Antarctic ice and snow surface temperature retrieval from MODIS and Landsat8

Ice surface temperature (IST) is of utmost importance to the ice sheet radiation budget and mass balance, which has been documented by many scientific researches.

This research firstly proposes an effective approach to retrieve IST in the Antarctic area by presenting a modified split-window algorithm (SWA) and introducing a polynomial fitting for atmospheric transmittance simulation. The effectiveness was quantitatively validated by a comparative study with a Moderate Resolution Imaging Spectroradiometer (MODIS) IST product (MOD29) and automatic weather station (AWS) data from Zhongshan Station and the Ross Ice Shelf from 2004 to 2013. From the algorithm validation and data comparison, it was found that: 1) The polynomial fitting can better describe the relationship between water vapor and atmospheric transmittance, with higher determination coefficients (0.99887 for band 31 and 0.99895 for band 32, respectively) and lower residual sum of squares (0.000373 for band 31 and 0.000234 for band 32, respectively). 2) Using the Zhongshan Station data set, the retrieved ISTs by the proposed method were more accurate than the MOD29 product, with a bias of −0.61 K and a root-mean-square error (RMSE) of 1.32 K; comparatively, the bias for MOD29 was −1.33 K and the RMSE for MOD29 was 1.81 K. 3) The proposed method also obtained the highest accuracy in the other experiment using the Ross Ice Shelf data set, in which the bias and RMSE for the retrieved ISTs were −1.62 K and 2.34 K, respectively; the corresponding accuracies for MOD29 were −2.54 K and 3.04 K, respectively. Overall, it was found that the proposed method shows a robust performance in Antarctic IST retrieval for MODIS data.

Besides, an improved single-channel (SC) algorithm is proposed for retrieving the ice surface temperature of polar regions from Landsat8 band10 in this study. The improved algorithm avoids using Taylor's theorem and eliminates Taylor approximation error. In addition, the atmospheric parameters suitable for polar regions are simulated and the effective mean atmospheric temperature is added to the fitting process. In order to maintain the advantage of the SC algorithm minimum input data requirements, the effective mean atmospheric temperature is obtained by using the existing parameters and iterative calculation. The results of sensitivity analysis show that the improved algorithm is not sensitive to atmospheric water vapor content but sensitive to the calibration precision of thermal infrared sensor. Theoretical verification results show that the RMSEs of the SC algorithm and the improved SC algorithm are 0.72 K and 0.33 K, respectively. Compared with MODIS land surface temperature product, the RMSE of the improved SC algorithm is 0.31K. Compared with the temperature of automatic weather stations, in the Antarctic, the RMSE of SC algorithm is of 1.48 K and the improved SC algorithm is 1.22 K. In conclusion, the improved SC algorithm performs better than SC algorithm in polar ice surface temperature retrieval.