Aerosols can have an important impact on human society. Remote sensing method plays an important role in the aerosol retrieval. The aerosol retrieval SARA algorithm isn’t required a lot of prior knowledge and look-up table comparing with the traditional aerosol retrieval algorithms. However, the retrieval results of SARA algorithm are affected by the acquisition time differences and instability of MODIS data and AERONET data. In order to improve SARA algorithm, this paper studies the nonlinear optimization methods to construct the three-parameter model and the five-parameter model. And used the model to optimize the input parameters to improve the accuracy of aerosol retrieval. The correlation of five-parameter optimization results with AERONET sites Beijing and XiangHe is 0.7454 and 0.701 respectively, which is better than that of three-parameter optimization model’s 0.6012, 0.6479, and both are higher than SARA’s 0.3844 and 0.526.