Estimates of Deforestation Rates in Rural Properties in the Legal Amazon

Abstract

Abstract This study aimed to assess the potential of artificial neural networks (ANN) as a tool to estimate deforestation rates in the municipality of São Félix do Xingu, PA, Brazil. The following input variables were used: deforestation rate until 2014, slope, altitude, Euclidean distance to roads and rivers, permanent preservation area (PPA), and property area. A total of 2,800 properties were used, of which 2,000 were used for training and 800 for validation of the networks. The input layer included nine neurons: six as quantitative variables and three as categorical variables. The output layer included a single neuron - the deforestation rate. The training results indicated high correlation (r = 0.92) and root mean square error (RMSE) of 12.4%. Validation of the model estimated RMSE = 12.9% and r = 0.91. The study results evidenced the high potential of ANN as a tool to estimate farm deforestation rates.

Keywords

• deforestation
• artificial intelligence
• validation