The use of satellite-derived skin temperature for soil moisture initialization in the Penn State/NCAR mesoscale model (MM5)

Previous numerical modeling studies of the influence of soil moisture on simulated atmosphere circulation have demonstrated that soil moisture has a relatively long temporal 'memory' compared to the atmospheric flow and thus a relatively slow temporal variation in the simulated climate. Thus, the initial soil moisture is very important for obtaining an appropriate depiction of atmosphere-surface exchanges in numerical weather prediction and climate models. But due to a lack of observation data, especially in high latitudes, the initial soil moisture is specified poorly in current numerical models.

In this study, we incorporate a satellite data assimilation method following the work of Jones et al.(1998) into the Penn State/NCAR MM5 modeling system to retrieve soil moisture. According to earlier studies that showed that the surface temperature is most sensitive to the soil moisture, relative to other surface factors, during the mid-morning hours, we assume that the difference between simulated skin temperature and the observed one in the mid-morning can be minimized by adjustments to the soil moisture. For the observed skin temperature, we derive a skin temperature from NOAA AVHRR data using an empirical relationship between the emissivity and the vegetation cover as well as a generalized split-window algorithmic approach. We will present results from case study simulation experiments over high latitudes to demonstrate the technique and its utility over a region not previously examined by other investigators.