No abstract

The Southern Terai plain area of Nepal is exposed to recurring floods. The floods, landslides and avalanches in Nepal cause the loss of lives of about 300 people and damage to properties worth about 626 million NPR annually. Consequently, the overall development of the country has been adversely affected. The flood risk could be significantly reduced by developing effective operational flood early warning systems. Hence, a study has been conducted to assess flood danger levels and determine the threshold runoff at forecasting stations of six major rivers of Nepal for the purpose of developing threshold-stage based operational flood early warning system. Digital elevation model data from SRTM and ASTER supplemented with measured cross-section data and HEC-RAS model was used for multiple profile analysis and inundation mapping. Different inundation scenarios were generated for a range of flood discharge at upstream boundary and flood threshold levels or runoffs have been identified for each river, thus providing the basis for developing threshold-stage based flood early warning system in these rivers.

The gravel-sand transition (GST) is commonly observed along rivers. It is characterized by an abrupt reduction in median grain size, from gravel- to sand-size sediment, and by a shift in sand transport mode from wash load–dominated to suspended bed material load. We documented changes in channel stability, suspended sediment concentration, flux, and grain size across the GST of the Karnali River, Nepal. Upstream of the GST, gravel-bed channels are stable over hundred- to thousand-year time scales. Downstream, floodplain sediment is reworked by lateral bank erosion, particularly during monsoon discharges. Suspended sediment concentration, grain size, and flux reveal counterintuitive increases downstream of the GST. The results demonstrate a dramatic change in channel dynamics across the GST, from relatively fixed, steep gravel-bed rivers with infrequent avulsion to lower-gradient, relatively mobile sand-bed channels. The increase in sediment concentration and near-bed suspended grain size may be caused by enhanced channel mobility, which facilitates exchange between bed and bank material. These results bring new constraints on channel stability at mountain fronts and indicate that temporally and spatially limited sediment flux measurements downstream of GSTs are more indicative of flow stage and floodplain recycling than of continental-scale sediment flux and denudation rate estimates.

Two important applications of rainfall-runoff models are forecasting and simulation. At present, rainfall-runoff models based on artificial intelligence methods are built basically for short-term forecasting purposes and these models are not very effective for simulation purposes. This study explores the applicability and effectiveness of adaptive neuro-fuzzy-system-based rainfall-runoff models for both forecasting and simulation. For this purpose, an adaptive neuro-fuzzy system with autoregressive exogenous input (ARX) structure is proposed and an application is presented for the modelling of rainfall-runoff processes in the Sieve basin in Italy.

In Nepal, as the spatial distribution of rain gauges is not sufficient to provide detailed perspective on the highly varied spatial nature of rainfall, satellite-based rainfall estimates provides the opportunity for timely estimation. This paper presents the flood prediction of Narayani Basin at the Devghat hydrometric station (32 000 km 2 ) using bias-adjusted satellite rainfall estimates and the Geospatial Stream Flow Model (GeoSFM), a spatially distributed, physically based hydrologic model. The GeoSFM with gridded gauge observed rainfall inputs using kriging interpolation from 2003 was used for calibration and 2004 for validation to simulate stream flow with both having a Nash Sutcliff Efficiency of above 0.7. With the National Oceanic and Atmospheric Administration Climate Prediction Centre's rainfall estimates (CPC_RFE2.0), using the same calibrated parameters, for 2003 the model performance deteriorated but improved after recalibration with CPC_RFE2.0 indicating the need to recalibrate the model with satellite-based rainfall estimates. Adjusting the CPC_RFE2.0 by a seasonal, monthly and 7-day moving average ratio, improvement in model performance was achieved. Furthermore, a new gauge-satellite merged rainfall estimates obtained from ingestion of local rain gauge data resulted in significant improvement in flood predictability. The results indicate the applicability of satellite-based rainfall estimates in flood prediction with appropriate bias correction.