Knowledge Resource Center for Ecological Environment in Arid Area
| 基于MODIS数据的中国陆地气溶胶光学厚度反演流程改进 | |
| 其他题名 | The Improvement of Aerosol Optical Depth Retrieval Process based on MODIS Data over Land in China |
| 何兴伟 | |
| 出版年 | 2013 |
| 学位类型 | 硕士 |
| 导师 | 薛勇 |
| 学位授予单位 | 中国科学院大学 |
| 中文摘要 | 自从MODIS传感器在TERRA 和AQUA 的卫星平台上运行以来，已经为全球气溶胶分布提供了大量的数据产品。NASA发布了基于MODIS数据的气溶胶反演产品。MODIS数据陆地气溶胶反演的经典算法为暗像元法，但该方法只能用于较低反射率的暗像元区域，且需要气溶胶类型等先验或假设的知识，故只能反演海洋和低反射率的陆地区域。针对高反射亮地表提出的深蓝算法已应用于 MODIS、SeaWiFS 等传感器，反演得到的AOD产品同样由于其反演算法的局限性，只在亮地表地区有其反演结果，且对吸收性气溶胶考虑不足。 为了解决亮地表区域气溶胶反演的难题，课题组开发了基于TERRA和AQUA双星MODIS 数据的协同反演算法(SRAP - Synergetic Retrieval of Aerosol Properties)，用于反演陆地上空大气气溶胶的光学厚度等信息。SRAP通过双星观测，构建了气溶胶反演的闭合方程，实现了复杂地表（如城市、干旱半干旱、沙漠等）上空的气溶胶反演，有效地扩展了AOD反演的覆盖范围。 SRAP算法为高反射率下垫面气溶胶反演提供了一种新方法，但其处理流程中还有许多方面需要完善，本文主要对SRAP反演流程进行改进，其主要研究内容包括： 利用National Centers for Environmental Prediction (NCEP)再分析数据对表观反射率进行气体吸收订正，消除H2O、CO2、O3 等气体的吸收影响，研究结果表明，经过气体吸收订正的SRAP算法反演得到的AOD精度更高，2007年5月测试结果表明，改进前，回归方程为：y = 0.670x + 0.110，SRAP AOD产品与AERONET真值的相关系数R2为0.564，改进后，回归方程为：y = 0.736x + 0.153，相关系数R2可达0.654。 改进云掩膜方法，计算可见光和近红外两个波段3×3窗口的绝对标准差，判断标准差和单个像元的反射率值是否在给定的阈值内，从而判断像元是否被云覆盖，研究表明，改进的算法得到的云掩膜结果与利用MODIS云产品得到云掩膜结果相比，前者能更好的剔除被云污染的象元，提高反演精度。 在以上工作的基础上，我们生产了2002年8月-2011年近10年的AOD数据集China Collection 2.0，重点分析了2007年的AOD结果，对2007年的结果进行了月平均和季平均分析，月平均结果显示3-8月份，中国中东部地区AOD都普遍偏高，其他月份AOD值一般在0.5以下。季平均结果显示四个季节的AOD高值区域基本上都分布在中国中东部、南部工业发达地区。冬季，中国西南部地区包括西藏、新疆南部、青海南部地区被云覆盖没有AOD结果。 为了验证反演结果的精度，我们将反演结果与AERONET、CARSNET地面观测数据进行了对比验证，结果表明SRAP反演的AOD值与地面站点实测数据相关性较好，其中与AERONET对比验证的回归方程为：y = 0.6483x + 0.0963，R2为0.654，RMSE为0.1299；与CARSNET对比验证的回归方程为：y = 0.6707x + 0.0618，R2为0.828，RMSE为0.1208。此外我们还将反演结果与MODIS气溶胶产品进行了交叉验证，结果表明，SRAP反演得到的AOD与MODIS官方气溶胶相比，两者具有很好的一致性，但是SRAP算法获得的AOD值偏低，而MODIS官方气溶胶产品通过验证在中国地区偏高。 |
| 英文摘要 | Since the operation of the MODIS sensor on NASA's TERRA and AQUA satellites, it has offered a wide range of data for the global distribution of aerosols products, NSA has released MODIS aerosol product. The operationalalgorithms for land aerosol inversion used by NASA is dark target method, which is only suitable for dark area, besides it also need prior knowledge or assumptionssuch asaerosol types. The AOD product provided by NASA is only for ocean and land areas with low reflectivity. For bright surface with high reflectivity, the DeepBlue algorithm was proposed and have been applied to sensors such as MODISand SeaWiFS. Deep blue algorithm also has its limitations.It has retrieval results only in bright surface area and did not give enough consideration to absorptive aerosol. In order to solve the aerosol retrieval problem over bright land surface, the Synergetic Retrieval of Aerosol Properties algorithm (SRAP) has been developed based on the synergetic use of the MODIS data of TERRA and AQUA satellite.SRAP is used to retrieve aerosol optical depth (AOD) over land and other information. Through the binary observation, SRAP construct the closed equation, retrieve AOD over complex surface (such as city, arid, desert), effectively expand the coverage of AOD. SRAP algorithm is suitable for the aerosol retrieval of underlying surface with high reflectivity, but it still needs some improvements. This article is to improve the SRAP retrieval process, and the improvements are as follows: 1. Using the National Centers for Environmental Prediction (NCEP) data to correct the effect of gas absorption correction on apparent reflectance, and eliminate gas absorption effect of H2O, CO2, O3. The research result shows that with the gas absorption correction, the precision of the retrieved AOD is higher than those before. Test results of May 2007 show that, before improvement, the regression equation is: y = 0.670 + 0.110 x, and the correlation coefficient R2 is 0.564. After improvement, the regression equation is: y = 0.736 + 0.153 x, and the correlation coefficient R2 is up to 0.654. 2. Improved the cloud mask method by calculating the absolute standard deviation in 3 x 3 window of visible and near infrared band, and judging whether the values of the standard deviation and the reflectance of single pixel in the given threshold value, thereby to determine whether a pixel is covered by clouds. Studieshave shown that compared to the results using the MODIS cloud mask products (MOD35), the new cloud mask method can better remove cloud-contaminated pixels and improve the retrieval precision. On the basis of above work, we produced AOD dataset-China Collection 2.0, from August 2002 to 2011. We analyzed the AOD results of 2007. The results are averaged monthly and quarterly. Monthly results show that AOD are generally high in China's eastern coastal region from March to August, and AOD is not more than 0.5 in other months. Season average result shows that the higher values of AOD are mostly distributed in eastern and southern China. There is no AOD result in winter in the southwest China as it is covered by clouds at that time. Inorder to validate theretrieved results, we compared the AOD results with AERONET, CARSNET. The compared results show that the correlation coefficient is high. Compare with AERONET data, the regression function is y = 0.6483x + 0.0963, with R-squared (R2) greater than 0.6 and root mean square error (RMSE) of 0.1299 at 0.55 μm. Compare with AERONET data, the regression function is y = 0.6707x + 0.0618, with R-squared (R2) greater than 0.8 and root mean square error (RMSE) of0.1208 at 0.55 μm. Inaddition, we also compare the AOD results with MODIS aerosol products (MOD04), the cross validation shows that the two AOD have good consistency. The SRAP_AOD value is lower than MOD04 which is validated overestimated in China. |
| 中文关键词 | 气溶胶光学厚度 ; 气体吸收 ; SRAP ; 云掩膜 ; MODIS |
| 英文关键词 | Aerosol optical depth Gas absorption SRAP Cloud mask MODIS |
| 语种 | 中文 |
| 国家 | 中国 |
| 来源学科分类 | 电子与通信工程 |
| 来源机构 | 中国科学院遥感应用研究所 |
| 资源类型 | 学位论文 |
| 条目标识符 | http://119.78.100.177/qdio/handle/2XILL650/287316 |
| 推荐引用方式 GB/T 7714 | 何兴伟. 基于MODIS数据的中国陆地气溶胶光学厚度反演流程改进[D]. 中国科学院大学,2013. |
| 条目包含的文件 | 条目无相关文件。 | |||||
| 个性服务 |
| 推荐该条目 |
| 保存到收藏夹 |
| 导出为Endnote文件 |
| 谷歌学术 |
| 谷歌学术中相似的文章 |
| [何兴伟]的文章 |
| 百度学术 |
| 百度学术中相似的文章 |
| [何兴伟]的文章 |
| 必应学术 |
| 必应学术中相似的文章 |
| [何兴伟]的文章 |
| 相关权益政策 |
| 暂无数据 |
| 收藏/分享 |
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。
