Abstract: As an important parameter forthe balance of water and energy in land surface and the atmosphere,the land surface temperature (LST)plays an important role in the study of global climate change. Visible Infrared Imaging Radiometer Suite (VIIRS)is a significant instrument equipped on the new generation earth observation satellite south-NPP. Compared to other sensors,VIIRS has a higher spatial and temporal resolution. It is fullydeveloped and can be used to retrieve the land surface temperature in a variety of methods. The split window algorithm, which takes the advantage oftwo thermal infrared channels,is the most commonly method used for retrieving LST. After years of researching and improving,a variety of split-window algorithmsare developed. The split window algorithms are derived from radiative transfer equation in thermal infrared region. The coefficients of the split window algorithms are determined in terms of the simulation data from MODTRAN5.2,the atmospheric radiation transfer software. The surface emissivity in the split window channels isestimated by using the NDVI-based threshold method. The channels M15 and M16 of VIIRS are used to calculatethe radiance and the brightness temperature of the atmosphere,and then the visible channel I1 and I2 are used to calculate NDVI for estimating land surface emissivity. Multiple forms of split window algorithms are used to acquire LST and then verified with thefield data of observation stations in Hailar field site. Those algorithms are compared and the one with the highest precision will be selected to improve LST retrieval accuracy for VIIRS. The error of the split window algorithm for retrieving land surface temperature is mainly caused by the instrument noise,surface emissivity and atmospheric properties,those uncertainty input factors will affect the accuracy of the LST retrieval. Through sensitivity analysis,the influence of these factors on the accuracy of the algorithm is determined. The total error of each algorithm is almost 0.75-1.5 K,the largest error is 1.42 K from the PR84 algorithm,and the minimum error of SR00 algorithm is 0.75 K. The land surface emissivity has a greater impact on the split window algorithm. The land surface temperature retrieves by different algorithms,and through the analysis and the verification of each algorithm, it can be seen that the algorithm containing two items has more accurate,and the algorithm accuracy can be improved if the algorithm contains the item of water vapor.The final result shows that the maximum error ofthe algorithm PR84 reaches to1.84 K,the maximum RMSE of the Surface temperatureretrieved by other algorithms is below 1.5 K. the algorithm BL95 has the highest accuracy about 1.23 K,which is more suitable for retrieving land surface temperature with the VIIRS thermal infrared data.

Key words: land surface temperature, split window algorithm, VIIRS