Quantitatively assessing cloud cover fraction in numerical weather prediction and climate models

摘要

Procedures are presented that exploit remotely-sensed satellite cloud data products to quantitatively assess the accuracy of cloud cover fraction (CCf) in datasets generated by numerical weather prediction (NWP) and climate models. These procedures are demonstrated with analyses created from the North American Mesoscale (NAM) Forecast System and forecasts generated with the WRF (Weather Research and Forecast) model. First VIIRS (Visible Infrared Imager Radiometry Suite) cloud data products are collocated within NAM gridded fields to identify grid cells for comparison against manually-generated cloud masks that are based upon VIIRS imagery and serve to characterize the accuracy of CCf fields in the NAM datasets. Next, short-range CCf forecasts are generated from these NAM datasets with the WRF model and the results are again compared to the manually-generated cloud mask datasets. Comparisons between the NAM CCf products and those in themanually-generated CCf fields reveal a systematic bias toward under-clouding in the NAM analyses which are important to cloud forecasts for air quality and solar energy applications as well as climate modeling. Poorer correlations were found in comparisons between the WRF cloud forecasts and the manually-generated CCf fields.