Solar Radiation Changes in the United States during the Twentieth Century: Evidence from Sunshine Duration Measurements

Abstract

Abstract Changes in sunshine duration (SS) measured in the conterminous United States during the past century were used as a proxy to explore changes in shortwave forcing at the earth’s surface when and where accurate measurements of global irradiance (Eg) were not available. Yearly totals of SS from the 106 Weather Bureau stations with 70 or more years of complete measurements between 1891 and 1987 were analyzed after establishing that the two changes in instrumentation during that period had not significantly influenced the measurements. Annual totals of SS were highly correlated (r2 = 0.86) with annual totals of global irradiance (Eg↓) measured at the 26 U.S. pyranometer stations during the 1977–80 period when the Solar Radiation Network (SOLRAD) was operating at its maximum accuracy. The linear relationship between annual totals of Eg↓ and SS was highly significant (P < 0.001), with each additional hour of sunshine duration equivalent to an increase of 0.0469 ± 0.002 W m−2 (or 1.48 ± 0.07 MJ m−2 solar radiation per year). The error term of annual values of Eg↓ estimated from SS was 5%. Almost half of the sunshine series showed significant linear time trends in SS. At 27 sites it increased significantly with time; at 21 sites it significantly decreased. Regionally, in the northwest quarter of the U.S. landmass (>36°N, >98°W), SS increased at nine sites and decreased at three; in the three other quarters of the United States, the numbers of sites with increasing and decreasing trends were equal. After 1950, a larger proportion of series showed decreases in sunshine duration, and more sites showing decreasing SS were found in the Northeast and in the West and South of the United States, but these regional differences were not significant. Normalized annual anomalies of SS averaged for all of the U.S. series showed no significant linear time trend during the last century, but the running 11-yr average values indicated clear peaks in the fourth and sixth decades of the last century and troughs in the first, fifth, and seventh decades; the peaks coincided with those reported for continental air temperature, and the troughs coincided with those for continental rainfall. A significant periodic component (with a median period of 10 yr) was found in half of the SS series; however the peak spectral density averaged for the United States, occurring at a period of 11.25 yr, was not significantly above that expected for the white noise level. An analysis of long-term records from outside the United States showed that the sensitivity of SS to Eg↓ was dependent on both astronomical and climatic factors, and the implications of this site dependence on the accuracy of this proxy relationship is discussed. A decline in SS followed major volcanic eruptions in North America. In the case of El Chichon, this change was calculated to have resulted in a negative shortwave radiative forcing of 6.4 W m−2 for the United States, some 3 times greater than the value calculated from the direct effect of the increase in aerosol loading. It is concluded that the U.S. sunshine duration database shows little evidence for a significant trend in solar forcing at the earth’s surface during the twentieth century. To reconcile this discrepancy with reports of decreases in Eg↓ measured in the United States during the last half century requires a more detailed understanding of the influence of clouds and aerosols on sunshine duration.