Predicting stormflow response of a degraded tropical grassland catchment using a spatially variable infiltration model

Abstract

<p>Reduced surface infiltration capacity (K<sub>sat</sub>), increased infiltration-excess overland flow (IOF) and soil loss after deforestation and subsequent surface degradation in the humid tropics are well-documented. However, attempts to predict concomitant increases in storm runoff using physically-based approaches or to relate infiltration model parameter values calibrated with observed hyetographs and hydrographs at the small catchment scale to point-based measurements of K<sub>sat</sub> are rare. We used measured rainfall intensity and stormflow rates at 5-min intervals for 37 separate events (receiving 5–154 mm of rain) from the 3.2 ha degraded fire-climax grassland Basper catchment (Leyte Island, Philippines) to evaluate the performance of a spatially variable infiltration (SVI) model. SVI relates actual infiltration rates to rainfall intensity and a spatially averaged infiltration parameter I<sub>m</sub> after an initial infiltration amount F<sub>0</sub> and has been used successfully to predict IOF at the plot scale at various tropical locations. Quickflow hydrographs were produced using the Hewlett & Hibbert straight-line separation method and actual infiltration rates were derived by subtracting 5-min quickflow rates from corresponding rainfall inputs. SVI-predicted actual infiltration rates were compared with observed rates to derive optimized values of I<sub>m</sub> and F<sub>0</sub> per event. Earlier work at Basper had revealed very low (near-)surface values of K<sub>sat</sub> (implying frequent IOF although there was reason to suspect that K<sub>sat</sub> was underestimated). No explicit measurement was made of hillslope IOF, but stable isotope mass balance computations and a high degree of stream-water dilution during times of rain suggested large contributions of ‘new’ water of low electrical conductivity that likely represented OF. Whilst SVI generally replicated individual quickflow hydrographs very well, values of I<sub>m</sub> and F<sub>0</sub> varied markedly between events. Using the median values of I<sub>m</sub> (46 mm h<sup>-1</sup>) and F<sub>0</sub> (6.8 mm) produced reasonable to good results (NSE > 0.6) for a subset of 15 (larger) events only. F<sub>0</sub> was positively related to maximum rainfall intensity over 15 or 30 min while I<sub>m</sub> was not significantly correlated to measured (mid-slope) soil water content or precipitation-based antecedent wetness indicators. However, I<sub>m</sub> exhibited a significant inverse correlation (Spearman r<sub>s</sub>=-0.617) with pre-storm baseflow rate Q<sub>b</sub> (notably for Q<sub>b</sub><0.5 mm d<sup>-1</sup>) suggesting foot-slope wetness status may be important for stormflow generation as well. The spatial distribution of K<sub>sat</sub>-values implied by SVI confirmed the suspected under-estimation of field-based K<sub>sat</sub> across the measured range, presumably reflecting a combination of macropore smearing (near-surface Amoozemeter measurements) and the limited size of the double-ring infiltrometer used for the measurement of surface infiltration rates.</p>