E0, ET, and Drought at Climate Scales
In this section, we do not attempt to predict how drought will evolve under climate change (see Chapter 4). Rather we summarize the state of the science with regard to the roles of ET and E0 in climate-scale drought analyses and forcing.
Under anthropogenic climate change heat is being added to the climate engine, and our general expectations are for increased E0 due to increased T and a resulting increase in ET in regions where moisture is available. In general, climate model projections indicate that wet areas will get wetter and dry areas drier, in line with latitudinal increases in subtropical dry zones (the descending arm of the Hadley cell); also projected are poleward shifts in the main mid-latitudinal storm tracks and changing Prcp seasonality (Trenberth et al. 2014).
The "Warming Is Drying" Message
The scientific community is not yet in agreement on the effects of these expected dynamics on drought. The main current issues with the estimation and use of ET and E0 in drought analyses at climate scales are exemplified in a recent arc of papers. Here, we start with one of the most influential papers on the topic: that of Dai et al. (2004), who ran the PDSI model across the globe at a 2.5-degree resolution from 1870 to 2002, forced by Prcp and T. Briefly, they reported a doubling of very dry areas (PDSI < -3.0) since the 1970s, and a decrease in very wet areas (PDSI > 3.0) during the 1980s, with regions at both extremes combined almost doubling in area since 1972. They concluded that the risk of drought had increased with global warming, and that this was likely the result of a higher water-holding capacity of the air due to warming, which raised E0 and hence ET. This "warming is drying" message resonated deeply within the community, particularly within the Intergovernmental Panel on Climate Change's (IPCC) Fourth Assessment Report (AR4) (Meehl et al. 2007), though it has been revisited by the IPCC Special Report on Extremes (Seneviratne et al. 2012).
However, the simple E0-parameterization at the basis of the evaporative driver of the PDSI bucket hydrology model used in Dai et al. (2004)—and operationally to this date—suffers from two issues: it does not incorporate changes in the other, non-T drivers of E0 (Rn, Uz, and humidity), and it cannot reflect changes in how plants will uptake carbon—determining the rate of the transpiration component of ET—in a CO2-enriched atmosphere.
To highlight the former issue, Hobbins et al. (2008) compared the effects of a T-based E0 to observed E0 (from Epan) on trends of water balance components of the PDSI at 35 stations across Australia and New Zealand from 1975 to 2004, finding that T-based E0 increased over the period almost everywhere (in line with T increases). Indeed, particularly in energy-limited regions, these trends were opposite to declines in observed E0, with the trends in E0 found to result, in the mean, from trends in Uz rather than in T (Roderick et al. 2007). When applied in the PDSI, the contrasting T-based and observed E0 trends resulted in SM trends that bore no relation to each other (Hobbins et al. 2008). Similarly, comparing global PDSI changes forced by E0 from the T-based
Thornthwaite (1948) equation traditionally used in the PDSI to those forced by the physically based Penman-Monteith ET0, Sheffield et al. (2012) found that, contrary to previous warnings (e.g., the IPCC AR4; Meehl et al. 2007), there had been little change in long-term (1950-2008) global average drought trends: the physically based E0-forced PDSI showed drying over only 58 percent of the globe, whereas the T-based E0 PDSI showed drying almost uniformly across the globe. Again, in energy-limited regions, differences in E0 trend directions translated into differences in wetting versus drying between the two E0 types (in water-limited regions, PDSI trends are driven by trends in Prcp, not in E0). These studies clearly demonstrate the importance of proper selection of E0 parameterization for climate-scale drought analyses.
The reliance on T as a proxy for E0 also has deleterious effects on the reconstruction of paleoclimate drought records (Sheffield et al. 2012). Generally, tree-ring data are scaled to T-based PDSI in their overlapping period. However, not only does this relationship assume that tree growth relates to T alone, and not, crucially, to atmospheric CO2, but also it breaks down at high elevations and latitudes, particularly during the recent decades of rapid warming. These assumptions lead to an overestimation of past changes and so to an underestimation of recent changes. Clearly, then, T-based param- eterizations of E0 in the PDSI—which effectively take a state variable (T) as a proxy for a flux (ET)—may be of use in the short term, but not in long-term analyses of drought or drought-relevant fluxes (e.g., ET, SM) of the future (Trenberth et al. 2014) or the past (Sheffield et al. 2012).
This is not to say that the sole source of the "warming is drying" message is misconceptions due to the use of T-based E0 parameterizations. Dai (2013) attempted to reconcile patterns of observed and modeled (from global climate models, GCMs) drying trends in the last part of the twentieth century using the self-calibrated PDSI forced by Penman-Monteith Ep (a fully physical measure). He concluded that while GCMs were able to capture ENSO's influence on drought and its recent trends, the differences between modeled and observed aridity changes result from natural SST variations not captured by the GCMs, and that more severe and widespread droughts should be expected over many land areas during the coming decades, due to either decreased Prcp or increased ET. This study implies that regional patterns important to infrastructure planning decisions to mitigate drought vulnerability are not being captured by GCMs. Further, Cook et al. (2015) used PDSI forced with Penman- Monteith Ep and Prcp from 17 GCMs running Representative Concentration Pathways (RCPs) 4.5 and 8.5 in June through August 2050-2099 and gave dire warnings of drought worse than those in the medieval climate anomaly (11001300) over the US Central Plains and Southwest—concluding that "megadroughts," or multidecadal droughts, will become more likely in both regions.
The question as to whether and where the future will become drier or wetter is becoming more nuanced—the IPCC Special Report on Extremes (Seneviratne et al. 2012) notes probable overestimation of drought and an overreliance on PDSI-based results—and its resolution will require more regional analyses.