Download the paper:

• Via DOI resolver
• From a password-protected webpage (send email to me at "rolf.reichle (at) nasa.gov")

Abstract:
The two main contributors to streamflow predictability at subseasonal to seasonal time scales in tropical regions are: (i) the predictability of meteorologic (particularly precipitation) anomalies, and (ii) the land surface soil moisture state at the start of the forecast period. Meteorological predictions at subseasonal time scale are usually fraught with error and may not be dependable. The accurate initialization of soil moisture, as obtained through real-time land data analysis, may provide skill in subseasonal to seasonal streamflow prediction, even when the prediction skill for rainfall is small.

A series of experiments using the Catchment Land Surface Model (CLSM) is performed to characterize the contribution of accurate soil moisture initialization to the skill of streamflow prediction in Sri Lanka at timescales up to 2 months. We find that at the monthly timescale, accurate soil moisture initialization provides between 10% and 60% of the total runoff prediction skill that could be obtained under a perfect prediction of meteorological forcing. Some contributions to streamflow forecast skill are also found for the second month of forecast.