Comparison of model predicted liquid water path with ground-based measurements during CLIWA-NET

Continuous cloud remote sensing measurements with microwave radiometers and various other instruments were conducted in a ground-based network as part of the EU-project CLIWA-NET with the focus on liquid water path (LWP). The observations, taken from a network of ground-based stations within the BALTEX modeling area, cover in total 6 months and were compared with short-term predictions from atmospheric models. Two models for numerical weather forecasting: the ECMWF model and the meso-scale Lokal-Modell operated by the German Weather Service, and two models for regional climate prediction: the Rossby Centre Atmospheric model and KNMI model RACMO, participated in the model evaluation. In view of the restriction that microwave radiometer measurements are unreliable when the instrument becomes wet, it is essential that events of precipitation be accurately identified in the observations and model values are effectively filtered out. Non-zero or even negative values of observed LWP in cloud free situations are effectively dealt with by a time-dependent correction based on information of supporting instruments for cloud detection. Temporal aggregation or, equivalently, conditional averaging in time, was used to bring the observations to scales matching the grid-box mean model predicted values. Mean observed LWP averaged over the scenes with water clouds present reduces from about 90 g/m(2) for temporal scales of 5 min to about 40 g/m(2) at scales of 1 h. All models, but in particular the RCA-model, tend to overpredict frequency and duration of precipitation. Different models differ most in their predictions of LWP illustrating the scarcity of LWP observations so far. The ECMWF model and, to a lesser extent, RACMO are found to overestimate mean LWP, while the RCA-model predicts values close to what has been observed. The LM-model tends to greatly underestimate the observed values. (c) 2005 Elsevier B.V. All rights reserved.