<![CDATA[Spatial information on deforestation is essential for supporting site-level forest management in areas experiencing land-cover change. This study aimed to detect deforestation in the Tepo Asa Aroa Forest Management Unit and to model its spatial probability and driving factors. Spectral changes between 2015 and 2026 were quantified using Change Vector Analysis, in which magnitude and direction were derived from multispectral bands and vegetation-related indices. These variables, together with spectral and index differences, were used as predictors in a Random Forest algorithm to detect deforestation. Binary logistic regression was then applied to model deforestation probability using distance from roads, elevation, slope, distance from settlements, distance from forest utilization permits, distance from rivers, population density, and distance from agricultural areas. The results showed that deforestation covered 1,956.23 ha during 2015-2026, with an average annual rate of 177.84 ha. All driving factors were statistically significant. Distance from roads, elevation, slope, distance from forest utilization permits, distance from rivers, and distance from agricultural areas reduced the odds of deforestation, whereas distance from settlements and population density increased them. The model showed good discrimination, with an area under the receiver operating characteristic curve of 0.889. The predicted probabilities were classified into low, moderate, and high deforestation susceptibility classes. These findings provide spatial evidence to support monitoring, field verification, rehabilitation planning, and sustainable forest management.]]>
Copyrights © 2026
