Prediction of the discharge in Pech River for the flood warning system, based on the time series of snow cover area and meteorological data

抄録

Floods in the arid areas provide fertile soil and water resources, enabling agricultural production. Based on its productivity, towns with a large population have developed along the riverbanks. On the other hand, large flood beyond the expectation will cause enormous human and economic damage. The Pech River, our study site, is a tributary of the Kunar River, which runs through eastern Afghanistan. The catchment area of the river is populated by 35.5% of the population and has 21% of the agricultural land of Kunar province.

The main source of the river discharge is melted snow in the upper mountainous areas, and the maximum discharge that appears in the following spring changes depending on the amount of snowfall in the previous winter. In the year of heavy snowfall in the upper mountainous areas, the residential areas and the agricultural lands along the river experienced severe inundation.

Therefore, it is necessary to predict the peak discharge in spring in advance and provide appropriate flood warnings so that the local residents can take evacuation and damage mitigation measures. In this study, we tried to develop a method for predicting the peak discharge based on the fluctuation of snow cover area evaluated with satellite images, temperature, and rainfall of the upper reaches of the Pech River.

First, the boundary of the river catchment was identified based on the DEM of 30 m mesh acquired by Advanced Land Observing Satellite (ALOS). Next, the Snow cover distribution data between 2008 and 2018 at weekly intervals was downloaded from the National Snow & Ice Data Center (NSIDC) database, and the catchment area was extracted. Daily temperature and precipitation data at the station nearby the catchment stored in the database of National Oceanic and Atmospheric Administration (NOAA) were expected as the parameter explaining the snow melting process in spring.

We tried to reproduce the discharge of the Pech River using these explanatory variables related to the process of snowfall and snowmelt with some statistical and stochastic methods. The multi-regression analysis and Neural Network were examined to reproduce the discharge fluctuation of Pech River.