Trenberth slams new Nature article on drought: “The conclusions of the paper are likely wrong.”
A flawed new article in Nature has a title that sums up its controversial conclusion, “Little change in global drought over the past 60 years.”
I generally judge an article at odds with the broad literature in two ways. How well does it cite and respond to the literature? What do the other leading experts in the field say? This new article comes up short in both areas.
Kevin Trenberth, former head of the climate analysis section of the National Center for Atmospheric Research, has sent me a strong critique which is printed below. NCAR’s Aiguo Dai also sent me a critique.
The new article simply ignores or dismisses a considerable amount of the drought literature and focuses instead on one narrow metric of soil moisture. But as I wrote last year in a Comment that reviewed much of the recent literature for Nature, “The Next Dust Bowl” (subs. req’d, full text here), climate change worsens droughts in three synergistic ways:
A basic prediction of climate science is that many parts of the world will experience longer and deeper droughts, thanks to the synergistic effects of drying, warming and the melting of snow and ice.
Precipitation patterns are expected to shift, expanding the dry subtropics. What precipitation there is will probably come in extreme deluges, resulting in runoff rather than drought alleviation. Warming causes greater evaporation and, once the ground is dry, the Sun’s energy goes into baking the soil, leading to a further increase in air temp- erature. That is why, for instance, so many temperature records were set for the United States in the 1930s Dust Bowl; and why, in 2011, drought-stricken Texas saw the hottest summer ever recorded for a US state. Finally, many regions are expected to see earlier snowmelt, so less water will be stored on mountain tops for the summer dry season. Added to natural climatic variation, such as the El Niño–La Niña cycle, these factors will intensify seasonal or decade-long droughts. Although the models don’t all agree on the specifics, the overall drying trends are clear.
There is simply little doubt that many dry areas have gotten drier and/or warmed up and/or seen earlier snowmelt.
I think it bizarre to claim that there is little change in global drought over the past 60 years when there are so many studies and analyses to the contrary directly linking severe droughts to climate change:
- NOAA Bombshell (Hoerling et al 10/11): Human-Caused Climate Change Already a Major Factor in More Frequent Mediterranean Droughts
- Study (10/10): Global warming is driving increased frequency of extreme wet or dry summer weather in southeast, so droughts and deluges are likely to get worse
- USGS Expert Explains How Global Warming Likely Contributes to East Africa’s Brutal Drought
The World Bank’s must-read new report, “Turn Down the Heat: Why a 4°C Warmer World Must be Avoided,” gets this right:
One affected region is the Mediterranean, which experienced 10 of the 12 driest winters since 1902 in just the last 20 years (Hoerling et al. 2012). Anthropogenic greenhouse gas and aerosol forcing are key causal factors with respect to the downward winter precipitation trend in the Mediterranean (Hoerling et al. 2012)…. East Africa has experienced a trend towards increased drought frequencies since the 1970s, linked to warmer sea surface temperatures in the Indian-Pacific warm pool (Funk 2012), which are at least partly attributable to greenhouse gas forcing (Gleckler et al. 2012). Furthermore, a preliminary study of the Texas drought event in 2011 concluded that the event was roughly 20 times more likely now than in the 1960s (Rupp, Mote, Massey, Rye, and Allen 2012).
You won’t find any of those studies referenced in the new Nature article. You can find Funk 2012 and Rupp et al 2012 in the The Bulletin of the American Meteorological Society Special Issue, “Explaining Extreme Events of 2011 from a Climate Perspective.”
I find it is especially surprising that Nature would publish this piece when three months ago its sister publication, Nature Climate Change, published a piece by Dai, “Increasing drought under global warming in observations and model” that is utterly at odds with it. There is simply no way both of these papers can be true — and yet the new Nature piece never discusses the Nature Climate Change piece.
Dai’s paper notes:
Historical records of precipitation, streamflow and drought indices all show increased aridity since 1950 over many land areas….
I conclude that the observed global aridity changes up to 2010 are consistent with model predictions, which suggest severe and widespread droughts in the next 30– 90 years over many land areas resulting from either decreased precipitation and/ or increased evaporation.
The new Nature paper dismisses previous work by Dai, unjustifiably according to Dai and Trenberth, but in any case, Dai’s Nature Climate Change goes much further in reconciling models with observations. Nature shouldn’t have published this new paper without a serious effort first to reconcile these two papers.
I think it unlikely that the new paper will stand up. Here is what Trenberth sent me:
A “drought” in good science
The new paper recently published in Nature by Sheffield et al “Little change in global drought over the past 60 years” [Nature 491 15 Nov 2012 435-440] has done some impressive work. But it should not have been published in Nature. It has 38 pages of dense supplementary material, for heaven’s sake. Was that reviewed? Probably not.
The conclusions of the paper are likely wrong. The paper re-examines the Palmer Drought Severity Index (PDSI) in different formulations and how it changes over time. However, this has been done before in references embedded in the paper:
24. Dai, A. Characteristics and trends in various forms of the Palmer Drought Severity Index during 1900–2008. J. Geophys. Res. 116, D12115 (2011).
25. van der Schrier, G., Jones, P. D. & Briffa, K. R. The sensitivity of the PDSI to the Thornthwaite and Penman–Monteith parameterizations for potential evapotranspiration. J. Geophys. Res. 116, D03106 (2011).
And dismissed with “Recent studies have claimed that there is little difference between the PDSIs that use the Thornthwaite and PM algorithms (PDSI_Th and PDSI_PM, respectively)24,25 but this can be attributed to inconsistencies in the forcing data sets and simulation configuration(see Supplementary Information).” Yes, but at least some of the supplementary material in this respect is wrong.
The simple PDSI_th relies on temperature to crudely estimate the atmospheric demand for moisture from the soil and thus evapotranspiration. It can be calibrated to each area and it works reasonably well but it is not perfect. It is known that actual evaporation depends on solar radiation and other energy sources, as well as surface wind speeds and humidity in addition to temperature. The problem is these fields are mostly not available in adequate quality, moreover daily fields are required.
The Sheffield et al Figs. S12-S14 suggest that the “little drying” conclusion is likely due to the use of the CRU precipitation data, which has fewer than 1500 raingauges for the recent years and which differs substantially from the GPCC and GPCP precipitation products that have many more gauge data for the last 10-20 years. The authors make a big deal of their findings, but in fact van der Schrier et al. have made similar conclusions in their previously published papers using the same CRU data set as the forcing. There are also major concerns about the reconstruction of the solar radiation data, which depends a lot on how clouds have changed.
Another key point is that while our previous results with PDSI have been compared with other related but independent records, such as soil moisture, streamflow, and GRACE satellite data, Sheffield et al. made only a detailed comparison of various forcing data for the PDSI calculations.
Another important factor not considered, is that precipitation on land is controlled to a large degree by ENSO: in general with La Niña, as experienced in recent years, there is more rain on land and so the past 2 years have been the wettest on record. That says nothing about whether the extent and intensity of drought is greater or not when it occurs, and so the ENSO signal should be removed before looking at trends associated with climate change.
Sheffield et al have uncovered some minor problems with Dai (2011) but none that explain the differences in the results. It is evident from all this that there are major issues with the “forcing” data for the more complex form of PDSI, and thus there remains a lot of merit in the simpler but self calibrated version of PDSI_th.
The bottom line is that climate change adds extra heat to the system and that much of heat goes into drying. A natural drought for whatever reason sets in quicker, becomes more intense, and possibly is longer lasting and more extensive as a result.
Dai sent me a long explanation of a major flaw in Sheffield et al, which I’ll summarize as “they normalized away the recent drying.” I think Nature needs to reconcile these papers, but until then, one should probably stick with the large and growing literature on the large and growing rise in aridity over many parts of the world.
Finally, this analysis does not undercut the large and growing literature on the threat to future generations of Dust-Bowlification — see “We’re Already Topping Dust Bowl Temperatures — Imagine What’ll Happen If We Fail To Stop 10°F Warming.”