Abstract: Observations of reference climatological stations plays an important role in climate change research and disaster warning, which requires the stations to be representative enough for their surroundings. Furthermore, representativeness is important in optimizing meteorological observation network and selecting locations for new meteorological stations. However, researches on meteorological stations representativeness are still limited, especially from regional point of view.A new method to quantify reference climatological stations representativeness based on remote sensing data is proposed. The representativeness of a reference climatological station is measured using its explained variances for the LST (land surface temperature) anomaly series extracted from different sizes of windows centered on that station. MODIS/LST products (MOD11A2, 1 km spatial resolution) from 2001 to 2007 are used. MOD11A2 is 8-day average composite of MODIS daily LST product (MOD11_L2) which is produced using split-window algorithm. The product accuracy is better than 1 km according to previous validation, providing quality guarantee. The selected window size ranges from 3×3 km² to 51×51 km², with a step of 2 km. For each window size, explained variances of all the 142 national reference climatological stations are calculated. Through investigating the trend of explained variances with window size increments, a threshold is selected, according to which the maximum area a station can represent is determined. When 0.75 is set as the threshold, about 41% stations have good representativeness, representing areas larger than 51×51 km² which are mainly located in the north regions. About 21% stations have low representativeness, representing areas less than 7×7 km² that mainly located in the south regions. Other stations with moderate representativeness are found to distribute in both north and south regions. In order to explore the factors influencing representativeness, two indices are defined. One is land cover diversity based on Shannon-Weiner index formula and retrieved from MODIS land cover product. The other one is terrain undulation, which is defined as the difference between the maximum and minimum elevation and retrieved from DEM data. It is found that as far as all reference climatological stations are concerned, land cover diversity and terrain undulation are negatively correlated with representativeness, and when window size increases this correlation strengthens accordingly. Land cover diversity has greater impacts on representativeness than terrain undulation. Using land cover diversity and terrain undulation as independent variables, linear regressions can model representativeness pattern of most stations very well. For several stations whose representativeness cannot be explained well, fast urban expansion maybe an important factor, which needs further research. Finally, the representativeness for climate zone to which a station belongs is also studied. The results reveal that in areas featured with complicated climate, such as southwest region and part of Xinjiang, the representativeness of reference climatological stations are low, suggesting more meteorological stations are needed.