exaggerated by about 0.1°C. An effect of similar magnitude may exist for Alaska. In the former Soviet Union, the population limit for inclusion of any station into the network was 10,000; in addition, however, no station could be within 1 km of any multi-story urban development. If the impact of the effect of urbanization on the DTR in the former Soviet Union is anything like that in the United States, the residual urban heat island effect on the DTR should be at least an order of magnitude smaller than the observed decrease of the DTR (nearly 1.5°C/100 yr).
In China and Japan, a number of tests were conducted to identify the impact of urbanization on the DTR. Three networks of stations were categorized on the basis of population. For China the categories included stations in proximity to cities with populations over 1 million, under 160,000, and in between these two threshold values. Categories in Japan were based on threshold values of 500,000 and 50,000. In the three population categories, proceeding from high to low, China had 23, 42, and 44 stations while Japan had 17, 71, and 66 stations. Figure 7 shows that the decrease of the DTR actually is more marked in China for stations in the lowest population category than for stations in the medium category, while the trend of the average temperature continues to decrease. This unexpected finding suggests that urban-
ization effects in China are dissimilar to those in the United States, since they seem not to have affected the maximum and minimum temperature trends in Chinese cities of 500,000 or less. In Japan, however, the impact of urbanization on the DTR is evident even in the lowest population category (less than 50,000), and is even more apparent for the average temperature. On the basis of these analyses, it seems that the impacts of urbanization in China (Wang et al., 1990) are unlikely to significantly affect the trends reported in Tables 1 and 2.
A previous paper by Jones et al. (1990) investigated the impact of increasing urbanization in the land data base used by the IPCC (1990, 1992) to calculate changes of global temperature. The conclusion from the work of Jones et al. (1990) was that any residual urban bias in the land-based average temperature records was about 0.05°C during the twentieth century. A comparison of the average temperature trends derived from the stations used in Tables 1 and 2 with the stations used by Jones (1988) revealed differences in trends from country to country, but virtually identical trends of temperature (within 0.02°C/100 yr) were found when all areas depicted in Figure 1 were considered. This similarity suggests that the degrees of urban-induced bias in these two data sets are of comparable magnitude over the past 40 years, despite the use of substantially different station networks.
It can be argued that increases in irrigation may account for the decrease in the DTR. The evaporation associated with soil moisture would convert sensible to latent heat and thus significantly reduce daytime temperature. In order to test this hypothesis, the correlation coefficient (both the Pearson product-moment and the Spearman rank) was calculated using the values of the trends of the DTR and the change in land area under irrigation from 1950 to 1987 (U.S. Department of Commerce, 1950, 1988) for each of the regions in the United States delineated in Figure 2. No relationship was found between the change in the DTR and the increase of irrigated lands, and in fact many of the largest decreases of the DTR were associated with areas with the smallest increases of irrigation. Considering that the United States has had significant decreases of the DTR over the past several decades, and has had relatively large increases of irrigation relative to other countries over the past 40 years, it seems unlikely that increases of irrigation can be regarded as a serious explanation for the decreases of the DTR.
The converse of the theoretical effects of irrigation would result from increased desertification. This might arise from