新型高光谱大气臭氧传感器大多计划搭载地球静止轨道平台, 其对大视场范围和时间范围内的臭氧总量提取提出了更高的需求。 TOMS V8算法在低轨卫星大气臭氧传感器得到广泛的应用和发展, 但其在大观测几何条件下的提取精度不足, 因此如何提高TOMS算法在新型传感器上的提取精度是当前亟待解决的问题。 利用MODTRAN大气辐射传输模型模拟了晴空场景TOMS算法标准廓线各观测几何条件下的地球紫外后向散射辐射, 分析了各观测几何条件下后向散射辐射亮度与臭氧总量拟合模型, 并根据模型拟合精度情况, 得到了不同观测几何条件下后向散射辐射随臭氧总量的变化关系, 提出了改进的臭氧总量初值估算模式。 对改进模式与传统模式的臭氧总量初估结果表明, 传统模式依赖于指数模型的拟合精度, 而改进模式依赖于对数模型的拟合精度, 指数模型和对数模型均表现出高的拟合精度, 但对数模型的拟合精度比指数模型在整个臭氧浓度范围内平均高约0.9%。 改进模式使臭氧总量初估的总体精度得到改善, RMSE降低约0.087%～0.537%, 并且在较大观测几何和臭氧总量低值区（175～275 DU）间则更加显著。 在臭氧总量低值区间以及较大观测天顶角和太阳天顶角条件下, 改进模式具有更高的估算精度和更大的适用范围。 该改进的臭氧总量初值估算模式可为日后TOMS算法的更新提供支持和参考依据。

New hyperspectral sensor to detect total ozone is considered to be carried on geostationary orbit platform in the future, because local troposphere ozone pollution and diurnal variation of ozone receive more and more attention. Sensors carried on geostationary satellites frequently obtain images on the condition of larger observation angles so that it has higher requirements of total ozone retrieval on these observation geometries. TOMS V8 algorithm is developing and widely used in low orbit ozone detecting sensors, but it still lack of accuracy on big observation geometry, therefore, how to improve the accuracy of total ozone retrieval is still an urgent problem that demands immediate solution. Using moderate resolution atmospheric transmission, MODTRAN, synthetic UV backscatter radiance in the spectra region from 305 to 360 nm is simulated, which refers to clear sky, multi angles (12 solar zenith angles and view zenith angles) and 26 standard profiles, moreover, the correlation and trends between atmospheric total ozone and backward scattering of the earth UV radiation are analyzed based on the result data. According to these result data, a new modified initial total ozone estimation model in TOMS V8 algorithm is considered to be constructed in order to improve the initial total ozone estimating accuracy on big observation geometries. The analysis results about total ozone and simulated UV backscatter radiance shows: Radiance in 317.5 nm (R317.5) decreased as the total ozone rise. Under the small solar zenith Angle (SZA) and the same total ozone, R317.5 decreased with the increase of view zenith Angle (VZA) but increased on the large SZA. Comparison of two fit models shows: without the condition that both SZA and VZA are large (>80°), exponential fitting model and logarithm fitting model all show high fitting precision (R2>0.90), and precision of the two decreased as the SZA and VZA rise. In most cases, the precision of logarithm fitting mode is about 0.9% higher than exponential fitting model. With the increasing of VZA or SZA, the fitting precision gradually lower, and the fall is more in the larger VZA or SZA. In addition, the precision of fitting mode exist a plateau in the small SZA range. The modified initial total ozone estimating model (ln(I) vs. Ω) is established based on logarithm fitting mode, and compared with traditional estimating model (I vs. ln(Ω)), that shows: the RMSE of ln(I) vs. Ω and I vs. ln(Ω) all have the down trend with the rise of total ozone. In the low region of total ozone (175～275 DU), the RMSE is obvious higher than high region(425～525 DU), moreover, a RMSE peak and a trough exist in 225 and 475 DU respectively. With the increase of VZA and SZA, the RMSE of two initial estimating models are overall rise, and the upraising degree is ln(I) vs. Ω obvious with the growing of SZA and VZA. The estimating result by modified model is better than traditional model on the whole total ozone range (RMSE is 0.087%~0.537% lower than traditional model), especially on lower total ozone region and large observation geometries. Traditional estimating model relies on the precision of exponential fitting model, and modified estimating model relies on the precision of logarithm fitting model. The improvement of the estimation accuracy by modified initial total ozone estimating model expand the application range of TOMS V8 algorithm. For sensor carried on geostationary orbit platform, there is no doubt that the modified estimating model can help improve the inversion accuracy on wide spatial and time range This modified model could give support and reference to TOMS algorithm update in the future.