Improving Drought Monitoring and Predictions Using Physically Based Evaporative Demand Estimates

Existing drought monitors rely heavily on precipitation (Prcp) and temperature (T) data to derive moisture fluxes at the surface, often using estimates of evaporative demand (Eo) based only on T to derive actual evapotranspiration (ET) from land surface models (LSMs). An example of this is the popular Palmer Drought Severity Index (PDSI). In the analysis of drought trends and dynamics, however, the choice of Eo-driver for LSMs is crucial: it significantly affects both the direction and magnitude of trends in estimated ET and soil moisture, particularly in energy-limited areas (in water-limited areas, ET and soil moisture trends are driven by Prcp trends). All else equal, in the long term, T-based Eo measures result in declining ET estimates (i.e., drying) as T rises, whereas using more appropriate, physically based Eo estimates will more accurately reflect observations of both wetting and drying under warming. With regard to the short-term variabilities more appropriate to monitoring ongoing droughts, we contend that, given that various requirements are met, using an appropriate Eo driver (i) as a drought metric in itself, (ii) to drive drought monitors' LSMs, and (iii) in combination with short-term Eo forecasts will enhance characterization of the evaporative dynamics of ongoing drought and permit more accurate predictions of drought development. The requirements of an appropriate Eo estimate are as follows: that at operationally appropriate time and space scales Eo is diagnostic of ET (i.e., ET and Eo co-vary in a complementary fashion); that the Eo formulation and driving data produce good estimates of Eo (i.e., the model is physically based in that it combines radiative and advective drivers, and produces Eo estimates that are accurate and unbiased with respect to observations from drivers that are available with limited latency on a daily basis) and at operational spatio-temporal resolutions; and that Eo can be forecast at operational time and space scales. We present our progress towards the implementation of such an Eo dataset. We test the first contention by deriving Eo from the Penman-Monteith formulation for reference crop evapotranspiration (ETrc). We generate a 30+ year reanalysis of daily CONUS-wide Eo driven by a combined North American Land Data Assimilation System/Real-Time Mesoscale Analysis dataset. From this Eo reanalysis, we have implemented a simple drought metric-the Evaporative Demand Drought Index (EDDI)-based on standardized accumulated ETrc anomalies. The basic premise is that ETrc increases under dry conditions at the land surface-atmosphere interface, due to the complementarity of Eo with regional-scale ET and a mixture of decreases in vapor pressure and increases in T, wind speed, and solar radiation. Extensive periods of positive accumulated ETrc anomalies (i.e., positive EDDI) then register as periods of drought; negative as wet periods. We compare our results to established drought measures across CONUS from 1979 to the present. In reanalyses of drought over CONUS the EDDI metric shows great promise, and the system is easily extensible to a global scale by driving the ETrc formulation with Global Data Assimilation System data.