Quantification of hillslope runoff and erosion processes before and after wildfire in a wet Eucalyptus forest

An experimental program was initiated following an intense wildfire in 2003 in NE Victoria Australia, to quantify an expected large increase in hillslope erosion by rill and interrill processes from a steep, wet, Eucalyptus forest. Water repellence, soil water content, infiltration capacity, infiltration-excess runoff generation, interrill erodibility, and rill erodibility were measured periodically for 3 years. Rill erodibility on 27° slopes increased immediately following the fire by a factor of 540 times, but decreased exponentially to very low background levels 2 years after the fire. Despite this increased erodibility, rill erosion was uncommon in the catchment, because soil saturated hydraulic conductivity (K(sat)) remained very high (100-1000 mm h-1) and simulated concentrated flows of 3 l s-1 were found to be “transient”, infiltrating into areas with high K(sat) a short distance downslope. Despite the very high K(sat) values, runoff ratios from 100 mm h-1 simulated rainfall as high as 65% were recorded 6 months after the fire due to very high spatial variability in K(sat). The runoff sediment concentration under 100 mm h-1 simulated rainfall increased 10-fold immediately following the fire and declined exponentially to pre-fire levels within 2 years as ground cover recovered to 90-95%. Considered together, the above results suggest that most infiltration-excess overland flow reaching streams in this catchment is produced from within several metres of the stream edge, and that the transported sediment is generated by interrill processes. This conclusion contrasts strongly with common perceptions/observations of widespread overland flow dominated erosion processes following fire, and perhaps indicates substantial functional differences in erosion processes between different Eucalyptus forests. Measurements in unburnt areas, intended for comparative purposes only, provided unexpected insights in their own right. These areas showed a large natural seasonal oscillation in water repellence (non-repellent in winter to strongly repellent in summer) and runoff generation from rainfall simulation (1% in winter to 42% in summer) as the soil gravimetric water content crossed a hysteresis-variable threshold value. These large seasonal fluctuations in runoff generation indicate that proportional increases in erosion over the unburnt condition are highly temporally variable; 18-fold immediately after the fire, increasing to 2240-fold at six months, and then reducing to only 2.5-fold by 12 months after the fire, perhaps explaining some of the diversity of reported results in the literature. The above results suggest erosion risk is greatest in the first winter following the fire. Both the runoff and erosion results show that conventional erosion modelling approaches in this catchment will fail (both in burnt and unburnt forests) because the dominant driving processes and soil properties are not represented.