Error estimates of spaceborne passive microwave retrievals of cloud liquid water over land

Cloud liquid water path (LWP) retrievals from the Special Sensor Microwave/Imager (SSM/I) and surface microwave radiometers are compared over land to assess the errors in selected satellite methods. These techniques require surface emissivity composites created from SSM/I and infrared (IR) data. Two different physical methods are tested: a single-channel (SC) approach (either 85.5-GHz channel); and a normalized polarization difference (NPD) approach (37 or 85.5 GHz). Comparisons were made at four sites in Oklahoma and Kansas over an 11-month period. The 85.5-GHz NPD method was the most accurate and robust under most conditions. An error analysis shows that the method's random errors are dominated by uncertainties in the surface emissivity and instrument noise. Since the SC method is more prone to systematic errors (such as surface emissivity errors caused by rain events), it initially compared poorly to the ground observations. After filtering for rain events, the comparisons improved. Overall, the root mean square (rms) errors ranged from 0.12 to 0.14 kg m/sup -2/, suggesting these methods can provide, at best, three categories of cloud LWP. It is anticipated that the techniques and strategies developed in this study, and prior related studies, to analyze passive microwave data will be requisite for maximizing the information content of future instruments.