Estimating surface heat fluxes and related parameters by combining satellite brightness temperature and a heat budget model over a semi-arid steppe

A time series and a spatial distribution of surface heat fluxes and related surface parameters over a semi-arid steppe were estimated using a technique that incorporating the thermal-infrared brightness temperature of a satellite into a land surface heat budget model. We studied the western part of the Kherlen River Basin in Mongolia, where a typical steppe dominates, including forest-steppe in the northern part and dry-steppe in the southern part of the basin. The typical annual average air temperature is around 0 degrees Celsius, and annual precipitaition is around 200mm in the basin. They significantly change year-to-year. The growing season of vegetation ranges middle May through early October, which also changes year-to-year. Our goal was to estimate the temporal change of surface heat fluxes and related parameters at a location, where was a surface flux station, in the typical steppe over a growing season in year 2003 (hereafter referred to as temporal estimate), and to estimate the spatial distribution of surface heat fluxes and related parameters over the subject region at a certain time of the growing season (hereafter referred to as spatial estimate). A surface heat budget model developed in this study was a two-layer model, which consisted a canopy layer and a bare soil surface. The model predicted diurnal changes of surface temperatures of the two layers, depending on five input variables, which were the incoming solar radiation, the downward longwave radiation, the air temperature, the specific humidity, and the wind speed. The model parameters that controled surface temperatures were optimized so as to minimize a sum of squares of differences between calculated surface brightness temperatures and observed brightness temperatures obtained by satellites. The archive of the brightness temperature of the Moderate Resolution Imaging Spectroradiometer (MODIS) was employed for this purpose. There were seven parameters with regard to the surface heat budget were optimized, which were the daily average bulk transfer coefficients for the canopy and the bare soil, the slope of the bulk transfer coefficients according to the wind speed for the canopy and the bare soil (that was introduced to take the atmospheric stability effect into account), the evaporation efficiency for the canopy and the bare soil, and the thermal inertia of the bare soil subsurface. The simplex method was employed for an algorithm for seeking the best combination of parameters. An observed time series of the brightness temperatures at the flux station in daytime of each day was used for the temporal estimate. On the other hand, a spatial distribution of satellite brightness temperatures was used for the spatial estimate to avoid a disadvantage that no sufficient time series of brightness temperatures of MODIS was available. The surface heat fluxes were reasonably reproduced on a daily basis, with an error of the sensible and latent heat fluxes within 27 Wm-2 of root mean squares error over the growing season in the temporal estimate. A high correlation between the thermal inertia and the volumetric soil water content of a shallow subsurface was indicated as well as between the evaporation efficiencies and the soil water content. This suggested the thermal inertia could be used as an indicator of water conditions in the shallow subsurface. On the spatial estimate, a diurnal change of surface heat fluxes of the nearest grid to the flux station was validated by observation. Spatial estimates of the latent heat flux and the thermal inertia after a rainfall on succesive summer days showed an exponential decay, which was consistent with in-situ observations of the subsurface soil moisture of the shallow layer.