大气校正已广泛应用于区域生态植被的动态监测, 但是不同校正方法和模型对遥感影像光谱和森林碳储量估算结果的影响不得而知, 同时这种差异在多时相遥感监测与应用时经常被忽略。 以多期Landsat影像为数据源, 借助植被指数MNDVI和野外实地调查的马尾松林样方数据, 进行马尾松林碳储量反演。 然后采用几种常用的大气校正算法: 6S, FLAASH(fast line-sight atmospheric), IACM(illumination and atmospheric)和QUAC(quick atmospheric correction), 并结合地面同步实测的光谱数据, 以评估其对马尾松冠层光谱曲线、 植被指数以及林分碳储量估算的影响; 同时从遥感动态监测角度出发, 分析了相对大气校正（pseudo-invariant feature, PIF）对多时相影像植被指数与碳储量反演结果的校正效果。 结果表明, 经大气校正后的影像波段反射率与实测光谱结果较为接近, 其中近红外和短波红外波段光谱反射率明显上升, 同时可见光波段减弱, NDVI(normalized difference vegetation index)增加明显。 不同大气校正模型对研究区马尾松林碳储量的遥感反演结果影响较大, 其中IACM与6S模型分别具有较高的精度和较低的误差。 此外, 经过PIF校正后不同时相影像的NDVI相对偏差降低了85.16%, 同时马尾松林碳储量反演模型精度得到明显提升, 表明辐射归一化处理对于多时相遥感影像的应用十分必要。 研究发现ICAM与PIF的大气校正模型组合可较好纠正大气效应, 适用于多时相遥感数据的森林碳储量反演与监测研究。

The atmospheric conditions will lead to the distortion of the ground radiance or reflectance recorded by satellite sensors which inevitably hampers the successful regional scale aboveground carbon density quantification which is critical to the understanding of forest contribution to the regional carbon cycles. Hence, the appropriate algorithms of atmospheric correction are necessary. The objective of this paper was to assess the utility of radiation correction algorithms for estimating aboveground forest carbon storage with the multi-temporal Landsat remote images(TM/OLI) and quantify the carbon storage in a forest plantation, and five atmospheric correction methods, two absolute modeling methods (6S, FLAASH), two absolute image-based methods (IACM, QUAC), and one relative method (PIF) were compared. Forest carbon storage was estimated using the viable biomass empirical statistical models. Parameters for the regression equation were determined by analyzing the relationship between the data of the selected vegetation indices derived from Landsat images and the field-measured data (height and diameter) using data from different tree stands in study area. In order to evaluate the accuracy of the carbon storage of Pinus massoniana forest derived from multi-temporal remote sensing images, we acquired the forest subcompartment survey data in 1997, 2002 and 2006 and conducted several field surveys in 2010 and 2013. Consequently, we found that the surface reflectance of Pinus massoniana forest decreased evidently after atmospheric correction in visible band, but the surface reflectance in nearinfrared band and shortwave-infrared band, as well as NDVIs had a significant increase. And different atmospheric correction models had significant different effects on the estimation of the carbon storage of Pinus massoniana forest. By studying the correlation between the field-measured data, the IACM-corrected and 6S-corrected MNDVI data were most suitable for estimating the carbon storage of Pinus massoniana forest in the study area, with an exponential regression model appeared to have the highest degree of agreement and the lowest relative error with the measured data. In addition, we also found that the relative error of NDVIs of Pinus massoniana forest in multi-temporal remote sensing images decreasd 85.16% after PIF correction, and the estimation accuracy of the forest carbon storage was improved simultaneously. The results suggested that more attentions should be paid to choose the appropriate atmospheric correction when remote sensing images were applied to quantitative analyzing and information collecting in field. And relative radiometric correction of remote sensing images is quite an important preprocessing technique, which can significantly improve the precision of the estimated results, especially in multi-temporal remote sensing monitoring. This study also confirmed that the combination of IACM and PIF could preferably reduce the atmosphere effect, which would be suitable for multi-temporal forest carbon storage estimation and other related quantitative remote sensing researches.