A trilateral cooperation for
the development of northeastern Brazil, II Klaus A.
Ulbricht DLR-Optoelektronik, 8031 Oberpfaffenhofen, Germany Harendra S. Teotia Univ. Federal de Paraiba, CCT-UFPB, 58100 Campina Grande, Brazil Daniel L. Civco,William C. Kennard University of Connecticut, Storrs, CT, 06269, U.S.A. Abstract: For a number of years there has been cooperation between the ‘Centro de Ciencias Tecnicas’ (CCT) of the ‘Universidade Federal do Paraiba’ (UFPB), Brazil, the ‘Institute fur Optoelektronik’ of the German Aerospace Research Establishment (DLR-OE), Germany, and the College of Agriculture and Natural Resources of the University of Connecticut (UCONN), USA, aiming at the improvement of the undeveloped and underdeveloped parts of the North-East of Brazil with help of remote sensing. Satellite imagery on computer compatible tapes has been evaluated with the help of a sophisticated laser beam recorder at DLR in Oberpfaffenhofen, Germany, to produce false colour images and single band evaluations from LANDSAT TM and SPOT satellites. Optical evaluation led to maps giving drainage charts for land irrigability, as well as morphophysical characteristics of the soil, geological landscape characteristics and major soil classes and soil association. Detailed digital image processing at the Laboratory for Remote Sensing of the University of Connecticut resulted in maps of land use and land cover classifications, soil associations, slope classes, soil mapping units, an erosion model for a particular area, and agro technical limitations. Several papers have been published at the more important earth reconnaissance conferences (1-7) and lectures and reports have been addressed to students and interested persons. A proposal for the installation of a remote sensing laboratory has been referred to a Brazilian University (8), and projects for training and education of students of higher academic level in Brazil, coupled with on the job training, have been submitted to funding organizations (9, 10) taking advantage of the experience of the German remote sensing center at DLR-Oberpfaffenhofen or the dedicated image processing at UCONN, and taking advantage of the more progressed education possibilities in Germany (Berlin) compared to Brazil. Last, but not least, limited financial sources for the pursuit of the work, including the purchase of tapes, photographic material, trips to Brazil, Germany, and the US, for information exchange and lectures, as well as training and teaching, had to be secured. Particular results of the cooperation, gathered from SPOT computer compatible tapes, are presented. Objectives Main purpose of the cooperation is the development of the Northeast of Brazil, to be acquired by
Investigated Area The research area in the northeast of Brazil was selected according to development necessity, personal experience, i.e. available ground truth, and coverage by LANDSAT and SPOT satellite. Computer compatible tapes (CCT) of cloudfree imagery investigated were:
As well as CCT’s of related areas. Additional information sources for the investigation were aerial photography and topographic maps. Photographic enlargements of false colour composites, real colour composites, and b/w prints of single bands were used for the optical interpretation. Particular digital investigations, correlated with ground truth acquisition, were centered around the Santa Luzia area in the state of Paraiba, as well as the Picos area in the state of Piaui. Methodology Optical interpretations were done using b/w recordings as well as false colour composites of LANDSAT TM and SPOT satellite spectral bands by DLR’s FIRE laser beam recorder. An enlarged paper print was placed on a light table and boundaries of homogeneous area were drawn in by hand. The mapping units this recorded were identified by.
A hybridized unsupervised-supervised classification approach was used for analysis of land use/cover and other earth resources information, and modified/corrected interactively. The classification approach was modeled after the USGS-System for use with remote sensing data, modified to account for locate conditions within the study area. Data in a later stage of the investigations were geometrically corrected and georeferenced to Universal Transform Mercator System. Auxiliary Means Information sources for the investigation, as well as instrumentation used in pursuit of the work, were: The land capability classification (12) Brazilian approximations (1979, 1981) Soil approximations of SUDENE 1972 (13) EMBRAPA Empresa Brasileira de Pesquisa Agropequaria Soil taxonomy of 1975 (14) Atlas Geographical do Estado da Paraiba of 1985 (15) Land use-cover acc. To J.R. Anderson et al. 1976 (11), and Computer compatible tapes of LANDSAT and SPOT satellites, Aerial photography, Topographic maps, Soil survey maps at the small scale of Brazilian and FAO/UNDP systems, Topographic maps of SUDENE (Superintendencia de Desenvolvimento de Nordeste) Ground truth aquired at the UFPB in Paraiba, Brazil, and The FIRE Laser Beam Recorder at DLR-Oberpfaffenhofen, Germany the ERDAS image processing system at the University of Connecticut, Storrs, USA the geographical information system correlated with ERDAS at UCONN, USA. Results/Optical Interpretation Because of limited space, only few examples of optically evaluated satellite imagery are presented (SPOT 728/364, S 07000’, W 36055’)
Parts of SPOT scenes were investigated by digital techniques. Training areas for supervised maximum likelihood classification were selected by interactive correlation (recoding) together with map- and ground truth verification. The classification served to identify land categories at different levels. Maximum likelihood classification could identify general land categories at level I, more detailed categories at level II, and very detailed (but less accurate) categories at level III (4, 5, 14). Depending (partially) on uniform training areas (to be controlled by computer printouts of grey values), results of maximum likelihood classification arrived at an overall classification accuracy of better than 90 %. With the help of UCONN’s ERDAS image processing system, maps of land use/cover, soil, land capability, and soil slope were prepared, trying to support local authorities in planning processes and decisions, if necessary and wanted. The production of maps was done using US Geological Survey (USGS) and US Department of Agriculture (USDA) classification systems. Main areas ‘investigated digitally’ were the Santa Luzia area in the state of Praiba (6), and the region of Picos in the state if Piaui (7). Furthermore, a tentative geographic information system has been developed, and an erosion model for the Santa Luzia municipality been produced with its help (3). These pilot results are the foundation for a development project for the northeast of Brazil, submitted to the proper authorities (9), comprising the training and working capacities of Brazilian scientists on remote sensing investigations of the northeast of Brazil for its development. Bibliography
|