由于沙尘气溶胶散射和吸收机制的复杂性，目前还没有成熟的可应用于沙尘天气的大气长波辐射反演算法。通过对比分析常见的沙尘、城市、乡村、海洋气溶胶对大气长波辐射强迫和MODIS通道辐亮度影响的基础上，提出利用气溶胶光学参数对线性模型进行修正，构建了适用于沙尘气溶胶条件下的大气长波辐射估算模型，并利用黑河流域4个观测站点（花寨子荒漠站、混合林站、黑河遥感站和张掖湿地站）实测数据对模型的应用能力进行检验。结果表明：4个站点遥感反演的大气长波辐射均方误差为17.1~20.4 W·m^-2，偏差为-12.3~-1.8 W·m^-2。由于考虑了沙尘气溶胶的光学厚度变化和辐射强迫效应，修正后的模型可显著提高沙尘气溶胶影响下大气下行辐射的反演精度，减少长波辐射应用中的不确定性，对干旱区地表能量收支研究提供了重要参考。

Downward surface longwave radiation (DSLR) is indispensable for research about surface radiation balance. DSLR is one of the components of the surface radiation budget. Most of the DSLR inversion methods with remotely sensed data are on the conditions of clear sky. This method has a limited application range and their performances can be severely degraded in complex weather conditions as the effects of different aerosol types and levels are rarely considered. However, the arid and semi-arid area in northwest China is a frequent area of sandstorms, especially in spring each year, millions of tons of dust can enter the atmosphere, which not only causes significant damage to human life and production activities but also affects climate and the environment due to the physical and chemical effects of dust aerosol. It is difficult to study the optical properties of dust aerosols at home and abroad. Due to the complex mechanism of aerosol scattering and absorption, there is no mature DSLR algorithm that can be applied to dust sky, and this research is a tentative exploration in this field. For this issue, DSLR and top of the atmosphere (TOA) channel radiances under seven main land cover types are simulated to establish a simulation database. Secondly, the model coefficients are grouped by view zenith angle (VZA), water vapor content (WVC), and aerosol optical depth (AOD) to form a coefficient lookup table. Finally, based on the analysis of the effects of various four aerosol types and levels on atmospheric longwave radiative forcing and MODIS channel radiances, an improvement of the linear model using aerosol optical parameters is proposed to establish a DSLR estimation model constructed to apply to dust aerosol conditions, and four sites in Heihe River Basin, Gansu Province, China were used to verify the application ability of the improved model. Overall, the root mean square error of DSLR between the retrieved value and the site value is from 17.1 W·m^-2 to 20.4 W·m^-2, and the bias is from -12.3 W·m^-2 to -1.8 W·m^-2 for the four stations. Due to considering the dust aerosol optical depth and radiative forcing of dust aerosol, the improved model is able to significantly increase the estimating accuracy of DSLR under dust aerosol conditions and can reduce the uncertainty of DSLR in the applications of radiation. From the results, the inversion accuracy of this model can meet the requirements of instantaneous DSLR research and serve as a reference for subsequent research. However, more case studies and comprehensive analysis are needed before the model is used to estimate DSLR in dust sky actually.