Mapeamento pedológico por meio de série histórica Landsat-5 TM e Biblioteca Espectral na Bacia do Rio Jardim (DF)

Knowledge of soils and their spatial variability are scare for the application of a rational management, promoting high productivity and low environmental impact. Among the information most used for planning and decision-making in agriculture, there is the soil map, which is a source of information on a number of attributes such as texture and chemical and morphological conditions. On the other hand, these maps are scarce or present in inadequate scale to support sustainable agricultural planning. The objective of this work was to map digitally soil classes of Jardim River Basin (DF) by remote sensing techniques, with the purpose of providing information more reliable soil and representativeness of the study area. Soil profiles we selected in five representative litho-toposequences located in Geomorphological Surface of Plans Intermediates, DF. Soil samples from horizon A we collected at 37 points and obtained reflectance readings. Spectra of 34 samples we grouped according to their main attributes, in 13 groups and calculated the average. A soil spectral library we elaborated with these 13 average spectra and resampled in the spectral ranges of the Landsat 5-TM bands. Ten images of the satellite Landsat 5-TM from 1984 to 2009 we selected to generate a composite image of bare soil. Analysis models of spectral mixture (SMA and MESMA) we applied to the composite image using 13 endmembers spectral library for the soil classes mapping of the study area. To compute the field truths, we used a rectangular grid of cells with 1.400 x 1.400 m. The performance of the MESMA we assessed using the kappa index. The soils of the five litho-toposequences showed similar morphological characteristics, differing mainly in color attribute. The soils were classified in 2nd and 4th category level of the Brazilian System of Soil Classification. The composite image showed 74% of the area composed of bare soil. Soil spectra classes we characterized showed standard reflectance variations mainly due to the variation in the color of the soil, the content of iron oxides and organic matter. The map we generated by the SMA model showed high confusion because of the limitations of this methodology. The digital map of soil classes obtained by the MESMA model presented Kappa index of 73%, showing an accurate distribution of the distribution of soil classes of the evaluated area.