Validating modeled soil moisture with in-situ data for agricultural drought monitoring in West Africa

The declaration of famine in Somalia on July 21, 2011 highlights the need for regional hydroclimate analysis at a scale that is relevant for agropastoral drought monitoring. A particularly critical and robust component of such a drought monitoring system is a land surface model (LSM). We are currently enhancing the Famine Early Warning Systems Network (FEWS NET) monitoring activities by configuring a custom instance of NASA's Land Information System (LIS) called the FEWS NET Land Data Assimilation System (FLDAS). Using the LIS Noah LSM, in-situ measurements, and remotely sensed data, we focus on the following question: How can Noah be best parameterized to accurately simulate hydroclimate variables associated with crop performance? Parameter value testing and validation is done by comparing modeled soil moisture against fortuitously available in-situ soil moisture observations in the West Africa. Direct testing and application of the FLDAS over African agropastoral locations is subject to some issues: [1] In many regions that are vulnerable to food insecurity ground based measurements of precipitation, evapotranspiration and soil moisture are sparse or non-existent, [2] standard landcover classes (e.g., the University of Maryland 5 km dataset), do not include representations of specific agricultural crops with relevant parameter values, and phenologies representing their growth stages from the planting date and [3] physically based land surface models and remote sensing rain data might still need to be calibrated or bias-corrected for the regions of interest. This research aims to address these issues by focusing on sites in the West African countries of Mali, Niger, and Benin where in-situ rainfall and soil moisture measurements are available from the African Monsoon Multidisciplinary Analysis (AMMA). Preliminary results from model experiments over Southern Malawi, validated with Normalized Difference Vegetation Index (NDVI) and maize yield data, show that the ability to detect a drought signal in modeled soil moisture and actual evapotranspiration was sensitive to parameters like minimum stomatal resistance, green vegetation fraction, and minimum threshold for transpiration stress. In addition to improving our understanding and representation of the land surface physics in agropastoral drought, this study moves us closer to confidently validating LSM estimates with remotely sensed data (e.g. MODIS NDVI), essential in regions that lack ground based measurements. Ultimately, these improved information products serve to better inform decision makers about seasonal food production and anticipate the need for relief, as well as guide climate change adaptation strategies, potentially saving millions of lives.