Dominant drivers of runoff in a slash-and-burn affected catchment in upland Eastern Madagascar

In many tropical areas slash-and-burn agriculture is an important driver of forest loss. In areas where slash-and-burn agriculture has been practiced for decades, land cover is typically a mosaic of patches of remnant forest, fields under active cultivation, fallows in various stages of regrowth (ranging from young shrub to semi-mature), and degraded fire-climax grasslands. Although runoff generation mechanisms are expected to be different for these different patches, little quantitative information is available in this regard, particularly at the catchment scale and over longer time-scales (i.e., multiple slash-and-burn cycles).

We re-instrumented a 31 ha catchment in upland Eastern Madagascar, where slash-and-burn agriculture has been practiced for more than 70 years in 2015; it had been monitored between 1963 and 1972 as well¹. We measured streamflow at two locations and overland flow and soil moisture for four hillside plots (0.05 – 1.93 ha): one plot under repeatedly coppiced and burned Eucalyptus and three plots under young shrub and tree fallows. One of the plots underwent rudimentary terracing in the past. We analysed the rainfall-runoff dynamics for 50 rainfall events (median 12 mm, maximum 71 mm).

For 60% of the events, the stormflow coefficient (minimum contributing area) was <3%, which is the proportion of valley-bottom wetlands and rice paddies in the catchment. Stable isotope sampling for five storm runoff events indicate a maximum total event-water contribution of 16%. However, instantaneous event-water contributions were as high as 66%. The hillside plot runoff response was dominated by saturation-excess overland flow and showed strong threshold behaviour in terms of the antecedent soil moisture storage in the upper 30 cm of the soil plus the event total rainfall amount (ASI + P). Average threshold values for overland flow occurrence ranged from 87 mm for the coppiced Eucalyptus to 137 mm for the young fallow plots (regardless of terrace presence). Stormflow also increased after an ASI+P-threshold was exceeded (100 mm based on the soil moisture sensors for the Eucalyptus plot and 150 mm for the sensors at the tree fallow plots).

These results indicate an increased hydrological connectivity between hillslopes and valley bottom under wetter conditions and that stormflow in the study catchment is strongly affected by variations in seasonal rainfall. The results will be used to validate a hydrological model to determine the net effect of concurrent changes in soil infiltrability and vegetation water use associated with forest loss and recovery on stormflow totals and the seasonal flow regime.

¹Bailly, C., de Coignac, G.B., Malvos, C., Ningre, J.M., and Sarrailh, J.M. (1974). Étude de l'influence du couvert naturel et de ses modifications á Madagascar. Expérimentations en bassins versants élémentaires. Cahiers Scientifiques, 4. Centre Scientifique Forestier Tropical, Nogent-sur-Marne, France, 114 pp.