New data acquisitions are used to examine recent global trends in maximum temperature, minimum temperature, and the diurnal temperature range (DTR). On average, the analysis covers the equivalent of 71% of the total global land area, 17% more than in previous studies. Consistent with the IPCC Third Assessment Report, minimum temperature increased more rapidly than maximum temperature (0.204 vs. 0.141°C dec−1) from 1950–2004, resulting in a significant DTR decrease (−0.066°C dec−1). In contrast, there were comparable increases in minimum and maximum temperature (0.295 vs. 0.287°C dec−1) from 1979–2004, muting recent DTR trends (−0.001°C dec−1). Minimum and maximum temperature increased in almost all parts of the globe during both periods, whereas a widespread decrease in the DTR was only evident from 1950–1980.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Minimum temperature increased about twice as fast as maximum temperature over global land areas since 1950, resulting in a broad decline in the diurnal temperature range (DTR [Folland et al., 2001]). Changes in cloud cover, precipitation, soil moisture, and atmospheric circulation likely accounted for much of the trend differential during the period [e.g., Dai et al., 1999; Przybylak, 2000; Braganza et al., 2004]. Changes in land use also impacted the DTR in some areas [e.g., Balling et al., 1998; Bonan, 2001; Small et al., 2001]. Unfortunately, data constraints have historically limited global-scale analyses of DTR trends and their causes; in particular, the most recent global assessment [Easterling et al., 1997] only covered about half of the land surface and ended in 1993. Consequently, in this study we use new data acquisitions to expand spatial coverage and update global trends in maximum temperature, minimum temperature, and the DTR through the period 1950–2004. For consistency with the IPCC Fourth Assessment Report, trends during the satellite era (1979–2004) are also discussed.
 Data for this study were compiled from 20 source datasets. The primary sources were the Global Historical Climatology Network [Peterson and Vose, 1997] monthly and daily databases (which contain most of the data used by Easterling et al.), and two editions of World Weather Records (1981–1990 and 1991–2000). These global datasets were supplemented with acquisitions from Argentina, Australia, Brazil, Chile, Cuba, Greece, Iran, New Zealand, and South Africa. CLIMAT reports were used to update about 20% of the stations after 1994. In addition, high-quality synoptic reports were included to fill recent gaps in about 10% of the stations (provided digital and manual checks indicated that the synoptic data closely matched historical monthly time series during periods of overlaps).
 The source data were quality assured using methods described by Peterson et al. . In brief, this involved identifying and merging duplicates, deleting stations with extreme changes in the mean and/or variance, and removing temporal and spatial outliers. The final, quality-assured dataset contained 7018 maximum and 6970 minimum temperature stations that had at least 20 years of “complete” data (i.e., 8 months per year). The DTR time series for each station was created by subtracting the minimum series from the maximum series.
 The approach of Menne and Williams  was then employed to produce adjustments for undocumented changes in station location, instrumentation, and observing practice. The first step entailed locating each station's six nearest neighbors and then compositing their annual anomalies into a reference series that represented the presumed “homogeneous” regional climate signal. Next, the reference series was subtracted from the candidate series, and the resulting difference series was checked for potential change points using two-phase regression and the Standard Normal Homogeneity Test. When both tests found a change point within two years of one another, an adjustment was applied to the candidate series.
 Unlike Easterling et al. , the final adjusted dataset was not stratified into urban and rural subnetworks because urban warming does not appear to significantly bias multidecadal trends over large areas. For example, Jones et al.  found that the impact of urbanization was roughly an order of magnitude smaller than the trend in mean temperature over global land areas during most of the 20th century. Easterling et al.  also noted that urban effects on global and hemispheric trends in maximum and minimum temperature and the DTR were negligible for the period 1950–1993. Peterson et al.  obtained only a slightly smaller global trend for a rural station network than for a blended rural-urban network over the period 1951–89, the difference being statistically insignificant. More recently, Parker  found that calm nights, which should in theory experience greater urban effects, had the same increases in minimum temperature as did windy nights for the period 1950–2004.
 The climate anomaly method [Jones and Moberg, 2003] was used to develop global, hemispheric, and grid-box time series for the period 1950–2004. The first step involved excluding stations that had less than 21 years of data during a base period of 1961–1990 and less than 4 years of data in each decade during that 30-year span. Base-period normals were then computed by month for the remaining 4280 maximum, 4284 minimum, and 4157 DTR stations. Next, each monthly temperature at each station was converted to an anomaly from its base-period mean. The station-based anomalies were then averaged into 5° by 5° latitude-longitude grid boxes for each year/month from 1950–2004. Finally, global and hemispheric means were computed by area-weighting each grid box by the cosine of the central latitude and averaging all of the weighted grid-box values in the given year/month. On average, the resulting dataset covers the equivalent of 71% of the total global land area, 17% more than in previous studies. Coverage exceeds 70% for all years during the base period, with a gradual decrease toward both ends of the record (to 54% by 1950 and 42% by 2004). Although sampling is relatively complete across the mid-latitudes, the tropical and polar land masses remain underrepresented because of a comparative lack of data.
 The climate anomaly method was also used to develop global, hemispheric, and grid-box time series for the satellite era. To maximize spatial coverage, a base period of 1979–2004 was used instead of 1961–1990 because the latter results in grid-box trend maps that cover 12% less of the land surface. (In general, the two base periods yield comparable results at large spatial scales; for instance, their global DTR trends for 1950–2004 differ by only 0.0013°C dec−1). A total of 3588 maximum, 3565 minimum, and 3360 DTR stations each have at least 21 years of data during the satellite period. On average, the resulting dataset covers the equivalent of 71% of the land surface. Coverage exceeds 70% from 1979–2000, dropping to 62% in 2003 and 48% in 2004.
4. 1950–2004 Trends
Figure 1 depicts the global annual time series for each variable for the period 1950–2004. In general, both maximum and minimum temperature increase from about the mid-1970s to present, with warming in the minimum during the 1950s as well. The DTR generally decreases during the period, though much of the change occurs in two periods (the 1950s and the early-1970s to early 1980s). From 1950–2004, the maximum temperature trend is 0.141°C dec−1, the minimum temperature trend is 0.204°C dec−1, and the DTR trend is −0.066°C dec−1 (all trends computed via least-squares regression). The maximum and minimum trends exceed those in Easterling et al.  by 0.050 and 0.018°C dec−1, respectively, whereas the DTR trend is less by 0.018°C dec−1. The larger maximum and minimum trends are generally consistent with the large positive global temperature anomalies observed in most years since 1993 [Levinson et al., 2005] while the smaller DTR trend likely reflects the accelerated rate of warming in the maximum. (Note that the DTR change does not appear to stem from differences in data coverage because the current dataset produces the same DTR trend as in Easterling et al. for the period 1950–1993.) In general, the trends for all variables are larger in the Northern Hemisphere, with the greatest warming in the boreal winter and spring (Table 1). Relatively speaking, trends exhibit little seasonality in the Southern Hemisphere.
Table 1. Annual and Seasonal Trends (°C dec−1) From 1950–2004 for Maximum Temperature, Minimum Temperature, and the DTR for Global Land Areas and the Northern and Southern Hemispheresa
Values in bold are not significant at the 5% level.
Figure 2 depicts the annual trend for each variable in each 5° by 5° grid box during the period 1950–2004. Boxes having less than 37 years of data (67% completeness) were excluded from the analysis. Maximum temperature increased in most regions except northern Mexico and northern Argentina, both of which cooled during the 55-year period. There was little net change in maximum temperature in northeastern Canada, the southeastern United States, and southern China. Minimum temperature increased in virtually all areas except Mexico, northeastern Canada, and parts of the western Pacific Ocean. The DTR generally declined in most areas, although the pattern was less spatially coherent than for its components, and increases were apparent in some regions (e.g., northeastern Canada, southern Argentina, eastern Africa, the western Pacific Ocean, southeastern Australia).
Figure 3 depicts the global annual time series for each variable for the period 1979–2004. In general, maximum and minimum temperature increase through most of the period whereas the DTR is basically trendless. The maximum and minimum temperature trends are nearly identical (0.287 versus 0.295°C dec−1), and both are comparable to the mean temperature trend over global land areas for the period (0.296 °C dec−1) as derived from the Global Historical Climatology Network database (J. Lawrimore, personal communication, 2005). Given the similarity between maximum and minimum temperature, the trend in the DTR (−0.001 °C dec−1) is not statistically significant at the 5% level. Although striking, the lack of a DTR trend is not without precedent (e.g., there was no trend from 1958–1976 either). Furthermore, the DTR does exhibit a significant decrease (−0.032°C dec−1) since 1976, which Folland et al.  defined as the start of the most recent warming period. On average, trends in maximum and minimum temperature are larger in the Northern Hemisphere, and both hemispheres have more warming in winter and spring (Table 2). A DTR increase is evident in the Southern Hemisphere winter (mainly in Australia), but all other DTR changes are insignificant.
Table 2. Annual and Seasonal Trends (°C dec−1) From 1979–2004 for Maximum Temperature, Minimum Temperature, and the DTR for Global Land Areas and the Northern and Southern Hemispheresa
Values in bold are not significant at the 5% level.
Figure 4 depicts the annual trend for each variable in each 5° by 5° grid box during the period 1979–2004. Boxes having less than 21 years of data (80% completeness) were excluded from the analysis. In general, maximum temperature increased in most regions except northern Peru, northern Argentina, northwestern Australia, and parts of the north Pacific Ocean. Minimum temperature increased in virtually all areas except southern Argentina, western Australia, and parts of the western Pacific Ocean. The DTR pattern was far less consistent, with increases in some areas (e.g., western North America, northern Eurasia, the Indian subcontinent, Australia) and decreases in others (e.g., northeastern Canada, the southeastern United States, Africa, parts of central and eastern Asia).
6. Summary and Conclusions
 In this study we used a suite of new data acquisitions to examine recent global trends in maximum temperature, minimum temperature, and the DTR. Consistent with Easterling et al. , minimum temperature increased at a faster rate than maximum temperature during the latter half of the 20th century, resulting in a significant decrease in the DTR for this period. In contrast, maximum and minimum temperature increases were roughly comparable during the satellite era, muting recent changes in the DTR. Maximum and minimum temperature increased in almost all parts of the globe during both periods, whereas a widespread decrease in the DTR was only evident from 1950–1980.
 The authors thank Tom Peterson, Phil Jones, and the anonymous reviewers for their insightful comments and suggestions on this article. Partial support for this work was provided by the Office of Biological and Environmental Research, U.S. Department of Energy (Grant number DE-AI02-96ER62276); and the NOAA Office of Global Programs, Climate Change Data and Detection Element.