E0, ET, and Their Physical Relationships to Drought
The relationship between E0 and ET is particularly important in terms of monitoring and forecasting drought. Separately, each estimate provides a unique contribution to the understanding of drought development; when used together, they provide a particular characterization of the role of the coupling between the atmosphere and land surface. From a theoretical standpoint, increased E0 precedes an observable land surface response (e.g., a decrease in SM), with increased E0 leading to stronger coupling between E0 and the land surface. While anomalies in E0 can be considered a leading indicator in drought development (potential for stress), an observation of the onset of vegetation stress, through an estimation of ET (actual stress), is still needed to properly characterize the role of the coupling between atmospheric demand and evolution of the land-surface response.
Evaporative Demand, E0
As noted earlier, E0 is defined as the maximal rate of ET that would occur if moisture availability at the evaporating surface were not limiting. This rate is constrained only by the drying power of the air, the energy available for evaporation, and the ability to bear evaporated moisture away. E0 is an umbrella term for a measure that takes on various conceptions that differ in their surface assumptions and/or measurement. E0 may be estimated either as a physical observation of Epan (generally daily from US Class-A evaporation pans) or as ET0 (ET from a well-watered, highly specified reference crop) at weighing lysimeters. More commonly, it is synthesized as either Ep or ET0 from observed radiative and meteorological drivers, with the prime difference between these two rates being their surface assumptions: Ep is more loosely defined as the ET rate that would pertain given ample moisture supply at the surface; ET0 is the ET rate given a well-watered and otherwise specifically defined, ideal surface, including a specific crop cover.
Estimators of these fluxes range in data requirements, adherence to physics and therefore in quality. Physically based approaches synthesize the effects of all physical drivers shown in Equation 11.2: these are Penman (1948)-based approaches, such as Penman-Monteith (Montieth 1965) or PenPan (Rotstayn et al. 2006). However, the development of the Ep concept (Penman 1948) arose when the stringent data and computational requirements of fully physical parameterizations (Equation 11.2) precluded their estimation on useful scales. Instead, simplified approaches to E0 developed at around the same time became popular, such as those based on T and Rn (e.g., the Priestley- Taylor equation for partial equilibrium evaporation, or evaporation from extensive wet surfaces; Priestley and Taylor 1972). Finally, many approaches are based solely on T, the most well-known being the Thornthwaite (1948) equation, which was developed as a tool to derive climate classifications and so involved mean conditions, not the deviations from the mean necessary for drought analysis. Other popular T-based methods include the Hamon (1961), Blaney-Criddle (Blaney and Criddle 1962), and Hargreaves-Samani (Hargreaves and Samani 1985) equations.
E0 parameterizations with physical underpinnings are crucial in drought analyses. For example, a rigorous decomposition of daily E0 variability (using Penman-Monteith ET0, widely considered the best estimator of E0 when all drivers are available and for nonsecular timescales) into its contributions from each of the drivers demonstrated that the drivers of daily variability—the most appropriate timescale for drought analyses—vary across the continental United States (CONUS) and with seasons (Hobbins 2016). Contrary to the basic assumptions of T-based approaches to E0, T is not the dominant driver for many regions across CONUS and many seasons: in summer, Uz dominates variability in the Desert Southwest and Rd in the Southeast; in winter, q dominates along the Eastern Seaboard. Similar conclusions were drawn by Roderick et al. (2007) in their decomposition of E0 trends across Australia: Uz
was found to dominate the mean trend at stations, with T playing little role. These studies underline the dangers of using T-based E0 in drought analyses, both in the short-term variability central to drought dynamics, and at secular and climate timescales (see Section 11.4).
The eradication of poor parameterizations of E0 must overcome significant paradigmatic inertia within the operational drought-monitoring community. This is demonstrated by the fact that new parameterizations of T-based E0 were being developed and/or installed as measure of E0 in new drought metrics (e.g., the PDSI as originally conceived by Palmer , or the original SPEI [Vicente-Serrano et al. 2010]) even as our understanding of the dynamics of both E0 and drought deepened, and long after data to support estimation of fully physical E0 became available. It is this inertia we seek to overcome here.