Publications

da Silva, HJF; Goncalves, WA; Bezerra, BG (2019). Comparative analyzes and use of evapotranspiration obtained through remote sensing to identify deforested areas in the Amazon. INTERNATIONAL JOURNAL OF APPLIED EARTH OBSERVATION AND GEOINFORMATION, 78, 163-174.

Abstract
The objective of this study was to improve the understanding of the spatial dynamics of evapotranspiration (ET) with a focus on the influence of the current level of Amazonian deforestation through comparative statistics of different land cover. The study area comprised the state of Rondonia, Brazil, which was subdivided into homogeneous regions regarding ET using cluster analysis. In addition, we analyzed the use of a logistic regression model to create deforestation maps in the Amazon based on ET fields. We used orbital data on ET and land cover type from the MOD16 product and the Amazon Forest Satellite Monitoring Project (PRODES), respectively, considering the period from 2000 to 2014. The cluster analysis results showed that for the study area, three homogeneous sub-regions were sufficient to represent the ET variability, mainly considering the intensity and seasonal pattern of this process. Regarding the impacts after changing from forest to deforested area, the analyses indicated that the ET of deforested areas decreased by an average of 28% in the dry period and increased by 4% in the rainy season. The effects observed in the rainy season were not significant at 5% significance according to the Student t-test, unlike the dry period, which presented statistical significance (p-value < 0.05). In general, the results indicated that MOD16 data can provide a satisfactory representation of the change in ET in large areas of the Brazilian Amazon. Logistic regression analysis showed that the spatial pattern of deforestation can be identified by biophysical factors such as ET with 87% accuracy, despite the spatial/temporal variability present in the region. These results support this approach as an effective tool for spatial identification of Amazonian deforestation.

DOI:
10.1016/j.jag.2019.01.015

ISSN:
0303-2434