UAV-based vegetation monitoring for assessing the impact of soil loss in olive orchards in Brazil

Abstract

Vegetation cover is one of the most critical factors in soil erosion processes. Notably, olive orchards have been cultivated in shallow and sloping soils, with low vegetation cover and increasing the soil exposure to raindrop impact. In the tropics, considerable care is required to adequately use cover crops to control water erosion in new frontiers of olive plantations. In this context, we proposed a new technique to correlate the cover-management factor (C-factor) with vegetation indices from images obtained by unmanned aerial vehicle (UAV) and evaluate soil erosion losses under natural rainfall. We studied the relationship between different cover indices (vegetation cover index, non-photosynthetic vegetation cover index, and total cover index) with the C-factor of the USLE/RUSLE. This study was carried out in standard erosion plots with different vegetation cover systems associated with olive cultivation. UAV images were classified by Random Forest algorithm, and soil losses were quantified by sampling after each erosive rainfall event. Results showed a good performance in UAV image classification: average user's accuracy of 94% for vegetation class and 91% for bare soil. The Total cover index presented a better performance in predicting soil loss and determining the C-factor for exponential model (R2 = 0.87). UAV-based imaging demonstrates promising potential in monitoring vegetation cover crops and their impact on soil erosion. Total cover index performs better in estimating C-factor and predicting soil loss. However, the result of response surface analysis suggested that the association between total cover index and rainfall erosivity using second-order model presented the best prediction (R2 = 0.98), positive correlation between rainfall erosivity and C-Factor, and negative correlation between C-factor and total cover index and rainfall erosivity.