UUUUpper PPPParaguay RRRRiver BBBBasin GISGISGISGIS DDDDatabase - Pilot Project I -

Upper Paraguay River Basin GIS DatabaseUpper Paraguay River Basin GIS DatabaseUpper Paraguay River Basin GIS DatabaseUpper Paraguay River Basin GIS Database ---- Pilot Project I Pilot Project I Pilot Project I Pilot Project I ----

Edited by Dawn Browne

Montserrat Carbonell Dick Kempka

Prepared by the �Upper Paraguay River Basin GIS Database Consortium�

BoliviaBoliviaBoliviaBolivia BrazilBrazilBrazilBrazil ParaguayParaguayParaguayParaguay Pamela Rebolledo Fabio Ayres Anibal Aguayo Heidi Resnikowski Luiz Benatti Rob Clay Julia Boock Claudia Mercolli Lindalva Cavalcanti Laura Rodríguez Gislaine Disconzi Oscar Rodas 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

July 2003

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: Browne, D., Carbonell, M. & Kempka,D. (Eds.) 2002. Upper Paraguay River Basin, Pilot Project I GIS Database, Final Report. 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. 2003

Contents

Upper Paraguay River Basin GIS Database ....................................................................... 1 I. Introduction ............................................................................................................ 1 II. Justification ............................................................................................................. 2 III. Objectives .............................................................................................................. 4 IV. Partner organizations............................................................................................. 5

Upper Paraguay River Basin GIS Database Pilot Project I .................................................. 6

V. Pilot Project I........................................................................................................... 6 VI. Methods................................................................................................................. 7

a. Meetings and Workshops .................................................................................... 7 b. Pilot Project Study Area ....................................................................................... 8 c. Image Selection and Data Acquisition.................................................................. 9 d. Fieldwork .......................................................................................................... 10

VII. Developing the database..................................................................................... 10 a. Metadata ........................................................................................................... 10 b. Data Serving...................................................................................................... 11 c. Summary of Pilot Project I Deliverables.............................................................. 12

VIII. Change Detection Methods ................................................................................ 13 IX. Seasonal Flood Extent Analysis ............................................................................ 15 X. Fire Scar Mapping ................................................................................................. 15 XI. Updating Roads................................................................................................... 17 XII. Partner Deliverables ............................................................................................ 18

a. Flood Analysis ................................................................................................... 18 b. NDVI Change Detection..................................................................................... 22 c. Multi-date Burnscar Delineation ........................................................................ 25 d. Updating Roads................................................................................................. 27 e. Field Work ......................................................................................................... 29 f. Additional Tasks ................................................................................................ 33 g. Technical Suggestions....................................................................................... 38 h. Hydrological Analysis for UPRB.......................................................................... 39 i. Landcover Classification..................................................................................... 43

Acknowledgements....................................................................................................... 47 Appendix I .................................................................................................................... 49 Appendix II .................................................................................................................. 67 Appendix III .................................................................................................................. 77 References and Other Related Literature ....................................................................... 91 Pantanal related websites.............................................................................................. 98

This page left blank intentionally

1

Upper Paraguay River Basin GIS DatabaseUpper Paraguay River Basin GIS DatabaseUpper Paraguay River Basin GIS DatabaseUpper Paraguay River Basin GIS Database

I. IntroductionI. IntroductionI. IntroductionI. Introduction The Pantanal is the vast floodplain of the Upper Paraguay River Basin (UPRB) that covers an estimated 496,000 km2 . Roughly one-third of the UPRB is the complex and vast low-altitude floodplain known as Pantanal, and two-thirds is the �Planalto� or highlands. These upland areas include the Grand Chaco of Bolivia and Paraguay and the Brazilian highlands (Swartz, 2000). The Pantanal is the world�s largest continuous freshwater wetland, approximately the size of Honduras, Nicaragua and El Salvador combined, with an estimated area of 150,000 km2, of which 110,000 km2 are wetland (Scott & Carbonell, 1986). Its boundaries extend across the borders of three countries: Bolivia, Brazil and Paraguay, but more than 70% of the Pantanal is located in Brazil (Dolabella, 2000). The headwaters of the Upper Paraguay River on its left margin, in Brazil, are clearly defined in the PCBAP report (1997). Here, the Paraguay River tributaries are permanent and the land presents variable elevations. For the tributaries on the right margin, WWF-Bolivia provides a definition (per. comm. H. Resnikovsky, 2002) while on the Paraguayan portion, south of the Río Negro, the headwaters are not clear. However, The Nature Conservancy (TNC) includes the wetlands of this area of the Paraguayan Chaco under a watershed division with no connection to the Pantanal/UPRB. The Programa Pantanal in Brazil (2001) considers both watersheds as one unit, and Mereles (2000) considers the southernmost limit of the UPRB Pantanal as the confluence of the Río Apa with the Rio Paraguay. The Pantanal is located between 12-24°S and 55-65°W. It has extensive seasonally inundated savanna areas, patches of humid deciduous forest, gallery forest and marshes that act like a sponge and prevent flash floods downstream as the Pantanal drains into the Paraguay River and out into the Atlantic Ocean through the Paraná River. The Pantanal is one of the world�s richest ecosystems. Due to its location in the center of South America, the Pantanal has fauna and flora typical of the Amazon, the Bosque Chiquitano, the Chaco, the Cerrado and the Atlantic Forest regions, which contribute to its rich biological diversity. There are at least 400 species of birds, 300 of fish, 120 of mammal, 170 of reptiles, 40 amphibians (Mittermier et al., 2002), and more than 2000 identified plant species recorded in the Pantanal (Seidl et al., 2001). The Pantanal also functions as a corridor through the dry Chaco, and the headwater wetlands of the Amazon and Paraguay rivers establish close hydrological contact at several points,

2

explaining why the Pantanal, a part of the Río de La Plata basin, has 60% of the fish species in common with the Amazon basin (Por, 2000). All three countries protect discontinuous areas of the Pantanal under different protection mechanisms (National Park, State Park, Forestry Reserve). They have also designated several portions of the Pantanal under the Convention on Wetlands of International Importance, especially as Waterfowl Habitat (Ramsar Convention, www.ramsar.org). However, much of this region is still unprotected and mostly in private hands. Until recently, preservation of the Pantanal�s pristine environment has been due in part to its inaccessible location in the center of South America. This situation is rapidly changing. Main economic activities include fishing, mining, tourism, and cattle ranching (Seidl, et al., 2001). Extensive cattle ranching by private landowners have been predominant for the last two centuries. Deforestation for pasture planting has extended to more than 500,000 ha (13% woodlands) over the past 25 years (Seidl, et al., 2001). In the Cerrado, soils are being depleted of their natural cover and replaced by agricultural and cattle ranching activities or urban development (Seidl, et al., 2001). Soil erosion and silting of rivers from indiscriminate farming practices are also causing changes in the landscape. Clearing of land for agriculture and cattle production, mining operations, construction of hydroelectric power plants, unplanned tourism, hunting, and the construction of gas and oil pipelines, and roads across the borders of the three countries are threatening the integrity of this unique ecosystem. The Hidrovia Project, aiming to make the Paraguay and Paraná rivers navigable for large commercial /transportation vessels that will bring agriculture and farming products directly to the ocean, could have irreversible impacts on this fragile environment if not planned and developed adequately. According to Allem e Valls (1987) the Pantanal could co-exist in relative harmony with activities such as extensive cattle ranching. Seidl, et al. (2000) indicate that the lack of predictability of the floods is the single major factor which prevents the transformation of the Pantanal into an immense soybean field or a boundless cattle ranch. II. Justification II. Justification II. Justification II. Justification Bolivia, Brazil and Paraguay are trying to respond with alternative options for development projects (i.e. Programa Pantanal in Brazil), make conservation-related decisions and manage the natural resources of the Pantanal and indeed the entire UPRB. To facilitate this, it is imperative to establish a Geographic Information System (GIS) capable of handling several data types in a common format in the same database. Remotely sensed data is very useful for studying inaccessible or remote areas. Satellite imagery used in conjunction with wildlife surveys and other GIS feature data can

3

produce habitat assessments. The data produced by these assessments can be used to model the effects of current and future land-use practices and determine, for example, boundaries of future protected areas or areas of priority action for restoration. It can also be used to make management decisions at sub-catchment levels and it offers planners and decision makers the tools necessary to provide sustainable alternatives to development projects. A GIS can store information about the UPRB through a system of thematic layers that can be linked together by geography. These thematic layers would be developed using satellite imagery, topographic maps, aerial photography, and fieldwork. A GIS database can establish a baseline environmental inventory including data layers such as vegetation, soils, roads, cities, protected areas, watersheds, flooding extent and detection of environmental changes. This data can be used to generate hard copy maps and GIS models of ecosystem associations. In turn, models can be developed to predict the impacts that development and land-use changes may have on the ecosystem and assist local interests in conservation efforts, management, restoration, delineation of protected areas and designation of Ramsar sites, among other possibilities. During the 7th Meeting of the Conference of the Contracting Parties to the Convention on Wetlands (Ramsar Convention) in Costa Rica, in May 1999, Ducks Unlimited (DU) and the USDA Forest Service organized and presented a GIS seminar where the GIS work of DU over the last 20 years was presented. DU has developed GIS models for over 30 wetland projects, including 81 million hectares in North America, similar to the one proposed here, and has developed data distribution networks and communication mechanisms for natural resource management and conservation applications. Because of this vast experience in the use of GIS for management and conservation of wetlands in North America, DU is in a unique position to work with partners in South America and transfer this technology. After consultation with government agencies, research institutions and individuals from the three countries, it became apparent that there is no comprehensive GIS database in place for the UPRB. Different projects have basic GIS data layers for the area they cover, but still lack the more detailed information that will allow for wetland management and conservation decision-making on a landscape level. None of them are international in scope, so their information relates only to individual countries and few, if any, share the formats and standards that will allow them to become useful regional scale tools. Several projects have been generated by organizations such as The Nature Conservance (TNC), Conservation International (CI), Programa Pantanal, and Plano de Conservação da Bacia do Alto Paraguai (PCBAP). These projects have produced important spatial information for the UPRB. However, each used different classification schemes and

4

varying formats, even within their own country, making data sharing and transfer extremely difficult. III. ObjectivesIII. ObjectivesIII. ObjectivesIII. Objectives The goal of the UPRB GIS Database project is to contribute to improved management and conservation of natural resources of the UPRB through the development of a GIS database and a data distribution network. The main objective of this project is to develop a standardized GIS database in an easy and compatible format that will enable users to exchange and access information. This will be a most powerful decision-making and management tool available for assessing the entire UPRB ecosystem.

Location of the Upper Paraguay River Basin and Pantanal Location of the Upper Paraguay River Basin and Pantanal Location of the Upper Paraguay River Basin and Pantanal Location of the Upper Paraguay River Basin and Pantanal ---- B B B Bolivia, Brazil and Paraguayolivia, Brazil and Paraguayolivia, Brazil and Paraguayolivia, Brazil and Paraguay (Source: USGS and Olson, et al.2001)(Source: USGS and Olson, et al.2001)(Source: USGS and Olson, et al.2001)(Source: USGS and Olson, et al.2001)

5

IV. Partner organizationsIV. Partner organizationsIV. Partner organizationsIV. Partner organizations The organizations that were actively involved in data development for Pilot Project I are listed in Table 1.

Table Table Table Table 1111. The partners involved in Pilot Project I and their roles or data development tasks.. The partners involved in Pilot Project I and their roles or data development tasks.. The partners involved in Pilot Project I and their roles or data development tasks.. The partners involved in Pilot Project I and their roles or data development tasks.

CountryCountryCountryCountry OrganizationOrganizationOrganizationOrganization Tasks in Pilot Project ITasks in Pilot Project ITasks in Pilot Project ITasks in Pilot Project I Bolivia World Wildlife Fund (WWF) Burnscar mapping, NDVI

analysis, flood analysis, digitizing topomaps

Brazil Ecotropica Aerial photo catalog, pilot

project II Brazil Empresa Brasilera de Pesquisa

Agropecuaria (EMBRAPA) Flood analysis, digitizing topomaps, fieldwork

Brazil Fundação Estadual de Meio Ambiente (FEMA-MT)

Fieldwork, pilot project II

Brazil Instituto Meio Ambiente -Pantanal (IMAP)

Fieldwork, aerialphoto mosaic

Brazil Universidade Católica Dom Bosco (UCDB)

UPRB hydrological flood analysis, burnscar mapping

Brazil Instituto Brasileiro do Meio Ambiente e dos Recursos Naturais Renováveis (IBAMA)

Roads, NDVI Analysis

Paraguay Fundación Moisés Bertoni (FMB) Burnscar mapping, NDVI

analysis Paraguay Guyrá Paraguay Flood analysis, roads, fieldwork USA University of Memphis Landcover classification USA University of Wisconsin (LICGF) Metadata development, data

gathering USA USDA Forest Service (USFS) Funding source USA US Fish & Wildlife Service (USFWS) Funding souce USA US Geological Survey Metadata serving USA/Canada Ducks Unlimited (US and Canada) Coordination, facilitation, data

serving, training, technical advice, fundraising

6

Partners have also taken the lead in the following aspects of the project: − Planning and management of remote sensing (e.g. satellite and airborne data) and

other mapping projects − Implementing GIS modeling applications − Conducting fieldwork to collect vegetation and land cover information − Designing and developing landcover mapping projects using remotely sensed data

and GIS data analysis − Manage assigned projects including budgets and timelines in conjunction with an

administrator

Upper Paraguay River Basin GIS Database Pilot Project IUpper Paraguay River Basin GIS Database Pilot Project IUpper Paraguay River Basin GIS Database Pilot Project IUpper Paraguay River Basin GIS Database Pilot Project I

V. Pilot Project IV. Pilot Project IV. Pilot Project IV. Pilot Project I While the main objective of the UPRB project is to develop a comprehensive database system of the entire basin, a full landcover classification and creation of seamless datasets would take much longer and would drastically increase the overall cost of the project. Therefore, because changes in landuse/landcover are considered a high priority for the conservation and management of the ecological integrity of the ecosystem, a pilot project was conducted: − to illustrate usefulness of the techniques proposed, standardize methodologies, and

fine tune solutions to problems encountered in a tri-national project − to contribute to the capacity building of governmental agencies as well as local non-

government agencies (NGOs) − to analyze the different alternatives for serving and distributing the data. The main objectives of the pilot project I included: − To detect historical landscape changes using repeatable, quantifiable methods − To determine seasonal water level changes, i.e. identify areas of seasonal flooding − permanent flooding, open water and dry land − Develop an application to serve data for the pilot project area.

7

VI. MethodsVI. MethodsVI. MethodsVI. Methods

a. Meetings and Workshops Several meetings have taken place through the duration of the project in order to coordinate activities, provide software training, or perform feasibility studies on future data serving partnerships. The agendas of these meetings are in Appendix III. Additionally, two discussion groups were established, one general list for exchange of information among persons interested in the conservation of the UPRB/Pantanal (http://groups.yahoo.com/group/pantanalGIS) and a second one for the technical group to discuss issues and progress related to the development of the database and the report. In December 1999, Brazilian government agencies and other potential users of a GIS database for the UPRB were visited and their needs for conservation planning and data management discussed, while an extensive review of existing information was carried out. A scoping meeting was arranged in Campo Grande (MS, Brazil) in April 2000, to bring together as many different governmental agencies, NGOs and community organizations within the three countries as possible. This participatory process had several purposes: − To determine the land-use and conservation planning needs with staff responsible

for natural resource management in this region − To determine information available, as well as guidelines and standards needed to

set up and deliver the GIS database, so all interested parties had a clear understanding of the final products

− To identify a pilot area − To establish which data sets are a priority for this system − To establish partnerships. A technical work plan was developed as a result of the Campo Grande meeting but adapted at later meetings as the project evolved. The work plan is not presented as a standalone document in this report but is incorporated throughout this final document. The technical design served as a template for delivery of the system among all organizations involved. It identified equipment, data, and staff needs. Methods and a task timetables were also assembled. One of the needs identified in the technical work plan was staff training in ArcView 3.2a, the software selected for GIS data development by the technical partners in the project. A training workshop was therefore organized and hosted by EMBRAPA-Pantanal in Corumbá (MS, Brazil) in December 2000.

8

In January 2001, a coordination and standardizing meeting was held in Memphis, TN (USA) and hosted by DU. The purpose of this meeting was to establish the primary spatial data needs for each country and to design basic technical methods for the change detection processing. Also, an effort was made to coordinate standards for compiling existing data sets between the three countries, and all the partners were assigned processing tasks based on the technical capacity of each organization. Training in ERDAS Imagine, an image processing software, was also considered an essential need among the partners therefore, a workshop was organized by Guyra Paraguay in Asunción (Paraguay) in May 2001. Also, there was further coordination on the methodologies and procedures for the change detection image processing and discussions on the progress made during the four months since the previous meeting. Participants were also trained in ArcView 3.2a Image Analysis. A workbook was distributed which provided illustrated, step-by-step instructions on the various methods of change detection in the above mentioned software. This meeting was followed by another coordination exercise in Puerto Suarez, Bolivia, and organized by WWF-Bolivia in November 2001. The meeting focused on setting metadata and data normalization procedures and provided standards for collecting field data. Topics that covered included GPS standards to ensure uniform data collection methods among the various countries, metadata standards, and designing a field data collection sheet. Partners from Brazil, Bolivia, and Paraguay had recently conducted fieldwork to provide GPS control points to verify the spatial accuracy of the Landsat imagery and other existing GIS data sets. Digital cameras with GPS units were distributed to partners that would be taking the lead in the fieldwork activities. A draft final report was presented at a meeting in October 2002 and was hosted by FEMA-MT in Cuiabá (MT, Brazil). Partners discussed improvements and suggestions before its final publication. The steps necessary for the next phase towards the development of the UPRB were also discussed and a draft technical work-plan was developed during the following months.

b. Pilot Project Study Area The UPRB pilot project covered an area that included all three countries - the region of the Nabileque (Brazil), Otuquis (Bolivia) and Río Negro (Paraguay). This area was selected for the following reasons:

− The core area remains intact, but surrounded by development pressure − The three countries have declared or are planning to declare protected areas − Paraguay and Bolivia have established Ramsar sites and Brazil has the intention − Bolivia has forest fire problems

9

− The area is rich in bird and fish fauna − Enough information existed to test the GIS database

c. Image Selection and Data Acquisition Satellite imagery was selected as the data source for change detection analysis due to the large overall size of the comprehensive UPRB. The use of this technology provided a cost-effective method for landscape level change analysis. Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper (ETM) was the first choice of data for the following reasons: - It covers a large regional area (185x170km/scene) - The spatial resolution (30x30meter/picture element) is sufficient - Scenes are captured and archived - Multi-spectral characteristics allow features such as vegetation, moisture and

inundation to be extracted from the data Based on river gauge and precipitation information gathered from several sources and provided by partner contacts, Landsat scenes and dates were selected for the pilot area. River height and/or discharge data was evaluated to determine the optimal timing for the satellite imagery for representing a range of flood conditions and vegetation phenology. The Pantanal has widely variable water regimes both seasonally and annually. It is very important to understand this variability when selecting imagery for change detection analysis. Precipitation data is important for the same reasons. Timing of rainfall must be well understood in order to apply it to the selection of imagery. The imagery dates used for this pilot project were:

Landsat TM � 9 June 1997 Landsat TM � 7 July 1998 Landsat TM � 19 November 1998 Landsat TM � 24 December 1999 Landsat TM � 23 November 1988 Landsat ETM+ - 14 November 1999

Landsat TM and ETM scenes were purchased from CONAE (Argentina, www.conae.gov.ar and from INPE (Brazil, www.inpe.br). All scenes were purchased at Level 1G and sent to Image Links, Inc. (a raster based image processing company) for geo-referencing. Image Links has developed software that uses feature correlation between overlapping areas and advanced statistical analysis techniques to calibrate and correct the satellite sensor model. The data is then accurately projected to the specified map projections and co-registered between and amongst scenes of different dates. When the images

10

were received back from Image Links they were checked for quality and distributed to the partners. Soils, transportation, proposed development projects, land ownership, protected areas, ecological units, rare and endangered plant communities are examples of other spatial data sets that may be useful components of the database. All publicly available data sets have been compiled by the partners and converted to a common format for use in the database.

d. Fieldwork The final, most detailed tier is field level information such as large-scale wetland identification, local land-use patterns, ecological zones, sample vegetation or wildlife transects conducted on-site. Due to financial restraints, it was not possible to include this level of ground data collection for the Pilot Project I. However, fieldwork was carried out in order to acquire aerial photographs and to gather GPS points to verify the spatial accuracy of the pilot area imagery and provide control points for rectification of future satellite imagery. Data collected through fieldwork was also integrated into the GIS database. VII. Developing the databaseVII. Developing the databaseVII. Developing the databaseVII. Developing the database All partners carried out separate tasks while DU has been in charge of receiving, quality checking, and storing all the data gathered and compiled by all partners (see below). DU has also been in charge of ensuring this data is in the standardized formats specified in the technical work plan. In order to provide technical support at the local level, a coordinator was hired part time. In addition to the data compiled and generated by all partners for Pilot Project, a hydrological analysis of the UPRB using AVHRR and river gauge data was carried out and a land cover classification scheme was initiated.

a. Metadata Metadata is detailed information describing the characteristics of a digital data layer. This includes information such as source organization, scale, format, projection, attributes/classes, etc. The availability of metadata greatly enhances the usefulness and validity of a GIS database. For this reason, it was decided during the scoping meeting at Campo Grande, Brazil, April, 2000, that the minimum metadata standards to document existing and newly created data would be the Federal Geographic Data Committee Level I (FGDC). MetaLite software was customized and distributed to each partner as the standard for compiling metadata for the pilot project datasets.

11

b. Data Serving An effective way to disseminate project results and attract more users of these data is over the Internet. Institutional capabilities of the partners for housing and serving the Pantanal GIS project data were evaluated as well as outside organizations who already have this capability and whose mission is to serve conservation data. There are many questions and issues related to how the database will be accessed including: - What data should be served? - Need for a relational database management system? - Paths for web-enabling for non-GIS users and GIS user? - Network Bandwidth? - Data compression? Typically, there are two models for serving data. One uses a �centralized� server and Database Administrator (DBA) to post and maintain data for all parties. The other is a �distributed� network that requires each organization to post, serve, and maintain their own data. There are computer equipment, software licensing, and staff benefits and disadvantages with each method of delivery. For instance a �central� fileserver is typically less expensive for computer hardware and software licensing since only a single device is needed. In the �distributed� model everyone needs its own hardware, software and DBA. Several meetings with organizations that could potentially serve the Pantanal GIS data and other project information (reports, bibliographies, etc.) were held. The organizations that were consulted were CES (Florida Center for Environmental Studies) in Palm Beach Gardens, FL; CIESIN (Center for International Earth Science Information Network) in Columbia, NY; USFWS (United States Fish and Wildlife Service) and USGS (United States Geological Survey) in Laurel, MD; Ducks Unlimited Canada in Stonewall, MB; and Ducks Unlimited, Inc., Memphis, TN. Each organization stated that part of their mission is to disseminate data or facilitate the exchange of data in various capacities. An evaluation of these organizations was made to determine which groups were most viable candidates in terms of their technical capabilities, staffing resources, and administrative partnership requirements or restrictions. The discussions included a review of the following developmental and administrative topics: - Budgets and Potential Funding Sources - Interface Functionality of Existing Internet Applications - Data Access Policy - Areas of Potential Collaboration

12

- Organizational Mandates - Commonalities of Existing Programs with Pantanal GIS Project - Organizational Structure The results of each meeting were discussed with the project partners at the August 2002 meeting and a decision on how to proceed with data serving partnerships was reached. At present there are four components of the project database currently being served or developed: - Metadata � now being served through the USGS clearinghouse

(http://130.11.52.184/FGDCgateway.html). - Data and literature catalogue � a data inventory and bibliography was compiled and

is available through a server at DU Canada (for partners only while it is being tested). - Internet Map Server � DU has created a web-based map server that will allow viewing

and downloading of project maps and images. - Complete database � DU will be the temporary central location for all project related

GIS data and documents until the internal capacity to serve or mirror the database is developed among the partners.

All four components will soon be accessible through DU�s Latin America and Caribbean Program web page at http://www.ducks.org/conservation/latinamerica.asp

c. Summary of Pilot Project I Deliverables - A GIS database for the pilot area along with general base information from satellite

imagery. - A new �change detection� dataset depicting areas with significant change between

the late 1980�s, 1990�s and present, (1) Vegetation/NDVI, (2) Seasonal Flood Extent, (3) Multi-date Burn scars.

- Analysis and map production at a landscape level. - Updated roads coverage for the Pantanal portion of each country within the pilot

area. - Landcover maps and other existing topographic maps for each country have been

compiled, reprojected and mosaicked where scale and format permitted. - A bibliography of related literature from each country. - A standardized metadata format has been completed for all new and existing

datasets and is currently being served through USGS. - A report documenting procedures and the contents of the database is in the final

stages of development.

13

- A team of local organizations with GIS capacity in Bolivia, Brazil, and Paraguay will continue to develop and maintain the integrated database for the entire UPRB.

In addition to the above contributions to the development of the Pantanal GIS database, the project has also produced the following benefits: - Establishment of a technical network of professionals specialized in wetland

management, GIS, remote sensing, and spatial data development. - A feasibility study for developing and maintaining the GIS database and server. - Building of alliances between institutions and countries sharing stewardship of the

Upper Paraguay watershed. - Coordination and standardization of applications and procedures among the three

countries for the development, maintenance and use of the comprehensive UPRB GIS database.

VIII. Change Detection MethodsVIII. Change Detection MethodsVIII. Change Detection MethodsVIII. Change Detection Methods One of the purposes of the change detection analysis for the pilot project area was to determine the most appropriate method for detecting landscape level changes, both natural and man-induced, for the Pantanal region. Several methods were used to identify temporal change in the following categories: - Hydrology

− Seasonal Flooded Area/Water Level Changes (Max/Min flooded area) - Historical Landscape Change

− Human-Induced (NDVI/Vegetation Change, roads expansion) − Natural (fires, regeneration)

A fundamental assumption of digital change detection is that there should exist a difference in the spectral response of a pixel on two dates if the biophysical materials within the Instantaneous Field of View (IFOV) of the sensor have changed between dates (Lillesand et al 1987). Another fundamental premise of using remotely sensed data to detect change is that variations in the object of interest will result in changes in radiance values that are significantly large compared to those caused by other factors, such as differences in atmospheric conditions, illumination angle and soil moisture. If the change detection technique is sensitive to these other factors, they need to be either corrected for or otherwise taken into consideration. Several digital change detection techniques require that the data be radiometrically corrected, calibrated, standardized and/or normalized. Some methods of radiometric normalization for multi-date imagery are usually performed before georeferencing. However, imagery for the pilot area was geometrically corrected by a contractor before

14

radiometric normalization between scenes could be applied. Also, performing radiometric normalization by linear regression techniques was beyond the capabilities of the software being utilized by the partners. Alternative techniques were used to reduce the occurrence of �false change� between dates due to atmospheric effects. Digital change detection techniques may be categorized into two basic approaches: - Comparative analysis of independently produced thematic labeling or classifications

of imagery from different dates - Simultaneous analysis of multi-temporal data sets. Therefore, the primary change detection processing methods used to identify change areas in the pilot project area were based on these two approaches.

a. Multi-date Post Classification Comparison This is the most commonly used quantitative method of change detection requiring rectification and classification of remotely sensed imagery. Image classification is the process by which image pixels are grouped into classes with similar spectral attributes and then each spectral class is assigned to an information class (Fig. 1). The accuracy of the �change file� is completely reliant on the quality of the classification process and the methods used. This data set will allow for the analysis of seasonal flood patterns and to track modifications of the natural hydrology. See Appendix II for technical steps in detail.

b. NDVI Band Subtraction and RGB Co-Display This method was performed on Normalized Difference Vegetation Index (NDVI) images that were calculated for each date. NDVI is a quantification of green biomass through the ratioing of the infrared bands of satellite imagery.

NDVI = R(4) - R(3)/ R(4) + R(3) The technique requires that each pixel value in the NDVI image be subtracted from its corresponding pixel value in another image. The resultant image represents the change in vegetation biomass between the two dates. The RGB Co-Display method is achieved by stacking three dates of NDVI and each date is assigned the red, green and blue color guns. A colorwheel is provided for visual interpretation of the enhanced image displaying various colors that represent the corresponding increase or decrease in vegetation biomass associated with each date.

15

The final product may be used as a digitizing backdrop or for regional scale interpretation. See Appendix III for technical steps in detail. IX. Seasonal Flood Extent AnalysisIX. Seasonal Flood Extent AnalysisIX. Seasonal Flood Extent AnalysisIX. Seasonal Flood Extent Analysis The Pantanal is a vast alluvial plain with a slight north-south gradient with very complex precipitation and river water level patterns. It provides an enormous natural control mechanism for the floodwaters resulting from the torrential rains in the UPRB. The River Paraguay overflows along with its tributaries and the water spreads out until the entire area is deluged to a depth of up to 3 or 4 meters (10 to 13 ft) except for higher ground. Bridges and roads are swept away around the periphery and cities such as Corumbá on the Brazilian border to Bolivia are seriously flooded. As the floodwaters recede and evaporate much of the Pantanal becomes a huge grassy plain on which the cattle are driven for fattening up for market. There are still many areas where there are lakes, lagoons, and saline pools, but these gradually diminish in size (Por 1995). Change detection analysis of seasonal flood extent and maximum and minimum flooded area was carried out in the pilot project area for Bolivia, Brazil and Paraguay using ArcView 3.2a Image Analysis and ERDAS Imagine 8.5 when available. In order to track these changes in seasonal hydrology, the Multi-date Post Classification Comparison method was applied to each date of the pilot area imagery. Each scene was classified (unsupervised) using the Iterative Self-Ordering Data Analysis (ISODATA) algorithm. All dates of imagery were classified in ArcView 3.2 Image Analysis and then target features extracted ( i.e. flooded areas). The image is recoded as a binary file representing target areas (0 = no water 1 = water). Binary files from different dates were subtracted and highlighted �change files� were produced (Appendix III). X. FirX. FirX. FirX. Fire Scar Mappinge Scar Mappinge Scar Mappinge Scar Mapping Another serious challenge for the Pantanal is the evident increase in erosion and sedimentation. Human activities accelerating this natural process include clearing the land for agriculture, the opening of new roads, logging, extensive burning, and so forth. Fires of human origin are particularly prevalent in the Brazilian Pantanal and highlands during September-October, and are used by ranchers to clear old pasture in order bring up tender, green shoots for their cattle or as an easy means of clearing the land for agriculture. Fire scars may be detected using satellite imagery because the spectral response pattern of a fire scar is markedly different from that of unburned vegetation. Burn scars are bare patches of ground where some of the plant cover has been burned off by either controlled or uncontrolled fires. Scars occur as isolated patches, as somewhat orderly

16

clusters, or as an interconnected series. Scars can be of any shape, and they range in size from small patches of less than an acre to large patches and streaks tens of kilometers long. Most edges are sharply defined, although the downwind border can be diffuse. During and for some time after a burn, the patch typically is darker than the surrounding area because it is covered with dark ash and soot. This dark tone persists until the wind or rain sweeps the surface clean (Fig. 2). Then the tones reverse and the exposed soil mantle appears brighter than the surrounding soil, vegetation, and ground litter. With time, the patch again may darken as vegetation is re-established. Consequently, aerial photos and satellite images commonly show a historical assemblage of burn scars that range in tone from very dark to very bright. Burn scar patterns on area satellite images, particularly clusters with a variety of tones, indicate the presence of people and grazing animals. Depending on soil type and condition (wet, dry, etc.), the exposed soil mantle of a recent burn scar can be softer and less stable than mantles of more heavily vegetated areas. Fire scars on the surface generally contrast strongly with unburned surfaces in the visible (0.4-0.7 mm), near-infrared (NIR)(0.7-2.0 mm), middle infrared (MIR) (2-4 mm) portions of the electromagnetic spectrum. A potentially useful source of remotely sensed information for fire-scar detection includes radiance reflected from the surface in the mid-infrared (MIR) portion of the electromagnetic spectrum. One widely used approach for fire-scar detection relies on time-series of the Normalized Difference Vegetation Index (NDVI), which is strongly correlated with green vegetation cover, photosynthetic activity, and primary production. Its application in fire studies is consistent with damage to green vegetation canopies that occurs as a result of fire. The ash and black carbon deposited on the surface after biomass burning result in a large decrease in NIR. Therefore, NDVI value drops substantially after a fire event, particularly if the fire has burned a large enough area of green vegetation within a pixel. A critical question with these and other studies of fire scars is the length of time they can be detected. The blackened, carbonized material deposited on the surface after fires may be transported by wind and covered quickly by canopy litter or regenerating canopies over a matter of days or weeks. The following approaches for fire scar detection were tested for the Pantanal region: - Time series NDVI using RGB Co-Display - NDVI differencing - Digitizing of burn scars using on screen interpretations of various band

combinations Time series NDVI and NDVI differencing appear to produce the best results for visually detecting landscape scale, clear-cut and burned areas in forest and heavily vegetated areas. A decrease in the infrared coupled with an increase in red leads to a large

17

decrease in the calculated NDVI for a burn scar compared to that of unburned vegetation. The rationale of this procedure is that it highlights areas showing a change in time, normally associated with fire damages and vegetation re-growth. The decorrelated data produced through this process were of great value in enhancing regions of localized change in NDVI. Further, the surface temperature of recently burned areas also increases due to the dark surface left behind after a fire and the absence of cooler vegetative material. In summary, when mapping burn scars using multi-spectral satellite imagery, we are looking for areas that show lower than expected near infrared and NDVI and higher than expected red and thermal values. The spectral contrast between burned and non-burned areas disappears with vegetation recovery, so burn-scar mapping should be carried out as soon as possible after the fire. In the case of the Pantanal, the purpose is not to map a single fire but rather to assess the area burned by hundreds of individual fires that start and end on different dates through several months or years. XI. Updating RoadsXI. Updating RoadsXI. Updating RoadsXI. Updating Roads There is just one major road in the Pantanal, the Transpantaneira - a raised dirt road that runs for 200 km, pocked by craters and punctuated by 114 precarious wooden bridges due south from the town of Poconé. Originally, the Transpantaneira was intended to extend further southwest to Corumbá, but lack of finance, technological problems as well as ecological considerations resulted in the road terminating at Porto Joffre, a small collection of buildings with a landing stage. The raised, dirt road was created by excavating either side, and the resulting huge ditches remain filled with water throughout the year providing refuge for wild life during the long dry period. Modifications of natural hydrological cycles in the Pantanal will allow the gradual expansion of trails and roads. Expansion of the cattle ranching industry is also a contributing factor. This human-induced landscape change was tracked by digitizing all discernable linear features on edge-enhanced satellite images from multiple dates (Fig. 3). The information recorded for each new digitized road included type, length, name, and country. The categories for �Type� are as follows: - Major roads (surfaced) - Secondary roads (surfaced and unsurfaced) Secondary roads may be (1) spur roads

from the principal roads (can be paved) and usually have less traffic volume than primary roads (2) unpaved, but have increased the right-of ways from tracks, which usually are no more than one lane based, literally, on tire tracks.

- Major tracks (unsurfaced -passable all year) - Minor tracks (unsurfaced - passable in the dry season only)

18

XII. Partner DeliverablesXII. Partner DeliverablesXII. Partner DeliverablesXII. Partner Deliverables The following section documents technical aspects of the data development tasks completed by each partner for the UPRB GIS pilot project. The partners adhered to agreed terms of file naming conventions, table structures, projection and metadata standards established by the tri-national group. The main data development tasks were 1) Flood Analysis; 2) NDVI Change Detection; 3) Multi-date burn scar delineation; 4) Updated roads coverage and 5) Fieldwork tasks. The pilot area was divided by country boundaries and each task assigned to a specific partner within each country for development. Additional tasks and data provided by the various partners has also been documented and included in the database. The description of the work and results for each of the main tasks for each participating country is described below.

a. Flood Analysis Bolivia-Flood WWF Bolivia was responsible for the flood analysis work for the Bolivian portion of the pilot area. A suitable country boundary for Bolivia was not available at the time so all georeferenced satellite images for the pilot area were subset in a rectangle that covers Bolivia with a 20 km buffer (Table 2). The subset area is approximately 1,600,000 hectares. This procedure considerably reduced the volume of information and processing time. Table Table Table Table 2222. Image extent of Pantanal study area in Bolivia.. Image extent of Pantanal study area in Bolivia.. Image extent of Pantanal study area in Bolivia.. Image extent of Pantanal study area in Bolivia.

CornerCornerCornerCorner XXXX YYYY Left Bottom 271131.630843 7743382.049088 Top Right 435516.030259 7870213.745705

Unsupervised classifications were run on the images dated 7/30/98, 11/19/98, 6/09/97 and 11/14/99 to analyze the seasonal change in water level and maximum/minimum flood extent. All bands of imagery were used in the classification except for the thermal band because of its course resolution. The visible bands of the November 1999 image were excluded due to heavy smoke in the area. The first classification attempt specified 32 output classes, however, several of the water classes included cloud shadow. Therefore, a second classification using 220 classes was performed to separate the spectrally confused classes. This resulted in more clearly defined flood classes on various land cover types. A disadvantage of this process is the large number of thematic classes resulting in a limited amount of image pixels per class

19

making water identification more difficult. The interpretation and selection of the classes that correspond to flooding is a complex process because the Pantanal is made up of a heterogeneous mosaic of vegetation with varying levels of flooding. Various band combinations and local knowledge of the area helped to identify the bodies of water and more humid areas. To lower the level of subjectivity, the Humidity Index was calculated based on the ratio of bands 5 and 2. This band ratio image provided a visual reference in the verification of water classes. After the water classes were defined, the data was recoded to binary with values of 1 for water and 0 for all other cover types. Afterwards, a comparison between dates was carried out to quantify the changes between seasons as well as dry years versus wet years. For this comparative analysis, the �Image Difference� function in ArcView was applied to compute total areas that increased or decreased between two dates. A comparison of average, wet and dry years show a notable difference in the extent of flooded area during 1997 and 1999. The maximum flooded area covers 573,955 ha in June 1997, inundating almost the entire Patanal within the subset area (Fig. 4). The minimum flooded area is 140,455 ha in November 1999. The Chiquitano Forest in the northeast portion of the subset area and the Chaco on the west side of the Negro River towards the west seem to be the natural borders of this process. In general, the minimum flooded area is concentrated to the north of the Paraguay River in Bolivia and is very much reduced towards the southeast. Between both dates, a decrease of 433,500 ha was observed between the maximum and minimum area flooded. In addition, an increase of 24,012 ha was identified mainly within the limits of the palm forests and the forest to the north (Fig. 5). The seasonal flooded area for the Bolivian portion of the project area in July 1998 was 213,069 ha, and reduced to 158,331 ha in November 1998. The reduction of flooded area was observed in the margins of the Paraguay and Negro Rivers and in the northeastern part of the Pantanal in Bolivia (Fig. 6). Brazil-Flood EMPRAPA, Brazil, was responsible for calculating the maximum, minimum and seasonal flood extent for the Brazilian portion of the pilot project area. The satellite images were processed and flooding was quantified using image processing techniques in ERDAS Imagine 8.4 software. Seasonal water level change analysis was completed according to the agreed methods. Five dates of satellite imagery were subset to the Brazil country boundaries. Principal components analysis (PCA) was performed on each date of imagery prior to classification. However, several problems were encountered with the ISODATA classification on the PCA images, so it was decided to run ISODATA directly on the

20

original 6-band image file. The water classes were identified and post classification cleaning was applied to the binary flood data to remove spectral confusion between some flooded areas, ash and terrain shadow. The image differencing process was performed in ERDAS to compare change in flooded area between 5 combinations of dates (Table 3). Table 3. Change in flooded area for the Brazil portioTable 3. Change in flooded area for the Brazil portioTable 3. Change in flooded area for the Brazil portioTable 3. Change in flooded area for the Brazil portion of the Pantanal pilot project area.n of the Pantanal pilot project area.n of the Pantanal pilot project area.n of the Pantanal pilot project area.

Change Detection [ha]Change Detection [ha]Change Detection [ha]Change Detection [ha] ComparisonsComparisonsComparisonsComparisons DecreaseDecreaseDecreaseDecrease UnchangedUnchangedUnchangedUnchanged IncreaseIncreaseIncreaseIncrease

111499-112388 dry year, low flood � dry year, low flood

13,611 446,023 146,996

111499-111998 dry year, low flood � average year, low flood

10,435 435,270 109,368

112388-111998 dry year, low flood - average year, low flood

92,610 432,797 109,368

073098-060997 average year, high flood � wet year, high flood

24,953 393,077 692,785

111998-073098 average year, low flood - average year, high flood

107,978 441,077 87,109

Paraguay-Flood GIS analysts at Guyra, Paraguay, carried out the image processing for flood extent analysis for the Paraguayan portion of the pilot project area. The methodological framework and procedures presented during the project workshops and in the associated technical support documents were followed. First, each full satellite image underwent processing, which although time consuming, had the advantage of facilitating the comparison of results between the three counties within the pilot project area. Next, a subset of each scene corresponding to Paraguayan territory was made, following the national border as delimited in maps obtained from the Rural Welfare Institute (Instituto de Bienestar Rural, or IBR). These are considered the most up-to-date and accurate maps available in Paraguay. The Paraguayan subset of the image was used in all subsequent analyses and calculations (for example, of the percentage of the Paraguayan Pantanal inundated in each scene). It should be noted that a slight difference was found to exist between the position of the northern border of Paraguay with Bolivia (the �dry frontier�, west of the río Negro) as shown by the maps of the IBR and those of the Military Geographic Service (Servicio Geográfico Militar, or SGM). The total extent of Paraguayan territory included within the pilot study area was calculated as 1,588,675 hectares.

21

To evaluate seasonal flood extent, an unsupervised classification with 24 classes was applied to all six Landsat images. The thermal band was excluded from the June 1997 image due to the low image quality and the presence of high levels of water vapor and smoke in the atmosphere at the time. To improve the accuracy of the identification and classification of water bodies, all clouds and shadows were removed manually (using Erdas Imagine). Such artifacts were identified through the comparison of different band combinations. Next, a supervised grouping of the 24 classes was undertaken, using both the experience of the interpreter and fieldwork data, to differentiate between water bodies and humid terrestrial classes. Finally, chronologically sequential paired sets of images were compared to obtain the increase or decrease in flooded areas, and to produce the corresponding maps and views. Comparisons were made between a total of 15 pairs of images (Fig. 7). Table 4 presents the extent of flooding in Paraguayan territory for each image and Table 5 presents the extent of flooding for the entire scene.

Table 4. Total extent of PTable 4. Total extent of PTable 4. Total extent of PTable 4. Total extent of Paraguayan territory within the pilot area araguayan territory within the pilot area araguayan territory within the pilot area araguayan territory within the pilot area = 1,588,674.6 ha= 1,588,674.6 ha= 1,588,674.6 ha= 1,588,674.6 ha

DateDateDateDate Extent of Extent of Extent of Extent of Flooding (ha)Flooding (ha)Flooding (ha)Flooding (ha)

% Area Flooded% Area Flooded% Area Flooded% Area Flooded

11-1988 12,021.1 0.76 06-1997 31,377.6 1.98 07-1998 25,706.2 1.62 11-1998 29,420.8 1.85 11-1999 6,869.5 0.43 12-1999 22,628.3 1.42

Table 5. ExteTable 5. ExteTable 5. ExteTable 5. Extent of flooded area in complete scenes with a common area between nt of flooded area in complete scenes with a common area between nt of flooded area in complete scenes with a common area between nt of flooded area in complete scenes with a common area between scenes of 2,974,633 hectares.scenes of 2,974,633 hectares.scenes of 2,974,633 hectares.scenes of 2,974,633 hectares.

DateDateDateDate Extent of Extent of Extent of Extent of Flooding (ha)Flooding (ha)Flooding (ha)Flooding (ha)

% Area Flooded% Area Flooded% Area Flooded% Area Flooded

11-1988 198,832.8 6.68 06-1997 789,246.8 26.53 07-1998 146,144.4 4.91 11-1998 158,316.8 5.32 11-1999 44,324.6 1.49 12-1999 87,688.8 2.95

Maximum flooding of the Paraguayan portion of the study area occurred during June 1997 (31,378 ha), at the same time the maximum extent of flooding was detected throughout the whole of the study area (789,247 ha). An almost equally large flood

22

event occurred during November 1998, when 29,421 ha were found to be inundated. The maximum extent of flooding within Paraguay covers approximately one third of the total area of grassland and palm savanna habitats in the vicinity of the río Paraguay (an area of approximately 90,657 ha). The extent of these habitats suggests that in exceptional cases, the whole of this area may be flooded. Analyses of images from periods of extreme flooding (for instance in 1982/1983) may prove revealing. Figure 8 depicts areas of concentrated flooding with the Paraguayan portion of the Pantanal. The minimum area flooded was detected during November 1999, when only 6,870 ha were inundated, again coinciding with a similar event throughout the whole of the study area. It is worthy of note that only one month later (December 1999) the flooded area had increased almost four times in extent, to 22,628 ha, underlining the rapidity of changes in the local flood regime.

b. NDVI Change Detection

Bolivia-NDVI WWF Bolivia completed the analysis of the changes in vegetation biomass for the entire pilot area as well as the Bolivian portion of the pilot area using the same rectangular subsets described in their flood analysis. The observations that follow include an overview of trends in biomass change for the Bolivian portion. The process was undertaken with NDVI data from November 23, 1988, November 19, 1999 and December 24, 1999. Later, these indices were compared to show the changes at 25% and 50% over a ten-year period and between seasons. A layerstack of the NDVI images for a three-year period (NDVI RGB Co-Display) was carried out in order to visualize areas of increased change in biomass and areas that have undergone less. Between November 1988 and 1999 a 25% reduction of NDVI was observed in the biomass mainly in the Paraguayan Chaco (Fig. 9). The anthropogenic causes include large cultivation and land clearing activities. This is also visible on the forest borders of the Pantanal and Chaco as well as south of the Paraguay River in Brazil. During a recent field visit, several cattle ranches along the southern edge of the Paraguay River in Brazil were observed which could possibly explain the change in vegetation patterns. It is possible that the reduction of biomass in areas of the Pantanal is due mainly to fires that occurred on a large scale during 1999. Bolfor (2000) reported in August 1999, about 2,872 hot spots in Bolivia and that 41.5% of the Otuquis National Park was burned. This can be spatially defined to the south of the Paraguay River where burned areas coincide with patches of reduced biomass. However, this relationship was less evident in the Bolivian Pantanal. There are also increases in biomass along the Negro River and around small lagoons in the Bolivian Pantanal. NDVI changes of 50% were

23

observed in the same areas described above and are generally less extensive. There are also increases in NDVI around the Negro River and small lagoons in Bolivia. This is possibly explained by the increased accumulation of floating vegetation along the river during times of year when the water level is low. Between the months of November 1998 and 1999 a 25% reduction of NDVI was observed in extensive areas in the Chaco, south of the Paraguay River on the border of the Pantanal with the Chaco and forest. Within the Otuquis area of the Pantanal there is less of a reduction in NDVI. The causes could be the draught from 1999 and fires south of the Paraguay River. Once again, an increase of 25% is found along the Negro River and lagoons to the north of its course. The changes of 50% are found in agricultural areas and areas that have been cleared in the Paraguayan Chaco as well as in the Pantanal on the eastern margin of the Negro River in Brazil and on the border of the Chaco � Pantanal. On a seasonal level, during 1999, there are large areas where there was a 25% increase in biomass covering almost the entire Pantanal both in Bolivia as well as in Paraguay and Brazil, and a large area in the Chaco and the borders between the Pantanal-Chaco and Pantanal-forest (Fig 10). Observing satellite images in various band combinations (3-2-1, 4-3-2, and 5-4-3) demonstrated that in December 1999 the area covered by pasture was greater than in November (a month in which there is extensive burning). Also, the humidity index for December was greater than those for November, causing regeneration in vegetation. An increase of 50% was found in cleared areas within the Chaco, the Pantanal in Otuquis, south of the Paraguay River in Brazil and on the border between the palm forests that neighbor the forest in Otuquis. Between November 1998 and December 1999 more increases than decreases in NDVI were noted for both 25% and 50% in the Bolivian Pantanal (Fig. 11). The reduction in biomass was concentrated on the border of the Pantanal � Chaco and area of human influence of the Paraguayan Chaco. In this case, the NDVI image is affected by the presence of smoke. Table 6 summarizes the areas in which biomass decreased or increased by 25% and 50% between the dates indicated (covering an area of approximately 1,624,000 ha). Table 6. Results of NDVI differencing for the subset area including the Bolivian portion of the Table 6. Results of NDVI differencing for the subset area including the Bolivian portion of the Table 6. Results of NDVI differencing for the subset area including the Bolivian portion of the Table 6. Results of NDVI differencing for the subset area including the Bolivian portion of the Pantanal pilot project.Pantanal pilot project.Pantanal pilot project.Pantanal pilot project.

25 % change in biomass25 % change in biomass25 % change in biomass25 % change in biomass 50% change in biomass50% change in biomass50% change in biomass50% change in biomass DateDateDateDate Increased area in haIncreased area in haIncreased area in haIncreased area in ha Decreased area in haDecreased area in haDecreased area in haDecreased area in ha Increased area in haIncreased area in haIncreased area in haIncreased area in ha Decreased area in haDecreased area in haDecreased area in haDecreased area in ha

11/23/88 � 11/14/99 25.148,16 382.058,10 5.368,95 250.928,57 11/19/98 � 11/14/99 99.193,23 589.830,84 86.967,51 301.126,87 11/14/99 � 12/24/99 808.154,28 32.393,61 188.724,24 25.026,93 11/19/98 � 12/24/99 219.856,86 44.208,72 68.728,95 10.230,84

24

Brazil-NDVI IBAMA carried out the image processing for NDVI change detection for the Brazilian portion of the pilot area. In order to subset the Brazilian area of the satellite image, three polygon shapefiles were created, i.e., corte_ndvi, corte_9899, and corte_982499. NDVI for each date listed above was computed followed by image differencing in ArcView Image Analysis. The difference of two dates of NDVI images were computed with change thresholds of 25% and 50%, with the following filenames: - ndvi_112388 � ndvi_111499 = sub_8899br25.img and sub_8899br50.img; - ndvi_111998 � ndvi_111499 = sub_9899br25.img and sub_9899br50.img; - ndvi_111499 � ndvi_122499 = sub_9999br25.img and sub_9999br50.img; - ndvi_111998 � ndvi_122499 = sub_98249925.img and sub_98249950.img; The resulting highlighted changes for each NDVI difference image was mainly due to burned areas, changes in areas of exposed soil, and flooded areas (Figures 12 and 13). Paraguay-NDVI Fundación Moisés Bertoni (FMB) completed the NDVI change analysis for the Paraguayan portion of the pilot area. A digital Paraguayan country boundary was used with a 3km buffer in order to establish an area of interest focusing exclusively on the Paraguayan side of the Pantanal. A mask was set prior to evaluating the properties of the NDVI images for each date. NDVI was calculated for images dates: 11/23/88; 11/19/98; 11/14/99 and 12/24/99. Next, the NDVI difference between dates was calculated, and a highlighted change image was produced. A NDVI RGB Co-display image was also created using the Layer Stack tool in the ArcView Image Analysis extension. The dates of the NDVI images used were 11/23/88, 11/19/98 and 11/14/99. Landcover verification was achieved by creating a randomly distributed point theme over the Paraguayan portion of the pilot area. The attribute table for the landcover verification includes: date, landuse, coordinates, and original landcover type. An Agricultural Use Map and a Vegetation Map of the Western Region of Paraguay were used to verify landuse and the original landcover. Also, photographs taken by Guyra Paraguay in the overflights made on December 8, 2001 were used as verification as well as those collected during the overflight coordinated by FMB and Guyra in March 22-23, 2002. The NDVI images for 11/23/88 and 11/19/98 show very little significant change in the Paraguayan portion of the pilot area. However, change in NDVI is apparent between 11/19/98 and 11/14/99. This could be attributed to the isolation of the area, as it was not until the late 1990's that the region started to be inhabited (not considering the indigenous settlements).

25

c. Multi-date Burnscar Delineation Bolivia-Burnscar urned areas were digitized by WWF Bolivia based on images from November 23, 1988, July 30, 1998, November 19, 1998, November 14, 1999 and December 24, 1999. The December 1999 image was used in order to visualize changes occurring in a one-month interval rather than seasonal variations. In order to identify burned areas, several band combinations were used that highlight ash and thermal anomalies on the landscape: 6-5-4, 3-2-1, 7-5-4, 5-4-3 and 4-5-7. The different polygons were delineated using the 6-5-4 combination with the �seed tool�, applying multiple values of �seed radius� depending upon the dimensions of the areas. A seed radius value was applied in different parts of the polygon until an adequate result was obtained. Determining the boundaries of a burned area was complex because the majority of the limits are often diffuse due to the continuous seasonal burning, wind and rain. Also, limitations in the manner in which ArcView displays the images at certain levels of zoom causes the inclusion or exclusion of certain areas depending on criteria employed by the technician. The burning in November 1999 was extensive and presented vague borders. In order to facilitate the identification of burned areas and digitizing of the polygons, a non-supervised classification was carried out on the thermal band. Classes were selected that coincided with the areas visibly burned in the 6-5-4 band combination. These classes were recoded and converted into polygons that were the base for the final delineation of the burns. It was observed that fires occurred in marsh areas where vegetation was burned superficially. Thus, the possibility should not be discarded that the thermal band is not 100% adequate for detecting burned areas. Further, the rapid regeneration of grass was observed on burned areas (the burn scars appear dark green). It was observed that large areas of grasslands were burned in the entire Pantanal as well as the Palm forest, particularly in the transitional part of the forest and Chaco within the Bolivian Pantanal. Table 7 summarizes the dimensions and statistics of the burns for the different dates within a subset rectangular area of approximately 1,545,802 has.

26

Table 7. Total area recently burned or actively burning in the subTable 7. Total area recently burned or actively burning in the subTable 7. Total area recently burned or actively burning in the subTable 7. Total area recently burned or actively burning in the subset area including the Bolivian set area including the Bolivian set area including the Bolivian set area including the Bolivian portion of the pilot area for each date of imagery.portion of the pilot area for each date of imagery.portion of the pilot area for each date of imagery.portion of the pilot area for each date of imagery.

DateDateDateDate Total Burned AreaTotal Burned AreaTotal Burned AreaTotal Burned Area HasHasHasHas

Number of Burned Number of Burned Number of Burned Number of Burned PolygonsPolygonsPolygonsPolygons

Mean AreaMean AreaMean AreaMean Area HasHasHasHas

11/23/88 4794.97 118 104.24 07/30/98 51248.16 128 400.38 11/19/98 53586.84 28 1913.81 11/14/99 114667.38 246 466.13 12/24/99 42119.21 46 915.63

Brazil-Burnscars Universidade Católica Dom Bosco (UCDB) completed the delineation of burn scars for multi-date imagery for the Brazilian portion of the pilot area. The occurrence of forest wild fires and intensive land burning in the pilot area was detected and area estimation was performed by using five dates of Landsat TM images covering path 227 row 74. Burn scars were classified as active or recent in the dataset. Actively burning areas were identified by the smoke plumes apparent in the 3-2-1 (RGB) band combination and dark brown areas with bright yellow/orange on the fire front. The areas identified as recent burns were dark brown in a 6-5-4 (RGB) composite image and did not have smoke. Total burned area was calculated for each burn scar polygon delineated in this process. In order to compare the burn scars temporal change, five Landsat images for different months and years were analyzed, including: 11/23/1988; 07/30/1998; 11/19/1998; 11/14/1999 and 12/24/1999. Total burnerd area calculations for the five dates were 30,805 has for 11/23/1988; 55,490 has for 07/30/1998; 11,749 has for 11/19/1998; 92,279 has for 11/14/1999 and 133,917 hectares for 12/24/1999. Figures 14 and 15 show the burn scars delineated from two of the image dates. It was observed that the total burned area increased from 30,805 has to 133,917 has � a difference of 103,112 has � between November 1988 and November 1999. It was also noted that an exceptionally low burn scar area of 11,749 hectares was delineated from the November 1998 image. The rainfall amount of 94 mm in October and of 285 mm in November of 1998 that was not exceptional. The results indicate that the land burning activities in the Mato Grosso do Sul state have been increasing rapidly. Currently, the state government is aware of negative impacts of burning on the environment and public health and is taking actions in educating farmers and citizens to prevent accidental fire occurrences and to adapt planned, controlled burning strategies. Paraguay-Burnscars Fundación Moisés Bertoni (Paraguay) produced the multi-date burn scar coverages for the Paraguayan portion of the pilot area. The seed tool in the ArcView Image Analysis

27

extension was used to delineate burn scars in the Paraguay portion of the Pantanal within path 224 row 74 for dates: 11/23/88, 07/30/98, 11/19/98, and 11/14/99. In the image dated 06/09/97 there was apparently no recent or active fires due to the high flood season. The fields used for the shapefile attribute table were: Date, Landuse, Type and Area. Photographs taken during a recent overflight and a map of Agricultural Use in the Western Region of Paraguay were used to define landuse. Most of the burn scars appeared in natural regions with only a few occurring in grazing areas. The image from 11/14/99 shows a large burned area in the northern portion of Paraguay that continues to Bolivia. Other significant burns occur next to the rio Paraguay and rio Negro. Although these spots were not mapped as �recent� or �active� because there was no detectable ash on the surface or smoke, it is important to note these areas were devoid of vegetation due to burning. A lengthy drought occurred in 1999 and caused extensive burns in the Western and Northern zones of Paraguay. The NDVI Differencing method highlighted these areas. Another image date that revealed a significant burned area was 07/30/98. As with the rest of the dates, most of the burns occurred in natural areas and only a few were recorded in farming or grazing areas. Table 8 shows the results obtained, including the landuse and area of the burn scars for each date.

Table 8. Results of burn scars of the Paraguayan portion of the Pantanal GIS Pilot Project.Table 8. Results of burn scars of the Paraguayan portion of the Pantanal GIS Pilot Project.Table 8. Results of burn scars of the Paraguayan portion of the Pantanal GIS Pilot Project.Table 8. Results of burn scars of the Paraguayan portion of the Pantanal GIS Pilot Project.

Area in hectaresArea in hectaresArea in hectaresArea in hectares DateDateDateDate Farm Farm Farm Farm GRGRGRGR NNNN AGAGAGAG EXEXEXEX TotalTotalTotalTotal

11/14/99 869 5984 8 6861 11/19/98 136 6048 34 6218 7/30/98 17 12031 12048 11/23/88 4794 4794

d. Updating Roads Bolivia �Roads WWF Bolivia updated digital roads and trails coverages for the Bolivian portion of the pilot area using images dated 11/14/99 and 12/24/99 (multispectral and panchromatic bands). The �Edge Detect� image enhancement utility in ArcView Image Analysis was used on both scenes. The digitizing was carried out while viewing the original band data with different band combinations simultaneously with the edge-enhanced images.

28

Within the Otuquis National Park in Bolivia a main trail was identified that runs from Puerto Suarez and nearly reaching Puerto Busch. There were also some secondary trails identified towards the north. On land it was possible to verify that these are simple dirt tracks covered by grass that are only passable in the dry season. Comparing the roads from the topographical chart of 1976 it was observed that the track from Puerto Suarez to Puerto Busch bordered the Negro River on the west. This road now runs through the center of the area and is probably producing some local drainage problems. In general, the roads in the southern portion of the Bolivian Pantanal are transitional and poorly defined � especially in areas that flood. In other instances, these roads in floodable areas seem to be cattle trails through grasslands � a common factor between the three countries. It is possible that geometric lines in the imagery are actually fences and not roads, making it necessary to carry out ground verification. Brazil�Roads IBAMA created a roads coverage for the Brazilian portion of the pilot area. A Landsat image of path 227 row 74 dated 12/24/99 was used to digitize new and existing roads. To facilitate the process, an edge-enhanced image was created and used as a background. A line shapefile was digitized with an approximate scale of 1:20,000 in ArcView. The road coverage was digitized inside the Pantanal complex within the pilot area and clipped to the Brazil country boundary with a 20 km-buffer (Figure 16). Initially, the Plano de Conservação da Bacia do Alto Paraguai (PCBAP) roads coverage was to be used and updated. However, a significant shift was apparent between the PCBAP roads coverage and the imagery and could not be corrected. Further, EMBRAPA, Brazil, will be submitting 10 digital maps of the pilot area that were digitized via a digitizing tablet and include roads information. The maps were georeferenced using existing scanned, georeferenced maps. Paraguay�Roads Guyra Paraguay conducted a detailed digitization of all roads and tracks for each satellite scene (Fig. 17). The final product is a file containing all roads present in the pilot area from 1988 to 1999. Additional separate files present the roads detected per image date. The original road classification, obtained from analysis of the satellite images, has since been refined using information gathered during fieldwork in July 2002. The new road classification file contains more than 1200 records. It is important to note that roads within the first category, of �Major Roads�, only exist within the two principal urban centers in the Paraguayan part of the project study area,

29

Bahía Negra and Fuerte Olimpo, both located along the río Paraguay. These �Major Roads� are short sections of �empedrado� (�crazy paving�). All other roads in the area correspond to the categories 2-4, with most in category 4 �Minor tracks�, known in Paraguay as �picadas� (Fig 18).

e. Field Work Bolivia-Fieldwork WWF�Bolivia completed fieldwork for the Bolivian portion of the pilot area. On March 1, 2002 an overflight was scheduled to acquire photographs of the Pantanal with coordinates using a digital camera and GPS. Several difficulties were encountered including the delay in collecting GPS coordinates while the plane was in motion as well as the camera�s sensitivity to the vibration of the plane. Photos were taken with the camera set on medium quality and high resolution. Approximately 122 photos were recorded of the pilot area and then connected to a digital coordinate file. This allows the visualization of the Pantanal landscape and documentation of some of the threats, such as fires and cattle ranching within the Otuquis National Park. On April 8-12, 2002 field data was collected along the Paraguay River starting in Puerto Suárez to Bahía Negra in Paraguay. The objective was to visit Otuquis National Park navigating along the Negro River. Different points visited along the trip were documented with the digital camera and GPS (including threats, vegetation formations, as well as verification of roads that were digitized from satellite imagery. More than 150 photographs were obtained with coordinates that are now linked to a GIS. Brazil-Fieldwork Instituto de Meio Ambiente � Panatanal (IMAP) coordinated fieldwork tasks with the technical support of IDATERRA and UCDB. Two fieldwork trips were completed using a GPS Garmin III Plus to collect ground control points (GCP) and to identify vegetation and landuse along the Paraguay River and its tributaries. Ground control points (GCP) and coordinates for roads and tracks were collected in Brazil and Paraguay. For each GCP, a digitally georeferenced photograph was recorded. The first trip was December 3-7, 2001 with technicians from IDATERRA, IMAP-SEMACT and UCDB. The trip departed from Campo Grande and ended in Corumbá and resulted in the collection of GCP�s from three countries (Brazil, Bolivia, Paraguay). The internal roads in this region are unpaved and made access difficult. Approximately 3,000 km were covered during the first five days of fieldwork.

30

A pair of Geodesic GPS 4.600 LS Trimbles and a Topographic GPS Pro XRS Trimble were used in the field. The coordinates were taken from the Network Brazilian Geodesic (IBGE) landmark placed at a water box at the facilities of EMBRAPA, Corumbá, using the Geodesic GPS 4.600 LS Trimble. This base served as the control point for the field data collection. The geodesic GPS served as the base for post-processing the data that were collected in field along with the topographic GPS Pro XRS Trimble. Chronology of Fieldwork: Monday, 03: Collected GCP�s from Campo Grande to Corumbá and meeting in

EMBRAPA; Tuesday, 04: Collected CGP�s in Bolivia and Brazil and then the Forte Coimbra base; Wednesday, 05: Collected GCP�s in the Nabileque region; Thursday, 06: Collected points in Corumbá-Bonito, passing by Bodoquena; Friday, 07: Collected GCP�s of Bonito to Porto Murtinho. Data was collected over

the border in Paraguay and then from Porto Murtinho to Campo Grande, passing by Nioaque, Jardim and Sidrolândia (these 3 cities are outside the pilot area).

Part of local roads and their intersections were photographed, positioned with GPS, identified and recorded on working maps with images using the Hotlink extension in ArcView. The second field trip identified vegetation type along the Paraguay River and it�s tributaries. This trip took place during May 14-17, 2002 with UCDB and logistical support of the patrolling staff of the Environmental State Military Police. Two groups crossed the Paraguay River with a motorboat. The choice to travel by river was made because it was the flood season and the roads were very wet, making it impossible to get a distribution of vegetation samples in the pilot area. The map produced from the fieldwork data includes the Paraguay River crossing one of the sources of the Nabileque River, and Corixo do Veado Gordo (perennial water body with a narrow channel which links adjacent bays). These areas were photographed, positioned with GPS, identified and recorded on working maps and images to provide information about vegetation type. The Pantanal vegetation classification system is very diverse and is based on the RADAMBRASIL/IBGE project, an extensive study of Brazil�s biomes/regions published in 1980. The study provided an integrated analysis of natural resources using remote sensing and designated a new denomination to Cerrado - the Savanna (PCBAP, 1994). Currently, local botanists have not accepted this denomination but regional official publications use either of both terms. To identify the pilot area vegetation, the regional

31

classification was used including the terminology �Cerrado�, but breaking it down into three physiognomies (light Cerrado, Dense Cerrado, and Cerradão). From a biological standpoint, there are several distinct biogeographical influences in the Pantanal. Around 70% to 80% of the sub-basins of Mato Grosso do Sul are located in Cerrado and contain characteristic vegetative formations such as Cerradão, Campo Sujo, Campo Limpo and others). However, the pilot project area is located in the subregion of Pantanal do Nabileque and holds a pronounced xeric vegetation which shows a strong floristic affinity to the western Paraguayan ecosystem (the Chaco) and the dry forests of Bolivia. In the pilot area the following type vegetations were identified: - Deciduous Forest covering residual mountains of Serra do Urucum-Amolar near

Forte Coimbra. - Alluvial Forest along Paraguay River. The margins of the rivers and swamps also have

distinctive woody vegetations types that vary from low scrubs to tall gallery forest depending on local conditions of the soil, flooding and topography.

- Cerrado - three types: light, dense and Cerradão; - Swamp vegetation types (along seasonal/or permanent flooded areas). - Parklands- Pioneer formation, encompassing large patches with predominance of

the Palm species, Carandá (Copernicia alba), with presence of xeric vegetation type, Aromita, Cereus, Opuntia, associated along the lower substrate. This formation has predominance along the pilot area in both the Paraguay River margins along Brazil and Paraguay borders, especially in the southern and western region around Corixo do Veado Gordo.

An abbreviated identification of vegetation type was carried out from field observations along roads and rivers. The group reviewed all Pantanal vegetation maps of Mato Grosso do Sul (RADAMBRASIL/1982, Macrozoneamento Geoambiental do MS/1986 and PCBAP/1994). These maps were used not to compare the local patterns of vegetation type (because the scale of all existing published maps was 1: 250,000), but to give support to the vegetation classification system that is being designed by other partners for the Pantanal pilot project (see Landcover Classification section). In the future, after finalizing the vegetation classification system derived from aerial photos from 1966, this product will contribute information on patterns of land cover change in the pilot area and also the entire URPB. Embrapa Pantanal also conducted fieldwork in January and February 2002 for the collection of ground control points using the Garmin III Plus GPS. The preparation and fieldwork took place as follows:

32

Using SPRING software, a Landsat image from WWF Brazil was used to design a sampling strategy along road intersections and to guide the GCP field data collection. Using ArcView 3.2, themes of main roads and points on roads intersections were digitized. The point theme and roads were re-projected from UTM WGS84 to Geographic WGS84, using Projection Utility Wizard and input in the GPS Garmin III+ using the DNR Garmin extension. Using the GPS, real-time navigation was possible and the road intersections were easily located. At each point location, the coordinates were recorded and the Estimated Position Error (EPE) and Dilution of Precision (DOP) error indices were taken. A digital photograph was also collected for each point location. The GCP�s were uploaded from the GPS to ArcView and reprojected to UTM, WGS84, Zone 21. Table 9 contains the information generated for each ground control point.

Table 9. Attributes for ground control points collected in the field.Table 9. Attributes for ground control points collected in the field.Table 9. Attributes for ground control points collected in the field.Table 9. Attributes for ground control points collected in the field.

The GCP�s were plotted over the panchromatic band of the Landsat 7 image from Nov. 14, 1999 in order to evaluate the image georeferencing accuracy (Fig. 19). There is a 1-4 pixel shift between the ground control points and the panchromatic image from Image Links. The ground control points data was sent to DU and forwarded to Image Links to be used for applying further corrections to the p227/74 pilot area imagery1. Fieldwork-Paraguay An overflight took place on March 22-23, 2002, and was coordinated by Guyra Paraguay, with the participation of the GIS technicians responsible for the tasks to be completed 1 It appears that there is a North/South shift of approximately 100-120 meters (3-4 pixels) and a East/West shift of approximately 50-80 meters (1-2.75 pixels). The distribution of the GCP data is at this point confined to the northeast portion of the pilot area. There are logistical difficulties with traveling to many areas within the scene due to flooding and lack of roads. However, preliminary plans for future fieldwork are underway and partners will reserve airplanes and vehicles if more ground control points are required.

33

by FMB. The flight was made in a Cessna high wing airplane, and photographs were taken with two cameras fixed to a platform; one vertical and the other in oblique position. The first day the flight consisted of passing over the Lagunas Saladas complex, taking photographs at an altitude of 3500 ft and a speed of 100 knots. The second day, photographs were taken during the flight over the Immakata Lagoon and along the Rio Negro, at 6000 ft altitude and 100 knots. In the Rio Negro area, a short, low altitude flight was made at 500 ft and then at 3000 ft for air photo collection. These photographs are developed and are still in the process of being scanned. Three fieldwork/overflight campaigns were undertaken by Guyra Paraguay: - December 2001: Georeferenced aerial photos were taken with a Kodak digital

camera during a commercial flight from Asunción to Bahía Negra. Georeferenced photos were also collected with a Kodak digital camera during a boat trip along the río Paraguay to the mouth of the río Negro, and then along the río Negro to the border with Bolivia (Hito 12). Further, georeferenced photos were taken in the vicinity of the Fortín Patria ranch, and along the banks of the río Negro.

- March 2002: Paired geocoded vertical and oblique aerial photos were taken with analogue cameras.

- July 2002: Standardized photos with coordinates and GPS-logged tracks of principal roads were taken during an overland fieldwork campaign.

f. Additional Tasks

In addition to the above outlined tasks, each organization was asked to compile and submit existing GIS data for the pilot area as well as for other portions of the Pantanal that would be useful for the UPRB database. The following section describes the additional tasks and spatial data that were provided by the various partners.

Additional Tasks-Bolivia LICGF-University of Wisconsin compiled and delivered the following data for the Bolivian portion of the pilot area: - Monthly precipitation data from 1943-2001 for the Puerto Suarez, Bolivia

monitoring station were gathered to support the imagery selection process. - An inventory of publicly available GIS data for Bolivia was acquired at the onset of

the project. - Publicly available data sets for Bolivia with metadata are included on the CD (see

Table 10).

34

Table 10. Publicly available GIS data for Bolivia.Table 10. Publicly available GIS data for Bolivia.Table 10. Publicly available GIS data for Bolivia.Table 10. Publicly available GIS data for Bolivia.

Data Layers for BoliviaData Layers for BoliviaData Layers for BoliviaData Layers for Bolivia Approx. ScaleApprox. ScaleApprox. ScaleApprox. Scale National Boundaries 1:1,000,000 Departments (states) 1:1,000,000 Provinces 1:1,000,000 Municipalities 1:1,000,000 Roads 1:250,000 Population Centers 1:250,000 Major Cities 1:100,000 Watersheds 1:5,000,000 Lakes 1:5,000,000 Rivers 1:250,000 Rain 1:5,000,000 Soils 1:5,000,000 Contour Lines (incomplete) 1:1,000,000 Vegetation 1:1,000,000 Forest Concessions 1:250,000 Protected Areas 1:250,000 Regional Parks 1:250,000 Forest Reserves 1:100,000 Agricultural Potential 1:5,000,000 Forestry Potential 1:5,000,000

Creating metadata, or documenting data, is vital to retain investment in data development and to provide institutional memory of project compontents. It is also a federal requirement from those agencies and organizations who receive funds from the US government. The documentation challege for this project was to find a tool that would support the three different languages used by participants in this project. LICGF-UW provided metadata training using MetaLite 1.7.5 software for the pilot project group at the Puerto Suarez meeting in Bolivia during November, 2001 and in Brazil for partners from EMPRAPA during July 2002. Selection of this metadata input tool was based on an evaluation of currently available tools. MetaLite was selected because of its easy to use graphical user interface, compliance with Federal Geographic Data Committee (FGDC) standards, it�s ability to support English, Spanish, and Portuguese, and it�s export funcion into .xml which allows the easy integration of metadata into the FGDC metadata search engine. A CD was made and distributed to metadata training participants with the following materials:

35

- MetaLite 1.7.5 auto-extracting software - MetaScan button for use in viewing MetaLite metadata in ArcView GIS software - General information file including a copy of metadata standards (in spanish),

glossary of terminology, and a list of other metadata resources. - Powerpoint files with training material and information about the development of a

Global Spatial Data Infrastructure.

Metadata from all three countries has been compiled, validated, and normalized so that it can be placed on an Internet-based search engine and may be viewed by all interested parties (see web page: http://130.11.52.184/servlet/FGDCServlet). This is a temporary node made available by U.S. Geological Survey (USGS) until a permanent home may be found. In general, the metadata submitted was of sufficient quality, but upon review, a few errors were found that may need to be adjusted to provide the best possible information for potential users. For example, some metadata creators did not validate their metadata before submission. This means that it cannot be put on the search engine without entering information for the empty fields. USGS will send back the metadata that are sub-standard for editing. Other details that need to be corrected/improved are: - Many times the field, �Use Constraints,� was filled out with the response of �None.� It

may be helpful to write something like, �Please credit data source when referencing.� - Sometimes the type of imagery used was omitted. The reader�s familiarity with the

WRS-2 Landsat path/row numeration should not be assumed. - Supplemental Information is weak on methods used. Perhaps a reference to the

methods document could be added. - The project parnters may want to consider developing a common phrase to be used

in the �Purpose� field so metadata creators may reference similarly that these data were created for the Pantanal GIS Pilot Project. This reference was made in most metadata, but not consistently or in the same place. Example: �Información originada por el Proyecto Piloto Pantanal SIG�

- Some organizations listed their web site in the �Online_Linkage:� field that we may consider using to link to the data server. This may be edited by adding the DU link either before or after their organizational link.

Because only one local organization in Bolivia was involved in the pilot phase of the project, it was particularly important to disseminate results of the project to potential stakeholders and to draw more participation in the wider project. Two workshops were held - one in Santa Cruz, Bolivia, to attract many of the NGO�s that are involved in conservation activities in and around the pilot project area and another in La Paz, Bolivia to attract government support for the project and involve other international conservation NGO�s that may be planning activities in the Pantanal. WWF Bolivia presented the initial results of the Pilot Project at both events. Over 20 participants

36

from 12 different organizations attended the two meetings. Based on suggestions from meeting participants it was generally thought important to involve the National Park Service and Superintendencia of Forestry and Agriculture at some level in the wider project since they are the main regulating bodies in the area and because such a large part of the Bolivian side of the basin is in parks. Municipalities or municipal consortiums such as the Mancomunidad Chiquitania are natural targets to involve in the wider project, as are the local NGO�s working in the area. In addition to the agreed tasks, WWF Bolivia digitized the planimetry of 30 cartographic maps from the Instituto Geográfico Militar at a scale of 1:50,000 for the pilot area and includes roads, trails, rivers, lagoons, lakes and populated areas. Although outdated (1976), this is the official information available from the Bolivian government and can be used as complimentary information that allows for, especially in the case of roads, a way to track the changes that have taken place in the last 30 years. (Table 11).

Table 11. Topographic maps digitized by WWF Bolivia as part of their data collection tasks.Table 11. Topographic maps digitized by WWF Bolivia as part of their data collection tasks.Table 11. Topographic maps digitized by WWF Bolivia as part of their data collection tasks.Table 11. Topographic maps digitized by WWF Bolivia as part of their data collection tasks.

DataDataDataData DateDateDateDate FontFontFontFont CodeCodeCodeCode Format and CharacteristicsFormat and CharacteristicsFormat and CharacteristicsFormat and Characteristics Topographic maps 1:50.000

Compiled in 1985 from aerial photos from 1967. Field verification 1970

IGM 7735 (I, IV) 7736 (I, II, III, IV) 7835 (I, II, III, IV) 7836 (I, II, III, IV) 7933 (I) 7934 (I, II, IV) 7935 (I, II, III, IV) 7936 (I, II, III, IV) 8033 (IV) 8034 (III) 8035 (III, IV) 8036 (III)

Paper and digital (just planimetry) Original Data: Spheroid International, UTM 21S.

Additional Tasks-Brazil EMBRAPA prepared river stage, river water discarge and precipitation data in ACCESS format and also compiled station data acquired from ANEEL (Agência Nacional de Energia Elétrica) (See Appendix IV). Topographic maps at scales of 1:100,000 and 1:250,000 were reprojected from UTM, datum Córrego Alegre to UTM, datum WGS84, Zone 21. EMBRAPA also generated metadata for the following datasets in Metalite: - 1:250,000 maps of vegetation and soils - 1:100,000 maps of vegetation/flood, drainage and relief - Landsat images from different dates for the pilot area - Ground Control Points and digital photographs

37

Also in Brazil, IMAP coordinated with Ecotropica to build a georeferenced photo index of IBAMA�s 1:60,000 aerial photos for the State of Mato Grosso�s Pantanal. To carry out this task, a partnership was made with UCDB to coordinate the following activities: - Mosaicing and georeferencing topographic maps of the pilot area; - Rasterization of orthophotos (USAID/1966) for the following pilot area sub-regions:

Rio Nabileque, Nabileque, Forte Coimbra, Baia Negra, Fundo da Baia Negra, Barranco Branco, Morro do Campo, Porto Esperança e Aldeia Tomazia;

- Build a georeferenced mosaic of approximately 300 photos within the pilot area.

In agreement with these tasks, a georeferenced digital mosaic of the Forte Coimbra and Nabileque sub-regions (the largest area) has been submitted to DU. The Centro de Sensoriamento Remoto (CSR) unit in IBAMA, Brazil, carried out the following additional tasks: - arc coverage converted to shapefile of boundaries of South America compiled from

Digital Chart of the World � DCW, in Geographic, Spheroid Clarke 1866 and units in decimal degrees;

- polygon coverage of Brazil country boundaries compiled from Instituto Brasileiro de Geografia e Estatística � IBGE, in Geographic, Spheroid SAD69;

- polygon coverage of boundaries of the Pantanal Complex from WWF/IBGE, projection UTM, Zone 21 S, Datum WGS84, Spheroid WGS84, units in meters;

- a point coverage of cities and villages inside the Pantanal boundaries from IBGE, projection UTM Zone 21 S, Datum WGS84, Spheroid WGS84, units in meters, and converted to shapefile;

- polygon shapefile of federal conservation units inside the Pantanal Complex from IBAMA, in projection UTM Zone 21 S, Datum WGS84, Spheroid WGS84, units in meters.

Additional Tasks-Paraguay The following additional data and metadata were delivered by Guyra Paraguay: - 1:100,000 digitized national topographic maps, from the Military Geographic Service

(Servicio Geografico Militar), Paraguay. - Rural Cadastre maps, from the Rural Welfare Institute (Instituto de Bienestar Rural),

Paraguay. - Recent Landsat 7 ETM + image from the pilot area (12 June 2001). - Paraguayan national borders, and the portion of Paraguayan territory included within

the project study area.

38

g. Technical Suggestions The following suggestions were received from various partners on ways to expand or improve upon the technical methods or to improve coordination between the participating countries. - For the larger Pantanal project and future imagery orders for the UPRB, ground

control points should be provided where possible. EMPRAPA has many ground control points that could be shared for this purpose.

- More ground truthing for the evaluation of floods and flooded vegetation should be carried out.

- Flooded areas are often spectrally confused with burn scars and dense vegetation. It would be very useful to include vegetation maps and burn scar data to better evaluate flooded areas.

- The standardization of the data should be led and coordinated by a group of technicians from various local organizations for proposing technically sound methods and providing technical support.

- The analysis of floods and fires should be supported by methodologies that reduce the subjectivity. For example, humidity and vegetation indexes, algorithms to detect heat sources that could be integrated in a more advanced image processing software such as ERDAS Imagine 8.5.

- Promote as much as possible the use of GIS modeling applications, as well as the digital processing of satellite images.

- Continue to coordinate fieldwork within the three countries. - It is important to consider the effects of smoke on NDVI as it could distort the

results for areas under intense burning. - For fire scar mapping, it would be beneficial to extend the dates of analysis to other

critical months in Bolivia such as August and September.

h. Hydrological Analysis for UPRB

Universidade Cprecipitation, aand Paraguay).boundaries of the project. Tcentral part of Bolivia (80,843from 1100 mmflat and forms

Graphic 1. Upper Paraguay River Basin bGraphic 1. Upper Paraguay River Basin bGraphic 1. Upper Paraguay River Basin bGraphic 1. Upper Paraguay River Basin boundaries and locationoundaries and locationoundaries and locationoundaries and locationof precipitation stations.of precipitation stations.of precipitation stations.of precipitation stations.

39

atólica Don Bosco (UCDB) carried out an analysis on the hydrology, nd other useful water related data sets for the entire UPRB (Brazil, Bolivia This information will be used to help improve the definition of the the UPRB and to characterize the flooding regime for the next phase of he UPRB has a catchment area of 484,970 km2, which is located at the South America, covering parts of western Brazil (372,501 km²), eastern

km²) and northeastern Paraguay (31,626 km²). The annual rainfall ranges to 1900 mm with a distinct wet and dry season. Most of the basin is very the world�s largest continental wetland called the Pantanal. The Pantanal

40

wetland covers an area of 138,000 km2. The region undergoes frequent floods during the rainy season and frequent drought during dry season. The combination of improper land management on sandy soils in the upper basin causes serious erosion and acceleration of sediment load in the flood plain, especially during the past thirty years. Besides the gradual loss of farmlands, unpredicted floods often result in serious cattle loss. In addition, due to the recent increase of river transportation, riverbank erosion during low water level periods is exacerbated. Therefore, the purpose of this study is to develop a river water level forecast model using NDVI (AVHRR) inferred soil moisture status for flood management and for river transportation traffic control, and to guide the satellite scene selection process for Phase II of the Pantanal GIS project. The UPRB River Water Level (RWL) data recorded at the Ladário hydrological station (Latitude: 19° 05´S; Longitude: 57° 30´W) for the period of January of 1981 to December of 2000 provided by the Brazilian Marine Corps at Corumba, Mato Grosso do Sul, were used in this study. For the same period, monthly precipitation (PCP) data from six rainfall stations, including Arenapólis (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 Melgaco (16.19°S, 55.95°W), provided by the Brazilian National Electrical Energy Agency (Agencia Nacional de Energia Elétrica, ANEEL) were used. The locations of the river water level and rainfall measurement stations are shown in Graphic 1. Averaged values of six station rainfall data were used to represent the rainfall amount received in the upper parts of the basin. The RWL monitors the water level of the drainage area indicated by the area above the dash line. Figures a to s in Annex I represent the time series plots of RWL, PCP and NDVI for a period of 20 years (1981-2000). An Excel spreadsheet was submitting showing historical monthly RWL maximum and minimum data during the period of 01/1981 to 12/2000 at the Ladário station. In order to compare inter-annual hydrological variability, four RWL conditions were assessed, including (Table 12): - wet/wet stands for the highest RWL in rainy season; - dry/dry stands for the lowest RWL in dry season; - wet/dry stands for the highest RWL in dry season; - dry/wet stands for the lowest RWL in rainy season.

41

Table 12. River water levels from 1988 to 2000.Table 12. River water levels from 1988 to 2000.Table 12. River water levels from 1988 to 2000.Table 12. River water levels from 1988 to 2000.

River Water Level ( m )River Water Level ( m )River Water Level ( m )River Water Level ( m )

________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ YearYearYearYear Conditions*Conditions*Conditions*Conditions* MaximumMaximumMaximumMaximum MonthMonthMonthMonth MinimumMinimumMinimumMinimum MonthMonthMonthMonth 1988198819881988 wet/wet 6.57 April 1.41 December 1995199519951995 wet/wet 6.50 April 2.02 December 1982198219821982 wet/wet 6.41 April 2.52 December 1989198919891989 wet/wet 6.10 May 2.10 December 1985198519851985 wet/wet 6.04 April 1.69 December 1999199919991999 dry/dry 4.58 June 1.19 November 2000200020002000 dry/dry 4.62 June 1.21 November 1986198619861986 dry/dry 4.22 July 1.32 December 1987198719871987 dry/dry 4.94 June 1.40 November 1988198819881988 dry/dry 6.57 April 1.41 December 1992199219921992 wet/dry 5.35 June 3.30 December 1111982982982982 wet/dry 6.41 April 2.52 December 1984198419841984 wet/dry 5.04 June 2.47 November 1991199119911991 wet/dry 5.38 June 2.37 December 1997199719971997 wet/dry 5.66 May 2.18 December 1994199419941994 dry/wet 3.91 July 1.42 November 1986198619861986 dry/wet 4.22 July 1.32 December 1990199019901990 dry/wet 4.45 June 1.95 December 1999199919991999 dry/wet 4.58 June 1.19 November 2000200020002000 dry/wet 4.62 June 1.21 November

* Four situations: - wet/wet indicates a maximum river water level in wet season (October � April); - dry/dry indicates a minimum river water level in dry season (May � September); - wet/dry indicates a maximum river water level in dry season (May � September); - dry/wet indicates a minimum river water level in wet season (October � April).

The results show that in the wet season the highest RWL of 6.57 m occurred in April 1988 and the lowest RWL occurred in July 1994. In the dry season the highest RWL of 3.30 m occurred in December 1992 and the lowest RWL of 1.19 m occurred in November 1999. It is interesting to note that in dry years, the highest RWL delayed two to three months to reach its maximum from April to June-July. This means that in the dry year, there was a reduced surface runoff water availability, but indicates that a slow moving underground water supply plays an important role in restoring RWL. In all the years studied, the lowest RWL occurred between November and December with little

42

variation. From the analysis of the time series plots of RWL, PCP and NDVI during the period of 1981 to 2000 shown in figures 2 through 22, it was observed that: a) In wet years: Minimum monthly RWL (mostly in November) occurred 5 months after the peak dry season (June-July), while the maximum monthly RWL (mostly in April) occurred right at the end of the wet season (November-March). From the comparison of RWL response to NDVI, it was observed that minimum RWL occurred 2 to 4 months after NDVI reached its minimum. In general, maximum RWL occurred 3 months after NDVI reached its maximum in January. It is important to note that NDVI reached a value of 0.60 mostly in November and remained above this value for a long period of 8 to 9 months (November to June or July). This suggests that vegetation reaches its maximum greenness right at the beginning of the rainfall season. Further increase of rainfall in the following months does not increase NDVI, which is around 0.60. NDVI maintains its maximum 2 to 3 months after the end of rainfall season since the flat basin maintains a suitable soil moisture condition for a longer period. One has to note that surface water absorbs both visible and near infrared bands such as AVHRR bands 1 and 2 that result in zero to negative NDVI values. Therefore, in the wet season, NDVI saturates at a low value of around 0.6. NDVI values lower than 0.6 might be due to high flood areas with larger areas of surface water. b) In dry years: The response of the minimum RWL to PCP with 5 months time lag was about the same as in wet years. But in dry years, maximum RWL to PCP occurred in June or July not in April, delayed by two to three months. This suggests that in drier years, the RWL reaches its maximum very slowly due to lower surface runoff water and slow recharge of underground water. The prolonged high NDVI values from November to July indicates a higher surface greenness as well as a prolonged higher soil moisture, which is attributed to the contribution of slow ground water recharge. This phenomenon is common in a large flat river basin such as the UPRB. Figure 22 shows monthly NDVI change in 1988 with a highest RWL in the wet season and in 1999 with a lowest RWL in the dry season. The NDVI images did not show well the difference between these two figures due to the fact that a large free water surface contributes lower NDVI values. But from the analysis of rainfall data, it was observed that the total annual rainfall of 1884 mm in 1988 was the highest rainfall amount of the 20 years studied and the year 1999 with 1396 mm was among 4 lowest rainfall years. AVHRR NDVI monthly maximum value composite data with a resolution of 8km by 8km for the period of August 1981 to December 2000, provided by the GSFC/NASA were used in this study. For each pixel, the maximum NDVI value within a certain month was selected to compose the NDVI monthly image. The NDVI monthly maximum value composite technique was applied to partially eliminate the cloud contamination (Holben, 1986). According to Eidenshink et al. (1997), these data have already been processed

43

with radiometric calibration using the NOAA standard method (Rao and Chen, 1995) and atmospheric corrections including Raleigh scattering using the methods of Gordon et al. (1988) and ozone absorption using the methods of Fleig et al. (1983). The NDVI value is calculated by taking the ratio of reflectance values of AVHRR channel 1 (Ch1: 0.58-0.68 µm) and channel 2 (Ch2: 0.725 -1.10 µm) and divided by the sum of them, which is expressed by the following equation: NDVI=(Ch2-Ch1)/(Ch2+Ch1).

i. Landcover Classification

The purpose of this task was to develop a land cover classification of the UPRB applicable to classification of remotely sensed imagery. Specific objectives consisted of: (1) identifying equivalent vegetation types with different regional designations; (2) proposing a classification scheme that organizes the land covers in a hierarchical system; and (3) classifying multi-date Landsat data to produce a land cover map of a specific study site within the UPRB, using the designations of the land cover classification system that is proposed. Methods A. Proposed landcover classification The literature review has included documents, maps, studies and other published materials from Brazil and Paraguay, and general regional phytogeographic information from South America. Vegetation types are characterized, when possible, by the following parameters: (1) structure (strata, height); (2) elevation/topography; (3) soil type; (4) climate; (5) aerial extent and spatial distribution; (6) floristic composition; and (7) phenology/seasonality. These data have been compiled mainly from reports accompanying maps and regional natural resource studies. A hierarchical classification system is proposed based on this review. B. Image processing Three Landsat images (227/74 scene) representing different seasons were selected for the analysis: - 06/09/1997 (all TM bands were used except the thermal), corresponding to the

flooded season of a high-flood year; - 07/30/1998 (all TM bands were used except the thermal), corresponding to the

flooded season of a low flood year; and - 11/14/1999 (all ETM + bands were used except the thermal and panchromatic),

representing the dry season of a low-flood year.

44

These images were acquired based on precipitation (Bahía Negra-Paraguay, and Corumbá-Brazil) and river gauge (Ladário-Brazil) data. The images represent different seasons that could provide explanation for the variability for the land cover characterization. Unsupervised classification (120 classes) was applied to a four-component image derived through principal component analysis from the three Landsat images. The 120 classes were regrouped to conform to the proposed land cover classification system. Accuracy assessment based on existing maps and aerial photographs from Bolivia, Brazil and Paraguay has been carried out and the results can be viewed in the final published thesis. Results A. Land cover classification system Based on literature review, the identified vegetation types are grouped physiognomically. The land uses and water bodies are added, along with the vegetation types, to the classification system scheme structured with the hierarchal criteria of (1) natural/modified and (2) flooded/non-flooded environments; (3) vegetation physiognomy; and (4) vegetation type (considering the vegetation parameters mentioned) (Graphic 2).

Graphic 2. Classification SysteGraphic 2. Classification SysteGraphic 2. Classification SysteGraphic 2. Classification System for UPRBm for UPRBm for UPRBm for UPRB

Natural

Modified

Permanently flooded

Floodable

Non floodable

Water

Forest GrasslandSavanna

Vegetation

FIRST LEVELFIRST LEVEL

SECOND LEVELSECOND LEVEL

THIRD LEVELTHIRD LEVEL Natural/modified environments The first level of the classification system separates natural and modified environments based on land cover rather than land use characteristics. The savannas and grasslands with relatively unmodified vegetation coverage are considered part of the natural environment although they might be associated with the land use of cattle ranching.

45

The first-level modified environments category is not subdivided and encompasses anthropogenic areas as well as modified natural formations. Hydromorphic environments Natural formations and water bodies were classified into (1) permanently flooded, (2) floodable, and (3) non-floodable. Water bodies are considered either or lotic or lentic regime; the existent semi-lentic regime was disregarded in the classification scheme. The category of floodable formations does not differentiate seasonal from occasional flood events. Vegetation The vegetation physiognomies differentiated for the study area were:

1. Forest formations: tree coverage of more than 50%; herbaceous stratum inconspicuous.

2. Arboreal-herbaceous formations: tree coverage between 5-70%; conspicuous shrubby and herbaceous strata may be present.

3. Herbaceous or herbaceous-shrubby formations: absent or insignificant tree coverage; conspicuous herbaceous stratum; shrubby stratum may be present.

The terrestrial vegetation types reviewed were allocated under these vegetation forms in the classification scheme. Forest�s subsequent levels, if data were available, indicate seasonality (evergreen, semi-deciduous or deciduous forests), water availability (xerophytic or mesophytic forests) and topography (lowland, submontane). Arboreal-herbaceous formations were separated floristically and then by tree density. Herbaceous formations were only differentiated floristically. Aquatic communities are implicit but not detailed in the classification scheme; ecotones were not considered. B. Image processing Based on visual interpretation and existing vegetation maps from Brazil and Paraguay, the 120 classes obtained were reassigned to eight classes: (1) bare soil, (2) permanent water bodies, (3) seasonally flooded grassland, (4) seasonally flooded savanna, (5) semi-deciduous forest, (6) forest (Bolivia), (7) grassland (Chaco)/pasture, (8) steppe savanna. From these classes, two do not conform to the proposed land cover classification system: bare soil, and a forest type located in Bolivia. In this preliminary assessment, subtypes of semi-deciduous forest and steppe savanna are not differentiated and the classes of pasture and grassland (Chaco) are merged because they are not mutually exclusive. Further processing to refine classes and accuracy assessment are pending. Discussion Because the proposed land cover classification system was intended to be applicable to (unsupervised) classification of remotely sensed data (Landsat TM), its vegetation

46

component was organized primarily on physiognomic features; issues that required extensive field work were overlooked. In both the proposed land cover classification system and the image processing, forest types were not consistently classified beyond their seasonality aspect due to the lack of data on water availability and topography. In the image classification process, three classes of forest were clearly discerned, but because their spatial distribution did not coincide with the maps consulted, two of them were included in the overlapping semi-deciduous category. The other forest type, which is located in Bolivia, was not assigned as a deciduous, semi-deciduous or evergreen forest because floristic data were unavailable at the time. The flooding seasonality detected in the image processing was based solely on the images available. Remotely sensed data from other years, and more detailed soils, topographic and vegetation data could change the assignment of floodable, non-floodable and permanent flooded classes. The final version of this report on landcover classification for the UPRB will be made available after completion and publication later in 2003.

47

AcknowledgementsAcknowledgementsAcknowledgementsAcknowledgements This project in its multiple stages would not have been possible without the help and support from many friends and colleagues. We have made every effort to include everyone here, and if by accident we have forgotten anyone we sincerely appologise. We are especially grateful to Rich Schulman of Resource Strategies, Inc. for coordinating the satellite image acquisition and technical advice. To Tim Leary of Image Links for the assistance with satellite image georeferencing and coregistering. To the staff in CONAE (Argentina) and in INPE (Brazil) for dealing with our Landsat scene requests, which was sometimes difficult because of communications problems. And to ERDAS and ESRI vendors in Bolivia, Brazil, Paraguay and the USA for providing their software at affordable prices to a conservation project. We also appreciate the help provided by Terry Rhea with the first version of the technical work plan, Stacey Shankle for help with translations, Terry Boyle of USGS for donating his literature collection related to Pantanal, and Steve Hamilton of Michigan State University provided an analysis of hydrology units of Pantanal. Chris Baker and Steve Veltman contributed to the technical design and development of the GIS database. DU Canada provided a web-enabled data catalogue and USGS is serving the metadata via the web. Many people have contributed to the organization of the different meetings, special thanks go to the hosts, Mario Dantas and Emiko Kawakami (former and present) directors of EMBRAPA-Pantanal, Frank Fragano and Alberto Yanosky (former and present) directors of Guyra-Paraguay, Roger Landivar director of WWF-Bolivia, and Federico Muller and Fátima Sonoda of FEMA-MT. Jesus Jemio assisted WWF-Bolivia with fieldwork and Nicole Martínez helped with the translation to english. Oscar Quiroga of Man and Nature reserve in Puerto Suarez, gave us an excellent introduction to the wildlife and conservation issues in that part of South America. Special recognition goes to the colleagues in the two organizations that have provided financial support: Val Mezainis, Jack Capp and Cindy Ragland at the International Programs of the USDA Forest Service for their trust in the Pantanal GIS project and their reiterated support through agreements # 99-G-191 and # 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 support to the training component of this project through agreement # 98210-1-G077.

48

This page left blank intentionally

49

Appendix IAppendix IAppendix IAppendix I

Figures 1-20

50

This page left blank intentionally

51

Figure Figure Figure Figure 1111. Portion of a classified . Portion of a classified . Portion of a classified . Portion of a classified satellite image dated December 24, 1999. The blue and satellite image dated December 24, 1999. The blue and satellite image dated December 24, 1999. The blue and satellite image dated December 24, 1999. The blue and cyan areas denote flooded vegetation.cyan areas denote flooded vegetation.cyan areas denote flooded vegetation.cyan areas denote flooded vegetation.

52

Figure Figure Figure Figure 2222. Change in burn scar patterns . Change in burn scar patterns . Change in burn scar patterns . Change in burn scar patterns between November 1998 (active burn) between November 1998 (active burn) between November 1998 (active burn) between November 1998 (active burn) and November 1999 (recent burn).and November 1999 (recent burn).and November 1999 (recent burn).and November 1999 (recent burn).

53

Figure Figure Figure Figure 3333. Edge enhancement was applied to satellite images for digitizing an updated . Edge enhancement was applied to satellite images for digitizing an updated . Edge enhancement was applied to satellite images for digitizing an updated . Edge enhancement was applied to satellite images for digitizing an updated roads and tracks coverage for the pilot project area.roads and tracks coverage for the pilot project area.roads and tracks coverage for the pilot project area.roads and tracks coverage for the pilot project area.

Figure Figure Figure Figure 4444. Maximum flooded area for the Bolivian por. Maximum flooded area for the Bolivian por. Maximum flooded area for the Bolivian por. Maximum flooded area for the Bolivian portion of the Pantanal pilot area.tion of the Pantanal pilot area.tion of the Pantanal pilot area.tion of the Pantanal pilot area.

54

Maximum and Minimum Flood Area

1,551,178.08

481,524.12

140,455.08

573,954.84

24,012.18

457,511.94

0

200,000

400,000

600,000

800,000

1,000,000

1,200,000

1,400,000

1,600,000

1,800,000

MaximumFlooded Area

(06/09/97)

MinimumFlooded Area

(12/24/99)

Differencebetw een dates

Decrease andIncrease

betw een areas

Subset Area

Has

Figure 5. The maximum and minimum flooded area in hectares for the Figure 5. The maximum and minimum flooded area in hectares for the Figure 5. The maximum and minimum flooded area in hectares for the Figure 5. The maximum and minimum flooded area in hectares for the subset pilot area including the Bolivian portion of the Pantanal.subset pilot area including the Bolivian portion of the Pantanal.subset pilot area including the Bolivian portion of the Pantanal.subset pilot area including the Bolivian portion of the Pantanal.

Seasonal Flood Area

1,545,802.39

51,196.50

213,069.06 158,331.24 156,546.18 105,349.68

0

200,000

400,000

600,000

800,000

1,000,000

1,200,000

1,400,000

1,600,000

1,800,000

Flooded Area inhigh w ater

season(07/30/98)

Flooded Area inlow w ater

season(07/30/98)

Differencebetw eenseasons

Decrease andincreasebetw eenseasons

Subset Area

Has

Figure 6. Variation in seasoFigure 6. Variation in seasoFigure 6. Variation in seasoFigure 6. Variation in seasonal flooded area for the Bolivian portion of nal flooded area for the Bolivian portion of nal flooded area for the Bolivian portion of nal flooded area for the Bolivian portion of the Pantanal pilot project.the Pantanal pilot project.the Pantanal pilot project.the Pantanal pilot project.

55

Figure 7. Flood "change" files showing increase and decrease in inundated areas between two dates.Figure 7. Flood "change" files showing increase and decrease in inundated areas between two dates.Figure 7. Flood "change" files showing increase and decrease in inundated areas between two dates.Figure 7. Flood "change" files showing increase and decrease in inundated areas between two dates.

Figure 8. Delineated areas of frequent, concentraFigure 8. Delineated areas of frequent, concentraFigure 8. Delineated areas of frequent, concentraFigure 8. Delineated areas of frequent, concentrated flooding within the Paraguayan ted flooding within the Paraguayan ted flooding within the Paraguayan ted flooding within the Paraguayan portion of the Pantanal pilot area on Landsat 7 image data 08/12/01.portion of the Pantanal pilot area on Landsat 7 image data 08/12/01.portion of the Pantanal pilot area on Landsat 7 image data 08/12/01.portion of the Pantanal pilot area on Landsat 7 image data 08/12/01.

56

Figure 9. Map created by WWF Bolivia depicting 50% increase or decrease in NDVI Figure 9. Map created by WWF Bolivia depicting 50% increase or decrease in NDVI Figure 9. Map created by WWF Bolivia depicting 50% increase or decrease in NDVI Figure 9. Map created by WWF Bolivia depicting 50% increase or decrease in NDVI between 1988 and 1999 for the Bolivian portion of the Pilbetween 1988 and 1999 for the Bolivian portion of the Pilbetween 1988 and 1999 for the Bolivian portion of the Pilbetween 1988 and 1999 for the Bolivian portion of the Pilot Project area.ot Project area.ot Project area.ot Project area.

57

Biomass Changes between November and December 1999

3,859.47 1,116.27

59,888.07

560,916.18

0

100,000

200,000

300,000

400,000

500,000

600,000

25% Decrease 25% Increase 50% Decrease 50% Increase

Percentage of variation in NDVI

Has

Figure 10. Temporal variability in NDVI for the study area that includes the Bolivian portion of the Figure 10. Temporal variability in NDVI for the study area that includes the Bolivian portion of the Figure 10. Temporal variability in NDVI for the study area that includes the Bolivian portion of the Figure 10. Temporal variability in NDVI for the study area that includes the Bolivian portion of the Pantanal pilot project for November 1999 and December 1999 (subset area is 1,620,416 Pantanal pilot project for November 1999 and December 1999 (subset area is 1,620,416 Pantanal pilot project for November 1999 and December 1999 (subset area is 1,620,416 Pantanal pilot project for November 1999 and December 1999 (subset area is 1,620,416 hectares).hectares).hectares).hectares).

Biomass Changes between 1998 and 1999

589,817.97

12,518.91

121,861.62

2,135.880

100,000

200,000

300,000

400,000

500,000

600,000

25% Decrease 25% Increase 50% Decrease Increase

Percentage of variation

Has

Figure 11. Temporal variability in NDVI for study area that includes the Bolivian portion of the Figure 11. Temporal variability in NDVI for study area that includes the Bolivian portion of the Figure 11. Temporal variability in NDVI for study area that includes the Bolivian portion of the Figure 11. Temporal variability in NDVI for study area that includes the Bolivian portion of the Pantanal pilot project for 1998 and 1999 (subset area is 1,545,802 hectares).Pantanal pilot project for 1998 and 1999 (subset area is 1,545,802 hectares).Pantanal pilot project for 1998 and 1999 (subset area is 1,545,802 hectares).Pantanal pilot project for 1998 and 1999 (subset area is 1,545,802 hectares).

58

Figure 12. Biomass changes based on 50% increase or decreaFigure 12. Biomass changes based on 50% increase or decreaFigure 12. Biomass changes based on 50% increase or decreaFigure 12. Biomass changes based on 50% increase or decrease in NDVI in between 1988 and 1999.se in NDVI in between 1988 and 1999.se in NDVI in between 1988 and 1999.se in NDVI in between 1988 and 1999.

59

Figure 13. Biomass changes based on 25% increase or decrease in between 1988 and 1999. Figure 13. Biomass changes based on 25% increase or decrease in between 1988 and 1999. Figure 13. Biomass changes based on 25% increase or decrease in between 1988 and 1999. Figure 13. Biomass changes based on 25% increase or decrease in between 1988 and 1999.

60

Figure 14. Burn scar polygons delineated over July 30, 1998 Landsat composite image with a Figure 14. Burn scar polygons delineated over July 30, 1998 Landsat composite image with a Figure 14. Burn scar polygons delineated over July 30, 1998 Landsat composite image with a Figure 14. Burn scar polygons delineated over July 30, 1998 Landsat composite image with a 7/5/4 (RGB) band combination.7/5/4 (RGB) band combination.7/5/4 (RGB) band combination.7/5/4 (RGB) band combination.

61

Figure 15. Burn scar polygons delineated over November 23, 1988 Landsat image using a 7Figure 15. Burn scar polygons delineated over November 23, 1988 Landsat image using a 7Figure 15. Burn scar polygons delineated over November 23, 1988 Landsat image using a 7Figure 15. Burn scar polygons delineated over November 23, 1988 Landsat image using a 7----5555----4 4 4 4 (R/G/B) band combination.(R/G/B) band combination.(R/G/B) band combination.(R/G/B) band combination.

62

Figure 16. Roads digitized by IBAMA for the Brazilian portion of the study area.Figure 16. Roads digitized by IBAMA for the Brazilian portion of the study area.Figure 16. Roads digitized by IBAMA for the Brazilian portion of the study area.Figure 16. Roads digitized by IBAMA for the Brazilian portion of the study area.

63

Figure 17. DigitizedFigure 17. DigitizedFigure 17. DigitizedFigure 17. Digitized roads for each date of imagery for the Paraguayan portion of the pilot area. roads for each date of imagery for the Paraguayan portion of the pilot area. roads for each date of imagery for the Paraguayan portion of the pilot area. roads for each date of imagery for the Paraguayan portion of the pilot area.

Figure 18. Examples of roads categories in the Paraguayan portion of the pilot area.Figure 18. Examples of roads categories in the Paraguayan portion of the pilot area.Figure 18. Examples of roads categories in the Paraguayan portion of the pilot area.Figure 18. Examples of roads categories in the Paraguayan portion of the pilot area.

64

Figure 19. Ground control points converted to a shapefile and plotted over satFigure 19. Ground control points converted to a shapefile and plotted over satFigure 19. Ground control points converted to a shapefile and plotted over satFigure 19. Ground control points converted to a shapefile and plotted over satellite imagery.ellite imagery.ellite imagery.ellite imagery.

Legend : Points_pan.shp � points over the road intersections pt22774_X.shp � ground control points collected in the field Pontos22774.shp � points digitized over the image

65

Figure 20. Monthly NDVI digital images of 1988, the year of mFigure 20. Monthly NDVI digital images of 1988, the year of mFigure 20. Monthly NDVI digital images of 1988, the year of mFigure 20. Monthly NDVI digital images of 1988, the year of maximum river water level at aximum river water level at aximum river water level at aximum river water level at Ladario in wet season.Ladario in wet season.Ladario in wet season.Ladario in wet season.

66

This page left blank intentionally

67

Appendix IIAppendix IIAppendix IIAppendix II

Change Detection Procedures

Seasonal Flood Analysis

NDVI Differencing Tracking Burn Scars

Roads Updating

68

This page left blank intentionally

69

Direct MultiDirect MultiDirect MultiDirect Multi----Date Classification for Seasonal Flood AnalysisDate Classification for Seasonal Flood AnalysisDate Classification for Seasonal Flood AnalysisDate Classification for Seasonal Flood Analysis

Step 1Step 1Step 1Step 1 • Start ArcView • Go to �File� � �Extensions� and check the Image Analysis and Geolink Extensions

• Click on �View� in the sidebar. Create a New view. • Click on �View�- �Properties� in the main tool bar of the new viewer. Change Map

Units and Distance Units to �meters�. • Input files called 73098_vi_sub.img and 60997_sub.img using the Add Theme

Button. Be sure to select �Image Analysis Data Source� in the Add Theme dialog box.

Step 2Step 2Step 2Step 2

• Select (click) the 73098_vi_sub.img file in the side bar. • Click on �Image Analysis� and �Categorize.

70

• Type in the number of classes. (Try 24 for this exercise). • ArcView will compute a thematic �classified� image with 24 classes and add it

automatically to the View. • Repeat these steps for file 60997vi_sub.img

Step 3Step 3Step 3Step 3

• Click on �GeoLink� in the main toolbar. Select �Create New Sibling and Link It�. • Dialog asks, �Would you like to add any current themes to the linked Viewer� �

click on Yes. • Select 73098vi_sub.img and click �ok�.

• GeoLink will create a new viewer with the multispectral file that is linked to the associated �classified� file Viewer. Click on �Window� � �Tile� and you will be able to have two or more Viewers of the same size displayed and linked.

• Double click on the �Categorization� for 73098vi_sub.img. This will open the

Legend Editor.

71

• Using the �P� (pan) button move around the two images and identify water.

• Use the Identity Tool to begin indentifying classes that represent water or flooded vegetation. Change all water classes to Blue in the Legend Editor.

• Cancel all other classes by highlighting the class in the Legend Editor and

clicking on . • Click �Save� in Legend Editor. • In the main toolbar click on �Theme� � �Save Image As� and select the �GRID�

option (make sure your Categorized Image for 73098_vi_sub.img is highlighted). • Repeat these steps for file 60997vi_sub.img. Make a water GRID and assign it a

different color. Step 4Step 4Step 4Step 4

• Highlight the two water GRIDs you just created in the sidebar. • Click � �Image Analysis� � �Image Difference� • Select the �Before� and �After� Water GRIDS to be differenced.

72

• Specify a change of 10% or greater • Select colors for output �change file� or leave colors as default. • Click �OK�. ArcView will produce a new change file and load it in the Viewer. • Note which color represents increase or decrease of 10%. • Try GeoLinking the Viewers and checking the change in flood for the two dates

by comparing the multispectral imagery with the change file. NOTES:NOTES:NOTES:NOTES: Repeat this process for all dates of imagery for the pilot area. If necessary, subset the Pilot Area scene so that the processing is done only within the boundaries of your country (ie. Create subset for Brazil, Bolivia or Paraguay). You may adjust the number of classes created in the �Categorize� step until you are satisfied that you have identified water correctly. Once the water GRIDS are created for each date, Difference Images can be computed between any two dates.

73

Image/Band DImage/Band DImage/Band DImage/Band Differencing using NDVIifferencing using NDVIifferencing using NDVIifferencing using NDVI Notes:Notes:Notes:Notes: Image Differencing using NDVI should only be performed on �ANNIVERSARY� date-images that were acquired at or near the same date on different years. Differencing NDVI images from different seasons will highlight mostly phenological changes rather than man-induced changes.

Step 1Step 1Step 1Step 1

• Create a new View. • Add two new files (111499vi_sub.img and 111998vi_sub.img) using the Add

Theme Button . • Compute NDVI for each date by highlighting one of the files and clicking on

�Image Analysis� � �Vegetative Index�. • Note that Bands 3 and 4 are identified correctly in the dialog box and click �OK�.

• A new NDVI image will be calculated and automatically added to the View. • Repeat steps for both files: 111499vi_sub.img and 111998vi_sub.img

Step Step Step Step 2222

• Select the two NDVI images in the View sidebar and Click on �Image Analysis� � �Image Difference�.

• Difference the two images using the same steps as in Method One but make the change threshold 20%.

Step 3Step 3Step 3Step 3

• Use the Geolink extension to compare NDVI from two dates and evaluate the �Change� file.

74

Seeding Burn Scar Polygons Seeding Burn Scar Polygons Seeding Burn Scar Polygons Seeding Burn Scar Polygons

• Open a new View.

• Load file 111499vi_sub.img . • Double click on the file in the side bar and change the band combinations to 3-

2-1. • Can you identify smoke plumes in the imagery? • Double click on the file in the side bar and change the band combinations to 6-

5-4. • Can you identify the burn scars? • Click on �View�-�New Theme�. • Create a new polygon theme, name it and save it on your directory. • Highlight the new theme in the side bar and click on �Theme� � �Start Editing�. • Find a burn scar in the imagery. Click on �Image Analysis� and �Seed Tool

Properties�.

• Set the seed radius to 5 pixels and deselect deselect deselect deselect the �island polygon� option.

• Click on the �Seed Button� in the toolbar and click your cursor inside a uniform area of a burn scar.

• The software will delineate a polygon around the burn scar that has a similar spectral profile to the 5 pixel radius seed you identified.

• Try creating several polygons around the burnscars or clearcut areas while adjusting the pixel radius to see how the polygon is affected.

75

Edge Enhanced Images for Updating Road CoveragesEdge Enhanced Images for Updating Road CoveragesEdge Enhanced Images for Updating Road CoveragesEdge Enhanced Images for Updating Road Coverages

Step 1Step 1Step 1Step 1::::

a) Open a new View. Name the view �Edge� and set the distance and map units to meters.

b) Load file 122499vi_sub.img with �Add Theme� button . c) Highlight the file in the sidebar and click on �Image Analysis� � �Edge Detect�. d) The edge-enhanced image will be computed automatically and added to the

View.

e) Double click on the Edge Enhanced image to open the Legend Editor. f) Change the band combination to 6-5-4 or whichever combination highlights the

road features best. Step 2:Step 2:Step 2:Step 2:

• Click on �Advanced� button in the Legend Editor. • Manipulate the histogram of the Edge image to get the maximum contrast on the

roads.

76

• This enhanced imagery can be used as a background for digitizing a Road shapefile or for updating existing coverages.

• Create a new �Line� theme (shapefile) and digitize some of the roads you are able to detect with this background image.

• Newly digitized roads should be attributed in the table using the steps outlined in Method 2.

− The Fieldnames should be Country (BR,BO,PY), Length (in meters), Name, and

Type. The following 4 attributes were agreed upon for the TYPE field for the roads shapefiles:

Major roads (surfaced) Secondary roads (surfaced and unsurfaced) Major tracks (unsurfaced -passable all year) Minor tracks (unsurfaced - passable in the dry season only)

77

Appendix IIIAppendix IIIAppendix IIIAppendix III

Project Meeting Agendas

78

This page left blank intentionally

79

Upper Paraguay River Upper Paraguay River Upper Paraguay River Upper Paraguay River Basin GIS Database Pilot ProjectBasin GIS Database Pilot ProjectBasin GIS Database Pilot ProjectBasin GIS Database Pilot Project ArcView Training WorkshopArcView Training WorkshopArcView Training WorkshopArcView Training Workshop

1-3 December 2000 Corumba, Brazil

HOSTS:HOSTS:HOSTS:HOSTS: Ducks Unlimited, Inc., EMBRAPA, US Forest Service PURPOSE:PURPOSE:PURPOSE:PURPOSE: This workshop is intended to take advantage of the fact that several people with an interest in Pantanal conservation will already be assembled in Corumba for the III Symposium on Natural and Socio-Economic Resources of Pantanal being organized by EMBRAPA (27-30 November 2000). This will provide a good forum to identify active partners for the Pantanal GIS database project which covers portions of Brazil, Bolivia, and Paraguay. The main objective of this workshop is: ! To provide training using ESRI Arcview and Arcexplorer software using data sets

covering the Pantanal region. Maps are very important to conservation planning, presently many exist in digital format for use in a GIS application. GIS technology allows several different map types to be integrated and viewed in a common setting. This training will familiarize users with the Pantanal Conservation GIS project technical plan. It will also teach attendees to load, view, query, and compose maps with low cost, user-friendly GIS software (Arcview and Arcexplorer) being used worldwide. ATTENDEES: ATTENDEES: ATTENDEES: ATTENDEES: Professionals involved in conservation planning in Pantanal (such as Government, NGO, University staff). COST:COST:COST:COST: No fee will be charged for the workshop. Selected participants will receive financial assistance. LANGUAGE:LANGUAGE:LANGUAGE:LANGUAGE: The workshop will be carried out mainly in Portuguese but Spanish and English will also be used. REQUIREMENTSREQUIREMENTSREQUIREMENTSREQUIREMENTS: 1. Brief summary (300 words or less) explaining why they are interested in Arcview and

Arcexplorer and how it will benefit their work in Pantanal. 2. Short CV (one page maximum) emphasizing previous work in Pantanal. Persons who have been actively involvement in the Pantanal GIS project will be given preference.

80

WORKSHOP ORGANIZERS:WORKSHOP ORGANIZERS:WORKSHOP ORGANIZERS:WORKSHOP ORGANIZERS: Dick Kempka, GIS Director, Ducks Unlimited, Memphis, TN, USA Carlos Padovani, GIS Coordinator, EMBRAPA-Pantanal, Brazil Kristine Kuhlman, Consultant, Natural Resource Management/GIS, Bolivia Fernando Troga, GEMPI/ESRI, Brazil PRELIMINARY PROGRAM:PRELIMINARY PROGRAM:PRELIMINARY PROGRAM:PRELIMINARY PROGRAM: Friday, 1 December 2000 Welcome, Workshop Structure and Personal Introductions Upper Paraguay River Basin/Pantanal GIS Project Background Introduction to Arcexplorer/Arcview Training in Arcexplorer Saturday, 2 December 2000 Training in Arcview Sunday, 3 December 2000 Training in Arcview (cont.) Workshop Conclusions and Closing Remarks

***************

81

Upper Paraguay River Basin GIS database Pilot Upper Paraguay River Basin GIS database Pilot Upper Paraguay River Basin GIS database Pilot Upper Paraguay River Basin GIS database Pilot ProjectProjectProjectProject Coordination and Brainstorming MeetingCoordination and Brainstorming MeetingCoordination and Brainstorming MeetingCoordination and Brainstorming Meeting

22-26 January 2001 Memphis, TN, USA

Participants:Participants:Participants:Participants: Core group of persons from Brazil, Bolivia, Paraguay and USA who over the last several months have indicated great interest in the project. Some have contributed to the Draft Technical Work Plan, others have provided useful information, and are willing to devote some of their time to the development of the database. Objectives: Objectives: Objectives: Objectives: To review the Technical Work Plan; to gather the technical staff and coordinate the methodologies and procedures for the development of the database; to discuss and anticipate potential technical challenges and find solutions before they are encountered during the development of the database; and to develop a set of tasks/responsibilities for each individual/institution to carry out during the following year. Sunday, 21st January 2001 - • Arrival of participants Monday, 22nd January 2001 - • Welcome message by Bruce Batt � DU Chief Biologist • Review of Project and Technical Work Plan � Montse Carbonell/Dick Kempka • Satellite Image Selection

- Criteria - Process � Carlos Padovani: graphs on precipitation and stream gauge - Raw versus processed satellite images - Problems � Clouds, data quality, different vendor / specifications, etc. − Loading and checking the quality of images - Geolink Viewers, co-display each image and check spatial accuracy of overlay. - Viewing individual bands for dropped lines and overall quality....

Tuesday, 23rd January 2001 - • Image rectification and co-registration, including maps and other data

- Image pre-processing providers- Image Links, etc. - Geometric correction in ERDAS - Radiometric correction, calibration and the Haze Reduction Utility in Imagine. - Swipe utility to check co-registration between dates. - Use of Imagine to view before and after images are treated with Haze Reduction

utility.

82

• Change detection - Final Categories needed - Optimal Band Selection - Viewing unprocessed scenes - Analyzing target features, e.g. water and/or flood extent, fire and fire scars,

clear-cutting, etc. - Profile tool - evaluate spectral signatures of water (in the infrared) and burned

ground (in the Thermal) - Overview of methods piloted and the possible variations depending on the target

of interest. # Direct Multi-date classification # Band Subtraction/Differencing/Ratioing # Principal Component Analysis

- Discussion on identifying various land-cover by using enhanced images (tasseled cap, ratioing).

Wednesday, 24th January 2001 - • Change detection (cont.)

− identifying change areas interpreting change detection images. − identifying burn scars by linking Burn Scar Index images with scenes colored

using various band combinations. − Principal component analysis on multidate and NDVI.

Thursday, 25th January 2001 - • Data conversion

- Assembling final products - Creating subsets of the target areas. - Producing maps in ArcView (map layout) or ERDAS (map composer). - Demonstrate ERDAS Mapsheets Utility

• Map production/template • Discussion on accuracy assessment. • Airborne videography/GPS. • Data Serving Friday, 26th January 2001 - • Review of conclusions and decisions made. • The US Forest Service contribution by Jack Capp. Saturday, 27th & Sunday, 28th January 2001 - • Participants return to their countries.

83

Upper Paraguay River Basin GIS database Pilot ProjectUpper Paraguay River Basin GIS database Pilot ProjectUpper Paraguay River Basin GIS database Pilot ProjectUpper Paraguay River Basin GIS database Pilot Project Change DetectioChange DetectioChange DetectioChange Detection Image Processing Meetingn Image Processing Meetingn Image Processing Meetingn Image Processing Meeting

7-11 May 2001 Asunción, Paraguay

Participants:Participants:Participants:Participants: Core group of persons from Brazil, Bolivia, Paraguay and USA who over the last several months have indicated great interest in the project and who are willing to dedicate time to the image processing component of the Pantanal change detection project. Objectives: Objectives: Objectives: Objectives: To review the Technical Work Plan; to gather the technical staff and coordinate the methodologies and procedures for the change detection image processing in ESRI ArcView Image Analysis and ERDAS Imagine; and to develop a set of tasks/responsibilities for each individual/institution to carry out for the completion of the change detection portion of the pilot project. Agenda: Agenda: Agenda: Agenda: Monday, 7 May 2001 − Welcome, Introductions, Purpose of Meeting − Review of Project, Technical Work Plan, Partner Matrix − DU�s Mississippi River Project − Equipment and staff available − Different Methods for image processing − Overview of three Change Detection Techniques:

- Image Subtraction/Ratioing - Post-Classification Comparison - �Seeding� Burn Scar Polygons with ESRI Image Analysis Extension

− Image pre-processing: - geometric correction (co-registration) -provided by vendor for Pilot Area. - radiometric normalization (histogram matching) - haze reduction utility in Imagine

• Change detection Method 1Change detection Method 1Change detection Method 1Change detection Method 1 Band Differencing/Ratioing

- Image Differencing ESRI ArcView Image Analysis (single band) - Band ratioing - NDVI - Differencing NDVI for different dates. - RGB Co-display using multiple dates of NDVI. - Creating Highlighted �Change� file for NDVI.

84

Tuesday, 8 May 2001 • Change detection Method 2 (PostChange detection Method 2 (PostChange detection Method 2 (PostChange detection Method 2 (Post----Classification Comparison)Classification Comparison)Classification Comparison)Classification Comparison)

- Categories to identify for pilot area: seasonal flood extent, max flood extent - Band Selection for water classification - Unsupervised Classification (ISODATA) - Geolinking viewers and extracting water/flooded vegetation. Load GeoLink

Extension. - Saving water files as a GRID. - Use ESRI Image Analysis to subtract thematic files for each date and create a

highlighted �change file�. Wednesday, 9 May 2001 • Change DChange DChange DChange Detection Method 3etection Method 3etection Method 3etection Method 3

- Updating Road Coverage with Edge Enhanced Images - Attributing Road Coverage - Seeding Burn Scar Polygons with ESRI Image Analysis - Attributing Burn Scar Polygons-load Calcacre script.

Thursday, 10 May 2001 Plotting maps in ArcView Discuss of Pros and Cons of using Image Analyst or Imagine Demonstration of ERDAS Imagine for:

• Improved Classifications • More complex ratios • New Change Detection Model in ERDAS • Accuracy Assessment �stratified random sampling points in ERDAS

Administrative Matters

− Partner Matrix: discuss task sharing and distribution. Who?What?When?HowMuch? − Agreements between DU and Forest Service and DU and partners in S.A. − Are we headed in the right direction?

o Local level fundraising o Local level training

Friday, 11 May 2001 - • Review of methods and decisions made • Sampling scheme and fieldwork • Set Final Product Generation Deadline Friday afternoon and Saturday, 11-12 May 2001 • Participants return to their countries

85

Upper Paraguay River Basin GIS Database Pilot ProjectUpper Paraguay River Basin GIS Database Pilot ProjectUpper Paraguay River Basin GIS Database Pilot ProjectUpper Paraguay River Basin GIS Database Pilot Project Data NorData NorData NorData Normalization and Field Data Collection Meetingmalization and Field Data Collection Meetingmalization and Field Data Collection Meetingmalization and Field Data Collection Meeting

11- 18 November, 2001 Puerto Suarez, Bolivia

November 13 November 13 November 13 November 13 (Tuesday)

• Project Progress Report o Software � update on who has received software/hardware. All have received

except for Partners in Brasil. In most cases the software/hardware has arrived in the country but is awaiting custom procedures.

o Change Detection Processing reports -Will set up technical workgroup at this meeting to discuss methods (via discussion group).

• Review of Change Detection Digital Products � Dawn Browne

o NDVI Differencing Change maps o NDVI RGB Co-Display o Final Digital Products for Change Detection o Plotting/Laminating Issues for Fieldwork o Review of roads coverages for each country o Review of suitability/practicality of 3 methods (flyover, boat, overland) o Sampling Strategy o Field Data Collection Sheet � construct as a group

• Metadata Standards � Kristine Kuhlman

o Metadata for managers and/or Metadata and its advantages Introduction to MetaLite Software Example of metadata with MetaLite

o Distribute copy of spreadsheet/program for metadata recording

November 14 November 14 November 14 November 14 (Wednesday) • Fieldwork Techniques - Pto Suarez (Asociacion Hombre Y Naturaleza (Association Man &

Nature Center) o GPS data collection standards o Digital Camera o Line Transect sampling method o Tour o Collect GPS pts, and record field data. o Compile GPS data, create shapefile in ArcView, add hotlinks

November 15 November 15 November 15 November 15 (Thursday)

• Future Project Coordination � Dick Kempka (a.m. only) o Expanding the project o Next Meeting - What additional partners should be invited and why?

86

o Schedule - Results/Products anticipated be March 2002 o Fundraising o Data distribution - How will it be done? Who will house? o Carlos or someone else who attended can provide a summary of the FOZ

meeting. • Data Normalization

o Projection standards o Standardized File Naming Conventions o Reprojecting in ArcView (vector) o Reprojecting in ERDAS (raster) o Creating Seamless Datasets in ERDAS o Distribute Hotlink script o Directory Structure Design o Collection of data from partners for database

NovemberNovemberNovemberNovember 16 16 16 16 (Friday)

• Review of Fieldwork Techniques • Presentation of results (Change Detection) • Landcover Classification System � Mario�s research • ½ day field work in Puerto Suarez

November 17November 17November 17November 17 (Saturday)

• Drive to Corumba for meeting at WWF-Brasil facilities • Introduction to Ducks Unlimited Pantanal Project if there are new organizations

involved • Presentation of some results (Change Detection) • Lunch and departures (Dawn, Heidy, Pamela back to hotel in Pto Suarez)

November 18November 18November 18November 18 (Sunday) participants depart

87

UPPER PARAUPPER PARAUPPER PARAUPPER PARAGUAY RIVER BASIN GIS DATABASE MEETINGGUAY RIVER BASIN GIS DATABASE MEETINGGUAY RIVER BASIN GIS DATABASE MEETINGGUAY RIVER BASIN GIS DATABASE MEETING Expanding the TriExpanding the TriExpanding the TriExpanding the Tri----National Pilot ProjectNational Pilot ProjectNational Pilot ProjectNational Pilot Project

11-15 August 2002 Cuiabá, Brasil

LOCATIONLOCATIONLOCATIONLOCATION Hotel Fazenda Mato Grosso Antônio Dorileo Street, 1200 Coxipó Tel. +55 (0xx65) 661 1200 Cuiabá, MT Fax. +55 (0xx65) 661 1276 Brazil http://www.hotelmatogrosso.com.br OBJECTIVESOBJECTIVESOBJECTIVESOBJECTIVES o Present the results of the tri-national Pantanal-GIS Database Pilot Project, a GIS

database developed for a pilot area in the Lower Pantanal, with the participation of 12 organizations from Bolivia, Brazil, Paraguay and the USA. A draft final report will be circulated at the meeting.

o Receive input to the methodology and approach used during the pilot project as we propose to expand our work to develop a comprehensive GIS database for the entire Upper Paraguay River Basin.

o Discuss the needs that administrators, managers, decision makers and funding agencies have relative to this GIS database in order to assist them with their responsibilities to protect this unique ecosystem. This is essential as we develop a tool that is required to be of easy access and friendly user across three countries with different needs, capabilities and interests.

o Identify steps that need to be taken in order to develop a draft Technical Work Plan for the comprehensive Upper Paraguay River Basin GIS Database. This work plan will require input from both technical and managers (administrators, decision makers, funding agencies) as the aim is to establish standards and methodologies that will respond to a broad audience�s needs.

The Final Report of the Pantanal-GIS Database Pilot Project and the Technical Report for the Upper Paraguay River Basin GIS Database will be edited and finalized approximately shortly after the meeting. The meeting is intended to be a working exercise and therefore only presentations directly related to the Pantanal-GIS Pilot Project and the Upper Paraguay River Basin GIS Project will be included. However, others willing to present their work in the Pantanal are requested to please contact the organizers so time can be allocated if possible.

88

PARTICIPANTSPARTICIPANTSPARTICIPANTSPARTICIPANTS The meeting is open to organizations and individuals interested in contributing to a tri-national effort aiming at developing a powerful GIS tool that will assist government agencies, NGOs, research institutions, community organizations and individuals working on different aspects of planning, conservation and management of the natural resources of the Upper Paraguay River Basin. The first two sessions will be attended by the agencies and the technical personnel who have participated in the development of the Pilot Project. Other persons interested are also welcome. Sessions three and four will be used to introduce the Pilot Project and set the basis for the discussion of the nest step necessary to initiate the Upper Paraguay River Basin GIS Database project. Participants in these two sessions will include technical personnel as well as managers, administrators, decision makers and funding agencies in need of a GIS database to assist them in the implementation of their conservation responsibilities. Input from everyone working in conservation issues in the region will be very valuable at this stage.

Saturday, 10 August 2002 Arrival of participants Sunday, 11 August 2002 08.30-12.30 Introductions and presentations Session 1 Presentation of results of Pantanal GIS Database Pilot Project 12.30-14.00 Lunch at Hotel 14.00-18.00 Discussion of results of Pantanal GIS Database Pilot Project Monday, 12 August 2002 Session 2 08.30-12.30 Discussion of technical issues related to Upper Paraguay River

Basin GIS database development based on Pilot Project 12.30-14.00 Lunch at Hotel 14.00-18.00 Discussion of technical issues related to Upper Paraguay River

Basin GIS database development based on Pilot Project Tuesday, 13 August 2002 Session 3 08.30-12.30 Presentation of results of Pantanal GIS Database Pilot Project 12.30-14.00 Lunch at Hotel Session 4 14.00-18.00 Discussion of Upper Paraguay River Basin GIS Database Wednesday, 14 August 2002 Session 4 (cont.) 08.30-12.30 Discussion of Upper Paraguay River Basin GIS Database

89

12.30-14.00 Lunch at Hotel 14.00-18.00 Discussion of Upper Paraguay River Basin GIS Database Conclusions and closing of meeting Thursday, 15 August 2002 Field trip 06.00 Departure for field trip by bus along Estrada Parque Late evening Return to Cuiabá Friday, 16 August 2002 Departure of participants

PROGRAMPROGRAMPROGRAMPROGRAM Introductions and Presentations: Introductions and Presentations: Introductions and Presentations: Introductions and Presentations: The meeting will start with an introduction by DU to the objectives of the Pantanal GIS Pilot Project and the expansion of the project. The representatives of the different partner organizations will present their organizations and describe why they are involved in the project. Presentations should not exceed 5 minutes. Session 1 Session 1 Session 1 Session 1 Pilot Project Results � Technical Group: The technical results of the following analyses will be presented for the pilot project area. ₋ Change detection � Dawn Browne (DU) ₋ Burn Scars � Heidy Resnikowsky (WWF Bolivia) ₋ Hydrological Analysis � Bill Lui (UCBD) ₋ Inundation/seasonal flood � Carlos Padovani (EMPRAPA) ₋ NDVI � Laura Rodriquez (M.Bertoni) ₋ Roads � Lindalva Ferreira (IBAMA) ₋ Field work � Sylvia Torrecilha (IMAP) ₋ Metadata � Kristine Kuhlman (U. Wisconsin / consultant) ₋ Serving data � Dick Kempka (DU) and Kristine Kuhlman (U. Wisconsin / consultant) ₋ Vegetation/Land Cover mapping � Mario Cardozo (DU /U. Memphis) Session 2Session 2Session 2Session 2 Next Steps � expansion of project to Upper Paraguay River Basin: The technical experience gained during the pilot project must now be discussed in the context of the Upper Paraguay River Basin. Session 3 Session 3 Session 3 Session 3 Pilot Project Results � General: In this session we will present the general results of the pilot project for participants outside the technical group and joining the meeting at this point. Session 4 Session 4 Session 4 Session 4

90

Upper Paraguay River Basin GIS Database project: : : : This session will be dedicated to a detailed discussion on specific needs for the Upper Paraguay River Basin conservation planning. DU will present the general context of the project. We will also ask that each participant present his/her organization, in maximum 5 minutes. Through the discussions during this session we hope to gain a good understanding of the following: ₋ Particular role of and/or projects conducted in the Pantanal (past and on-going) by

different organizations ₋ What conservation planning questions do the different organizations need to

answer? ₋ What data are needed to assist in answering these questions? ₋ What are the highest priority areas for the different organizations in the Upper

Paraguay river Basin? ₋ Do most organizations have internal GIS capabilities or do they need to hire a

contractor? ₋ Basic data analyses and data layers for conservation planning (examples will be

provided) ₋ Which important data sets are missing? ₋ Where (geographically) should the Upper Paraguay River Basin Project start? ₋ Where are the data residing? How to access? ₋ Are the different organizations willing to share data and be partners in this project? ₋ Staff availability to work on the proposed expanded Pantanal project? ₋ What data sets are best used to be integrated into the existing Pantanal database? ₋ How should the database be accessed? Where should the database reside and who

will maintain it? Please feel free to bring further documentation if you cannot cover all your information in the time allocated

NOTE: If your organization has GIS capabilities, please bring the following information (on cd-rom, diskette or printed): ₋ The GIS capability of the organization� equipment (NT, UNIX, etc), software (GIS,

Remote Sensing, RDBMS), and staff ₋ How is your GIS currently used? Does it provide service within your organization?

And/or does it provide service outside your organization? ₋ What existing data sets will help answer questions put forth by managers in your

organization?

91

References and Other Related LiteratureReferences and Other Related LiteratureReferences and Other Related LiteratureReferences and Other Related Literature Abdon, M., J. da Silva, V. Pott, A. Pott e M. da Silva. 1998. Utilização de dados analógicos do Landsat-TM na discriminação da vegetação de parte da sub-região da Nhecolândia no Pantanal. Pesquisa Agropecuária Brasileira 33: 1799-1813. Abdon, M., V. Pott, S. e J. da Silva. 1998. Avaliação da cobertura por plantas aquáticas em lagoas da sub-região da Nhecolandia no Pantanal por meio de dados Landsat e SPOT. Pesquisa Agropecuária Brasileira 33: 1675-1681. Achutuni, R., R.A. Scofield, N. Grody, e C. Tsai. 1996. Monitoring global large flood área using DMSP SSWI suggested humidity index. Procceeding of 8° conference of satellite meteorology and Oceanography, Atlanta, GA, January 29 � February 2, 1996, AMS, Boston, 455-459.

Adámoli, J. 1986 A dinâmica das inundações no Pantanal. In: I Simpósio sobre recursos naturais e sócio-econômicos do Pantanal, EMBRAPA Pantanal, Corumbá:51-61.

Adámoli, J. 1999. Los humedales del Chaco y del Pantanal. In: Malvárez, A., Tópicos sobre grandes humedales templados y subtropicales de Sudamérica. Universidad de Buenos Aires.

Adámoli, J. 1999. Previsão do médio prazo dos níveis do Rio Paraguai em Ladário - MS. In: II Simpósio sobre recursos naturais e sócio-econômicos do Pantanal, EMBRAPA Pantanal, Corumbá: 59-72.

Adámoli, J. e A. Pott. 1999. Estudo fitossociológico e ecológico do Pantanal dos Paiaguás. In: II Simpósio sobre recursos naturais e sócio-econômicos do Pantanal, EMBRAPA Pantanal, Corumbá: 215-225.

Adámoli, J. e A. Pott. 1999. Unidades de vegetação do Pantanal dos Paiaguás. In: II Simpósio sobre recursos naturais e sócio-econômicos do Pantanal, EMBRAPA Pantanal, Corumbá: 183-202. Allem, A. e J. Valls. 1987. Recursos forrageiros nativos do Pantanal Mato-grossense. Brasília, CENARGEM/EMBRAPA-CPAP, Documento 8. Alfonsi, R. e M. de Camargo. 1986. Condições climáticas para a região do Pantanal mato-grossense. In: I Simpósio sobre recursos naturais e sócio-econômicos do Pantanal, EMBRAPA Pantanal, Corumbá: 29-42.

92

Bignelli, P., M. Abdon, U. Palme e J. da Silva. 1998. Avaliação preliminar de dados radar ERS-1 para estudos do Pantanal brasileiro. Pesquisa Agropec. Brasileira 33:1691-1701. Cabrera, A. y A. Willink. 1973. Biogeografía de América Latina. OEA: Washington. CAF, Bolfor, Geosystems 2000, �Bolivia: Determinación del Daño Causado por los Incendios Forestales Ocurridos en los Departamentos de Santa Cruz � Beni en los meses de Agosto y Septiembre de 1999.� Informe CAF. Calheiros, D. e S. Hamilton. 1998. Limnological conditions associated with natural fish kills in the Pantanal wetland of Brazil. Verh. Internat. Verein. Limnol. 26: 2189-2193. Carrón, J. M. 2000. The Pantanal and the Paraguay River Basin: from the technical focus to the political option. In: F. A. Swarts (Ed.) The Pantanal of Brazil, Bolivia and Paraguay: selected discourses on the world's largest remaining wetland system. Hudson MacArthur Publishers, Gouldsboro. Carvalho, N. 1986. Hidrologia da Bacia do Alto Paraguai. In: I Simpósio sobre recursos naturais e sócio-econômicos do Pantanal, EMBRAPA Pantanal, Corumbá: 43-49.

Cowardin, J. and V. C. LaRoe. 1979. Classification of wetlands and deepwater habitats of the United States. US Fish and Wildlife Service. Department of the Interior. Da Silva, J. e M. Abdon. 1998. Delimitação do Pantanal brasileiro e suas sub-regiões. Pesquisa Agropecuária Brasileira 33: 1703-1711.

Da Silva, J., M. Abdon, A. Boock e M. da Silva. 1998. Fitofisionomias dominantes em parte das sub-regiões do Nabileque e Miranda, Sul do Pantanal. Pesquisa Agropecuária Brasileira 33: 1713-1719.

Da Silva, J., M. Abdon, M. da Silva e H. Romero. 1998. Levantamento do desmatamento no Pantanal brasileiro até 1990/91. Pesquisa Agropecuária Brasileira 33: 1739-1745.

Da Silva, T. 1986. Contribuição da geomorfologia para o conhecimento e valorização do Pantanal. Anais do 1º Simpósio sobre Recursos Naturais e Sócio-econômicos do Pantanal: 77-90.

De Salis, S., V. Pott e A. Pott. 1999. Fitossociologia de formações arbóreas da Bacia do Alto Paraguai, Brasil. In: II Simpósio sobre recursos naturais e sócio-econômicos do Pantanal, EMBRAPA Pantanal, Corumbá: 357-374.

93

Dolabella, A. L. 2000. The Brazilian Pantanal: an overview. In: F. A. Swarts (Ed.) The Pantanal of Brazil, Bolivia and Paraguay: selected discourses on the world's largest remaining wetland system. Hudson MacArthur Publishers, Gouldsboro. Dhakal, A., T. Amada, M. Aniya and R. Sharma. 2002. Detection of areas associated with flood and erosion caused by heavy rainfall using multitemporal Landsat TM data. Photogrammetric Engineering & Remote Sensing 68: 233-239. Eidenshink, J.C. and J.L. Faundeen, 1997. The 1-km AVHRR global land data set: first stages in implementation. International Journal of Remote Sensing. 51:39-56. Florenzano, T. 1998. Imagens TM-Landsat e HRV-SPOT na elaboração de cartas geomorfológicas de uma região do Rio Taquari, MS. Pesquisa Agropecuária Brasileira 33: 1721-1727.

Filho, J. 1986. Aspectos geológicos do Pantanal Mato-grossense e de sua área de influência. In: I Simpósio sobre recursos naturais e sócio-econômicos do Pantanal, EMBRAPA Pantanal, Corumbá: 63-76.

Filho, P., F. Ponzoni e M. Pereira. 1998. Mapeamento da fitofisionomia e do uso da terra de parte da Bacia do Alto Taquari mediante o uso de imagens TM/Landsat e HRV/Spot. Pesquisa Agropecuária Brasileira 33: 1755-1762.

Filho, Z. 1986. Solos do Pantanal Mato-grossense. In: I Simpósio sobre recursos naturais e sócio-econômicos do Pantanal, EMBRAPA Pantanal, Corumbá: 91-103. 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.

Frazier, P. and K. Page. 2000. Water body detection and deliniation with Landsat TM data. Photogrammetric Engineering & Remote Sensing 66: 1461-1467. Frey, R. 1995. Flora and vegetation of �Las Piedritas� and the margin of Lacuna Caceres, Puerto Suarez, Bolivian Pantanal. Bulletin of the Torrey Botanical Club 122: 314-319. Galdino S. e Robin T. Clarke, 1997. Probabilidade de ocorrência de cheia no Rio Paraguai, em Ladário. MS � Pantanal. 1997. Circulação Técnica N° 23, EMBRAPA-CPAP, Corumbá, MS.

94

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. 1999. Potential effects of a major navigation project (Paraguay-Paraná Hidrovía) on inundation in the Pantanal floodplains. Regulated Rivers: Research & Management 15: 289-299. Hamilton, H., S. Sippel, D. Calheiros and J. Melack. 1997. An anioxic event and other biochemical effects of the Pantanal wetland on the Paraguay River. American Society of Limnology and Oceanography 42: 257-272. Hamilton, H., S. Sippel, and J. Melack. 1995. Oxygen depletion and carbon dioxide and methane production in waters of the Pantanal wetland of Brazil. Biogeochemistry 30: 115-141. Hamilton, S., S. Sippel and J. Melack. 1996. Inundation patterns in the Pantanal wetland of South America determined from passive microwave remote sensing. Arch. Hydrobiol. 137: 1-23. Hayes, D.J., Sader, S.A. 2001. Comparison of change-detection techniques for monitoring tropocal forest clearing and vegetation regrowth in a time series. Photogrammetric Engineering and Remote Sensing; Vol. 67, No. 9; 1067-1075. Heckman, C. 1994. The seasonal succession of biotic communities in wetlands of the tropical wet-and-dry climatic zone: I. Physical and chemical causes and biological effects in the Pantanal of Mato Grosso, Brazil. Internationale Revue der gesamten Hydrobiologie 79: 397-421. Instituto Brasileiro de Geografia e Estadística. 1992. Manual técnico da vegetação brasileira. Séries Manuais Técnicos em Geociencias. Número 1. Juárez, R.N. and Liu, W.T. 2001. NDVI FFT analysis of spatial climatic variation in northeast Brazil". International Journal of Climatology, Vol. 21(14):1803-1820. Jiménez-Rueda, J., J. Pessotti e J. Tavares de Mattos. 1998. Modelo para o estudo da dinâmica evolutiva dos aspectos fisiográficos dos pantanais. Pesquisa Agropecuária Brasileira 33: 1763-1773. Kandus, P. 1999. El concepto de sucesión vegetal y su aplicación en sistemas de humedales deltaicos. In: Malvárez, A., Tópicos sobre grandes humedales templados y subtropicales de Sudamérica. Universidad de Buenos Aires. 162-177.

95

Kidwell, K. B., 1995. NOAA Polar Orbiter Data Users Guide. Satellite Data Service Division, NESDIS/NOAA, Washington D.C. Liu, W.T.H. Ayres F.M., Salles A.T. and Padovani C., 2002. Upper Paraguay River water level prediction using NOAA AVHRR NDVI. Proceedings of the 1º International Symposium �Recent Advances in Quantitative Remote Sensing�, Valencia, Spain. Liu, W.T.H., O. Massambani and C. Nobre 1994. Satellite Vegetation response to drought in Brazil. Int. Jounal Climatology, 14:343-354. Liu, W.T.H. and F. Kogan. 1996. Monitoring regional drought using vegetation condition index. Int. Journal of Remote Sensing, 17:2761-2782. Liu, W.T. and R. N. Juarez. 2001. ENSO drought Prediction of Northeast Brazil Using NDVI. International Journal of Remote Sensing, Vol. 22(17): 3483-3501. Lunetta, R. and M. Balogh. 1999. Application of multi-temporal Landsat 5 TM imagery for wetland classification. Photogram. Eng. & Remote Sensing 65:1303-1310. Lyon, J. 2001. Wetland Landscape Characterization: GIS, Remote Sensing, and Image Analysis. Ann Arbor Press, Chelsea. Margolis, M. 1995. Treasuring the Pantanal; International Wildlife Federation. Matovani, A. e S. Amaral. 1998. Avaliação preliminar da utilização de imagens AVHRR/NOAA na detecção de desmatamento no Pantanal. Pesquisa Agropecuária Brasileira 33: 1683-1690. Melack, J., L. Hess and S. Sippel. 1994. Remote sensing of lakes and floodplains in the Amazon basin. Remote sensing reviews 10: 127-142. Ministerio de Agricultura y Ganadería (MAG) y Bundesanstalt fuer Geowissenschaften und Rohstoffe (BGR). 1999. Proyecto de Sistema Ambiental del Chaco. Asunción. Ministério de Minas e Energia. 1982a. Projeto RADAM Brasil: programa de integração nacional. Levantamento dos recursos naturais. Vol. 26. Ministério de Minas e Energia. 1982b. Projeto RADAM Brasil: programa de integração nacional. Levantamento dos recursos naturais. Vol. 27. Ministério do Meio Ambiente, dos Recursos Hídricos e da Amazônia Legal. 1997. Diagnóstico dos meios físico e biótico: meio biótico. Plano de Conservação da Bacia do

96

Alto Paraguai (Pantanal). Vol. II, tomo III. Mittermeier R.A., I.G. Camara, M.T.J. Padua and J. Blanck. 1990. Conservation in the Pantanal of Brazil. Oryx 24: 103-112. Mittermeier, R.A., C.G. Mittermeier, P. Robles Gil, J. Pilgrim, G. Fonseca, T. Brooks and W.R. Konstant. 2002. Wilderness: Earth's last wild places. CEMEX, Mexico. Muchoney, D., J. Borak and A. Strahler. 2000. Application of the MODIS global supervised classification model to vegetation and land cover mapping of Central America. International Journal of Remote Sensing 21: 1115-1138. Neiff, J. 1999. El régimen de pulsos en ríos y grandes humedales de Sudamérica. In: Malvárez, A.I. (Ed.) Tópicos sobre grandes humedales templados y subtropicales de Sudamérica. UNESCO, Montevideo. Nordemann, D. 1998. Periodicidades, tendencias e previsão a partir da análise spectral dinâmica da série dos níveis do Rio Paraguai, em Ladário (1900/1995). Pesquisa Agropecuária Brasileira 33: 1787-1790. Olson, D.M., E. Dinerstein, E.D. Wikramanayake, N.D. Burgess, G.V.N. Powell, E.C. Underwood, J.A. D'Amico, H.E. Strand, J.C. Morrison, C.J. Loucks, T.F. Allnutt, J.F. Lamoreux, T.H. Ricketts, I. Itoua, W.W. Wettengel, Y. Kura, P. Hedao, and K. Kassem. 2001. Terrestrial ecoregions of the world: A new map of life on Earth. BioScience 51: 933-938. PCBAP. 1994. Conservation Plan for the High Paraguay River Basin. Por, F.D. 1995. The Pantanal of Mato Grosso (Brazil), Kluwer Academic Publishers, Dordrecht. Pott, V., A. Cervi, N. Bueno e A. Pott. 1999. Dinâmica da vegetação aquática de una lagoa permanente da Fazenda Nhumirim, Pantanal da Nhecolândiam - MS. In: II Simpósio sobre recursos naturais e sócio-econômicos do Pantanal, EMBRAPA Pantanal, Corumbá: 228-235.

Pott, A. and V. Pott. 1997. Plants of the Pantanal. EMBRAPA, Brasilia. Pott, A. e V. Pott. 1999. Flora do Pantanal � listagem atual de fanerógamas. In: II Simpósio sobre recursos naturais e sócio-econômicos do Pantanal, EMBRAPA Pantanal, Corumbá: 297-325.

97

Pott, A. e V. Pott. 2000. Plantas aquáticas do Pantanal. EMBRAPA, Brasilia. Ramella, L. y R. Spichiger. 1989. Interpretación preliminary del medio físico y de la vegetación del Chaco Boreal: contribución al estudio de la flora y de la vegetación del Chaco. Conservatoire et Jardin Botaniques de Genève 44: 639-680. Rao, C.R.N., and Chen J. 1995. 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. Ribeiro, J., e B. Machado. 1998. Fitofisionomias dos bioma cerrado. In: Cerrado: ambiente e flora. EMBRAPA, Planaltina. Seidl, A. F., J. S. V. de Silva, and A. S. Moraes. 2001. Cattle ranching and deforestation in the Brazilian Pantanal. Ecological Economics 36: 413-425. Scott, D.A. and M. Carbonell (Comps.). 1986. A Directory of Neotropical Wetlands. IUCN, Cambridge and IWRB, Slimbridge. Silva, J. dos S. V. da; Abdon, M. de M. , 1998. Delimitação do Pantanal Brasileiro e suas sub-regiões. Pesquisa Agropecuária Brasileira, Brasília, v. 33, Número Especial, p. 1703 � 1711. Shrestha, T. 1999. Land cover classification using satellite-sensed imagery and its texture values: an accuracy assessment based on the Florida land use and cover classification system. University of Florida. Swarts, F. 2000. The Pantanal in the 21st Century: for the planet's largest wetland, uncertain future. In: F. A. Swarts (Ed.) The Pantanal of Brazil, Bolivia and Paraguay: selected discourses on the world's largest remaining wetland system. Hudson MacArthur Publishers, Gouldsboro. Tso, B. and P. Mather. 2001. Classification methods for remotely sensed data. Taylor and Francis, Inc., London. Universidad Nacional de Asunción (UNA) y Deutsche Gesellschaft für Technische Zusammenarbeit (GTZ). 1991. Vegetación y uso de la tierra, Región Occidental del Paraguay (Chaco), 1986-1987. San Lorenzo. Welcomme R.L. 1995. River fisheries. FAO Fisheries Technical Paper N° 262.

98

Pantanal related websitesPantanal related websitesPantanal related websitesPantanal related websites

http://bolfor.chemonics.net/novedades/INDEX.HTM http://earth.esa.int/symposia//participants/data/henebry1/index.html http://life.csu.edu.au/bushfire99/papers/kitchin/ http://www.cla.sc.edu/geog/rslab/rsccnew/fmod8.html http://www.csc.noaa.gov/products/sccoasts/html/ccapguid.htm http://www.ducks.org/conservation/latinamerica.asp http://www.ngdc.noaa.gov/paleo/igbp-dis/frame/publications/wp_21/sc_wp_21.html http://www.wwf.or.id/forestfires/DOUGMICH.html