A minimal model for soil-erosion connectivity in drylands
Abstract. Soil erosion is a major contributor to the degradation of soil resources underpinning ecosystem functioning across terrestrial systems worldwide. Yet existing erosion models often require extensive parameterisation and computational resources, limiting their spatiotemporal generalisability. Here, we adapt a dissipative sandpile model, incorporating spatially heterogeneous vegetation and microtopography, to reproduce erosion patterns and provide mechanistic insights into connectivity-driven dynamics in drylands. We benchmark our model against a process-based erosion model and field observations, using spatial connectivity patterns and event-scale sediment exports. We introduce a structural connectivity (SC) measure to predict erosion hotspots and find strong correlations (ρ ≈ 0.88) with the landscape's functional connectivity (FC), particularly in shrubland. Using generated landscapes, we disentangle the contributions of vegetation-microtopography coupling, vegetation cover, and patchiness. We find that this coupling, associated with shrub encroachment, can counteract and reverse the effect of vegetation loss alone on connectivity and the propensity for large erosion events. Besides, SC-FC correlations are inversely related to the avalanche scaling exponent (mean Pearson correlation ≈ 0.6), indicating improved predictability of erosion hotspots in landscapes prone to large-scale geomorphic feedback. Our results demonstrate how a minimal sandpile framework can isolate mechanisms underlying connectivity-driven erosion while providing a computationally efficient complement to detailed process-based models.