Binary Logistic Regression applied to erosion susceptibility mapping in the Southern Amazon
Palavras-chave:Susceptibility, Logistic regression, Soil erosion on Amazon
Problems with soil erosion by water wind in the Brazilian Amazon are intensifying as the forest is replaced by agricultural production. Deforestation, burning, logging, and advancing the agricultural frontier have altered the soil-vegetation balance. In this context, the analysis of soil erosion susceptibility is one of the most significant challenges to developing long-term sustainability strategies and policies. Based on the above, the present study used the principles of Statistical Modeling - Logistic Regression - to develop and validate a model for analysis of susceptibility to soil erosion using 14 environmental factors. The study was carried out in a hydrographic sub-basin with 330 km2, located in the south of the State of Rondônia in the western Amazon, which combines characteristics of intense anthropic activity, loss of fertile soil, gullies, and silting of rivers. The study area has rainfall above 2000 mm year-1, they are transcurrent shear zones, predominant relief forms are flat to slightly convex tops, drainage networks are dendritic in an exorheic system, vegetation cover is composed of areas of forests and natural or regenerated forest fragments, agriculture is destined to annual crops. Livestock is extensive, with a predominance of small rural properties. The logistic regression model showed satisfactory results with an AUC of 0.888 and global accuracy was 0.77. The variables with the most significant effect on the equation were NDVI, erosivity, and TST. The mapping found that 57.71% of the study area is in places susceptible to soil loss due to water events.
Este trabalho está licenciado sob uma licença Creative Commons Attribution-NonCommercial 4.0 International License.
Autor(es) conservam os direitos de autor e concedem à revista o direito de primeira publicação, com o trabalho simultaneamente licenciado sob a Licença Creative Commons Attribution que permite a partilha do trabalho com reconhecimento da autoria e publicação inicial nesta revista.