Relating flood characteristics to vegetation dynamics in the Geruwa floodplain in the Terai Arc Landscape, Nepal

Floods significantly influence land cover such as vegetation in a floodplain. Land cover changes with change in flood characteristics such as frequency, duration, extent, flow depth, and flow velocity. Riparian vegetation type, distribution, and patterns provide habitat for many species. The Geruwa River is a bifurcate of the Karnali River emerging out of the Himalayan foothills in the Terai Arc Landscape. The Geruwa River is part of the Bardiya National Park in western Nepal, which is one of the most important and biodiverse nature reserves in Nepal. Its channel-floodplain system provides, for example, habitat for endangered Gangetic dolphins (Platanista gangetica), bare riverbanks for crocodiles, and riverine grassland-forest mosaics for elephants and tigers. Since the intense monsoon season in 2009, the Geruwa discharge has reduced gradually, initiated by the deposition of coarse sediment over the upstream end of the branch during this monsoon season. The Geruwa branch, which used to be the dominant branch before 2009, now receives only about 20 percent of the Karnali river discharge during the peak flows and about 5 percent during low flows.

Understanding how changing flood characteristics influence the dynamics of floodplain vegetation can help us better manage this important wildlife habitat. Our objective is to quantify the relationship between flood characteristics and vegetation cover in the Geruwa floodplain. We fit a multinomial logistic regression model, where the response is land cover (Forest, Grassland, Agriculture, Bare sediment, and Water) class derived from remote sensing and predictors are flood characteristics derived from hydrodynamic simulations. Model performance is evaluated using spatially stratified cross-validation to reduce bias from spatial autocorrelation. The fitted model will be used to extrapolate vegetation cover under future flood regimes influenced by climate change or anthropogenic activity. To support this analysis, we compiled and constrained flood discharge and duration over the last 63 years. We have developed a two-dimensional flow model to simulate similar floods to derive spatially distributed flood characteristics. We extract vegetation cover over the last 15 years using remote sensing (after channel switch). Preliminary analysis indicates that the decreased flood discharge in Geruwa is followed by an increase in vegetation cover (grassland and forest) in general in the Geruwa floodplain after 2009. Reduced floods in future may increase the succession of grasslands into forests.