A methodology is presented for estimating downward longwave irradiance at the ocean surface from satellite radiance data. The downward longwave irradiance is computed with a fast and accurate radiative transfer model as a function of temperature, water vapor, ozone and carbon dioxide mixing ratios, fractional cloud coverage, emissivity of clouds, and cloud top and cloud base altitudes. A sensitivity study is performed to assess the relative importance of the model input parameters and devise strategies regarding their retrieval. Ozone and carbon dioxide mixing ratios are consequently fixed at their climatological values, whereas the other parameters, highly variable in space and time, are determined from satellite data. Temperature and water vapor mixing ratio are obtained from NOAA Tiros operational vertical sounder data, and cloud parameters are retrieved from GOES visible and infrared spin scan radiometer data. Several methods are investigated to retrieve the cloud parameters. In the most refined method, cloud base altitude is deduced from cloud top altitude and liquid water path, assuming a vertical liquid water distribution within the clouds. In the other methods, simplifying assumptions are introduced, which include directly relating liquid water path to cloud geometrical thickness, fixing the cloud geometrical thickness to its climatological value, and, finally, parameterizing the cloud effects only as a function of fractional cloud coverage. Satellite‐derived irradiances are compared to those measured in situ during the Mixed Layer Dynamic Experiment, conducted in October–November 1983 off the central California coast. The results indicate that the satellite methods perform similarly, with standard errors of estimate ranging from 21 to 27 W m−2 on a half‐hourly time scale and from 16 to 22 W m−2 on a daily time scale. These errors correspond to 6 to 8% and 4 to 6% of the average measured values, respectively. When compared with techniques based on empirical formulas that employ conventional surface data, the satellite methods also exhibit similar standard errors of estimate. The satellite methods, however, are favored, since they are generally less biased and globally applicable.
Read full abstract