Knowledge of spatial and temporal distribution and monitoring of its dynamic evolution are the most important factors for the study and management of wetlands. The aim of this study was to characterize, to map, to compare the spectral response of coastal wetlands, as well as detect the changes on the eastern shore of Marajo, State of Pará, using images from TM Landsat-5 reflectance. To this end, the images were geometrically and radiometrically corrected. Were collected in the reference image (2008), at least 20 polygonal samples (5x5 pixels) for each type of land cover. The analysis of variance (ANOVA), Tukey HSD Test and an index of spectral separability of pairs of regions of interest (ROIs) were calculated. The mapping was generated from the supervised classification Spectral Angle Mapper, and validating data, outlined by the Confusion Matrix. Thus were recognized the following units: MAN - mangrove, MAD - degraded mangrove, PRD – beaches and dunes, VSI – initial secondary vegetation, VSA - advanced secondary vegetation, ACS – water with sediment, ASS - water without sediment, OCH - human occupation and CAM – marshes. The result of ANOVA showed that there are significant differences between the average reflectance in all classes and at least one pair of means, for all bands (1-5 and 7) image. The Tukey HSD test found that the smallest difference between two means of bands 1 and 2 occurs in pair VSAMAN, the band 3, VSI-MAN, the band 4, OCH-MAN, the band 5, OCH-PRD, and band 7, ASS-ACS. The function of spectral separability of pairs of ROIs highlighted a low value for the pair of classes OCH-CAM. The calculation accuracy of the mapping showed acceptable values. Was also applied the technique of spectral mixture model to determine the fractions – green vegetation, water and/or shade, soil and vegetation not photosynthetically active – in reflectance images in years from 1988, 1994, 1998, 2004, 2006 and 2008. The detection of changes over the years was done with the help of the tri-linear diagram of Thompson, used to analyze the evolution of Mangrove environments, Salt Marshes with a predominance of soils, Grassland Salt Marshes, Beaches and Dunes, Wet Sediment, Water with Sediment and Water without Sediment. The results showed that 80% of the samples from mangrove environment, maintained their spectral characteristics, ie, the environment has not changed over time, and that 15% of possible paths indicated a gradual change for the formation of mangrove until 2008. Of the total number of pixels analyzed for this environment, 93% corresponded to the actual mangrove, confirming a relative stability of this environment in the study area. Salt Marshes showed an unusual condition in the analysis of the image of the reference year (2008), with spectral response similar to the soil characteristics, this is not checked in years gone by. In recent years, there was a predominance of Grassland Salt Marshes. As for the environment Beaches and Dunes revealed a considerable increase of its surface (66.7%) in the 20-year period (1988-2008). Overall, the results suggest that the study area has undergone the process of progradation of sediments along the shoreline and stabilization in the increase of surface Mangrove since 2006. For the Salt Marshes detect the variations in the abundance values among the three fractions (Green Vegetation, Shade/Water, Non-photosynthetically Active Vegetation and Soil) generated by the spectral mixture model, mainly due to seasonal climate in the region. The method of change detection applied fostered the development of a model spectral dispersion diagram for coastal wetlands. This model should assist in future research about the monitoring of the impacts of the humid coastal environments from the increase in sea level, the location and spectral characterization, and change detection.
|
