Combining field data and a spatially distributed model to understand the effects of land cover, soil degradation, and climate variability on the hydrological response of a meso-scale catchment in Eastern Madagascar

Land cover in areas undergoing long-term shifting cultivation typically represents a mosaic of agricultural fields, fallows in different stages of regrowth, remnant forest, and degraded grasslands. The Ankeniheny Zahamena corridor in eastern Madagascar represents a case in point. Previous research revealed major differences in hydrological response between forests, fallows or reforested sites, and degraded grasslands. We used these field data in the physically-based RoGeR_Dyn model[1] to examine the effects of topography and land cover on hydrological processes at the meso-catchment scale (58 km²). Using historic land-cover maps, predicted deforestation rates, and a theoretical reforestation rate, we created eight land-cover scenario’s with forest fractions ranging from 25 to 97% and combined these with five rainfall scenario’s (range: 1050 – 2100 mm/year).

The simulations showed that total evapotranspiration increased with rainfall and was ~20% less for the most degraded land-cover scenario than for a land cover dominated by (non-degraded) forest in all rainfall scenario’s. Topsoil infiltration capacities were reduced following soil degradation but still exceeded maximum hourly rainfall intensities. The lower vegetation water use in the degraded scenario produced wetter soil conditions, leading in turn to moderate increases in saturation-excess overland flow as well as deep percolation (baseflows). Deep percolation increased >2.4 times when annual rainfall was increased two times (regardless of land cover). The total stormflow in the dry scenario was more than three times larger for the most degraded land cover than under full forest (48 mm vs. 14 mm) and increased by an order of magnitude when rainfall was doubled (359 mm vs. 220 mm, respectively). Restoring forest on degraded soils is thus seen to change runoff processes, mostly by reducing overland flow and stormflow, and to a lesser extent by decreasing deep percolation due to increased evapotranspiration.

Above all, the simulations show that a hydrological model driven by measurements at the point- or plot-scale and very limited calibration, can provide useful information on how climate, land cover and soil degradation together affect the occurrence of high and low flows.

[1] Schwemmle, R., Leistert, H., Steinbrich, A., and Weiler, M.: RoGeR v3.0.5 – a process-based hydrological toolbox model in Python, Geosci. Model Dev., 17, 5249–5262, https://doi.org/10.5194/gmd-17-5249-2024, 2024.