MULTITEMPORAL REMOTE SENSING APPROACH FOR MAPPING WETLANDS SHALLOW LAKES IN THE PARANÁ RIVER FLOODPLAIN

BORRO, MARIA MARTA; MORANDEIRA, NATALIA; GONZALEZ TRILLA, GABRIELA; SALVIA, MARÍA MERCEDES; MINOTTI, PRISCILA; KANDUS, PATRICIA

Orlando

Congreso; 9 th INTECOL International Wetlands Conference "WETLANDS IN A COMPLEX WORLD"; 2012

One of the main features of South America floodplains is the presence of shallow lakes (SL) included in the wetland matrix, which are key elements to sustain their high biological diversity. Mapping of SL is critically required to understand wetland dynamics and for ecosystem monitoring. However, problems arise when trying to identify their boundaries, due to the fragmented nature of wetland landcover types, the coverage of macrophytes in the floodplain, and the spatial and temporal variability of the actual area occupied by the SL. In order to discriminate SLs from the wetland matrix in the Paraná River Delta floodplain, we used 77 NDVI scenes (Path 226, Row 83) derived from Landsat TM, from 1987 to 2010, representing different hydrometric levels. A NDVI threshold of 0.3 was estimated based on pixel samples of different types of SL and surrounding wetlands landscape elements. Individual scenes were segmented into binary SL images and a SL frequency image was calculated for the whole dataset. We applied a mask to exclude the river courses. The same analysis was performed for set of images corresponding to high, medium and low water level. To evaluate the minimum number of images needed to reduce the variability in SL cover we run a sensitivity analysis with stacks composed of different numbers of SL images (5 to 30) taken at random. We found it requires at least 20 images to obtain a trustful result of SL cover. The resulting coverages were contrasted to an existing map of SL and ground data and all the Kappa indices showed high accordance (> 0,85). SLs represent circa 0,1% of the studied region. For the whole 1987 - 2010 period, the maximum coverage occurred for images with high water levels (205,690.59 ha). The lowest SL coverage (59,227.92 Ha) occurred for low water levels of the dry period between 2000 - 2010. The use of time series of LANDSAT images helped distinguish the diversity of SL immersed in the floodplain wetland mosaic, for its spatial resolution allowed the detection of SL with an area less than one hectare which are the predominant in the Paraná River Delta region.