Climate change and sectors of the surface water cycle In CMIP5 projections

Hydrol. Earth Syst. Sci., 18, 5317–5329, 2014 www.hydrol-earth-syst-sci.net/18/5317/2014/ doi:10.5194/hess-18-5317-2014 © Author(s) 2014. CC Attribution 3.0 License. Climate change and sectors of the surface water cycle In CMIP5 projections P. A. Dirmeyer, G. Fang, Z. Wang, P. Yadav, and A. Milton George Mason University, 4400 University Drive, MS 6C5, Fairfax VA 22030, USA Correspondence to: P. A. Dirmeyer () Received: 27 June 2014 – Published in Hydrol. Earth Syst. Sci. Discuss.: 25 July 2014 Revised: – – Accepted: 24 November 2014 – Published: 19 December 2014 Abstract. Results from 10 global climate change models are synthesized to investigate changes in extremes, defined as wettest and driest deciles in precipitation, soil moisture and runoff based on each model’s historical 20th century simulated climatology. Under a moderate warming scenario, regional increases in drought frequency are found with little increase in floods. For more severe warming, both drought and flood become much more prevalent, with nearly the entire globe significantly affected. Soil moisture changes tend toward drying, while runoff trends toward flood. To determine how different sectors of society dependent on various components of the surface water cycle may be affected, changes in monthly means and interannual variability are compared to data sets of crop distribution and river basin boundaries. For precipitation, changes in interannual variability can be important even when there is little change in the long-term mean. Over 20 % of the globe is projected to experience a combination of reduced precipitation and increased variability under severe warming. There are large differences in the vulnerability of different types of crops, depending on their spatial distributions. Increases in soil moisture variability are again found to be a threat even where soil moisture is not projected to decrease. The combination of increased variability and greater annual discharge over many basins portends increased risk of river flooding, although a number of basins are projected to suffer surface water shortages. 1 Introduction The suite of climate model simulations from the Coupled Model Intercomparison Project Phase 5 (CMIP5) offers a wealth of information about the potential for future climate change across a range of emission/mitigation scenarios. The CMIP5 simulations have suggested that hydrologic feedbacks of the land surface to the atmosphere are likely to intensify and the spatial and temporal extent of the regions of strong feedbacks will expand in the 21st century (Dirmeyer et al., 2013, 2014). Land–atmosphere interactions are studied because of their potential implications for climate predictability and their role in hydrologic extremes. These previous results motivate us to examine how extremes in the surface water cycle are projected to change in the next century, and how they may affect specific sectors of society. Recent studies have used the output of a small number of CMIP5 models to drive an additional suite of sector models to assess changes in the likelihood of flood (Dankers et al., 2014), drought (Prudhomme et al., 2014), significant water resource impacts (Schewe et al., 2014) and agriculture (Rosenzweig et al., 2014). In such a two-step modeling approach, versions of the same land surface model are sometimes employed twice (once in the CMIP5 climate model and again as a sector model) or different land surface models are convolved where inconsistencies can amplify errors (cf. Koster et al., 2009). It is arguably a cleaner comparison to examine the CMIP5 output directly, even if sector behaviors are poorly represented or absent in the models themselves. By comparing relevant climate outputs superposed on secondary sector data sets such as crop distributions and hydrologic catchments, key drivers may be assessed in an alternative way. In this study, we examine projected changes in the extremes within aspects of the surface water cycle and their potential impacts as directly represented by the CMIP5 models. We have characterized three sectors of water cycle extremes for this Published by Copernicus Publications on behalf of the European Geosciences Union. 5318 P. A. Dirmeyer et al.: Surface water cycle In CMIP5 projections study: meteorological (precipitation), hydrological (runoff) and agricultural (soil moisture). Section 2 describes the data sets used and the specific analyses applied to those data in this study. Changes in the occurrence rates of extremes are presented in Sect. 3. Section 4 categorizes projected hydrologic changes in terms of the sectors defined above. Rainfall changes are assessed based on the spectrum of various climate indices. Soil moisture changes are composited against the global coverage of various types of crops, suggesting possible impacts on agriculture. Runoff changes are integrated over large river basins to assess impacts on water resources. A discussion of caveats for this study is given in Sect. 5, and a summary is presented in Sect. 6. 2 Data and analyses Monthly mean fields of precipitation, total runoff and moisture in the uppermost 10 cm of the soil from CMIP5 simulations (Taylor et al., 2012) are examined, along with averages of interannual standard deviation (IASD) based on monthly means. Data are taken from single simulations of 10 different climate models, the first ensemble member in each case (r1i1p1). The 10 models are listed in Table 1. Three different climate cases are considered: the transient 20th century (historical) case as well as two of the future climate scenarios – net radiative forcing scenarios of 4.5 Wm−2 (RCP45) and 8.5 Wm−2 (RCP85) (van Vuuren et al., 2011). Output from the last 90 simulated years of the three different climate cases of each model are detrended before interannual variances are calculated, as described in Dirmeyer et al. (2014). All analyses are performed on each model’s native grid on a monthby-month basis, and then aggregated to standard 3-month seasons by averaging the results across the 3 months, or aggregated across growing seasons for the crop-based analysis as explained below. Only multi-model statistics are shown – all multi-model means are a simple average of the indicated statistic across the 10 models. Historical thresholds for extremes are defined for each model and month at each land surface grid box. Specifically, the extreme deciles are calculated from the last 90 years of un-detrended data from the historical case; the thresholds would be the values for the ninth driest and ninth wettest value for each month at each point. This defines for this study the contemporary 20th century definitions of “drought” and “flood” as represented by each model for each month of the year. These thresholds are then compared to the distributions derived from the RCP4.5 and RCP8.5 cases; that is, we determine how many years out of 90 exceed the historical thresholds. Large changes indicate areas (...truncated)