Object‐oriented soil erosion modelling: A non-stationary approach towards a realistic calculation of soil loss at parcel level.

• Monthly soil loss estimation at farmland level delivered highly realistic results. • Joint spatiotemporal R-factor and C-factor delineation in the same application. • C-factor was based on a spatially explicit land use dataset. • Erosion variability, hotspots amid risky seasons, critical land uses identification. • Upscaling and reproducibility potential for larger scale time-varying modelling. The spatiotemporal variability of rainfall and vegetation play a key role in the soil erosion process. However, most research efforts tend to overlook their strong spatial diversification and time-varying character. The parameters are often expressed in a stationary manner focusing on long-term (usually annual) averages, and/or at rather rough spatial analysis. Apparently, a spatially definite non-static parameterization in the same modelling application leads to more realistic soil loss results. In this context, the study attempted to quantify temporal land degradation at the Sperchios river watershed using the RUSLE model. The dynamic R-factor and C-factor coefficients were delineated on a monthly time step accounting for the climatic and biomass seasonality, respectively. Furthermore, C-factor was estimated at farmland level based on a highly detailed land use/land cover (LULC) dataset that provided explicit definition of its cultivated classes (as to holdings demarcation and crop type identification). Intra-annual soil loss variability, hotspots in conjunction to high-risk seasons, and critical land uses were successfully identified. The outputs will assist agronomists and stakeholders to implement targeted (hence cost-/labour-effective) mitigation measures and optimum erosion control strategies, especially amid the riskiest period-area coupling. The methodology's reproducibility potential merits the upscaling of dynamic erosion simulation at larger spatial units. [ABSTRACT FROM AUTHOR]