1. Introduction
The U.S. state of California has immense municipal and agricultural water demands, with over 38 million residents and the largest farming economy in the United States (U.S. Department of Agriculture 2012). California receives most of its precipitation from October through March (Cayan and Roads 1984), hereinafter referred to as the “wet season.” Annual precipitation totals are highly variable (Dettinger 2011; Mitchell and Blier 1997). Since the population is expected to grow beyond 50 million by 2060 (California Department of Finance 2013), satisfying future water demands requires an understanding of how precipitation may vary throughout the twenty-first century due to climate change.
While temperatures in California are widely expected to increase under climate change (Pierce et al. 2013b; Cayan et al. 2008; Duffy et al. 2006; Hayhoe et al. 2004), the associated precipitation signal is more ambiguous. Several studies have found only modest projected changes in mean precipitation, which are small compared to natural variability (Pierce et al. 2013b; Cayan et al. 2008; Duffy et al. 2006; Hayhoe et al. 2004). But even these modest signals in the mean may be accompanied by significant changes in precipitation extremes. Recent studies that examined projections in phase 3 of the Coupled Model Intercomparison Project (CMIP3) found an increase in extreme precipitation over California. For example, Das et al. (2013) found significant end-of-century increases in 3-day flood magnitudes across the entire California Sierra Nevada mountain range according to all 16 projections they analyzed. Pierce et al. (2013a) also found evidence in model projections that daily precipitation intensity will increase, particularly in the northern half of California. Using projections in phase 5 of the Coupled Model Intercomparison Project (CMIP5), Polade et al. (2014) showed that an increased number of dry days along with higher frequencies of very heavy precipitation days combine to increase interannual precipitation variability over California.
These prior studies of extreme precipitation changes over California focused on high-frequency weather events. The purpose of this study is to analyze changes in extreme precipitation on interannual time scales. Such prolonged wet and dry periods can cause significant stress to ecosystems, agriculture, and urban water supply. The potential role of climate change in the recent low precipitation period of 2013/14 is an especially relevant aspect of this topic. Using 34 CMIP5 global climate models (GCMs), the study first examines future trends in average precipitation over California. Then it assesses the statistical significance of simulated changes in both extremely wet and dry wet seasons by the middle and end of the twenty-first century. Through these analyses, as well as an examination of the observed trend in precipitation over the region since 1900, an argument attributing the very low precipitation of 2013/14 to natural variability is presented.
2. Observations and models
This study takes advantage of California’s uniquely dense and long-term observational record of precipitation across the state (http://cdec.water.ca.gov/index.html). Monthly precipitation totals from 1900 through 2014 are available from 103 stations across the state (Fig. 1a). To understand the spatial structure of the precipitation fields sampled by this observational network, we correlate time series of wet season (October–March) precipitation anomalies at each station to the station-averaged anomalies (Fig. 1b). All stations in the southern third of California, along with several stations scattered in the northern two-thirds of the state, have correlations below 0.7. Our objective is to obtain a cohesive observation network, so we exclude stations with correlations below 0.7 and recalculate correlations between the remaining stations and the new station average. This process is repeated twice until all stations have a correlation greater than 0.7, yielding a final, “strongly cohesive” (average r = 0.85) network of 75 stations (Fig. 1c). Therefore, the station average is highly representative of precipitation variations at the scale of the northern two-thirds of California, as well as smaller spatial scales. This makes it an appropriate metric to compare with output from GCMs, whose gridbox scales are comparable to the span of this observational network. Stations are located primarily throughout the Sierra Nevada, Central Valley, and coastal regions of California and approximately sample where the main water resources are found in California (i.e., the northern two-thirds of the state). Monitoring water resources over the Sierra Nevada is particularly important since more than 60% of the state’s water supply originates from these mountains (Water Education Foundation 2011). Station elevations range from 4 to 2940 m, with an average of 967 m. Fourteen stations recorded wet season precipitation totals at the start of the twentieth century, growing to the current number of 75 stations by 1984/85. Measurement gaps are found throughout the time series (particularly 1981/82), although at least 30 stations are active from 1904/05 to the present (Fig. 1d).
Observational network. (a) Location and height above sea level (m) for all 103 stations of precipitation observations. (b) Correlation between time series of station and station-averaged wet season precipitation anomalies for all 103 stations. (c) Correlation between time series of station and station-averaged wet season precipitation anomalies for “strong” (r > 7.0) stations. (d) Number of active stations in the “strong” network from wet seasons 1900/01 through 2013/14.
Citation: Journal of Climate 28, 16; 10.1175/JCLI-D-14-00624.1
GCM outputs of monthly precipitation, geopotential height, and sea ice from the CMIP5 archive (http://pcmdi9.llnl.gov/esgf-web-fe/) are also employed in this study (Table 1). We use information from the first realization of two standardized experiments (historical and RCP8.5). RCP8.5 represents a “business as usual” future forcing scenario where greenhouse gas emissions continue to increase throughout the twenty-first century. GCMs are regridded to a 1.5° × 1.5° horizontal resolution (using a spherical harmonic regridding function within the NCAR Command Language), allowing for uniform area averages throughout the study. In the GCMs, “California” is defined as the continuous set of grid cells spanning the observational network in Fig. 1c. A map showing these grid cells is provided in the inset of Fig. 2.
CMIP5 models used in this study and their corresponding institutions. (Expansions of acronyms are available online at http://www.ametsoc.org/PubsAcronymList.)
Observed and simulated wet season precipitation totals: 1900/01–2013/14 wet season totals (cm) according to station-averaged observations (green line; see Fig. 1c for station locations); 1900/01–2099/2100 wet season totals averaged over California in 34 GCMs (thin gray lines; the California averaging zone is shaded yellow in the upper left of the plot); and GCM-mean 1900/01–2099/2100 wet season totals averaged over California (thick black line). Also overlaid are 2005/06–2099/2100 trends (RCP8.5 forcing scenario) for each GCM. These trends are colored according to their sign (blue for positive trends, red for negative trends).
Citation: Journal of Climate 28, 16; 10.1175/JCLI-D-14-00624.1
3. Results
a. GCM precipitation trends under climate change
Here we examine twentieth- and twenty-first-century GCM simulations of California precipitation using the historical (1900/01–2004/05) and RCP8.5 (2005/06–2099/2100) experiments. RCP8.5 is the most extreme forcing scenario in the CMIP5 framework and therefore the one most likely to be associated with statistically significant anthropogenic trends. This may also represent the most realistic scenario since current emissions already exceed those represented in RCP8.5 (Peters et al. 2013). A total of 34 GCMs are included in the analysis, allowing us to robustly characterize intermodel spread and compute the ensemble-mean outcome. Shown in Fig. 2 are each GCM’s wet season accumulations from 1900/01–2099/2100 in thin gray lines, along with the ensemble mean in a thick black line. Also shown is the observed time series from 1900/01–2013/14 in green. As seen in the figure, California’s observed precipitation is extremely variable on interannual time scales (Dettinger 2011). In fact, the observed distribution is notably non-Gaussian with a heavy wet tail, and less than half of the wet seasons are within 25% of the median. From 1900/01 to 2013/14, the average wet season accumulation is 62.2 cm, with a standard deviation of 20.6 cm. Generally speaking, simulated California precipitation is also highly variable. The average of each model’s 1900/01–2013/14 standard deviation is 18.1 cm. As a group, the models appear to have a slightly low bias (12%) in precipitation variability. Langford et al. (2014) also found that CMIP5 GCMs simulate the average and standard deviation of historical precipitation over California fairly accurately.
Looking at the 2005/06–2099/2100 trends (blue and red lines in Fig. 2), we see that GCMs disagree on the sign of projected changes in precipitation, as 21 models show positive trends and 13 show negative trends. This disagreement was found in other recent studies (e.g., Das et al. 2013; Neelin et al. 2013; Berg et al. 2015) and can be interpreted within the framework of the “rich get richer” or “wet regions get wetter and dry regions drier” effect (Chou and Neelin 2004; Held and Soden 2006; Trenberth 2011; Durack et al. 2012). Specifically, California is located at the node of oppositely signed large scale precipitation changes (van Oldenborgh et al. 2014), whereby northern mid- and high-latitude regions are projected to become wetter, whereas southern subtropical areas are projected to become drier. GCM projections of precipitation over California tend to disagree on the sign of change, depending on whether they place the region more in the northern or southern regime of large-scale change.
Regardless of sign, the projected precipitation trends throughout the twenty-first century are generally insignificant. The average positive and negative trend is +1.04 and −0.75 cm decade−1, respectively, while the ensemble-mean is +0.35 cm decade−1. These trends are very small compared to both simulated and observed interannual standard deviations of 18.1 and 20.6 cm per wet season, respectively. Only the most extreme cases (+2.68 and −1.84 cm decade−1) could lead to a significant shift in hydroclimate when the variability is taken into account. Taken as a whole, Fig. 2 provides little evidence from the GCMs that wet season totals will greatly change throughout the twenty-first century, even under a very aggressive warming scenario. The most likely case is that the region’s very large variability will continue to dominate any long-term trends due to anthropogenic forcing.
b. Extreme precipitation changes in the twenty-first century
Despite the small projected mean precipitation trends in Fig. 2, large changes in extremes may still occur (e.g., Das et al. 2013; Polade et al. 2014). To quantify how the frequency of wet seasons with extreme wetness and dryness may change throughout the twenty-first century, we examine extreme wet season totals within each GCM. The thresholds for extreme events are defined according to detrended simulated historical (1900/01–2004/05) precipitation distributions. An extreme wet event is one where the wet season total exceeds the 95th percentile in the model’s own historical period climatology, while an extreme dry event is one where the wet season total is less than the 5th percentile. The results of this exercise are nearly identical if one instead estimates natural variability levels using preindustrial simulations within the subset of GCMs that show minimal model drift (not shown). Results are also nearly identical if one instead examines extreme changes using water year totals (October through September), rather than wet season totals (not surprising since most precipitation falls during the wet season months). Based on our definition of extremes, we expect one wet season to be extremely dry and one to be extremely wet by chance about every two decades. Bearing this level of natural variability in mind, we examine changes to extremes in the twenty-first century in Fig. 3 and Table 2.
Counts of projected wet season precipitation totals that are extremely (a) dry or (b) wet in each GCM during two-decade intervals throughout the twenty-first century. Only counts greater than the expected number by chance (one extremely dry and wet event per two decades) are colored. GCMs are ordered wettest to driest from top to bottom in each panel based on their twenty-first-century climatology of extremes (see section 3b for more details).
Citation: Journal of Climate 28, 16; 10.1175/JCLI-D-14-00624.1
Number of wet seasons in each GCM that are extremely dry and wet during 2020/21–2059/60 (mid-twenty-first century) and 2060/61–2099/2100 (end of twenty-first century). Increases that are statistically significant at the 95% level, calculated using a Monte Carlo resampling technique, are noted with an asterisk. GCMs are ordered wettest to driest from top to bottom based on their twenty-first-century climatology of extremes.
Figures 3a and 3b show the number of wet seasons that are extremely dry and wet, respectively, within a given two-decade interval in the twenty-first century for 34 GCMs. (The color scheme of Fig. 3 is chosen so that any box with shading in it represents a number of extreme events greater than the one event expected by chance.) Models are roughly ordered wettest to driest from top to bottom based on their twenty-first-century climatology of extremes. Specifically, we calculate the difference between each model’s total number of twenty-first-century wet and dry extremes shown in Fig. 3. The model with the largest difference, GISS-E2-H, is ranked the wettest model, while MIROC-ESM-CHEM has the smallest difference and is ranked as the driest model. For discussion purposes, it is useful to divide the latter eight decades of the twenty-first century into two periods, 2020/21–2059/60 (mid-twenty-first century) and 2060/61–2099/2100 (end of twenty-first century). Testing of statistical significance on the number of extremes in each of these periods is performed with a Monte Carlo resampling technique. For each GCM, 40 wet seasons (matching the length of each future period) are randomly selected from its detrended historical (1900/01–2004/05) time series and the numbers of dry and wet extremes within that synthetic time series are recorded. This is repeated 1000 times to produce two distributions showing the expected number of dry and wet extremes within a 40-yr historical slice. We then extract the percentiles of these distributions corresponding to the numbers of mid- and end-of-twenty-first-century dry and wet extremes (i.e., boxes in Fig. 3). If a percentile is greater than the 95th historical percentile, a significant (p < 0.05) increase in extremes is said to occur. The results of this exercise for all 34 GCMs are shown in Table 2.
Focusing on the mid-twenty-first century in Table 2, only 2 out of 34 models project statistically significant increases in wet seasons that are extremely dry (average of 10 during these 40 years). Across the model ensemble, the average and median occurrence of extremely dry wet seasons is found to be 2.2 and 2.0, respectively. Since the region would experience around two extremely dry wet seasons during four decades by chance, this information indicates that the frequency of extremely dry wet seasons is unlikely to change between 2020/21 and 2059/60. On the other hand, 16 out of 34 models have statistically significant projections of more than two wet seasons that are extremely wet during this time period (Fig. 3b; Table 2), with an average of 6.8 per 40 years. The ensemble-mean average and median occurrence is 4.2 and 4.0, respectively. Thus nearly 50% of models significantly project a tripling of the frequency of wet extremes by the mid-twenty-first century. However, the ensemble mean suggests an increase of only two additional wet extremes, which is statistically significant at the 95% level.
Extreme changes become more pronounced and better model agreement is found when analyzing the 2060/61–2099/2100 period. For instance, 10 out of 34 models have statistically significant projections of more than two extremely dry wet seasons during this 40-yr period (Fig. 3a; Table 2). When averaged across these models, a mean occurrence of dry extremes during these four decades is found to be 8.4. Across all models in the ensemble, the average and median occurrence are found to be 3.6 and 3.0, respectively (see also middle column of Table 3). While over one-quarter of the models project a significant increase in the frequency of dry extremes 4 times greater than the historical levels, the ensemble mean projects an additional one or two more extremely dry wet seasons per two decades by the end of the twenty-first century (not statistically significant at the 95% level). This effect is also visually apparent in the increasing redness of the boxes in Fig. 3a. A majority of the models (19 out of 34) project statistically significant increased occurrences of wet seasons that are extremely wet during 2060/61–2099/2100. The average occurrence of extremely wet years across models with significant increases for this period is 9.5 or roughly 4–5 times the expected frequency. The ensemble-mean average and median occurrence is 6.4 and 5.5, respectively (middle column of Table 3), indicating a statistically significant tripling of the frequency of wet extremes compared to the historical frequency by the end of the twenty-first century.
Comparison between the historical (1900/01–2004/05) and 2060/61–2099/2100 (end of twenty-first century) average/median frequencies of extreme wet seasons. There are two cases for 2060/61–2099/2100. In the center column, each GCM’s 2060/61–2099/2100 time series of precipitation is unaltered. This allows for simulated mean and variability changes to influence extreme changes during 2060/61–2099/2100. In the right column, each GCM’s 2060/61 – 2099/2100 precipitation time series is modified so that its mean is equal to the 1900/01–2004/05 (historical) mean. This allows for only simulated variability changes to influence the number of extreme years during 2060/61–2099/2100.
Interestingly, only three models overlap [CESM1(CAM5), CMCC-CESM, and EC-EARTH] within the 10 out of 34 and 19 out of 34 model subsets that project significant increases in dry and wet extremes, respectively, by the end of the century (Table 2). While models generally project increases in either wet or dry extremes, some models appear to project increases in both. Multiple scenarios could help explain this model behavior. Of course, changes in mean precipitation without any change in variability would by itself produce an increase in extremes by simply shifting the distribution in one direction and fattening that direction’s tail. This may explain the increases in either wet or dry extremes in a particular model. However, the increases in both wet and dry extremes in some models indicate changes in variability that amplify extremes in one tail may also be at work.
To unravel the influences of mean and standard deviation changes to extreme changes, we first show each GCM’s twenty-first- minus twentieth-century mean and standard deviation change in Fig. 4. The mean change is fairly skewed toward moistening (19 positive and 15 negative), an expected result given the information shown in Fig. 2. Moreover, a great majority of models (28 out of 34) show positive changes in the variability of wet season totals during the twenty-first century. Some models that project statistically significant increases in wet extremes [e.g., GISS-E2-H, MRI-CGCM3, and BCC_CSM1.1(m)] are accompanied by large positive changes in both the mean and standard deviation. On the other hand, there are several other models that also project significantly increased wet extremes [e.g., CanESM2, and CESM1(BGC)] but are associated with negative mean changes and positive standard deviation changes. Similar nonuniform behavior is seen in the subset of models that project significant increases to dry extremes by the end of the century. One model associated with significant increases in dry extremes during this period (MIROC5) is accompanied by large decreases to the mean and standard deviation. Other models that project increased dry extremes, such as GFDL-ESM2G, show decreases in the mean but increased variability. So both changes in the mean and changes in variability may conspire to produce an increase in extremes in many models.
Twenty-first- (2000/01–2099/2100) minus twentieth-century (1900/01–1999/2000) changes in the mean and standard deviation according to 34 GCM simulated California wet season precipitation totals. The 1900/01–1999/2000 and 2000/01–2099/2100 time series are detrended before calculating the standard deviation change. Models are ranked wettest-to-driest based on their twenty-first-century climatology of extremes.
Citation: Journal of Climate 28, 16; 10.1175/JCLI-D-14-00624.1
Table 3 separates out these two effects by comparing how extremes change during 2060/61–2099/2100 with and without an accompanying mean change in that time period. In the center column of Table 3, numbers of extreme wet seasons for the 2060/61–2099/2100 period are shown including the mean precipitation change. In the right column, each GCM’s end-of-twenty-first-century (2060/61–2099/2100) mean is reset to equal the corresponding mean for the historical period (1900/01–2004/05). This is accomplished by adding the difference between historical and end-of-twenty-first-century means to each wet season total during 2060/61–2099/2100. In essence, this compares how many extreme wet seasons occur when both the simulated mean and variability are allowed to change (center column) and when only variability changes (right column). Table 3 shows that the increase in precipitation extremes occurs even when the mean change is removed. Wet seasons with extreme wetness only decrease from around 6 to 4 per 40 years, while the frequency of extremely dry wet seasons is approximately unchanged. A decrease in wet extremes when removing the mean change is consistent with the moderate moistening tendency in the majority of the GCMs (Fig. 2). Removing the mean change effectively “dries” the 2060/61–2099/2100 precipitation time series for most models, and reduces the number of extremes in the wet tail. In sum, Fig. 4 and Table 3 reveal that, as a group, 1) half of the simulated increase in wet extremes can be attributed to increased variability, while the other half is due to increases in the mean, and 2) increased dry extremes are almost entirely due to increased variability.
4. Low precipitation of the 2013/14 California drought
These results showing increases in extreme wet season precipitation are especially relevant given recent drought conditions over California. The state has received below average wet season precipitation for three consecutive years, 2011/12–2013/14. Water shortages prevented irrigation in many instances, leading to projected economic losses of nearly $2.2 billion for 2014 alone (Howitt et al. 2014). The drought’s severity and socioeconomic impacts led many to question whether the conditions of 2013/14 could be related to climate change (Swain et al. 2014; Wang et al. 2014; Diffenbaugh et al. 2015). Here we reexamine Figs. 2 and 3 and Table 2 to gain a better understanding of the causes behind the low precipitation associated with the 2013/14 drought.
We first place the 2013/14 wet season in historical context using the station-averaged observations (Fig. 1). 2013/14 recorded 32.3 cm of precipitation, representing just above 50% of the average over the past 114 wet seasons. By this measure, the 2013/14 wet season precipitation, while severe, is not unprecedented in California. In fact it ranks as the fifth driest since 1900/01, following the wet seasons of 1976/77 (19.4 cm), 1923/24 (26.1 cm), 1975/76 (31.3), and 1930/31 (32.0 cm). Moreover, even though the 1900/01–2013/14 period ends with a dry anomaly, the trend in wet season accumulation over this period is only −0.044 cm decade−1 (−0.14 to +0.05 cm decade−1 at the 95% confidence level), statistically not different from zero and negligible compared to the standard deviation of observed variability (20.6 cm) during this period. Recent variations in California precipitation, including the 2013/14 wet season, would appear to be highly consistent with natural variability.
Although the wet season of 2013/14 is only the fifth driest on record (and the 2013/14 water year the third driest), the calendar year of 2013 is the driest of the whole record [also found in Swain et al. (2014)]. This is an interesting fact, but should be interpreted with caution. Given California’s Mediterranean climate regime, calendar years actually measure two half wet seasons of precipitation, a metric that does not have a natural physical interpretation. Nonetheless, the discrepancy in rankings between the 2013/14 wet season/water year and the calendar year 2013 can be resolved by focusing on just January through March (JFM) precipitation accumulations. JFM 2013 was the all-time driest, recording just 24% of the JFM average. However, JFM 2014 was only modestly dry at 71% of the average. This prevented the 2013/14 wet season and water year totals (physically based measures of single “annual” amounts) from becoming record lows.
While the observed time series does not appear to contain any obvious anthropogenic signatures, the GCM projections of increasing extremes accompanied by small or no trends in mean precipitation do raise the possibility that the low precipitation of 2013/14 is at least partly anthropogenic. However, even in the future 2020/21–2059/60 period, there is no model consensus that extreme dryness will occur more frequently than the region already experiences (Fig. 3a; Table 2). Only by the end of the century does a signal of enhanced frequency of dry extremes emerge past natural variability levels. Thus this analysis finds no model evidence that the 2013/14 low wet season precipitation, and perhaps the remaining extremely dry wet seasons over California that will surely occur throughout the first half of the twenty-first century, can be clearly linked to climate change. Interestingly, a signal of increased wet extremes is more robust in terms of model consensus than the corresponding dry extremes signal, and is associated with a much larger change in frequency, and an earlier emergence from the noise in the twenty-first century. In sum, evidence from both the observed historical precipitation time series and future precipitation projections supports the idea that the 2013/14 low wet season precipitation totals can be very likely attributed to natural variability, even if the GCMs suggest that extremely dry wet seasons may become somewhat more common by the end of the twenty-first century.
The CMIP5 ensemble can also be used to investigate previously identified mechanisms leading to an increased tendency toward decreased precipitation in California. A study by Sewall and Sloan (2004) found some relationship between anthropogenic warming and reduced precipitation over California. Specifically, the authors forced a GCM (CCSM, version 1) with prescribed declines in Arctic sea ice fraction and corresponding sea surface temperatures and found wintertime [December–February (DJF)] precipitation reductions of ~30% over California by 2050. The authors attribute this reduction to increased 500-hPa geopotential heights over the Gulf of Alaska, which has the effect of steering more storms away from California. This increased ridging is suggested to have formed from planetary wave pattern disturbances originating over the Barents and Kara Seas due to sea ice loss and increased ocean exposure in that region. To test the link between Arctic sea ice loss and California precipitation decreases in the CMIP5 ensemble, we calculate end-of-twenty-first-century DJF changes (2080/81–2099/2100 average minus 1980/81–1999/2000 average) in Arctic sea ice fraction and California precipitation within the first realization of 28 CMIP5 GCM historical and RCP8.5 simulations. While nearly every model projects significant Arctic sea ice fraction declines between these two time periods, there is no consensus in the sign of precipitation change over California (as seen in Fig. 2). In addition, the correlation between these two changes is essentially zero (r = −0.004) and not significant (p > 0.5). While the mechanism identified by Sewall and Sloan (2004) may be operating in the GCM they analyzed, it does not appear to play any significant role in shaping future precipitation changes over California in the CMIP5 ensemble as a whole.
5. Conclusions
This study analyzes how simulated mean and extreme precipitation during the wet season change in California as GCMs respond to an aggressive twenty-first-century warming scenario. Mean precipitation trends and changes to extremes are quantified using historical (1900/01–2004/05) and RCP8.5 (2005/06–2099/2100) output from 34 CMIP5 GCMs. Models tend to disagree on the sign of the mean precipitation trends during the twenty-first century, and the magnitudes of the trends are generally small compared to natural variability levels. There is a slight tendency toward moistening, with the ensemble-mean trend being positive but close to zero (+0.35 cm decade−1).
Extreme precipitation changes are also examined over two time periods in the twenty-first century, 2020/21–2059/60 and 2060/61–2099/2100. Extremes are based on thresholds in detrended simulated historical (1900/01–2004/05) data. Specifically, dry extremes are those whose totals fall below the 5th historical percentile, while wet extremes are those whose totals exceed the 95th percentile. During the 2020/21–2059/60 period, there is no model consensus that extremely dry wet seasons will increase in frequency beyond the two per 40 years expected by chance. However, nearly 50% of models project statistically significant increases in wet seasons with extreme wetness, with an ensemble-mean increase of two wet extremes over this time period (statistically significant at the 95% level). Signals become stronger and better model agreement is found when analyzing the 2060/61–2099/2100 period. For instance, the ensemble mean projects an increase of about one to two additional extremely dry wet seasons beyond the number expected by chance (not statistically significant at the 95% level). A majority of models also project enhanced frequency of wet extremes during 2060/61–2099/2100, with an ensemble-mean, statistically significant increase equal to a tripling of the historical frequency.
We tested to see whether these changes to extremes are biased by our choice to include all 34 CMIP5 GCMs in the analyses (e.g., Knutti 2010). For example, it is possible to imagine constructing a smaller ensemble based on a criterion of model quality in simulating the region’s climate. Swain et al. (2014) identified 12 CMIP5 GCMs that best simulate 500-hPa geopotential heights over the northeastern Pacific Ocean, a climatic feature that shapes California precipitation. We found similar changes in extremes when analyzing only these 12 models, indicating that the major conclusions of this study are not sensitive to reasonable assumptions about which models to include in the ensemble.
Examining simulated changes in the mean and standard deviation between the twentieth and twenty-first centuries sheds light on the source of increased frequency of extremes during 2060/61–2099/2100. Distributions of twentieth- and twenty-first-century precipitation in the GCMs show that a large majority of models (>80%) project increases in the standard deviation of interannual precipitation totals [also found in Polade et al. (2014)]. Since changes in the mean can also lead to changes in extremes, the contributions of changes in the mean and standard deviation to increased extremes are separated. We do this by recalculating changes to extremes in the 2060/61–2099/2100 period in each GCM but forcing the mean precipitation to be equal to its 1900/01–2004/05 value. When no mean change is allowed in the models, wet extremes still double in frequency. Extremely dry wet seasons increase in frequency to the same degree as when the mean change is included. Therefore, enhanced precipitation variability fully accounts for the fact that the models project modest increases in the frequency of dry extremes. Since including the mean change leads to a tripling of the frequency of wet extremes, changes to the mean and variability each account for half of the increase in wet seasons that are extremely wet.
Increased interannual precipitation variability suggests a change in the frequency and/or intensity of storms traversing California. These changes could manifest themselves, for example, through the low-frequency components of increased variability of the Northern Hemisphere jet stream and associated storm tracks (Neelin et al. 2013; Lorenz and DeWeaver 2007; Yin 2005) or enhanced column water vapor (Held and Soden 2006). Through atmospheric teleconnections, changes in Northern Hemispheric low-frequency circulation and thermodynamic patterns (and California precipitation on annual time scales) may reflect future shifts in El Niño–Southern Oscillation (ENSO). For instance, “extreme El Niño events,” associated with extraordinary precipitation along the eastern Pacific Ocean, are projected to significantly increase in the twenty-first century within the CMIP5 ensemble (Cai et al. 2014). More specific to the regional focus of this study, Seager et al. (2012) showed that ENSO-driven precipitation minus evaporation anomalies significantly increases over Northern California during the twenty-first century. Maloney et al. (2014) also found a strengthening of California precipitation patterns associated with ENSO in their analysis of CMIP5 GCMs under RCP8.5.
Wang et al. (2014) also investigated the mechanisms behind enhanced variability in one GCM they analyzed. The authors showed that upper-level storm activity, as measured through a dipole of 250-hPa geopotential height over the northeastern Pacific Ocean and north of the Great Lakes in the United States, has increased in variance since 2000. As we have noted, it takes decades for the increased variance in precipitation to emerge as a statistically significant climate change signal in the GCMs. Wang et al. (2014) may have correctly identified a mechanism enhancing precipitation variance; but if this is the case, the CMIP5 models would indicate that the mechanism does not have a noticeable effect on precipitation variance until the latter half of the twenty-first century. We calculated the correlation of the dipole index of Wang et al. (2014) with our California precipitation time series. It is statistically significant but rather low (r = −0.45, p < 0.01), consistent with this line of reasoning. Even a significant increase in the variability of this dipole would translate into a proportionally much smaller increase in the variance of precipitation. Still it is worth pursuing the validity of this mechanism in the GCM output.
Regarding the low precipitation of the 2013/14 California drought, we present several pieces of evidence showing that its origins can be understood in terms of natural variability. Using a network of rain gauges across the state, the severity of the 2013/14 wet season precipitation total is found to lie well within the range of historical totals and in fact ranks as only the fifth driest since 1900/01. Moreover, the observed trend in 1900/01–2013/14 wet season precipitation is statistically zero. This absence of any signal that could be construed as anthropogenic is entirely in line with expectations from the GCMs. When analyzing future GCM projections of extreme precipitation, we find that extremely dry wet seasons occur about as frequently as expected from natural variability levels even as late as the middle of the twenty-first century. Only by the end of the twenty-first century does a small signal of increased extreme dryness emerge past natural variability levels. Based on these multiple lines of evidence, it seems difficult to escape the conclusion that the 2013/14 low precipitation of the California drought can be attributed to natural variability of the climate system.
Although 2013/14 precipitation levels over California are very likely attributed to natural variability, the region’s recent drought (i.e., considering all water cycle processes) was likely exacerbated by anthropogenic factors (Diffenbaugh et al. 2015). In fact, 2014 was the warmest year on record over California (http://www.ncdc.noaa.gov/sotc/national/2014/13), in line with observed and projected anthropogenic warming trends over the region (van Oldenborgh et al. 2014). These record warm temperatures, along with low precipitation, greatly amplified evaporative fluxes and lead to 2014 attaining the lowest Palmer drought severity index score of the last 1200 years (Griffin and Anchukaitis 2014). Future droughts are also projected to become significantly more severe over the Southwest due in part to increased evaporative demands in the twenty-first century (Cook et al. 2015). Along with these anthropogenic climate stressors to the region’s water budget, large statewide population growth (California Department of Finance 2013) and rapidly depleting groundwater reservoirs (Famiglietti 2014) will pose additional severe challenges to the management and mitigation of twenty-first-century California droughts.
Acknowledgments
Support for this study has been provided by the Annenberg Foundation and the U.S. Southwest Climate Change Center. Constructive comments from two anonymous reviewers greatly strengthened the findings of this study.
REFERENCES
Berg, N., A. Hall, F. Sun, S. Capps, D. Walton, B. Langenbrunner, and D. Neelin, 2015:Twenty-first-century precipitation changes over the Los Angeles region. J. Climate, 28, 401–421, doi:10.1175/JCLI-D-14-00316.1.
Cai, W., and Coauthors, 2014: Increasing frequency of extreme El Niño events due to greenhouse warming. Nat. Climate Change, 4, 111–116, doi:10.1038/nclimate2100.
California Department of Finance, 2013: New population projections: California to surpass 50 million in 2049. CDOF Press Release, 10 pp. [Available online at http://www.tularecog.org/DocumentCenter/View/374.]
Cayan, D. R., and J. O. Roads, 1984: Local relationships between United States West Coast precipitation and monthly mean circulation parameters. Mon. Wea. Rev., 112, 1276–1282, doi:10.1175/1520-0493(1984)112<1276:LRBUSW>2.0.CO;2.
Cayan, D. R., E. P. Maurer, M. D. Dettinger, M. Tyree, and K. Hayhoe, 2008: Climate change scenarios for the California region. Climatic Change, 87, 21–42, doi:10.1007/s10584-007-9377-6.
Chou, C., and J. D. Neelin, 2004: Mechanisms of global warming impacts on regional tropical precipitation. J. Climate, 17, 2688–2701, doi:10.1175/1520-0442(2004)017<2688:MOGWIO>2.0.CO;2.
Cook, B. I., T. R. Ault, and J. E. Smerdon, 2015: Unprecedented 21st century drought risk in the American Southwest and central plains. Sci. Adv., 1, e1400082, doi:10.1126/sciadv.1400082.
Das, T., E. P. Maurer, D. W. Pierce, M. D. Dettinger, and D. R. Cayan, 2013: Increases in flood magnitudes in California under warming climates. J. Hydrol., 501, 101–110, doi:10.1016/j.jhydrol.2013.07.042.
Dettinger, M., 2011: Climate change, atmospheric rivers, and floods in California—A multimodel analysis of storm frequency and magnitude changes. J. Amer. Water Resour. Assoc., 47, 514–523, doi:10.1111/j.1752-1688.2011.00546.x.
Diffenbaugh, N. S., D. L. Swain, and D. Touma, 2015: Anthropogenic warming has increased drought risk in California. Proc. Natl. Acad. Sci. USA, 112, 3931–3936, doi:10.1073/pnas.1422385112.
Duffy, P. B., and Coauthors, 2006: Simulations of present and future climates in the western United States with four nested regional climate models. J. Climate, 19, 873–895, doi:10.1175/JCLI3669.1.
Durack, P. J., S. E. Wijffels, and R. J. Matear, 2012: Ocean salinities reveal strong global water cycle intensification during 1950 to 2000. Science, 336, 455–458, doi:10.1126/science.1212222.
Famiglietti, J. S., 2014: The global groundwater crisis. Nat. Climate Change, 4, 945–948, doi:10.1038/nclimate2425.
Griffin, D., and K. J. Anchukaitis, 2014: How unusual is the 2012–2014 California drought? Geophys. Res. Lett., 41, 9017–9023, doi:10.1002/2014GL062433.
Hayhoe, K., and Coauthors, 2004: Emissions pathways, climate change, and impacts on California. Proc. Natl. Acad. Sci. USA, 101, 12 422–12 427, doi:10.1073/pnas.0404500101.
Held, I. M., and B. J. Soden, 2006: Robust responses of the hydrological cycle to global warming. J. Climate, 19, 5686–5699, doi:10.1175/JCLI3990.1.
Howitt, R. E., J. Medellín-Azuara, D. MacEwan, J. R. Lund, and D. A. Sumner, 2014: Economic analysis of the 2014 drought for California agriculture. Center for Watershed Sciences, University of California, 20 pp. [Available online at https://watershed.ucdavis.edu/files/biblio/DroughtReport_23July2014_0.pdf.]
Knutti, R., 2010: The end of model democracy? Climatic Change, 102, 395–404, doi:10.1007/s10584-010-9800-2.
Langford, S., S. Stevenson, and D. Noone, 2014: Analysis of low-frequency precipitation variability in CMIP5 historical simulations for southwestern North America. J. Climate, 27, 2735–2756, doi:10.1175/JCLI-D-13-00317.1.
Lorenz, D. J., and E. T. DeWeaver, 2007: Tropopause height and zonal wind response to global warming in the IPCC scenario integrations. J. Geophys. Res., 112, D10119, doi:10.1029/2006JD008087.
Maloney, E. D., and Coauthors, 2014: North American climate in CMIP5 experiments: Part III: Assessment of twenty-first-century projections. J. Climate, 27, 2230–2270, doi:10.1175/JCLI-D-13-00273.1.
Mitchell, T., and W. Blier, 1997: The variability of wintertime precipitation in the region of California. J. Climate, 10, 2261–2276, doi:10.1175/1520-0442(1997)010<2261:TVOWPI>2.0.CO;2.
Neelin, J. D., B. Langenbrunner, J. E. Meyerson, A. Hall, and N. Berg, 2013: California winter precipitation change under global warming in the Coupled Model Intercomparison Project phase 5 ensemble. J. Climate, 26, 6238–6256, doi:10.1175/JCLI-D-12-00514.1.
Peters, G. P., and Coauthors, 2013: The challenge to keep global warming below 2°C. Nat. Climate Change, 3, 4–6, doi:10.1038/nclimate1783.
Pierce, D. W., and Coauthors, 2013a: The key role of heavy precipitation events in climate model disagreements of future annual precipitation changes in California. J. Climate, 26, 5879–5896, doi:10.1175/JCLI-D-12-00766.1.
Pierce, D. W., and Coauthors, 2013b: Probabilistic estimates of future changes in California temperature and precipitation using statistical and dynamical downscaling. Climate Dyn., 40, 839–856, doi:10.1007/s00382-012-1337-9.
Polade, S. D., D. W. Pierce, D. R. Cayan, A. Gershunov, and M. D. Dettinger, 2014: The key role of dry days in changing regional climate and precipitation regimes. Sci. Rep., 4, 4364, doi:10.1038/srep04364.
Seager, R., N. Naik, and L. Vogel, 2012: Does global warming cause intensified interannual hydroclimate variability? J. Climate, 25, 3355–3372, doi:10.1175/JCLI-D-11-00363.1.
Sewall, J. O., and L. C. Sloan, 2004: Disappearing Arctic sea ice reduces available water in the American West. Geophys. Res. Lett., 31, L06209, doi:10:1029/2003GL019133.
Swain, D. L., M. Tsiang, M. Haugen, D. Singh, A. Charland, B. Rajaratnam, and N. S. Diffenbaugh, 2014: The extraordinary California drought of 2013/2014: Character, context, and the role of climate change [in “Explaining Extreme Events of 2013”]. Bull. Amer. Meteor. Soc., 95, S3–S7, doi:10.1175/1520-0477-95.9.S1.1.
Trenberth, K. E., 2011: Changes in precipitation with climate change. Climate Res., 47, 123–138, doi:10.3354/cr00953.
U.S. Department of Agriculture, 2012: California agricultural statistics, 2012 crop year. 94 pp. [Available online at http://www.nass.usda.gov/Statistics_by_State/California/Publications/California_Ag_Statistics/Reports/2012cas-all.pdf.]
van Oldenborgh, G. J., M. Collins, J. Arblaster, J. Christensen, J. Marotzke, S. B. Power, M. Rummukainen, and T. Zhou, 2014: Annex I: Atlas of global and regional climate projections. Climate Change 2013: The Physical Science Basis, T. F. Stocker et al., Eds., Cambridge University Press, 1311–1393. [Available online at http://www.ipcc.ch/pdf/assessment-report/ar5/wg1/WG1AR5_AnnexI_FINAL.pdf.]
Wang, S.-Y., L. Hipps, R. R. Gillies, and J.-H. Yoon, 2014: Probable causes of the abnormal ridge accompanying the 2013–2014 California drought: ENSO precursor and anthropogenic warming footprint. Geophys. Res. Lett., 41, 3220–3226, doi:10.1002/2014GL059748.
Water Education Foundation, 2011: Looking to the source: Watersheds of the Sierra Nevada. Water Education Foundation, 27 pp. [Available online at http://www.sierranevada.ca.gov/our-region/docs/waterreport.pdf.]
Yin, J. H., 2005: A consistent poleward shift of the storm tracks in simulations of 21st century climate. Geophys. Res. Lett., 32, L18701, doi:10.1029/2005GL023684.
