UAV-BASED SOIL ORGANIC CARBON ASSESSMENT IN COMPLEX TERRAIN: A REVIEW OF SENSING PATHWAYS, TERRAIN CONTROLS AND MODELLING STRATEGIES

Abstract

Soil organic carbon (SOC) is distributed unevenly across landscapes because carbon inputs, decomposition, erosion, deposition, soil moisture and management seldom vary independently of one another. Unmanned aerial vehicle (UAV) platforms have made field-scale SOC assessment technically feasible, yet variable terrain alters both the true carbon pattern and the remote-sensing signal from which it is inferred. This review develops a terrain-explicit framework for UAV-based SOC assessment by synthesising published work on UAV multispectral, hyperspectral, RGB and LiDAR sensing, satellite-assisted mapping and terrain-augmented digital soil mapping. It departs from earlier SOC remote-sensing reviews by treating terrain as an organising variable instead of one predictor among many. Studies from erosion-prone cropland, hummocky ground moraine, semi-arid farmland, sloping vegetable fields, mountainous tobacco fields, coastal wetlands, tropical forest, desertified land and permafrost gullies show that the best UAV models cannot be identified by sensor type alone. Their performance depends on sampling design, bare-soil timing, surface moisture, crop residue, canopy obstruction, topographic covariates and whether the validation scheme tests transfer beyond the calibration field. Random forest, XGBoost, support vector methods, partial least squares regression, deep neural networks and hybrid geostatistical models each perform well under particular data conditions, but none of them removes the need for pedological reasoning. We propose a workflow that links UAV observation, terrain stratification, laboratory calibration, model validation, uncertainty mapping and land-management interpretation. Terrain-aware UAV-SOC assessment is most reliable when spectral information is read together with geomorphic position, hydrological setting and soil redistribution, particularly where erosion and deposition create sharp SOC contrasts within a field.