Forests as 'sponges' and 'pumps': Assessing the impact of deforestation on dry-season flows across the tropics

Forests act as 'pumps' through their evapotranspiration (E-tot) and as 'sponges' by enhancing soil infiltration capacity and moisture retention. Tropical deforestation and poor post-forest land management generally result in lower E-tot, but also reduce infiltration. Strongly diminished infiltration is typically accompanied by enhanced overland flow and can cause reduced groundwater recharge and baseflows. A grid-based land surface hydrological model (W3RA LUM) was tailored to incorporate the trade-offs between the 'pump' and 'sponge' effects to investigate where deforestation can be expected to have the greatest impacts on dry-season flows. Streamflow sensitivity analyses for scenarios with or without full forest cover and/or good versus poor surface infiltration were performed for: (i) selected tropical catchments with documented changes in streamflow after deforestation; and (ii) the tropics at large (23.5 degrees N to 35 degrees S, to include important seasonal montane forests). The catchment sensitivity analyses showed that W3A LUM captured the streamflow response to imposed deforestation and changes in surface conditions reasonably well. For the tropics as a whole, an increase in mean annual streamflow of 18% was obtained for forest conversion (to pasture) only, versus 26% after an additional imposed reduction in surface infiltration. Much of the inferred flow increases concerned water-limited regions. A reduction in dry-season flows was predicted for nearly one-fifth of all grid cells (despite lower E-tot after forest conversion) after impaired infiltration, marking potential 'hot spots' of hydrological change after deforestation and reduced infiltration due to surface degradation. The affected grid cells shared the following key characteristics: (i) strong seasonality; (ii) high infiltration capacity under forested conditions; (iii) sufficient wet-season precipitation to recharge deep soil- and groundwater stores; (iv) sufficient soil water storage under forested conditions to 'carry over' infiltrated wet-season rainfall into the subsequent dry season; and (v) slow groundwater recession maintaining baseflow throughout the dry season. Our results demonstrate that forest removal in highly seasonal tropical catchments, whilst typically increasing mean annual water yield, can indeed decrease dry-season flows, depending on pre- and post-forest removal surface conditions and groundwater response times.