Dry me a river: regional and global drought in a warming climate

The American southwest is known for being hot and dry, and for the millions of residents flocking to desert metropolises like Phoenix and Las Vegas, this relatively stable, snow-free climate is one of its major appeals. For those managing its water resources, fighting wildfires, or growing crops, however, the threat of climate change to America's desert landscapes is a serious concern. It's no secret that global temperatures are on the rise, and snowbird states like Arizona have felt the impact (Fig. 1). For the brave few that enjoy Arizona's sauna-like summers, this may not sound so bad. But higher average temperatures and a warmer global climate has detrimental implications for water resources in the American west. Basic physics and climatology tell us to expect the following long-term hydrological impacts in a warming climate:

1) More heat means more evaporation. With all else constant, that means less water staying in soils and feeding rivers/aquifers.

Figure 1: Trend in mean annual air temperature for AZ.
Image from Climate Central's interactive database.

2) Warmer winters mean less snow at higher elevations. Since snowpack is the major source of runoff to rivers and aquifers, that means less water available to plants and humans alike.

3) Warmer oceans means greater storm energy, which results in more water falling in fewer events. Downpours in the southwest produce rapid runoff, which tends to infiltrate less effectively than steady rain/snow. 

4) Warming climate means an expansion and strengthening of the subtropical high pressure belt, which currently contributes to the semi-arid climate of the American southwest. A similar phenomenon during the Medieval Climate Anomaly contributed to the 'megadroughts' (Fig. 2) partly responsible for sudden collapse of several Native American populations, such as the Anasazi.
5) Finally, a more indirect link: global warming has caused a steady decline in Arctic sea-ice volume and extent since the 1970's. A reduction in Arctic sea-ice extent tends to weaken the polar vortex, which means stronger meridional (north–south) atmospheric circulation during winter. Counterintuitively, melting Arctic sea ice results in more frequent winter chills for much of the continental U.S. (as we saw in recent months), but it also means fewer Pacific storms reaching the American southwest (i.e. drier winters). Apart from the monsoon-driven regions, we can expect a significant reduction in effective moisture for the southwest (including California) from this factor alone.

Confirmation of these most basic predictions is available through multiple proxies of drought, which have been summarized in the recent Assessment of Climate Change in the Southwest United States. The authors conclude that despite relatively stable levels of annual precipitation, rising temperatures and reduced snowpack are characteristic of recent decades, of which the most recent (2001–2010) was one of the driest/hottest in over a century. Since ~1980, these trends became distinguishable from the natural background climate variation, according to Barnett et al. (2008), who attributed ~60% of drought trends to forcing by greenhouse gases. Drought reconstructions by Damberg and AghaKouchak (2013) confirm the trend toward drought in the American southwest, which MacDonald (2010) and Dai et al. (2011) note is consistent with the modeled response to global warming and may only get worse.

Figure 2: Last 1,200 years of drought in the southwest. Figure taken from the
Assessment of Climate Change in the Southwest United States.

While the American southwest is no stranger to extreme drought (Fig. 2), the possibility of returning to the climate of the Medieval Climate Anomaly is all but comforting to those managing its water resources. There will always be wet and dry years, but the increasingly robust forecast is that we can expect fewer wet to combat the dry.

Additional climate filters: PDO and ENSO

Figure 3: PDO index since 1950, from NOAA.

One cannot address the question of climate change in the southwest without considering the most prominent, natural thermostats lying just off the coast. Year-to-year climate of North and South America varies substantially due to oscillations in sea-surface temperature (SST) in the Pacific Ocean. This variability is due to the fact that Pacific SST is not geographically homogenous. In some years, it's colder in the east Pacific than in the west (La Niña), while in others, it's warmer in the east Pacific than in the west (El Niño). The Pacific Decadal Oscillation (PDO; Fig. 3) is described by a temperature gradient in the northern Pacific ocean from the middle of the sea to the California coast. The positive index reflects relatively warm waters off the California coast and a weakened high-pressure cell. Weakening of the high-pressure cell allows more storms to penetrate the continent from the cool waters of the northern Pacific. Notice that the PDO tended to be positive during much of the 80's and 90's (Fig. 3). This positive phase was accompanied by wetter conditions in the American west/southwest, refilling of dammed reservoirs, and rapid, optimistic growth in cities like Las Vegas/Phoenix.

The El Niño/Southern Oscillation (ENSO) is the PDO's slightly more chaotic cousin. Comparing figures 3 and 4, however, you can see that they are positively correlated. In fact, the positive trends in ENSO and PDO from 1950–1998 explain why most regions in the southwest have enjoyed slightly more precipitation during the same interval (McCabe et al., 2010), while the negative trends since 1998 partially explain the current 'megadrought'. When precipitation trends (McCabe et al., 2010) are plotted alongside temperature trends (i.e. a measure of potential evaporation and transpiration), it becomes apparent that recent global warming has exacerbated the drought and that future warming will impede drought recovery in decades to come.

Figure 4: ENSO variability since 1950, from the NOAA ESRL.

Given the high variability of naturally occurring climate oscillations like PDO and ENSO, it's not surprising that non-climatologists like Dr. Jay Wile are skeptical of global warming's impact on recent droughts. In my opinion, Dr. Wile has oversimplified the science of climate change and therefore expects a simple correlation between CO2, temperature, rainfall, and drought. But if global warming does affect drought in the southwestern US, that trend will be superimposed on the ENSO/PDO variability shown by figures 3 and 4, and researchers agree that is indeed the case. Dr. Wile cited McCabe et al. (2010) to suggest that the American southwest has not seen more drought, but that study only addressed trends in annual precipitation. Annual precipitation has changed little for the southwest since 1950, and in many parts has increased slightly, but the type of precipitation falling (snow vs. rain) and the timing of that precipitation has changed systematically in response to global warming (Barnett et al., 2008). When all factors are considered, therefore, a more dismal outlook emerges, so McCabe et al. (2010) themselves do not share Dr. Wile's skepticism of global warming or its impact on drought in the southwest.

Global trends in drought: is the Earth as a whole drying up?

In 2013, the IPCC reiterated their long-held assessment (p. 205–206) that in response to global warming, dry regions tend to get drier and wet regions tend to get wetter (see also Seager et al., 2010; Trenberth et al., 2014). There are no shortage of studies exploring the impact of global warming on drought, so I am only addressing a fraction of what might be said. Nonetheless, models (i.e. basic physics and climatology) generally predict that with time, we can expect a larger percentage of land area on Earth to be characterized by drought conditions. This seems counterintuitive, because a warming atmosphere simultaneously results in a more intense hydrological cycle (warmer air holds more moisture, for example). However, most of that extra moisture is confined to the equatorial region, high-latitude continental regions (including the Arctic and Siberia), and certain coastal regions, due to the way moisture is transported around the globe. Most models do not predict this change to be catastrophic or sudden, but suggest rather that it will take decades or even a century for average drought conditions to affect some 5–20% more of global land area (e.g. Dai et al., 2013 is a high-end estimate; see Fig. 5). For some regions, it is predicted that drought will increase from reduced precipitation, whereas in other regions, drought will increase primarily from enhanced evaporation as air temperatures continue to rise, or from enhanced SST, which affects pressure patterns and atmospheric circulation.

Figure 5: Slide 36 from a presentation by Aigou Dai (PDF here), illustrating modeled
drought response to enhanced SST under global warming scenarios.

Has global drought increased over the past century in response to global warming? It may sound like this is an easy prediction to test, but it's not. The key factor lies in how 'drought' should be quantified. One classical formulation, called the Palmer Drought Severity Index (PDSI) calculates the effective water balance by comparing total precipitation to the solar energy available to evaporate that water. Several researchers have criticized the use of PDSI to quantify drought solely because of its simplicity (i.e. it doesn't take into account actual soil moisture, wind velocity, and other factors that affect evaporation). In other words, although PDSI can track drought from a basic climate perspective, it may overestimate actual evaporation and thus real risk of drought to agriculture.

To correct this bias, researchers like Hao and AghaKouchak (2013) have devised a more comprehensive metric for drought to forecast crop shortfalls. This metric was applied to the globe by Hao et al. (2014), who captured historical droughts over the past 30 years. No significant trends emerged from the reconstruction, which may suggest that global drought has not increased with global temperature (of course, neither has it decreased). I would caution against over-interpreting historical trends from their plot, however, since it was not their intent to answer how global drought has responded to global warming trends. In fact, their data cover a very short period of time (1982–2012), during which a shift from a predominantly El Niño to La Niña conditions explains why global drought appears slightly more extensive at the beginning of their record. Furthermore, the driest regions on Earth were omitted from their reconstruction, due to their high sensitivity to changes in precipitation. Finally, given their emphasis on soil moisture in the drought metric, it is likely that extensive crop irrigation and management will artificially mitigate the level of drought severity in some major watersheds. The areal extent of irrigation (which affects semi-arid regions already susceptible to drought) has increased steadily from 1982 to present, and since humans irrigate and rotate crops in response to short-term climate variability, human activity on land can blur long-term trends according to this measurement of drought.

On the other hand, Sheffield et al. (2013) also criticized the simplicity of classic PDSI formulations and argued that global drought has changed little in response to global temperature rise. To accomplish this, Sheffield et al. (2013) devised a better formulation of PDSI, which incorporated more variables to estimate real evaporation, and reconstructed global drought trends for the past 60 years (Fig. 6).

Figure 6: Global drought over the past 60 years, from Sheffield et al. (2013),
using the classic PDSI metric (blue) and a more rigorous calculation (red).

The blue lines in Figure 6 are similar to the reconstruction by Dai (2011), who demonstrated that PDSI has decreased substantially (i.e. more prevalent drought) over the past 50 years on a global scale. This conclusion by Dai (2011) cannot easily be ignored, since is corroborated by evidence of enhanced evapotranspiration (Wang et al., 2010) and reduced streamflow (Dai et al., 2009) during the modern warming period. The latter is a more direct measurement of effective moisture over land, since it relies less on computer models to reconstruct trends (keep in mind, all reconstructions of global drought are modeled interpolations of historical data, and so all have intrinsic uncertainties). One important proxy not considered by any of these models is the shift from winter snow to winter rain at high elevations, and the shift toward earlier dates of snowmelt (e.g. Barnett et al., 2008). As less precipitation falls as snow and that snow begins to melt earlier in the year, available water resources will diminish, even if PDSI and soil moisture do not change significantly. Taking all of these factors into consideration, we can say with moderately high confidence that over the past 50-60 years, drought has increased on a global scale. It's only a question of how much.

Despite the conflict with Dai (2011), Sheffield et al. (2013) did not entirely negate their conclusions regarding global drought. It is apparent from Figure 6 that the last three decades were characterized by more extensive drought than from 1950–1977, both in terms of PDSI and total land area in drought conditions. When I plotted the revised data myself (the red line in Fig. 6a), I obtained a downward (drying) trend, which is statistically significant at 99.4% confidence. Therefore, it is important to note that the title of Sheffield et al. (2013)—"Little change in global drought over the past 60 years"—does not mean "no change in global drought".

The most important revision by Sheffield et al. (2013) is that oversimplified calculations of PDSI resulted in an overestimate of drought response to global warming and, therefore, an overestimate of future drought risk from climate change. One key difference (overlooked by Dai, 2011) is the cooling effect of evaporation on air temperatures over land. Like a giant air conditioner, enhanced evaporation works as a negative feedback that slightly mitigates rising air temperatures. Amid the academic controversy, however, these authors worked together on a more recent synthesis (Trenberth et al., 2014) that affirmed drought has and will increase in response to global warming, though the response is more complex and less uniform than previously stated (Fig. 7).

Figure 7: Modeled reconstructions of global drought from historical data,
according to Trenberth et al (2014). Note the downward (drying) trend.

Three out of four global drought reconstructions by Trenberth et al. (2014) indicate that drought has become more prevalent over the land surface since global temperatures rose significantly in response to greenhouse-gas forcing. This conclusion is similar to Damberg and AghaKouchak (2013), who also noted reconstructed a trend toward enhanced drought (though their trend is statistically significant only for the southern hemisphere and individual watersheds on land). The initial predictions by the IPCC assessments, as well as Dai (2011) have been refined, but not negated.

Conclusion

Despite existing uncertainties in quantifying the response of drought to global warming, whether on a regional or global scale, the general consensus is that drought has already become more extensive and severe, and that future warming will only exacerbate the current situation. This relationship is more dreary from a human perspective, since growing population will only increase water-resource and agricultural demands, and no studies imply a wetting trend. Although we have much to learn about how global warming impacts the hydrological cycle, Dr. Wile's skepticism is premature and misguided, at best. Perhaps the best indication of this lies with the fact that none of the researchers cited (Dai,  Trenberth, Sheffield, Damberg, Hao, AghaKouchak, McCabe, Barnett, or their co-authors) share Dr. Wile's skepticism regarding global warming.

And neither do I, but I'm just a lowly paleoclimatologist. ;)


Note: This controversy has reached the Senate floor in Washington D.C. as well. Go here for a interesting discussion on the misperceptions of what climate scientists are/are not saying regarding global and regional drought.