Dados do Trabalho

Título

IS IT POSSIBLE TO IDENTIFY SOIL EROSION BY SATELLITE IMAGE SPECTRAL SENSING?

Resumo

Brazil loses an estimated 616.5 million tons of soil annually due to erosion. The aim of this study is to identify erosion sites using VIS-NIR satellite imagery. The area of interest covers the state of São Paulo, Brazil, in which we selected 16 known erosion sites for analysis. A Synthetic Soil Image (SySI), generated by the Geospatial Soil Sensing System (GEOS3) using Landsat was the source of reflectance data. The system creates an image where only the bare soil appears in a certain period. We identified the eroded and non-eroded areas at each site using QGis software comparing the spectral curves. In addition, a temporal soil color observation served to assess both locations using a system in Google Earth Engine. This allowed a comparison of eroded and non-eroded soil over time and space. The observation of the two techniques allowed the detection of erosion sites. The spectra of eroded spots were different from non-eroded ones. The spectral signature was different mainly in the range where the organic matter acts, becoming the spectra of eroded areas convex shaped. Another specific band was 5 of the Landsat related to the performance of quartz under spectral patterns. Generally, the eroded areas showed a higher reflectance peak and intensity in this SWIR 2 band, from 1900 nm, indicating that there was loss of the A horizon, more common in ultisol areas. Spectral curves were able to detect eroded areas reaching 50% of accuracy. Temporal Soil color reached about 20-30% accuracy. Detecting eroded areas is complex and requires new and quick technologies development. The main limitation is the spatial resolution with 30 m. Spectral sensing has potential, but still needs more extensive work, in this way future technology may bring better resolution to sensors than we have nowadays.   

Palavras-chave

Sysi; Google Earth Engine; soil health; soil quality.

Instituição financiadora

The State of São Paulo Research Foundation (FAPESP) 2021/05129-8.

Agradecimentos

The authors thanks GeoCis Research Group for sampling and analyses.