Thursday, 13 February 2003
Estimation of land surface temperature on the Tibetan Plateau using GMS data
The Tibetan Plateau has been thought to play an important role in the progress of Asian summer monsoon as an elevated heat source/sink protrudes into the middle of troposphere. During the GAME (GEWEX Asian Monsoon Experiment) IOP surface sensible and latent heat fluxes were intensively measured at some sites for more than four months with eddy covariance technique. Surface parameters relevant to asses the surface energy balance such as land surface temperature, soil moisture, net radiation, surface and albedo were also continuously monitored. Although the land surface atmosphere interaction has experimentally revealed at these sites, the use of satellite remote sensing is essential to upscale the patch scale information to the understanding. Furthermore, it is required to use continuous data observed by a geostationaly satellite, since the diurnal variation is very strong over the Plateau. For this purpose, we developed a methodology to retrieve one of important surface parameters, the land surface temperature, using GMS data. We started from the split window algorithm originally developed for NOAA/AVHRR. The atmospheric transmittance for GMS/VIHRR sensors were calculated using a radiative transfer code with atmospheric profiles obtained in the Intensive Observation Period of GAME/Tibet. The asymptotic formula of transmittance is obtained as a function of both the precipitable water and the satellite zenith angle. For the operational use for this algorithm, the precipitable water is retrieved from the water vapor channel (6.7 Ám) data. Since the altitude of the plateau surface is about 4000 m on average, the atmospheric layers have about 700hPa thickness. The weighting function of the water vapor channel almost covers this layer with the maximum sensitivity near 400hPa level. Thus, the radiance observed this channel is a good indicator of the precipitable water over the Tibetan Plateau. We have also developed a new algorithm for cloud removal. Since the amplitude of diurnal change of the plateau surface temperature is comparable to that of annual, we employed a variable threshold technique. This threshold, which is determined from the surface measurements data, varies in both day of the year and the local time, so we can effectively exclude low clouds in the daytime and get low surface temperature data in the night times. We have already retrieved the annual change of surface temperature over the plateau for 1998 and are still continuing to detect the inter-annual changes. The comparison of retrieved surface temperature with in-situ measurements shows high correlation coefficient of over 0.9. Using this data, we have got the result that the diurnal range over the western plateau is about 5 K larger than that of the eastern plateau during pre-monsoon period in 1998, which is consistant with Yanai et al. (1992).