Translating Atmospheric Predictability to Hydrology in Flood Forecasts for South Asia

South Asia is a flashpoint for natural disasters with profound societal impacts for the region and globally. Half the world’s population depends on the region’s great rivers, the Indus, Ganges, and Brahmaputra. The frequent occurrence of floods, combined with large and rapidly growing populations with high levels of poverty, make South Asia highly susceptible to humanitarian disasters. The 2007 Brahmaputra floods affecting India and Bangladesh, the 2008 avulsion of the Kosi River in India, and the 2010 flooding of the Indus River in Pakistan exemplify disasters on scales almost inconceivable elsewhere. The challenges of mitigating such devastating disasters are exacerbated by the lack of rain/gauge measuring stations and transboundary data sharing, and perhaps most importantly, by the limited application of ensemble numerical weather predictions applied to the regional flood forecasting problem.

The availability and application of weather forecasts from ensemble prediction systems (EPS) have transformed river forecasting capability over the last decade. In this talk, we focus on how atmospheric predictability as measured by an EPS, is transformed into hydrometeorological predictability for South Asia, while also discussing recent developments in remote sensing data that are greatly increasing the potential of providing skillful long lead river flood forecasts for the region. We will also frame the discussion of atmospheric-to-hydrologic predictability in the context of one particular potential benefit of an EPS, which is its capacity to forecast its own forecast error through the ensemble spread-error relationship. In practice, an EPS is often quite limited in its ability to represent the variable expectation of forecast error through the variable dispersion of the ensemble. We will examine the ensemble skill-spread relationship of a precipitation forecast ensemble constructed from the TIGGE (THORPEX Interactive Grand Global Ensemble) dataset of global forecasts, as it is transformed into river flow predictability for the region, and discussing one methodology that utilizes the ensemble’s ability to self-diagnose forecast instability to produce forecasts with informative skill-spread relationships.