Understanding the causes of observed regional temperature trends is essential to projecting the human influences on climate, and the societal impacts of these influences. In their recent study, Christy et al. (2006, hereafter CRNG06) hypothesized that the presence of irrigated soils is responsible for the rapid warming of summer nights occurring in California’s Central Valley over the last century (1910–2003), an assumption that rules out any significant effect due to increased greenhouse gases, urbanization, or other factors in this region. Their interpretation is based on an apparent contrast in summer nighttime temperature trends between the San Joaquin Valley (∼+0.3 ± 0.1°C decade⁻¹) and the adjacent western slopes of the Sierra Nevada (−0.25 ± 0.15°C decade⁻¹). Here, we question the interpretation of the difference in temperature trends between the valley and the Sierra Mountains (and other regions of California), as well as the amplitude, sign, and uncertainty of the Sierra nighttime temperature trend itself.

Regarding the veracity of the apparent Sierra nighttime temperature trend, CRNG06 generated the valley and Sierra time series using a meticulous procedure that eliminates discontinuities and isolates homogeneous segments in temperature records from 41 weather stations. This procedure yields an apparent cooling of about −0.25 ± 0.15°C decade⁻¹ in the Sierra region. However, because removal of one of the 137 Sierra segments from the most elevated site (Huntington Lake, 2140 m) results in a zero trend (CRNG06), the apparent cooling of summer nights in the Sierra regions seems, in fact, largely uncertain, which should be expressed with a larger estimate of the error bar in Fig. 8 (in CRNG06).

Evidence of a large impact of irrigation, as suggested in CRNG06, should be captured in other high-quality observational datasets with a summer nighttime temperature trend in the valley that contrasts with those observed in the Sierra Mountains and other parts of California. We have analyzed those trends from four widely used gridded meteorological datasets [University of Washington (UW; Hamlet and Lettenmaier 2005; Mote et al. 2005), Vegetation/Ecosystem Modeling and Analysis Project (VEMAP; Kittel et al. 2001), Climate Research Unit version 2.0 (CRU2.0; Mitchell et al. 2004), and CRU version 2.1 (CRU2.1; Mitchell and Jones 2005)] for the period 1910–2000 (CRU2.0 and CRU2.1) or for the longest period of coverage within 1910–2000 (UW: 1915–2000; VEMAP: 1910–93). Additionally, we have also analyzed the 1910–2000 temperature trends from stations of U.S. Historical Climatology Network (USHCN), using observations that have been adjusted for biases induced by changes in instrumentation and station relocations and that contain estimated values for missing and outlier observations (Karl et al. 1990). While the gridded datasets are not truly independent because they include many of the same observations, the differences in resolution (1/8° for UW, 1/2° for the others) and the diverse processing and aggregation techniques applied make it valuable to analyze them all. For instance, most datasets (CRU2.0, CRU2.1, and UW) are developed to facilitate long-term trend analysis in incorporating a method to adjust for temporal inhomogeneities while VEMAP is not. The coverage of California stations employed also varies. Datasets such as CRU2.0 and CRU2.1 are predominantly based on long-term, adjusted HCN records. In these datasets, the Sierra region as defined by CRNG06 is only represented by one station in the Yosemite Valley. However, the station coverage in UW and VEMAP datasets is of good quality in the entire state and over the time period of consideration because they have used both HCN and cooperative network observations. In the UW dataset, some of the early observations were discontinued, but the number and spatial distribution of stations through time were however fairly constant after about 1950, and not radically different from those available prior to 1950 (A. Hamlet 2006, personal communication).

All gridded datasets (except for VEMAP) show that summer nighttime (i.e., daily minimum) temperatures are rising in the Central Valley (Fig. 1), at a rate that is significant and comparable to the one found in CRNG06. However, in contrast to CRNG06, none of the gridded datasets shows a nighttime cooling signal in the adjacent Sierra. Rather, daily temperature minima are found to be rising in the mountains at a rate similar or faster than in the valley region. Similar results are found using USHCN observations. We are not aware of a systematic source of error that could artificially enhance the spatial homogeneity of the nighttime temperature trends in California (individual changes in instrument, location, or local time of measurements usually deteriorate this homogeneity with the sign of biases varying from station to station). Nevertheless, it is possible that a lack of a gradient of temperature trends between the valley and the Sierra could result from an incomplete coverage of underlying stations in the Sierra (particularly in the CRU datasets). This would however not explain the similarity of trends between the valley and other adjacent regions that contain better station coverage: although there is a large variety of significance patterns, all datasets consistently show a rise in daily minimum temperatures that occurs across the entire state, from the western coast to the eastern mountains, and that affects all elevations (exceptions exist in the areas around Electra and Lake Spaulding; USHCN, Fig. 1). In fact, strong positive trends in nighttime temperature are a common feature throughout the western United States (A. Hamlet 2006, personal communication). If nighttime warming were a consequence of irrigation, it should be more rapid in the valley than in other regions, and this is not the case. Furthermore, rapid nighttime warming is observed in many other parts of the globe that are not irrigated. In contrast, the significant decrease in summer maximum temperatures is prevalent in the Central Valley but not always in adjacent regions (Fig. 1) and likely is related to radiative cooling over irrigated areas (as suggested in CNRG06).

CRNG06 assume that any difference in temperature trends between the Central Valley and Sierra regions, if confirmed, must be due to irrigation in the valley. Yet other factors can cause regional trend differences. Regional model simulations, for example, show greater warming response to increased CO₂ in the mountains than in the valley, even though these simulations do not represent irrigation or other forms of land use change (Duffy et al. 2006). Thus, differences in mountain versus valley meteorology can result in different temperature trends. Other spatially heterogeneous forcings (due to, e.g., aerosols or urbanization) could contribute to the trend differences. In the Central Valley, the population growth, the urban development (Bereket et al. 2005), and an increase in dirt and asphalt roads to access agricultural fields might also be at the origin of some increases in nighttime surface temperature. The magnitude of the changes attributable to such anthropogenic effects are, however, still unknown.

One way to study the effect of irrigation on California climate, while avoiding potential biases induced by elevation, different climatological regimes, and other spatially heterogeneous forcings, is to focus on the seasonality in observed temperature trends in the valley alone, since irrigation is more common in summer months. For instance, CRNG06 found a strong seasonality in daytime temperature trends in the San Joaquin Valley. The largest cooling, which occurs during summer while trends in other seasons are small, suggests again a reduction in sensible heating of the lower atmosphere due to increased evapotranspiration. The seasonality in minimum temperatures is quite different, however. Minimum temperatures rose during all four seasons in the valley, most rapidly during autumn. The uncertainties in these trends are such that the nighttime warming rate could have been identical in all seasons, which is inconsistent with these trends resulting from irrigation.

In summary, while a cooling of summer daytime temperatures due to irrigation in California’s Central Valley seems plausible, the interpretation that irrigation explains the rise in nighttime temperature does not seem supportable. Neither the results from gridded or meteorological station datasets nor the seasonality of the trends can support this hypothesis. According to the observational datasets used here, the rise in minimum temperatures has occurred across the entire state (although it is not significant everywhere), affected all elevations (Fig. 1), and accelerated during the second half of the twentieth century, which suggests a large-scale influence on California climate. A valuable hypothesis, disregarded by the authors, is that various human-induced factors (greenhouse warming and urbanization) act in concert to rise the temperature of summer nights in the Central Valley, while irrigation mitigates greenhouse warming of the Central Valley during the day.

1. Datasets

The UW gridded meteorological dataset was obtained from the Surface Water Modeling group at the University of Washington from its Web site (http://www.hydro.washington.edu/Lettenmaier/Data/gridded/index_hamlet.html). The CRU2.0 and CRU2.1 gridded datasets were developed by the CRU of the University of East Anglia. The VEMAP dataset was developed through the Vegetation/Ecosystem Modeling and Analysis Project Phase 2. The USHCN dataset has been developed to assist in the detection of regional climate change (Karl et al. 1990) and is commonly used for trend studies in the United States. It is maintained at the National Climatic Data Center (NCDC) and the Carbon Dioxide Information and Analysis Center (CDIAC) of the Oak Ridge National Laboratory.