Accuracy of cloud optical depth retrievals from ground-based pyranometers

Abstract

Errors in cloud optical depth retrieved from pyranometer irradiances are estimated using a fractal model of cloud inhomogeneity. The cloud field is constructed from a two-dimensional array of pixels. For each of the pixels, which are 200 x 200 m2 in size, the radiative transfer is calculated using the independent pixel approximation. If cloud cover is 100%, the retrieval bias can be positive or negative for individual 10-min averaged transmittances, depending on the position of cloud inhomogeneities with respect to the pyranometer. The mean bias is always negative. Increasing the averaging time to 40 min reduces the scatter in the bias, although the mean bias remains -1.0, a value that depends on the choice of fractal model. If cloud cover is less than 100%, but there is no independent means to omit partly cloudy periods from the irradiance records, the negative retrieval bias will increase with reduced cloud cover and optical depth. Below optical depths of 5, the retrieval errors are so large that no meaningful results are obtained despite the fact that retrievals may appear to be reasonable. The simulations herein cannot take account of three-dimensional photon transport. The results of this study demonstrate that it is essential to measure cloud fraction and the variability of the cloud structure if optical depth is to be retrieved from pyranometer observations. Extra instruments recommended for ground-based remote sensing of cloud optical depth are a cloud lidar, powerful enough to probe the entire troposphere, and a microwave radiometer to measure the total column liquid water.