The response of a general circulation model to a change in its treatment of cloud solar forcing is investigated. Radiation field data from the forecast model of the Navy Operational Global Atmospheric Prediction System for five Julys (1979--1983) are presented in an investigation of the effect of a change from grid cell averaged clouds to maximally overlapping clouds in the model's solar radiation scheme. The model results are compared with Nimbus 7 Earth Radiation Budget top of the atmosphere (TOA) solar and longwave irradiances and with derived surface solar irradiance data. Although the maximal overlap scheme performs considerably better than the grid cell averaging scheme (reducing maximum deficiencies in TOA and surface solar irradiance by over 100 W m-2), significant errors remain. The simulated correlation between TOA net solar and longwave irradiance improves at low latitudes in the northern hemisphere, with little change at higher latitudes. This improved correlation is consistent with the greater consistency between the treatments of solar and longwave cloud radiative forcing brought to the model by the new solar radiation scheme. The change in the radiation treatment is shown to have the greatest direct effect on solar radiation over convective regions, a consequence of the scarcity of optically thick clouds produced by the model's cloud parameterization in other regions.

The model responds with an increase in convective activity over land and an increase in the flux of moisture from sea to land. Planetary cooling over the oceans increases because of a decrease in cloud cover. From mid to high latitudes in the northern hemisphere, there are scattered regions of increased cloud water content associated with increased tropospheric temperatures. Over land the model response in terms of TOA downwelling solar irradiance tends to counter the increase in solar irradiance caused by the model change in all latitudinal zones in the northern hemisphere. This response is caused primarily by changes in the cloud fields, which thus act as a negative feedback following the change in cloud solar forcing.

The significance of this response is examined with respect to the perturbation in solar irradiance represented by the model change. An estimate of this perturbation is obtained by taking the difference in solar irradiance diagnosed by the two cloud solar forcing treatments for simulations employing the grid cell averaging scheme. The response is significantly greater in magnitude in the tropics than at midlatitudes, both in an absolute sense and as a percentage of this perturbation. Because TOA longwave irradiance exhibits a positive response in the tropics, and a negative response at midlatitudes, however, the percentage response in net TOA downwelling irradiance is actually greater than in magnitude atmidlatitudes. In number of regions the cloud feedback is very large, showing the importance for cloud field prediction of improvements in the treatment of cloud solar forcing. Such cloud feedback also explains the small improvement seen here in the prediction of TOA solar irradiance in certain regions. Increases in surface sensible heating and longwave cooling are generally considerably less than increases in surface latent heating, though a notable exceptions occurs in arid central Asia. A large ground temperature increase in that region is strongly correlated at low levels with the atmospheric temperature increase observed at midlatitudes in the northern hemisphere. ¿American Geophysical Union