Intercomparison between forest cover mapping for two areas in Santa Catarina (Brazil)

Land cover mapping, especially with regards to forest cover, resulting from different methodologies will likely disagree in forest cover proportions. This study evaluates and compares consolidated thematic maps, in vector and raster formats, derived from national and international research efforts at larger regional or global extents. The municipalities of Blumenau (BL), Rio do Oeste and Presidente Getúlio (RO/PG), in the state of Santa Catarina (Brazil), were selected as representative of both a high percent and a low percent forest cover within the subtropical evergreen rainforest. In this characterization of the forest cover response at local scales when using generalized products, the raster products provided the highest scores: e.g., 97% in BL and 83% in RO/PG for rasters versus 70% and 46% for vectors respectively. The minimum scores attained were respectively 65% and 34% (rasters) versus 52% and 18% (vectors). All in all, on the basis of fotointerpretation and field work estimates of >50% of forest surface in BL and <30% of forest in RO/PG, too divergent over/under-rated scores are found. As expected, overlap of maps resulted in reduced coincident forest surface: agreement of less than 50% (e.g., 35% for BL), decreased considerably to a 5% in RO/PG. Nonetheless, vector-vector and raster-raster comparisons work better (e.g., agreement in RO/PG vector 10.4% and 13.5% raster). Although the diversity of methods and forest definitions lie beneath the discrepancies found, the need of homogeneous accurate workflows and products in forest cover estimation and monitoring at regional and local scale is proved.