Abstract: A compensation scheme on shallow convection clouds including both cloud cover and hydrometeors has been designed and tested in the GRAPES-Meso model on the basis of the shallow convection parameterization scheme modified by Wan, et al (2015). The purpose of this work is to improve simulation deficiencies on the low-level cloud cover and cloud radiation effects of shallow convection. Results of this study showed that:(1) The introduced diagnostic method performs reasonably on the simulation of shallow convective cloud cover; there is corresponding cloud cover in the triggering areas of shallow convection, and the deviation of the model simulated cloud cover from satellite observation is reduced by the improved scheme. (2) Compensations of cloud cover and cloud hydrometeors exist in a shallow layer that is about 0.5-4 km above the ground, corresponding to the active layer of shallow convection; the maximum of cloud compensations occurs at the height of 1-1.5 km, and the proportion of compensated cloud hydrometeors (the sum of cloud water content and rain water content) to the total is about 20%-55%. (3) There is a reasonable response of the cloud optical depth to the compensations of cloud hydrometeors in the lower troposphere; the vertical profile of the increment in the cloud optical depth and its variation trend with integration time are very similar to those of the cloud hydrometeor compensation,and the variation of the cloud optical depth depends more on the change in cloud water content than on the change in rain water. (4) In the daytime, the cloud radiation effects from the shallow cloud compensation relevant to shallow convection are clearly reflected by decreases in surface total solar radiation and 2 meter air temperature. As a result biases of these variables in the model simulation decrease. The positive effect of this improved scheme on reducing biases of the GRAPES-Meso simulation of cloud cover, surface total solar radiation,and 2 meter air temperature has been confirmed by verification results of a 1-month forecast experiment.