SNOW with the DTR (Figure 8b) suggests that the length of night relative to the TRAD is a significant factor in the impact of changes of snow cover on the DTR. In the northern United States, around the winter solstice, TRAD is relatively low. Snow on the ground at this time of year is important because of its excellent insulating properties (it reduces heat flow from the soil), which help lower the nighttime minimum. During the daytime the TRAD is already low, so the amount of solar radiation reflected by the snow cover is no longer so important. As a result, snow cover at this time of year at these latitudes leads to an increase in the DTR. This is not the case as the season progresses or the latitude decreases, as is reflected by the negative partial correlations (lower values of DTR with snow cover) associated with the highest category of TRAD, where there were still nearly 1,000 cases of snow cover. The data suggest that snow-cover ablation will not necessarily lead to an increase of the DTR.
From the above analysis it is apparent that there are many factors, often intricately related, that affect the DTR. Many of these variables are very much related to a greenhouse effect, not necessarily one anthropogenically induced. Overall, however, the two variables related to changes in cloudiness, sky cover, and ceiling height explain the greatest portion of the variance of the DTR. Changes in cloudiness should be one of the first considerations in searching for an explanation of the observed decrease of the DTR (Figure 2). Indeed, when large continental scales are considered, the relationship between cloud amount (or sky cover) with the DTR is quite impressive (Figure 9). Plantico et al. (1990) have already demonstrated that the decrease in the DTR
over the United States is strongly linked to an observed increase in daytime and nighttime cloud cover and to a lowering of cloud ceilings.
Is there a general increase in cloud cover over much of the globe? Empirical evidence by Henderson-Sellers (1986, 1989, 1992) and Jones and Henderson-Sellers (1992) suggests that this may be the case over Canada, the United States, Europe, the Indian subcontinent, and Australia. Analyses of cloud cover changes over China from a network of 58 stations are inconclusive, but there is evidence for a decrease in the sunshine (a 2 to 3 percent decrease in sunshine from the 1950s to the 1980s). The quality of the cloud data (and perhaps the sunshine data) is questionable because the correlation between monthly anomalies of sunshine and cloudiness at many sites is not high. Analyses of changes in cloudiness over the former Soviet Union by Kaiser and Razuvayev (1995) reveal a general increase of cloud cover (about 5 percent) over the period 1936 to 1986, with cirrus increasing in frequency by nearly 20 percent. They also found considerable interannual and interstation variability, so the quality of these data could also be called into question. Nonetheless, the trend over the former Soviet Union is consistent with a decrease in the DTR. Frich (1992) and Bücher and Dessens (1991) also found that decreases in the DTR in Denmark and at the Pic du Midi de Bigorre Observatory occurred concurrently with an increase in cloudiness. An analysis of changes in cloud cover over Japan, using many of the same stations selected for the analysis of the maximum and minimum temperatures, indicates that cloud cover may have increased on an annual basis by nearly 1 percent since 1951, but there is no apparent response in the DTR. Changes in sunshine in Japan are not altogether consistent with the increase in cloud cover, perhaps because a new instrument was introduced into the network in 1986.
Interest in the possible change of the DTR with increasing anthropogenic greenhouse gases has prompted several modeling groups to publish information from their models regarding the change in projected DTR when CO2 is doubled. Table 5 summarizes the results of these models. For the GCM experiments, the magnitude of the decrease in DTR is small relative to the overall increase of mean global temperature. Moreover, these experiments reflect a level of CO2 increase well in excess of present-day values, so even smaller changes would be expected in the observed-temperature record. Outside of the Rind et al. (1989) study, only Cao et al. (1992) focus specifically on the changes of the DTR over land areas. Cao et al. show that their model can reasonably simulate the range of the present-day diurnal