Abstract
The dependence of pixel-scale cloud-cover frequency distribution on cloud size and pixel size is examined through Monte Carlo simulations. A shape parameter, which describes the shape of the frequency distribution, is found to be a simple function of the ratio of cloud size to pixel size. The form of the frequency distribution changes from U shape to bell shape when the ratio decreases. It becomes uniform when the ratio is about 0.8–0.9 for the cases where the regional-scale cloud cover is 0.5. It is shown that on average the cloud cover in partially cloudy pixels increases with increasing regional-scale cloud cover when the ratio of cloud size to pixel size is small. It becomes insensitive to regional-scale cloud cover when the ratio becomes large. It is also shown that, in comparison with the results from Monte Carlo simulations, the grid-scale cloud-cover frequency distribution obtained using a threshold method tends to be more U shaped, and that obtained using a method that assigns 50% cloudiness to partially cloudy pixels tends to be less U shaped, particularly for subgrid cloudiness. A possible way of retrieving cloud size is suggested. It is found that the difference between a simulated cloud field, where clouds are uniformly distributed, and a real cloud field, where clouds may not be uniformly distributed, biases cloud size retrieval. Investigations on how clouds are distributed in a real cloud field are recommended.
