Microwave Remote Sensing Over Land: Application to SSM/I

Microwave retrieval over land is potentially very useful for numerical weather prediction, climate studies and other weather related studies. The major problem however, is the high variability of the surface emissivity over land. We used a physically-based approach to retrieve the land surface temperature LST, the total precipitable water TPW and the cloud liquid water CLW over land. The emissivity is retrieved as part of the geophysical vector. To demonstrate the feasibility of the microwave retrieval over land, an attempt has been made to retrieve the main geophysical parameters using the Special Sensor Microwave Imager data. In order for the UR to be successful over land a priori information is needed to constrain the retrieved surface emissivities. Before to proceed the retrieval, the SSM/I brightness temperature measurements are optimally interpolated to the 19 GHz footprint. To validate this unique retrieval method, we have undertaken few comparison studies for each parameter. The LST retrievals are compared to a statistically based algorithm, and to weather shelter temperatures. The results show very good agreement, about 2 to 3 Kelvin rms error and almost no bias for retrievals using descendant (morning) passes. The UR land emissivity retrieval is tested over North America for the one-week period from 1 October to 6 October 1995. The emissivity is retrieved at the SSM/I frequencies 19.35, 37.0, and 85.5 GHz for both vertical and horizontal polarization and for the 22.235 GHz at the vertical polarization. The retrieval results show that the emissivities are stable during the entire one-week period, except for 3rd October, for all frequencies and polarization. On 3rd October a significant decrease in retrieved surface emissivity is observed in some areas and particularly in Texas. This decrease is observed in all channels. An examination of the weather radar reflectivities for the period approximately one hour before the F11 overpass, indicated that a heavy rain was falling in the region where the emissivity decrease is observed. The heavy rain significantly wetted the soil and likely produced large areas of surface water. Since water has a much lower MW emissivity than land this scenario is consistent with the lower land-surface-emissivity retrievals produced by the UR. The total precipitable water amounts over land from the UR are checked by comparing them to analyses from the National Center for Environmental Prediction 's (NCEP) Medium Range Forecast model (MRF). The UR consistently captures the main features of the MRF water vapor fields. Another by-product of the UR is the rain scattering detection. Indeed, the cost function of the rainy cases is high (above 1Kelvin) which could be used as rain detector. The comparison with scattering index shows very good agreement. The advantage of the UR residual-based detection algorithm is its very low false alarm rate.