与混合像元的地表温度相比，植被和土壤的组分温度具有更明确的物理意义。因此，本文提出了一种从具有广泛应用的极轨卫星地表温度产品中分离出植被和土壤组分温度的算法。该算法使用温度日变化模型作为桥梁连接极轨卫星一日内的两次观测，利用多像元数据进行模型求解，从而得到过境时刻的地表植被和土壤组分温度。论文针对MODIS数据开展了地表组分温度的反演，并利用实测站点数据和高分辨率卫星数据对反演结果进行了验证。结果表明，该算法可以提供合理的植被和土壤组分温度信息，反演温度的误差变化范围为1.4 K到2.5 K。此外，对观测时刻组合方式的分析表明该算法只需要一次白天观测和一次夜晚观测就可以得到精度较好的分离结果，并且两次观测可以来自于不同传感器，进一步表明了算法具有良好的可操作性。

The component temperature encapsulates more physical meaning than Land Surface Temperature (LST) and better meets the requirements of estimating evapotranspiration, monitoring drought and other studies. The polar-orbit satellites can observe the entire globe with a high spatial resolution and a modest temporal resolution from 1980 to present, and therefore have more wide applications than geostationary satellites. For these reasons, the study focuses on the methodology for estimating vegetation and soil component temperatures from polar-orbit satellite data.To meet operational and accurate requirements, the study proposed to use multi-temporal and multi-pixel data to separate the vegetation and soil component temperature. Specifically, a well-studied Diurnal Temperature Cycles (DTC) model was applied to link the two observations on one day, and then the moving-window technology was used to add available observations for solving the retrieval model. In addition, a spatial weighting matrix was adopted to improve the limitation of using multi-pixel data.The proposed algorithm was implemented by using Moderate Resolution Imaging Spectroradiometer (MODIS) data, and was evaluated by using in-situ measurements on Skukuza site and high-resolution Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data, respectively. In the case of the validation of field data, the separation accuracy of component temperatures is about 2 K, and RMSEs of daytime vegetation, nighttime vegetation, daytime soil, and nighttime soil are 2.3 K, 2.5 K, 1.5 K and 1.9 K, respectively. The better performance at daytime is resulted from the fact that DTC model cannot describe the temperature decrease at night well. Regarding with the validation of ASTER data, the separation accuracies of the vegetation and soil component are 1.4 K and 1.7 K, respectively. The vegetation component is slightly overestimated (bias = 0.3 K) while the soil component is slightly underestimated (bias = -0.7 K), which is because of the systematic error between MODIS LST and ASTER LST. Moreover, this study also analyzed the influence of different time groups. Firstly, the combinations of one daytime moment and one nighttime moment can provide same estimation with high accuracy while the performance of the combination of two daytime moments is worse. The result is expected because two daytime moments are close to the maximum temperature moment, and therefore more sensitivity to temperature variation. Secondly, the performance of the time group from two sensors or one sensor is basically same, indicating that the time group is not limited by the sensor.This study proposed an algorithm for separating vegetation and soil component temperatures from polar-orbit satellite land surface temperatures. The practical method need only two observations from single or different sensors, i.e., one in daytime and the other one in nighttime, which makes it available for almost all sensors. The validation of field data and high-resolution data indicated that the separation accuracy is about 2 K and the best up to 1.4 K. Considering its accuracy, operationality and robustness, the proposed method would be an effective tool for separating component temperatures.