Abstract
Using radiometer data collected during the Canadian Atlantic Storms Program, we have investigated five different methods of estimating the path-integrated, or columnar, cloud liquid water. The methods consist of one- and two-channel physical retrievals, the standard method of linear statistical inversion using two channels, and two statistical methods that proceed from an initial determination of several empirical regressions between measured and computed quantities. Though differing in details and complexity, the methods gave estimates of cloud liquid that did not deviate greatly from one another. We assessed the accuracy of the methods by simulation. Using hypothetical profiles of cloud liquid in archival soundings, we calculated the atmospheric emission and thus the brightness temperatures that would be measured in the two channels of the radiometer. These values were taken as data for the five methods, and the amount of liquid was calculated. Results showed that the three statistical methods were more accurate than the physical methods, but no one of the three was significantly better than the others. In the four methods requiring measurements in two channels, the columnar water vapor is computed as part of the retrieval procedure. A comparison of the computed with the actual vapor amounts showed that one of the statistical methods employing empirical regressions was the most accurate for vapor retrieval. For this optimum method, the rms deviation of the measured columnar liquid from its actual value was 0.159 mm and the rms deviation of the columnar vapor was 0.867 mm. As fractions of the overall average liquid and vapor in the simulations, these deviations amount to 37% and 8.7% respectively. If cases are excluded in which the liquid amount is small or nonexistent, the fractional deviation of the liquid estimates decreases and that of the vapor increases.
