Faults in power distribution feeders cause damage to power utilities due to the deterioration of reliability and power quality indexes and the displacement of field maintenance teams to replace or repair power grid equipment. Additionally, consumer units have energy supply interruptions for an undetermined time. Studies in specialized literature usually detect, classify, and locate faults after they occur. In contrast, preventing faults by estimating areas vulnerable to them is crucial to mitigate all inconveniences and additional costs after they occur. Tree vegetation is an essential factor contributing to faults. In this sense, an enhanced method for tree vegetation mapping by areas is developed using multilayer perceptron neural networks trained on high-resolution images from Google Earth. A geographic space is incorporated to estimate the regions vulnerable to failures due to tree vegetation. Geographically weighted spatial analysis is applied from local variables aggregated by areas. Spatial data analysis is used to real faults and tree vegetation data from a medium-sized Brazilian city via QGIS and R programming environments. As a result, thematic maps are produced with the areas whose feeders are vulnerable to faults, where there is a moderate positive correlation by regions between the faults in distribution transformers and tree vegetation in the northeast and southwest areas of the city under study.