Abstract:  The spatial distribution of the daily average surface air temperature is vital for many hydro-ecological applications. The air temperature usually recorded at fixed-point stations provides little distribution information， and easily suffers from the scarce amount and uneven distribution of the stations in the data sparse regions. In this study，a method based on multi-source spatial data was developed to decrease the dependence on the conventional stations
observations in the daily average surface air temperature estimation， especially for data sparse regions. The method consists of two parts： step1， instantaneous surface air temperature when the satellite（Terra）passed by was retrieved using the Klemen method on the basis of Moderate Resolution Imaging Spectroradiometer（MODIS）products together with Shuttle Radar Topography Mission（STRM）products； step2， the instantaneous data from step1 was transformed to daily mean value using time scale transformation equations built up using the NCEP/NCAR re-analysis temperature data. The instantaneous surface air temperature simulations were evaluated against in situ measurement from a field test site（from August 2009 to September 2009， at Ahyz） ， and the derived daily average surface air temperature was validated using observations from Zhaosu station（2006-2012） . The results indicate as follows： （1）the derived instantaneous surface air temperature using Klemen method on the basis of muti-source spatial data show
good consistence with the field measurements， the root mean square error （RMSE） of the simulations is 2.33℃ and the R² between the derived and observed values is 0.78； （2）the R² and RMSE of the statistical equations built up for surface air temperature time scale transformation in this study is 0.98 and about 2.00 ℃ respectively，demonstrating the validity and effectiveness of this transformation method； （3）the overall R² and RMSE of the sim-
ulated daily mean surface air temperature is 0.90 and 4.63 ℃， and the simulation has a higher accuracy at high temperatures（above 0 ℃）compared with the simulations at low temperatures（below 0 ℃） ， which may result from the characteristic of the Klemen method. More work could be done to improve the simulation accuracy at low temperatures（below 0 ℃）to achieve better overall accuracy， and the re-calibration of the coefficients in Klemen method
and the adjustment of the simulation at low temperatures（below 0 ℃）from Klemen method are possible ways. The method developed in this study is designed for the spatial-temporal distribution estimation of the daily average surface air temperature， and it is useful for the reason that the spatial-temporal distribution of surface air temperature is significant in various hydrological and ecological modeling or applications. This method benefits from its independence from the station observations， and is alternative for large area daily average temperature simulation， especially data sparse regions.

Key words: air temperature, muti-source spatial data, data sparse region, time scale transformation