0.624 was obtained for RWL in responding to after 3 months. Table 3 also shows that the highest r-value of 0.579 was obtained for NDVI in responding to PCP with a time lag of 2 months.

79

Table 2. Occurrence of maximum and minimum River Water Level (RWLmax and RWLmin) at Ladário Hydrological Station for the period of January 1981 to 2000. (Data

Sources: Brazilian Navy at Corumbá, MS)

Year RWLmax (m) Month RWLmin (m) Month 1981 5.43 5 1.95 12 1982 6.41 4 2.52 12 1983 5.35 5 2.39 12 1984 5.04 6 2.47 11 1985 6.04 4 1.56 13 1986 4.22 7 1.32 12 1987 4.94 6 1.40 11 1988 6.57 4 1.41 12 1989 6.10 5 1.87 13 1990 4.45 6 1.95 12 1991 5.40 5 2.37 12 1992 5.35 6 3.25 13 1993 5.13 5 1.42 12 1994 3.91 7 1.42 11 1995 6.50 4 2.02 12 1996 5.05 6 1.97 12 1997 5.66 5 2.13 13 1998 4.61 6 2.17 11 1999 4.58 6 1.19 11 2000 4.62 6 1.21 11

Table 3. Correlation Coefficients (r) of PCP x RWL; NDVI x RWL and PCP x NDVI with 1

month time lag of either RWL or NDVI for each step of correlation.

Correlation Coefficient (r ) Time Lag (month) RWL x NDVI RWL x PCP NDVI x PCP

Parameter RWL RWL NDVI 0 0,083 -0,553 0,277 1 0,364 -0,200 0,494 2 0,572 0,189 0,579 3 0,624 0,543 0,531 4 0,500 0,769 - 5 - 0,807 - 6 - 0,652 -

80

River Water Level PredictionRiver Water Level PredictionRiver Water Level PredictionRiver Water Level Prediction

The parameters with a correlation coefficient value higher than 0.5 were considered suitable for model construction. By examining the correlation coefficients presented in Table 3, three correlation coefficients of PCP and three of NDVI (Table 3) were used as independent variables. By applying the Stepwise Multiple Linear Regression technique, the model with the highest R2 was selected as a candidate model, which is presented in the following equation: Equation 3. RWLt = 1.19 + 2.89PCP t-6 + 3.30PCP t-5 + 6.05PCP t-4 + 1.59NDVI t-4 (3) Where: RWLt = River water Level of month t (m); PCP = monthly precipitation (m); NDVI = monthly NDVI (dimensionless); Subscripts t-6, t-5 and t-4 represent 6, 5 and 4 months before the month t respectively. The model was constructed using RWL, PCP and NDVI data for the period of 1981 to 1993. The model had a R2 value of 0.79 with MSE of 0.36 and SD of 0.60 (P < 0.002). An averaged absolute error of 15.22% was obtained from model simulation. Figure 7 shows the comparison of observed and simulated RWL for the period of November 1981 to April 1994. The model simulated the occurrences and magnitudes of maximum and minimum RWL relatively well, but failed to simulate the occurrence of peak RWL in 3 years. The peaks of RWL in 1982, 1985 and 1988 occurred in April as opposed to the simulated peaks that occurred in July. Based on the RWL data, it was observed that these 3 years had maximum RWL over 6 m. This indicates that a larger volume of surface runoff water caused by intense summer precipitation may contribute significantly to the RWL peak value 1 to 3 months earlier than in normal rainfall years. The model also failed to simulate peak RWL in 1984, 1989 and 1993. Instead of a peak in July, the simulated RWL dropped and then rose again in the following month. A lower NDVI caused by a larger flood area may have contributed to this error. Therefore, further research on rainfall patterns should be conducted in order to detect the anticipation of peak RWL in high rainfall years and correct the abnormal drop of RWL caused by a low NDVI in the peak rainy season.

81

0

1

2

3

4

5

6

7

nov/

81

mar

/8

jul/8

2

nov/

82

mar

/8

jul/8

3

nov/

83

mar

/8

jul/8

4

nov/

84

mar

/8

jul/8

5

nov/

85

mar

/8

jul/8

6

nov/

86

mar

/8

jul/8

7

nov/

87

mar

/8

jul/8

8

nov/

88

mar

/8

jul/8

9

nov/

89

mar

/9

jul/9

0

nov/

90

mar

/9

jul/9

1

nov/

91

mar

/9

jul/9

2

nov/

92

mar

/9

jul/9

3

nov/

93

mar

/9

July/1981 - April/1994 (month)

Riv

er W

ater

Lev

el (m

)

Observedsimulated

Figure 7. Comparison of observed and simulated river water level for the period of July,

1981 to April, 1994 at Ladário Hydrological Station.

RWL Model ValidationRWL Model ValidationRWL Model ValidationRWL Model Validation The independent data set of 1995 to 2000 was used for RWL prediction model validation. The results of model validation showed an average absolute error of 14.45%. This indicates that the RWL model was relatively stable since this value was within the simulated error (15.22%). Figure 8 shows the comparison of observed and predicted RWL for the period of May 1995 to September 2000. There were 39 out of 65 months with absolute error within 10%, and 19 months with absolute error between 10 to 30%. The remaining 7 months had error higher than 30%. In 1999, a lowest RWL of 1.19 m was recorded due to minimal rainfall from May to August and low rainfall recorded in September (40 mm) and October (80 mm). This demonstrates that the model was not able to predict this extreme drought event

82

0

1

2

3

4

5

6

7

mai/95

jul/95

set/9

5no

v/95

jan/96

mar/96

mai/96

jul/96

set/9

6no

v/96

jan/97

mar/97

mai/97

jul/97

set/9

7no

v/97

jan/98

mar/98

mai/98

jul/98

set/9

8no

v/98

jan/99

mar/99

mai/99

jul/99

set/9

9no

v/99

jan/00

mar/00

mai/00

jul/00

set/0

0

May/1995 - September/2000 (month)

Riv

er W

ater

Lev

el (m

)ObservedPredicted

Figure 8. Comparison of observed and predicted river water level for the period of May, 1995 to September, 2000 at Ladário Hydrological Station.

The RWL model tends to predict the yearly maximum RWL with one to three months delay depending on the maximum RWL value. This error is inevitable since the model was based on a statistical linear regression approach, which selected the best correlation of RWL to PCP and RWL to NDVI at fixed months. Table 2 demonstrates that the maximum RWL month was closely related to the increase of RWL. Galdino and Clarke (1997) have also observed this relationship. Therefore, it is possible to improve the prediction of the maximum RWL occurrence month using RWL data. The occurrence of annual maximum RWL month as a function of annual maximum RWL value was then included to improve the model prediction under higher maximum RWL year, which is presented in Equation 4 as following: Equation 4. Mmax = integer (11.4143 – 1.14167RWL) (4) Where: Mmax = the month in which annual maximum RWL (month) occurs; RWL = annual maximum River Water Level (m) Equation 4 was constructed using maximum RWL data for the period of 1981 to 2000. The model had a R2 value of 0.8674 with MSE of 0.1237 and SD of 0.3518 (P < 0.00001). The calculated Mmax value was rounded up to an integer value to obtain the maximum RWL occurrence month. Table 4 shows the comparison of the observed and predicted annual maximum RWL month at LHS. The results show that for 12 out of 20

83

years, the equation correctly predicted the annual maximum RWL month. It must be noted that for the remaining 8 years, the equation’s estimation was one month early, and none of the estimated maximum RWL months was later than the observed month. Therefore, the equation can be applied safely to predict the occurrence of annual peak RWL months. Considering the complexity of water flow between sub basins in this considerably flat river basin, the combined use of the RWL prediction model (equation 3) and the maximum RWL month occurrence model (equation 4) works sufficiently for predicting the maximum RWL month. This simple approach may provide a useful tool to predict RWL and, hence, prevent flood damage in high RWL periods as well as control river transportation traffic in order to prevent riverbank erosion during low RWL periods.

Table 4. Comparison of observed and predicted annual maximum RWL month as a function of annual maximum RWL at Ladário Hydrological Station

Year RWLmax (m) Month obs Month pred Error

01-12(1981) 5,43 5 5 0 01-12(1982) 6,41 4 4 0 01-12(1983) 5,35 5 5 0 01-12(1984) 5,04 6 5 -1 01-12(1985) 6,04 4 4 0 01-12(1986) 4,22 7 6 -1 01-12(1987) 4,94 6 5 -1 01-12(1988) 6,57 4 4 0 01-12(1989) 6,1 5 4 -1 01-12(1990) 4,45 6 6 0 01-12(1991) 5,4 5 5 0 01-12(1992) 5,35 6 5 -1 01-12(1993) 5,13 5 5 0 01-12(1994) 3,91 7 6 -1 01-12(1995) 6,5 4 4 0 01-12(1996) 5,05 6 5 -1 01-12(1997) 5,66 5 4 -1 01-12(1998) 4,61 6 6 0 01-12(1999) 4,58 6 6 0 01-12(2000) 4,62 6 6 0

Inundation Area PredictionInundation Area PredictionInundation Area PredictionInundation Area Prediction

Hamilton et al. (1996) presented the inundation area estimation model (equation 1), which estimated well the inundation area using two month averaged RWL at LHS after

84

the inundation had occurred. The RWL prediction model (equation 3), using PCP and NDVI, can be applied to predict RWL three months before the event occurs. Therefore, with the combined use of these two models, the inundation area can be predicted one month before it occurs. Figure 9 shows the comparison of the inundation area calculated from observed and predicted RWL for the period of 1981 to 2000. An averaged absolute error of 23.65% was obtained. The error was acceptable since it was a sum of two estimation errors, that of the RWL prediction model (14.45%) and the other from the inundation area estimation equation (12%).

0

20000

40000

60000

80000

100000

120000

nov/

81

jun/

82

jan/

83

ago/

83

mar

/8

out/8

4

mai

/85

dez/

85

jul/8

6

fev/

87

set/8

7

abr/8

8

nov/

88

jun/

89

jan/

90

ago/

90

mar

/9

out/9

1

mai

/92

dez/

92

jul/9

3

fev/

94

set/9

4

abr/9

5

nov/

95

jun/

96

jan/

97

ago/

97

mar

/9

out/9

8

mai

/99

dez/

99

jul/0

0

Time Period (month)

Pant

anal

Inun

datio

n Ar

ea (k

m²)

Inundation Area ObservedInundation Area Predicted

Figure 9. Comparison of Pantanal inundation area estimation by Hamilton Method for the period of November 1981 to July 2000, using observed and predicted RWL at Ladário

Hydrological Station.

CCCCONCLUSION AND ONCLUSION AND ONCLUSION AND ONCLUSION AND SSSSUGGESTIONSUGGESTIONSUGGESTIONSUGGESTIONS This study presented a RWL prediction model using rainfall and NDVI data and applying statistical regression approaches. The RWL prediction model had an averaged absolute error lower than 15%, which is considered acceptable. An annual maximum RWL occurrence month model was constructed to improve the prediction of maximum RWL occurrence month. This study also demonstrated that the inundation area extent could be predicted one month before its occurrence through a combined use of RWL prediction model (Liu et al., 2002) and the inundation area estimation model proposed by Hamilton et al. (1996). For further study, it is suggested that several sub basin models should be constructed to evaluate the consistency of the method as well as to improve the prediction accuracy. Moreover, a high resolution Digital Elevation Model is needed to better understand the detailed interflows between sub-basins caused by

85

spatial variation of rainfall. In order to better understand this complicated hydrology, a high accuracy digital elevation model and a high precision rainfall forecasting system such as a radar system are needed to fulfill a real time flood advancing prediction. Before an adequate hydrological simulation model is developed to predict Pantanal inundation on an operational scale, the statistical approach presented here may provide a useful tool to predict RWL and, hence, prevent flood damage in high RWL period as well as control river transportation traffic in order to prevent riverbank erosion during low RWL period. AAAACKNOWLEDGEMENTSCKNOWLEDGEMENTSCKNOWLEDGEMENTSCKNOWLEDGEMENTS This study is in part research under taken in association with the Ducks Unlimited Pantanal-GIS Pilot Project financed by the US Forest Service International Program. The authors wish to thank the Brazilian Marine Corp at Corumbá and National Water Agency (ANA, Agencia Nacional de Aqua) for providing the river water level and precipitation data. The authors also wish to thank the Distributed Active Archive Center (Code 902.2) at the Goddard Space Flight Center, Greenbelt, MD, 20771, for producing the data in their present form and distributing them. The original data products were produced under NOAA/NASA Pathfinder program, by a processing team headed by Ms. Mary James of the Goddard Global Change Data Center; and the science algorithms were established by the AVHRR Land Science Working Group, chaired by Dr. John Townshend of the University of Maryland. Goddard's contributions to these activities were sponsored by NASA's Mission to Planet Earth program. RRRREFERENCESEFERENCESEFERENCESEFERENCES DNMET, 1992. Normais Climatológicas (1961-1990). Departamento Nacional de Meteorologia, Ministério da Agricultura e Reforma Agrária, Brasília, Brasil. 83pp. Eidenshink, J.C. and Faundeen, J.L. 1997. The 1-km AVHRR global land data set: first stages in implementation. International Journal of Remote Sensing. 51:39-56. Fleig, A. J., Heath, D. F., Klenk, K. F., Oslik, N., Lee, K. D., Park, H., Bartia, P. K., and Bartia, D., 1983, User’s Guide for the Solar Backscattered Ultraviolet (SBUV) and the Total Ozone Main Spectrometer (TOMS) RUT-S and RUT-T Data Set: October 31, 1978 to November 1980. NASA Reference Publication Nº 1112, 51p. Galdino, S. and Clarke, R.T. 1997. Probabilidade de ocorrência de cheia no Rio Paraguai, em Ladário. MS – Pantanal. 1997. Circulação Técnica N° 23, EMBRAPA-CPAP, Corumbá, MS. 58p.

86

Gordon, H.R., Brown, J.W. and Evans, R.H. 1988. Exact Rayleigh scattering calculations for use with the Nimbus-7 coastal zone colour scanner. Applied Optics, 27:2111-2122. Hamilton, S.K., Sippel, S. J. and Melack, J. M. 1996. Inundation patterns in the Pantanal wetland of South America determined from passive microwave remote sensing. Arch. Hydrobiology. 137: 1-23. Kidwell, K. B. 1995. NOAA Polar Orbiter Data Users Guide. Satellite Data Service Division, NESDIS/NOAA, Washington D.C., USA. 435p. Liu, W.T.H., Massambani, O. and Nobre, C. 1994. Satellite Vegetation response to drought in Brazil. International Journal Climatology, 14:343-354. Liu, W.T.H., Ayres, F.M., Salles, A.T. and Padovani, C. 2002. Upper Paraguay River water level prediction using NOAA AVHRR NDVI. Proceeding of the 1º International Symposium “Recent Advances in Quantitative Remote Sensing”, Valencia, Spain, 2002. Mittermeier, R.A., Camara, I.G., 1990. Padua, M.T.J., and Blanck, J. Conservation in the Pantanal of Brazil. Oryx 24: 103-112. Rao, C.R.N. and J., 1995. Chen Inter-satellite calibration linkages for the visible and near-infrared channels of the Advanced Very High Resolution Radiometer on the NOAA-7, -9, and -11 spacecrafts. International Journal of Remote Sensing. 16, 1931-1942, 1995. Silva, J. dos S. V. da; Abdon, M. 1998. Delimitação do Pantanal Brasileiro e suas sub-regiões. Pesquisa Agropecuária Brasileira, Brasília, v. 33, Número Especial, p. 1703 – 1711. Welcomme, R.L. 1995. River fisheries.-Food and Agriculture Organization of the United Nations, Rome, Italy: FAO Fisheries Technical Paper N° 262.

87

3. FURTHER CONSERVATION EFFORTS IN THE UPRB

3333AAAA.... BBBBOLIVIAOLIVIAOLIVIAOLIVIA:::: RRRRESULTS OF THE ESULTS OF THE ESULTS OF THE ESULTS OF THE PPPPANTANAL ANTANAL ANTANAL ANTANAL PPPPROGRAM IN ROGRAM IN ROGRAM IN ROGRAM IN BBBBOLIVIA OLIVIA OLIVIA OLIVIA (WWF(WWF(WWF(WWF BBBBOLIVIAOLIVIAOLIVIAOLIVIA)))) HEIDI RESNIKOWSKI1 AND PAMELA REBOLLEDO2

RRRRESULTSESULTSESULTSESULTS − Copies of the UPRB GIS Database Development Project Phase I final report have been distributed to the Forestry Superintendence, National Protected Areas Service, Agrarian Superintendence, Noel Kempff Mercado Natural History Museum, Municipalities, Cattle Federation and private companies that are interested in investing in the Puerto Busch project. − The flooded area grids and final report have been used as reference in the Management Plan of the Otuquis National Park. − The project results will be used as reference to complete a project proposal to be presented to the International Tropical Timber Organization (ITTO), called “Fire Fight – Sur Project,” developed by the World Conservation Union (IUCN) in collaboration with WWF International. − Three papers have been presented:

a) Resnikowski H. (2002). “Dynamics of Change in the Landscape of the Pantanal Otuquis National Park Caused by Fires, Biomass Changes, Roads and Fluctuations of the Flood Levels Between 1988, 1997, 1998 and 1999”. Paper presented at the X International Symposium SELPER (Society of Latin American Experts in Remote Sensing), Cochabamba, Bolivia. November 2002.

b) Resnikowski H. (2003). “Upper Paraguay River Basin GIS Database – Pilot Project I” Invited to give a presentation at the X Latin-American GIS Users Conference. Santa Cruz, Bolivia. September 10.

c) Resnikowski H. (2003). “Fires in the Bolivian Pantanal”. Presentation provided as a consultant to the WWF Bolivia’s Pantanal Program at the Workshop Fires in the Bolivian Pantanal. Puerto Suárez, Bolivia. January 16.

− Flooded area maps and digitized roads were presented in the Forum called Puerto Busch.

1 WWF Bolivia, Santa Cruz, Bolivia; hresnikowski@wwfbolivia.org 2 WWF Bolivia, Santa Cruz, Bolivia; prebolledo@wwfbolivia.org

88

− The results will be distributed to the members of the Pantanal Coordinator (“Coordinadora Pantanal”) as reference information. − Since the project was finished, the compilation of metadata is a common process in the GIS work developed in WWF Bolivia. − A technical proposal to expand the flooded area map was sent to SWS Ramsar Support Framework Funding. The same methodology was suggested.

89

3333BBBB.... BBBBRAZILRAZILRAZILRAZIL:::: PPPPLANNING AND LANNING AND LANNING AND LANNING AND MMMMANAGEMENT OF THE ANAGEMENT OF THE ANAGEMENT OF THE ANAGEMENT OF THE “N“N“N“NASCENTES DO ASCENTES DO ASCENTES DO ASCENTES DO RRRRIO IO IO IO TTTTAQUARIAQUARIAQUARIAQUARI”””” SSSSTATE TATE TATE TATE PPPPARK ARK ARK ARK SYLVIA TORRECILHA3, WILLIAM T. LIU4 AND FÁBIO M. AYRES5

IIIINTRODUCNTRODUCNTRODUCNTRODUCTIONTIONTIONTION As part of the strategy of the state government of Mato Grosso do Sul to develop a system for conservation units through the “Secretaria Estadual de Meio Ambiente” (State Ministry of the Environment) and the “Instituto de Meio Ambiente Pantanal” (Institute of the Environment for the Pantanal), studies were conducted and a management plan was implemented to guarantee biodiversity conservation in the state’s most significant ecosystems, especially through the establishment of new conservation units. In order to select locations representative of the environmental (geological, edaphic, climatic and biological) and socioeconomic diversity of the state’s territory, studies in the area of the river Taquari headwaters were prioritized, in the municipalities of Costa Rica and Alcinópolis. As a result, the state park “Nascentes do Rio Taquari” was created, which constitutes the first state conservation unit established in the section of the Upper Paraguay River Basin included in Mato Grosso do Sul. This park is strategically located for the implementation of the corridor of biodiversity Cerrado-Pantanal, in one of the core zones of the Pantanal Biosphere Reserve (Figure 1). PPPPURPOSEURPOSEURPOSEURPOSE The purpose of this study was to characterize and analyze the structure of the study area’s landscape (including physical, biological and ecological aspects) for the environmental zoning of the state park “Nascentes do Rio Taquari,” using geographic information systems (GIS) and remote sensing techniques for the development of a formal management plan. The conservation parameters considered in this study include biodiversity, recreation, environmental education, ecotourism, watershed protection, scientific research, and monitoring of environmental quality. The study area was subdivided taking into account physical and natural characteristics, but also considering cultural, recreational and research factors. The management plan of the park was based on the guidelines established by IBAMA and official documents such as the law of number 9,985/2000 and its regulating decree of number 4,340 from August 22, 2002. Additionally, land surveys, data from the “Conservation Plan of the Upper Paraguay River Basin” (PCBAP), interpretation of a Lansat-7 image, botanical characterization conducted by the “Ecological Corridor Cerrado-Pantanal” project and field data were also considered. Workshops were conducted in Campo Grande (August 21, 2003), Costa

3 Universidade Católica Dom Bosco, Campo Grande, MS, Brazil; sylviat@ucdb.br 4 Universidade Católica Dom Bosco, Campo Grande, MS, Brazil; will@ucdb.br 5 OIKOS (Cooperativa de Trabalho Sócio-ambiental), Campo Grande, MS, Brazil; fabio_oikos@pop.com.br

90

Rica (September 18, 2003) and Alcinópolis (September 19, 2003) to discuss the management planning of the park with all the interested stakeholders.

Figure 1. Biosphere reserve of the Pantanal (text in Portuguese).

SSSSTUDY TUDY TUDY TUDY AAAAREAREAREAREA The State Park “Nascentes do Rio Taquari,” created by the decree of number 9,662 on October 19, 1999, encompasses 30,619 hectares (ha) located in the state of Mato Grosso do Sul between the coordinates 17º 59’-18º 15’ S and 53º10’-53º 26’ W, with 26,850 ha in the municipality of Alcinópolis and 3.769 ha in the municipality of Costa Rica (Figure 2). The park is 50 kilometers (km) from the urban center of Costa Rica and 60 km from that of Alcinópolis. The park’s buffer zone not only includes Costa Rica and

91

Alcinópolis, but also, to the north, part of the municipality of Alto Taquari (from the state of Mato Grosso).

Figure 2. State Park “Nascentes do Taquari” (text in Portuguese).

MMMMETHODOLOGYETHODOLOGYETHODOLOGYETHODOLOGY Data were digitized, processed and analyzed using the programs ArcView 3.2a (with extensions Image Analysis and Spatial Analysis) and Spring 3.6.03. To improve the visualization and delineation of features in the vectorization process using raster data, contrasting techniques were employed. Supervised and unsupervised classification procedures were applied to characterize and quantify land cover. The classification results were integrated to other GIS data (obtained from topographic maps and field observations) to develop the environmental zoning of the study area. The combination of these datasets allowed an integrated analysis and the assessment of the applicability of GIS for future actions in the evaluation and monitoring of the park’s

92

natural resources. In addition to the environmental zoning, the generated maps include hydrography, hypsometry, slope, vegetation and land use. RRRRESULTSESULTSESULTSESULTS The buffer zone of the park encompasses the headwaters of three of the streams forming the Taquari River, namely the Furnas, Mutum and Engano, located near the borders with the states of Mato Grosso and Goiás. The park is close to the headwaters of the Araguaia River and National Park “Emas,” and includes scarps of the western border of the “Planalto Central Brasileiro” (central Brazilian plateau region). Also included in the park are “cuesta” formations. The park presents high topographic relief; elevations (above sea level) vary from 890 meters (m) in the Chapadão dos Baús, to 390 m, at the headwater of Ribeirão Engano, which is a tributary of the Taquari River in the pre-Pantanal depression, with Canyon formations and abundant geological monuments. The vegetation in the park area is represented by physiognomic formations characteristic of the Cerrado biome: submontane seasonal semideciduous forest and “cerradão” (Cerrado forest) are found along cuestas, which also present “cerrado sensu stricto” (Cerrado savanna) and “campo sujo” of Cerrado (Cerrado grassland with few scattered shrubs and trees) along scarps in the low areas of the plain. In the vegetation map (Figure 3), similar physiognomic formations were grouped in the same category, which resulted in three classes: (1) seasonal forest with cerradão (36.57% of the total area of the park); (2) cerrado sensu strictu with campo sujo of Cerrado and “campo rupestre” (grassland) (45,74%); and (3) pasture with Brachiaria predominance with small patches of agricultural fields (17,70%). This vegetation map is the product of a supervised classification of Landsat imagery, which followed a preliminary unsupervised classification procedure. The main result of the integration of the several GIS datasets generated is the map of environmental zoning in which six classes are distinguished (Figure 3). Flexible in use are: (1) zone of especial use (0.9%), (2) zone of recuperation (35.7%), (3) zone of extensive use (2.90%), and (4) historical/cultural zone (0.48%); while those more restrictive in use include: (5) intangible zone (13,12%), and (6) pristine zone (46,90%). Considering the intensive agricultural occupation surrounding the park, a buffer zone subdivided in two sections was proposed (Figure 4). It is suggested that the buffer zone section that is adjacent to the park (2 km wide) be restricted to agroforestry systems, while the outer section of the buffer zone (1 km wide) may include pastures.

93

Figure 3. Vegetation and environmental zoning of the park (text in Portuguese)

94

Figure 4. Buffer zone subdivision (text in Portuguese)

AAAACKNOWLEDGEMENTSCKNOWLEDGEMENTSCKNOWLEDGEMENTSCKNOWLEDGEMENTS This study was conducted in association with the “Instituto de Meio Ambiente Pantanal” (institute of environment for the Pantanal), the municipalities of Costa Rica and Alcinópolis. Funding was provided by the project GEF/Alto Paraguai (PNUMA/OEA).

95

3C. BBBBRAZILRAZILRAZILRAZIL:::: FFFFLOOD LOOD LOOD LOOD MMMMONITORING IN THE ONITORING IN THE ONITORING IN THE ONITORING IN THE PPPPANTANAL ANTANAL ANTANAL ANTANAL WWWWETLANDETLANDETLANDETLAND1111 CARLOS R. PADOVANI2, JORGE SAMBRANA3, MARIANE L. DA CRUZ4 AND MANUEL FERREIRA5

IIIINTRODUCTIONNTRODUCTIONNTRODUCTIONNTRODUCTION The flood pulse is the driving force of ecological processes and economic activities in the Pantanal wetland. Wildlife distribution, including that of caimans, marsh deer, capybaras and waterfowl species, is strongly dependent on the occurrence and frequency of flooding in wetland areas. The floodplains in the Pantanal play an important role in water purification, filtering the pollutants coming from the highlands, which increase every year. These floodplains constitute habitat for fish communities and encompass significant food and shelter resources that play an important nursery role for many other species. Embrapa Pantanal has mapped flooded areas in a variety of Pantanal locations using different satellite imagery (Padovani et al., 2003; Padovani et al., 2004), but a long-term ecological project such as the Brazilian “Programa Ecológico de Longa Duração” (Ecological Long-Term Program) (PELD) requires a continuous and long-term monitoring program. The main goals of this flood monitoring program are to: (a) quantify and map the flood dynamics of the Paraguay River floodplain; (b) establish the relationship between water level, flooded area and stress through the spatial analysis of flooding, wildlife distribution, water chemistry and other biotic and physical elements of the system; and (c) perform an integrated analysis of how the inter-annual variations in flooding relate to the abundance and distribution of terrestrial and aquatic organisms. MMMMETHODSETHODSETHODSETHODS The reflectance bands, NDVI, Enhanced Vegetation Index (EVI) (with 250 m of spatial resolution) and surface temperature band (with 1 km of spatial resolution) of the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor have been processed for flooded area quantification from February 2001 to the present. A MODIS imagery database that is updated monthly has been created, and an automatic procedure for image processing has been developed. Paraguay River level data from “Ladário NAVI” harbor was used to characterize flood events (Figure 1). The study area encompasses 1 Poster presentation. International Symposia on Long-Term Ecological Projects, July 2004. Manaus, Amazonas, Brazil 2 Empresa Brasileira de Pesquisa Agropecuária (Embrapa) Pantanal, Corumbá, MS, Brazil; guara@cpap.embrapa.br 3 Universidade Federal de Mato Grosso do Sul, Corumbá, MS, Brazil 4 Universidade Federal de Mato Grosso do Sul, Corumbá, MS, Brazil 5 Universidade Federal de Goiás, GO, Brazil

96

41.53 km2 that correspond to the sub-regions of Corixo Grande (CORI), Paraguay (PARA) and Nabileque (NABI) as defined by Hamilton (1996).

Figure 1. River level data from Ladário.

RRRRESULTS AND ESULTS AND ESULTS AND ESULTS AND DDDDISCUSSIONISCUSSIONISCUSSIONISCUSSION The results of a classification procedure to quantify the area of lowest and highest flood events in the studied period are presented in Figure 2.

Image difference

October 2001Dry period

June 2003High-flood period

Figure 2. River level data from Ladário Flood classification for the dry and high-flood periods in the area of influence of the Paraguay River. The difference image shows the

areas where the floods increase (pink) or decrease (gold).

97

The maximum flooded area, which occurred in June 2003, was 11.1 km2; while the minimum flooded area, which occurred in October 2001, was 8.1 km2. This difference of 3 km2 represents the total range in variation of flooded area for the period of 2001 to 2003. In some regions, unlike what was expected, a decrease in flooded area occurred during the dry-to-flooded season. It is well known that because of the extensive length (~680 km), sinuosity and “sponge” effect of the Paraguay River, the flood wave takes 3 to 4 months to reach the city of Ladário, where the river level measurement station used in this study is located. Further analysis of rain data integrated to river water level data will be conducted to better understand the flood dynamics of the region. As expected, the flood dynamics of the Paraguay River region are influenced by more than one water source at different times of the year. With this integrated continuous long-term monitoring program, the contribution of each factor influencing flood dynamics in the Pantanal will be better understood. AAAACKNOWLEDGEMENTSCKNOWLEDGEMENTSCKNOWLEDGEMENTSCKNOWLEDGEMENTS To Dr. Rob Longman from the ALTERRA institute, Netherlands, whose project has partially supported this activity. To Gerard Groenveld for the help with GIS. RRRREFERENCESEFERENCESEFERENCESEFERENCES Hamilton, S. K, S.J., Sippel and Melack. 1996. Inundation patterns in the Pantanal wetland of South America determined from passive microwave remote sensing. Archives of Hydrobiology 137 (1): 1-23. Padovani, C. R., , Moraes, A. S., Resende, E.K. 2003. Banco de dados geográfico da Estrada-Parque-Pantanal, MS: enfoque na atividade de pesca esportiva. Corumbá: Embrapa Pantanal (Documentos / Embrapa Pantanal ISSN 1517 – 1973; 44). Padovani, C.R., Assine, M., Vieira, L.M. 2004. Inundações no leque aluvial do rio Taquari. Unpublished technical report. Embrapa Pantanal.

98

3333DDDD.... BBBBRAZILRAZILRAZILRAZIL:::: RRRREEEE----DRAWING THE DRAWING THE DRAWING THE DRAWING THE PPPPANTANAL WETLAND DELIANTANAL WETLAND DELIANTANAL WETLAND DELIANTANAL WETLAND DELINEATION IN NEATION IN NEATION IN NEATION IN BBBBRAZILRAZILRAZILRAZIL,,,, BBBBOLIVIA AND OLIVIA AND OLIVIA AND OLIVIA AND PPPPARAGUAYARAGUAYARAGUAYARAGUAY6666

CARLOS R. PADOVANI7, MARIANE L. DA CRUZ8 AND JORGE SAMBRANA9

PPPPROJECT ROJECT ROJECT ROJECT SSSSUMMARYUMMARYUMMARYUMMARY Wetland delineation is an important issue in ecological and hydrological studies, and conservation policymaking. Several authors have delineated the Pantanal wetland. Silva and Abdon (1998) proposed limits for the Brazilian Pantanal based on previous work, Landsat imagery, and vegetation, soils and topographic maps at the 1:250,000 scale. Hamilton et al. (1996) delineated the entire Pantanal (Bolivia, Brazil and Paraguay) using passive microwave thermal bands (with a spatial resolution of 25 km). Currently, new remote sensing products are available and actively employed in wetland delineation. We propose to review the Pantanal limits for Bolivia, Brazil and Paraguay using MODIS products such as NDVI, Enhanced Vegetation Index (EVI) and reflectance bands (with 250 m of spatial resolution), and the surface temperature product (with 1 km of spatial resolution). MODIS products have the advantage of being composite images free of clouds, which permits maximum flood events to be quantified periodically. The Shuttle Radar Topographic Mission (SRTM) is also a very useful sensor for the delineation of floodable areas because of the detailed relief data it provides. Imagery from the new satellite China Brazil Earth Resources Satellite 2 (CBERS-2), which has a higher spatial resolution (20 m), has also been assessed for use in the quantification of flooded areas. The first results obtained applying MODIS, SRTM and CBERS-2 data for wetland delineation in the study area show their great potential for improving previously established Pantanal limits. In future research efforts, flood analysis previously conducted by Embrapa Pantanal and its partners using Landsat and SAC-C imagery will also be employed. The following figures (Figures 1 to 6) illustrate the data and preliminary procedures in use in this project for the re-delineation of the Pantanal.

6 International Symposia on Long-Term Ecological Projects, July 2004. Manaus, Amazonas, Brazil. 7 Empresa Brasileira de Pesquisa Agropecuária (Embrapa) Pantanal, Corumbá, MS, Brazil; guara@cpap.embrapa.br 8 Universidade Federal de Mato Grosso do Sul, Corumbá, MS, Brazil 9 Universidade Federal de Mato Grosso do Sul, Corumbá, MS, Brazil

99

-52.47-13.89

-64.41-26.20

-52.47-13.89

-64.41-26.20

Figure 1. SRTM digital elevation model of the UPRB (Bolivia, Brazil and Paraguay)

-53.30-14.00

-61.00-22.00

-53.30-14.00

-61.00-22.00

Figure 2. SRTM digital elevation model of the UPRB in Brazil

100

-56.85-15.45

-59.40-17.10

-56.85-15.45

-59.40-17.10

Figure 3. SRTM image of the northwestern portion of the Pantanal showing topographic data in brown tones (highest height is 200 m). Indicated are the delineation of Hamilton

et al. (1996) (in blue), and that of Silva et al. (1998) (in red).

-56.85-15.45

-59.40-17.10

-56.85-15.45

-59.40-17.10

Figure 4. SRTM image of the northwestern portion of the Pantanal combined with a

Landsat image at 30 % of transparency.

101

-56.85-15.45

-59.40-17.10

-56.85-15.45

-59.40-17.10

Figure 5. SRTM image of the northwestern portion of the Pantanal combined with the

reflectance MODIS image at 30% transparency.

-60.30-22.30

-54.60-15.45

N

-60.30-22.30

-54.60-15.45

-60.30-22.30

-54.60-15.45

NN

Figure 6. SRTM image of the Pantanal (Bolivia, Brazil and Paraguay) combined with the reflectance MODIS image at 30% transparency.

102

AAAACKNOWLEDGEMENTSCKNOWLEDGEMENTSCKNOWLEDGEMENTSCKNOWLEDGEMENTS To Dr. Rob Longman from the Research Institute for our Green Living Environment (ALTERRA), Netherlands, whose project has partially supported this activity. To Gerard Groenveld for the help with GIS. RRRREFERENCESEFERENCESEFERENCESEFERENCES Hamilton, S. K, Sippel, S.J. and Melack. 1996. Inundation patterns in the Pantanal wetland of South America determined from passive microwave remote sensing. Archives of Hydrobiology 137 (1): 1-23. Silva, J.S.V. and Abdon, M. 1998. Delimitação do Pantanal Brasileiro e suas sub-regiões. Pesquisa Agropecuária Brasileira 33 (Número Especial): 1703–1711.

103

3E. BBBBRAZILRAZILRAZILRAZIL:::: FFFFIRE MONIIRE MONIIRE MONIIRE MONITORING AND ANALYSIS TORING AND ANALYSIS TORING AND ANALYSIS TORING AND ANALYSIS FOR THE FOR THE FOR THE FOR THE BBBBRAZILIAN RAZILIAN RAZILIAN RAZILIAN PPPPANTANALANTANALANTANALANTANAL10101010 CARLOS ROBERTO PADOVANI11, MARIANE L. DA CRUZ12 AND JORGE SAMBRANA13

IIIINTRODUCTIONNTRODUCTIONNTRODUCTIONNTRODUCTION The use of fire for pasture management in rural areas is a common practice in the Brazilian Pantanal. Many areas, in the dry period, present a significant amount of dry material, which represents available fuel for the ignition and spread of fires. Generally, farmers intend to use fire in small areas only, but sometimes these small fires accidentally spread and, as a result, affect large areas. Uncontrolled fires can be disastrous for wildlife and vegetation in terrestrial and aquatic environments. Continuous fire monitoring and the development of strategies for fire prevention and control are new objectives for the management plan for PELD site 2, in the Pantanal. MMMMETHODSETHODSETHODSETHODS Data on occurrence of fire were obtained from the “Instituto Nacional de Pesquisas Espaciais” (National Institute for Spatial Research) (INPE), Brazil. These data were derived from AVHRR/NOAA imagery as focal points of fire occurrence. Only data from July to October of each year were considered in this study because this is the period of greatest incidence of fire. The imagery was subset using the limits of the Brazilian Pantanal as established by Silva and Abdon (1998). The present study was conducted to determine fire density (number of fires in relation to area) considering political units (municipalities and states) and vegetation types. RRRRESULTS AND ESULTS AND ESULTS AND ESULTS AND DDDDISCUSSIONISCUSSIONISCUSSIONISCUSSION The spatial distribution of fires in the Pantanal is better explained with relative measurements such as fire density (Figure 1). The highest fire density occurs in the southwestern section of the Pantanal, which is, surprisingly, one of the most frequently flooded areas in the region. Fire in the Pantanal is an environmental problem that must be considered in a policymaking context. Because the majority of fires are anthropogenic in origin, the establishment of preventive and controlling measures

10 International Symposia on Long-Term Ecological Projects, July 2004. Manaus, Amazonas, Brazil. 11 Empresa Brasileira de Pesquisa Agropecuária (Embrapa) Pantanal, Corumbá, MS, Brazil; guara@cpap.embrapa.br 12 Universidade Federal de Mato Grosso do Sul, Corumbá, MS, Brazil 13 Universidade Federal de Mato Grosso do Sul, Corumbá, MS, Brazil

104

(including the implementation of specialized laws and regulations) implies actions in policy decision-making at the municipality and state levels (Figures 2 and 3).

Figure 1. Fire density in the Brazilian Pantanal

105

MS 20152 88866 0.23 46MT 12905 49152 0.26 54

TOTAL 33057 138018 100

State Fire number Fire/area %Area of the state in the Pantanal (Km²)

Figure 2. Fire distribution and states

106

Aquidauana 2405 12559 0.19 4Barão 3675 11004 0.33 7

Bodoquena 7 48 0.15 3Cáceres 3327 14326 0.23 5Corumbá 12744 62207 0.20 4Coxim 662 2022 0.33 7Itiquira 702 1850 0.38 8Ladário 13 49 0.27 6Lambari 106 279 0.38 8Leverger 2215 6915 0.32 7

Livramento 265 1019 0.26 6Miranda 844 2166 0.39 8Poconé 2595 13760 0.19 4

Porto Murtinho 1919 4418 0.43 9Rio Verde 1376 4734 0.29 6Sonora 210 693 0.30 7TOTAL 33065 138049 100

%Municipality Fire number

Municipality area (km²) in the Pantanal Fires/Area

Figure 3. Fire distribution and municipalities

107

The media (newspapers and television) usually reports fires employing absolute values, which cause certain misunderstandings. In the news, the municipality of Corumbá is frequently mentioned as the location with the highest incidence of fires in the state of Mato Grosso do Sul. However, if the number of fires were reported in relation to area, Corumbá would generally hold the fifth position in fire occurrence, not the first. Vegetation structure, phenology and productivity, and their relation to climate, relief, flood regime and human activities, are the main factors that determine fire occurrence and spatial distribution in the Pantanal. Unfortunately, because the AVHRR data present an error of approximately 5 km in georeferencing, only major vegetation types at coarse scales can be used for spatial analysis. In the Pantanal, the highest number of fires occurred in the “�honolo estépica” (steppe savanna from the Chaco region) and “formações pioneiras” (secondary formations) (Figure 4). These vegetation types occur mainly in the southwestern section of the Pantanal as shown in the fire density map developed (Figure 4). The open structure, dense herbaceous stratum, high primary productivity and strong phenological seasonality that characterize the vegetation in this area can be the main cause for fire occurrence and spread. CCCCONCLUSIONONCLUSIONONCLUSIONONCLUSION The spatial distribution of fire density in the Pantanal is heterogeneous when presented in relation to major vegetation types. If political units are considered, such as municipalities and states, the spatial distribution of fire density appears relatively homogeneous. RRRREMARKS AND EMARKS AND EMARKS AND EMARKS AND FFFFURTHER URTHER URTHER URTHER AAAANALYSISNALYSISNALYSISNALYSIS The remote sensing laboratory of Embrapa Pantanal has been building for many years a GIS database encompassing information on the main factors that determine fire occurrence, which include vegetation types, phenology and productivity, and climate, relief, flood regime and human activities. This database constitutes an important tool/resource for the prevention and control of extensive fires. Fire departments usually interact with Embrapa Pantanal for assistance in fire control analysis. This database is constantly growing, gradually improving Embrapa Pantanal’s capabilities for GIS support in issues of fire control and prevention. Currently, the remote sensing laboratory of Embrapa Pantanal is working on the quantification of flooded areas and vegetation phenological variability using MODIS imagery. In comparison to AVHRR data, fire detection using MODIS imagery is more accurate. The spatial analysis being conducted with these data sources and techniques will be essential in the preparation of precise risk maps for fire prevention.

108

Main vegetation typesMain vegetation typesMain vegetation typesMain vegetation types Number of Number of Number of Number of f iresfiresfiresfires

Vegetation area Vegetation area Vegetation area Vegetation area (km(km(km(km2222 ))))

Fires/areaFires/areaFires/areaFires/area %%%%

Ecotono 2347 17577 0.13 6Encrave 1392 5361 0.26 11

Floresta Estacional Decidual 71 223 0.32 13Floresta Estacional Semidecidual 1927 7004 0.28 12

Savana (Cerrado) 18474 84622 0.22 9Savana Estepica (Chaco) 4857 10153 0.48 20

Formações Pioneiras 1364 3894 0.35 15USO 2313 7058 0.33 14

TOTAL 32745 135892 100

Figure 4. Fire distribution and vegetation

AAAACKNOWLEDGEMENTSCKNOWLEDGEMENTSCKNOWLEDGEMENTSCKNOWLEDGEMENTS To Dr. Rob Longman from the ALTERRA institute, Netherlands, whose project has partially supported this activity. To Gerard Groenveld for the help with GIS.

109

RRRREFERENCESEFERENCESEFERENCESEFERENCES Silva, J.S.V. and Abdon, M.M. 1998. Delimitação do Pantanal Brasileiro e suas sub-regiões. Pesquisa Agropecuária Brasileira 33 (Número Especial): 1703–1711.

110

3F. PPPPARAGUAYARAGUAYARAGUAYARAGUAY:::: UUUUSE OF SE OF SE OF SE OF RRRREMOTE EMOTE EMOTE EMOTE SSSSENSING TO ENSING TO ENSING TO ENSING TO PPPPLAN LAN LAN LAN WWWWATERBIRD ATERBIRD ATERBIRD ATERBIRD CCCCONSERVATION IN THE ONSERVATION IN THE ONSERVATION IN THE ONSERVATION IN THE CCCCENTRAL ENTRAL ENTRAL ENTRAL PPPPARAGUAYAN ARAGUAYAN ARAGUAYAN ARAGUAYAN CCCCHACOHACOHACOHACO

ROB P. CLAY14, OSCAR RODAS15, HUGO CABRAL16, ARNE LESTERHUIS17 AND RODRIGO ZÁRATE18

IIIINTRODUCTIONNTRODUCTIONNTRODUCTIONNTRODUCTION Despite being a landlocked country, two-thirds of Paraguay’s borders are formed by rivers. Associated with these rivers, and particularly with the Paraguay River, are numerous and extensive wetlands, such that between 30 and 40% of the country is considered to comprise wetlands (Burgos, 2004). Paraguay’s wetlands support a rich and diverse avifauna: 115 species or 17% of the Paraguayan avifauna are waterbirds and at least a further 68 species (10%) are ecologically dependent on wetlands (Clay et al. 2004). However, in comparison with many other South American countries, the wetland avifauna of Paraguay has been little studied, despite the prominence of waterbirds as a natural resource and the role they play in Paraguayan culture (Clay et al., 2004). One exception is the wetland avifauna of the upper watershed of the Yacaré Sur River and associated wetlands in Presidente Hayes department. Recent studies have shown this wetland complex to be of key importance for waterbirds, periodically holding one percent or more of the global population of eight species (Table 1).

Table 1: Maximum counts of waterbird species in the Yacaré Sur watershed

SpeciesSpeciesSpeciesSpecies 1 % Threshold1 % Threshold1 % Threshold1 % Threshold Max. CountMax. CountMax. CountMax. Count Chilean Flamingo Phoenicopterus chilensis 2000 4000 Coscoroba Swan Coscoroba coscoroba 250 500 Ringed Teal Callonetta leucophrys 1000 1500 South American Stilt Himantopus melanurus 10,000 10,000 Collared Plover Charadrius collaris 350 500 White-rumped Sandpiper Calidris fuscicollis 4000 4000 Pectoral Sandpiper Calidris melanotos 1000 3000 Wilson´s Phalarope Phalaropus tricolor 15000 25000

14 Guyra Paraguay, Asunción, Paraguay; rob@guyra.org.py 15 Guyra Paraguay, Asunción, Paraguay; oscar@guyra.org.py 16 Guyra Paraguay, Asunción, Paraguay; hugocabral54@hotmail.com 17 Guyra Paraguay, Asunción, Paraguay; arne_j_lesterhuis@yahoo.co.uk 18 Guyra Paraguay, Asunción, Paraguay; rzarate@guyra.org.py

111

The upper watershed of the Yacaré Sur River lies in the central Paraguayan Chaco, an area typified by xerophytic thorn-scrub forests, interspersed by large bare saline depressions and riverbeds that form shallow lakes after heavy rains (Figure 1). However, this area has been heavily modified by the agricultural practices of the Mennonite-dominated local communities, and the agricultural frontier continues to expand, with the result that virtually all natural vegetation has been replaced by crops, and most watercourses have been dammed. During the 1990s, 9% of the central Chaco thorn-scrub forest was cleared for agriculture (Alstatt et al., 2003). The effects of the resulting changes (such as a marked increase in salinity) on local waterbird populations are unknown.

Figure 1. Study area

112

In 2000, Guyra Paraguay initiated periodic waterbird surveys throughout the upper Yacaré Sur watershed, and has recently begun using remote sensing techniques, working with Landsat and MODIS imagery, to help identify key areas for waterbirds, to assist with the development of population monitoring protocols, and to document and monitor changes within the watershed, with a focus on vegetation cover and flood cycles. MMMMETHODOETHODOETHODOETHODOLOGYLOGYLOGYLOGY Landsat ETM imagery was used as the primary data source for change detection analysis. Dates of Landsat imagery were selected through a review of daily MODIS imagery, which permitted the identification of maximum and minimum flood events. The Landsat images used were: Landsat ETM+ 227-76 – 18 August 2002 Landsat ETM+ 228-76 – 2 August 2002 Landsat ETM+ 227-76 – 26 February 2003 Landsat ETM+ 228-76 – 8 May 2003 The interpretation and classification of Landsat images was conducted with ERDAS Imagine 8.6 and verified with vertical aerial photographs and videography obtained during overflights on 22 March 2002, 27 September 2002 and 14 July 2004, in addition to terrestrial photos and personal observations made during numerous visits to the area since 2000. An analysis of changes in vegetation cover was conducted in cooperation with the Global Land Cover Facility, University of Maryland (Alstatt et al., 2003). Maximum and minimum flooded areas were identified using a multi-date post-classification comparison. A preliminary assessment of salinity levels was carried out using variation in the spectral response of water in infrared (assessed in all four Landsat images). Four water classes were identified: freshwater, brackish, saline and highly saline. Freshwater gave a spectral response close to the minimum level (0), and highly saline close to the maximum level (255), effectively representing a thin cap of water over a salt deposit. This initial classification was partially verified through a visual analysis of vertical aerial photos and videography to identify accumulations of salts and the presence of plant communities associated with different salinity levels. Land cover was identified initially through an unsupervised classification, with the resulting 150 classes then regrouped into 10 classes reflecting a three-tier classification system (Natural/modified environments, hydromorphic environment, vegetation type). The ten classes were: (1) agriculture, (2) areas affected by fire, (3) permanent water bodies, (4) waterlogged soils, (5) seasonally inundated palm savanna, (6) seasonally

113

inundated grasslands, (7) marsh vegetation, (8) semi-deciduous forest, (9) low forest and dense palm savanna, and (10) xerophytic thorn-scrub. The land cover classification was used to identify those water bodies bordered by natural vegetation (as opposed to those bordered by agricultural land). RRRRESULTSESULTSESULTSESULTS The analysis of changes in vegetation cover combined with the land cover classification revealed the upper watershed to have been largely converted to agriculture (Figure 2). Although most forest cover in the upper watershed was cleared prior to 1990, large blocks were still being cleared throughout the 1990s (Figure 3). The first large scale clearances in the lower watershed were also detected during this period. Only the middle watershed remains largely unaffected by deforestation. Natural shoreline vegetation is no longer present at many water bodies within the Mennonite colony in the upper watershed. However, there are some notable exceptions, in particular the Laguna Capitán, Laguna Ganso and Campo León complexes. From the edge of the Mennonite colony eastward, the waterbodies are largely bordered by natural vegetation (Figure 4).

Figure 2: Natural communities of the Natural communities of the “Riacho Yacaré Basin”

114

Figure 3. Deforestation (1990s) in the Riacho Yacaré Basin.

115

Figure 4. Shoreline vegetation.

116

Figure 5. Maximum and Minimum flood.

117

No clear pattern of flood was found throughout the watershed. Instead, water levels appear to be more dependent on local conditions (presumably rainfall and other changes in the local watertable). This is best shown by variation in water level in the Laguna Salada-Campo María complex (Figure 5). The maximum and minimum flood at Laguna Salada and adjacent Campo María was found to be the inverse of that in the preceding branch of the Yacaré Sur immediately to the west. In the lower watershed, the flood cycle is closely tied to that of the Paraguay River. When the latter is in flood the net water flow in the Yacaré Sur appears to be upstream. In general, salinity levels were found to be high throughout the waterbodies in the upper watershed. However, there were some notable exceptions, with freshwater bodies in the vicinity of Campo María and Laguna Capitán. Freshwater was also found to predominate in the middle and lower watershed, with an abrupt change from saline to freshwater immediately to the east of Laguna Salada (Figure 6). Small numbers of waterbirds were found on virtually all waterbodies throughout the upper Yacaré Sur watershed. Figure 7 shows the areas were waterbirds were found with greatest regularity. However, large congregations were regularly recorded only at Campo María-Laguna Salada (Figure 7, points 5 and 7) and at Laguna Capitán (Figure 7, point 8), and with less regularity in the Campo León (Figure 7, point 3) and Laguna Ganso (Figure 7, point 9) areas. Although insufficient quantitative data is available to statistically analyze the relationship between waterbird numbers, flood levels, salinity levels and surrounding land cover, a pattern does appear to be present: all four of the above mentioned areas combine the permanent presence of water with natural shoreline vegetation, land cover which is predominantly natural and water bodies with a variety of salinity levels (particularly Campo María-Laguna Salada and Laguna Capitán).

118

Figure 6. Salinity levels.

Figure 7. Waterbirds observed.

119

CCCCONCLUSIONSONCLUSIONSONCLUSIONSONCLUSIONS The data and observations presented here are preliminary, particularly with regard to salinity levels, and require verification through more detailed studies including the sampling of water (e.g. for electrical conductivity) throughout the watershed. However, these preliminary data do allow for the construction of hypotheses regarding the factors of importance for waterbirds within the watershed, and also provide tools for the design of monitoring protocols. Guyra Paraguay has used the results to identify key sites to be incorporated within terrestrial and aerial waterbird monitoring regimes, and has developed protocols with an adaptive site-selection component based on real-time analysis of MODIS imagery. The land cover and change detection analyses have also been used to define the boundaries of key sites within the watershed (which together form two Important Bird Areas: Upper Yacaré Sur and Campo María-Laguna Salada) and priority areas for restoration of natural vegetation cover. AAAACKNOWLEDGEMENTSCKNOWLEDGEMENTSCKNOWLEDGEMENTSCKNOWLEDGEMENTS The technical experience and capacity required for the change detection analyses was largely acquired through the Pantanal GIS pilot project funded by the USDA Forest Service and coordinated by Ducks Unlimited. Fieldwork, aerial surveys and change detection analyses within the Yacaré Sur watershed were made possible through projects funded by National Fish and Wildlife Foundation, USFWS (NMBCA) and USAID and TNC through the Fundación para el Desarrollo Sostenible del Chaco. LLLLITERATURE ITERATURE ITERATURE ITERATURE CCCCITEDITEDITEDITED Burgos, S.M. 2004. Localización de los humedales en Paraguay. Pp. 9-12 in: D. Salas-Dueñas, F. Mereles & A. Yanosky (Eds) Humedales del Paraguay. Comité Nacional de Humedales: Asunción: Paraguay. Clay, R.P., Lesterhuis, A.J. & López, L. 2004. Aves de los humedales del Paraguay. Pp. 175-190 in: D. Salas-Dueñas, F. Mereles & A. Yanosky (Eds) Humedales del Paraguay. Comité Nacional de Humedales: Asunción: Paraguay. Alstatt, A., Kim, S., Townshend, J., Tucker, C., Rodas, O., Musinsky, J., Clay, R.P., Curry, T. & Yanoksy, A. 2003. subtropical deforestation: Paraguay in the 1990s. http://glcf.umiacs.umd.edu/index.shtml

120

4. ACKNOWLEDGEMENTS The expansion of the pilot project and the ancillary activities carried out by different partners in the UPRB GIS database consortium was a real challenge as each became involved in numerous independent projects. In addition to the friends and colleagues who helped in many ways during the pilot project (see the final report of this project) the editors of this volume are especially grateful to Fábio Ayres, Hugo Cabral, Rob Clay, Mariane da Cruz, Wolf Eberhardt, Manuel Ferreira, Bill Liu, Arne Lesterhuis, Theodore Mealor, Carlos Padovani, Pamela Rebolledo, Heidy Resnikowsky, Oscar Rodas, Jorge Sambrana, Sylvia Torrecilha and Rodrigo Zárate for responding so generously to our request for summaries of the projects they are now involved in and which benefited from the activities carried out during the pilot project. Special thanks goes to Mario Dantas, Gislaine Disconzi, Kristine Kuhlman and Alberto Yanosky for their interest in spreading the word about the project and trying to secure further funding for a comprehensive database for the UPRB. In spite of his demanding new role as Director of Energy and Technology Partnerships, Dick Kempka contributed generously with his advice and support in all stages of this phase of the project. Valmir Ortega deserves not only recongition for helping get the project started in 2000, but also an apology for we forgot to acknowledge his support in the final report of the pilot project. Once again, two organizations –with very special friends- provided financial support to allow us to complete this work: Val Mezainis, Jack Capp and Michelle Zweede at the International Programs of the USDA Forest Service for their trust in the Pantanal GIS project and their reiterated support through agreement # 02-DG-11132762-128. Herb Raffaele, Gilberto Cintrón and Frank Rivera at the Division of International Conservation of the US Fish and Wildlife Service, for their extended support to the training component of this project through agreement # 98210-1-G077. Thanks also to ESRI and ERDAS Imagine/Leica Geosystems for software support for training workshops.