Evaluation of Flood Risk in Afghanistan using Retrospective Land Surface and Routing Models via the Land Information System

Using the Land Information System (Kumar et al 2008), we employ daily historical flow data from 1959-1980 in forty river reaches in Afghanistan in order to select and calibrate appropriate land surface and routing models for this region. We also assess the effects of using two different sets of meteorological forcing data (GDAS and Princeton-GPCP) on modeled river flows for the years 2006-2008. Finally, we aim to evaluate the extent to which current flood outlooks might be improved by using near-real time retrospective models, addressing recent concerns on the part of the Afghan government in the Panjshir district. We will use satellite observations of flooded areas from the Dartmouth Flood Observatory to select flow metrics that convey flood potential for several locations and discuss the nature and magnitude of uncertainties surrounding this analysis. Results from previous work in snowmelt-dominated watersheds suggest that even when the Noah v3.3 land surface model has only a neutral Nash-Sutcliffe efficiency with routing calculated via HyMAP (Getirana et al 2012), it can significantly improve overall baseflow representation when used in ensemble with the Catchment Land Surface Model Fortuna 2.5. These techniques could be used to provide flood outlooks for humanitarian and agricultural purposes in data-sparse regions of the world subject to seasonal, snowmelt-induced floods.