A SMAP-Based Drought Monitoring Product

A sustained lack of soil moisture over time results in failure of crops, death of livestock, and ultimately death of humans. Our food security relies on developing measures and products that monitor agricultural drought to help reduce societal vulnerability to drought, stressing risk management rather than crisis management. Soil moisture information over time and space provides critical insight for agricultural management, including both water availability for crops and moisture conditions that affect management practices such as fertilizer, pesticide applications, and their impact as non-point pollution runoff. Since April of 2015, NASA’s Soil Moisture Active Passive (SMAP) mission has retrieved soil moisture using L-band passive radiometric measurements at a 8 day repeat orbit with a swath of 1000 km that maps the Earth in 2-3 days depending on locations. Of particular interest to SMAP-based agricultural applications is a monitoring product that assesses the SMAP soil moisture in terms of probability percentiles for dry (drought) or wet (pluvial) conditions. SMAP observations do result in retrievals that are spatially and temporally discontinuous. Additionally, the short SMAP record length provides a statistical challenge in estimating a drought index and thus drought risk evaluations.  

We describe a SMAP drought index for the CONUS region based on near-surface soil moisture percentiles. Because the length of the SMAP data record is limited, we use a Bayesian conditional probability approach to extend the SMAP record back to 1979 based on simulated soil moisture of the same period from the Variable Infiltration Capacity (VIC) Land Surface Model (LSM), simulated by Princeton University. This is feasible because the VIC top soil layer (10 cm) is  highly correlated with the SMAP 36 km passive microwave during 2015-2016, with more than half the CONUS grids having a cross-correlation greater than 0.6, and over 0.9 in many regions. Given the extended SMAP record, we construct an empirical probability distribution of near-surface soil moisture drought index showing severities similar to those used by the U.S. Drought Monitor (from D0-D4), for a specific SMAP observation. The analysis is done for each of the 8,150 SMAP grids covering the CONUS domain. Comparisons between the SMAP drought index and those from the VIC LSM, Washington University-UCLA multimodal Surface Water Monitor, and a similar product produced by the National Centers for Environmental Prediction are presented for selected recent drought events, with particular attention to persisting drought in the southern half of California. Issues such as seasonality, robustness of the fitting, regions of poor SMAP-VIC correlations, and extensions to the global domain will be discussed.