微波遥感土壤湿度误差估计与水文数据同化

发布时间：2018-01-31 21:48

本文关键词： 微波遥感土壤湿度数据同化误差估计周期性误差集合卡尔曼平滑半分布式水文模型　出处：《武汉大学》2016年博士论文　论文类型：学位论文

【摘要】：近年来,人口增长与全球变化大环境下,干旱、洪水、热浪等极端天气事件频发,气候灾害不断加剧。从全球尺度来看,地球的气候系统与水循环过程,十分庞大且极其复杂。作为大气与陆表过程耦合的重要状态变量,土壤湿度的精准时空刻画对于加深地球气候系统认知起着关键作用。随着卫星传感器与计算机技术的日臻完善,基于卫星平台的遥感手段能够探测或反演多种区域气候变量,而主被动微波遥感已被广泛应用于全球尺度的土壤湿度观测。尽管目前全球多源卫星微波土壤湿度产品能够提供长达近四十年的长时序数据集,然而数据质量参差不齐,针对多源卫星微波土壤湿度反演的误差评估与精度刻画显得十分必要。同时,观测仅能够提供瞬时真值,对于水文过程研究与预报预警而言,仍需结合水文模型。由此,本文针对卫星土壤湿度观测精度以及水文数据同化框架展开理论与方法的研究,主要内容如下：(1)以数据同化框架为主线,从观测、模型、算法三大数据同化基本要素出发,分别系统地总结当前卫星微波土壤湿度观测、水文过程数值模型、数据同化算法三个方面的国内外研究现状与基础理论方法。通过分析其各自在实际应用所存在的优势与不足,从而引出本文的研究目标与基本任务。(2)多源卫星微波遥感土壤湿度来源丰富,通过卫星升降轨能够提供当地时上午与下午两个时刻的观测,然而具体应用时往往主要选用晚间或凌晨时段所获取的观测。本文首先针对多源卫星微波遥感土壤湿度的观测精度进行研究,包括主动与被动微波来源。以观测获取时刻为基本划分,分析不同时刻地表温度与植被冠层温度差异、植被含水量等观测条件对于土壤湿度反演误差的影响。选取美国大陆为研究区域,将主动、被动微波以及陆面过程模式所分别得到的表层土壤湿度观测组合在一起,采用空间大尺度的三重组合法,分析观测时刻对于误差的影响。进一步结合地面站点量测,运用直接对比法进行交叉验证。通过上述两组独立估计方法,揭示卫星微波遥感土壤湿度观测误差与获取时刻、遥感手段、地表覆被及反演算法间的关联关系,呈现不同地表覆被条件下多源卫星土壤湿度观测精度的基本空间分布。(3)卫星数据验证与实际运用时,通常假定土壤湿度观测仅具有高斯分布的随机误差,例如水文数据同化。然而受地表覆被空间异质性、卫星周期性采样与产品处理方式等因素的影响,基于卫星的土壤湿度反演可能存在周期性误差。针对这一问题,本文从频率域角度分析被动微波土壤湿度周期性误差产生的基本物理机制。通过模拟实验、单点验证、全球评估等多角度探究周期性误差出现的原因,结合站点土壤湿度观测与高空间分辨率的土地利用数据,揭示周期性误差与地表覆被空间异质性间的关系。选取当前连续观测时段最长的被动微波土壤湿度数据作为研究对象,运用平均周期图法估计土壤湿度高频与低频的功率谱密度,采用本文所提出的高频峰值检测算法,从而获取全球周期性误差的空间分布图。根据土壤湿度反演算法的物理机制,以卫星直接获取的亮温观测为数据源,提出基于亮温观测派生参量且能够刻画地表覆被特性的空间异质性指数,从而对周期性误差可能存在的区域进行预测分析。(4)水文数据同化通过将观测与具有物理机制的水文模型相结合,能够缓解多源异质来源数据的不确定性,获取具有物理一致性、时空连续的陆表水文过程状态估计与预测。尽管目前集合卡尔曼滤波及其变种算法己在数据同化领域得到充分的实验论证,具体应用中仍存在一定问题。本文选取中国黑河上游八宝河流域作为研究区域,通过组织模型基础地理与驱动输入数据,校准水文模型相关敏感性参数,建立数据同化的模型基础。通过不同来源驱动数据,包括站点观测与天气预报模型再分析资料,构建观测系统模拟试验框架。针对集合卡尔曼平滑算法与复杂水文模型的数据同化问题,以半分布式水文模型为模式依托,引入膨胀因子与局地化改进方法,改善背景场误差协方差矩阵估计,提升集合卡尔曼平滑算法处理高维状态估计的效用。通过表层土壤湿度观测的数据同化,改进深层与根区土壤湿度以及关键水文变量的估计值。联合分析空间异质的输入数据与参数以及数据同化改进算法,揭示多重因素对数据同化效果的影响,包括降水、土壤类型以及土地利用数据等。
[Abstract]:In recent years, population growth and global change environment, droughts, floods, heat waves and other extreme weather events and frequent climate disasters intensified. From the view of global scale climate system and the water cycle process of the earth is very large and extremely complex. As an important state variable atmospheric and land surface process coupling, the precise temporal soil moisture the characterization plays a key role in the global climate system. With the deepening cognition of satellite sensor and computer technology is improving, the means of satellite platform remote sensing can detect or inversion of multiple regional climate variables based on soil moisture observation and passive microwave remote sensing has been widely used in the global scale. Despite the current global multi satellite microwave soil moisture products to provide the long time-series data for nearly forty years, however, the uneven quality of data, aiming at Multi-source Satellite Microwave soil moisture retrieval It is necessary to describe the evaluation accuracy and error. At the same time, the observation can only provide instantaneous true value for research and prediction of hydrological process, still need to combine with the hydrological model. Thus, the research of satellite soil moisture observation accuracy and hydrological data assimilation framework theory and method, the main contents are as follows: (1) to the data assimilation framework as the main line, from the observation model, three algorithms based on data assimilation of basic elements, were systematically summarized the current satellite microwave soil moisture observation, numerical model of hydrological processes, the three aspects of Data Assimilation Algorithm Research Status and basic theory methods. Through the analysis of their respective advantages and existed in the practical application which leads to insufficient, and basic tasks of the research goal of this paper. (2) multi satellite microwave remote sensing of soil moisture rich source, via satellite to provide local rail lift The observation of the morning and in the afternoon the two time, however, specific applications are often observed using the evening or early morning hours are obtained. This paper firstly according to the observation accuracy of multi satellite microwave remote sensing of soil moisture, including active and passive microwave sources. To observe the time for obtaining the basic division, to analyze the different time land surface temperature and vegetation canopy the temperature difference effect of vegetation water content observation conditions for soil moisture inversion error. Select the United States, as the study area, the active, passive microwave and land surface models were obtained from the surface soil moisture observations together with the three combination method of large scale space, analysis of the influence of observation time for the error. Combining with the ground station measured by direct comparison method of cross validation. Through the above two groups of independent estimation methods, revealing the satellite microwave Time, and access to remote sensing soil moisture observation error of remote sensing, land cover and the relationship between the inversion algorithm, different land cover distribution of space observation accuracy of soil moisture under the condition of multi satellite. (3) satellite data validation and practical application, usually assume that the soil moisture observation has only the random error of Gauss distribution for example, the hydrological data assimilation. However the land cover spatial heterogeneity, the influence factors of periodic sampling and processing satellite products such as satellite, soil moisture inversion may exist periodic error based on. To solve this problem, the basic physical mechanism based on frequency domain analysis of passive microwave soil moisture cycle error through the simulation experiment, single point verification, global assessment of multiple perspectives to explore the causes of periodic error, combined with soil moisture observations with high spatial resolution The land use data, reveal periodic error and ground cover relationship between spatial heterogeneity. Selection of passive microwave soil moisture data is currently the longest continuous observation period as the research object, using the average periodogram estimation of soil moisture in the high-frequency and low-frequency power spectral density, high frequency peak detection algorithm proposed in this paper, space in order to obtain the distribution map of global periodic error. According to the physical mechanism of soil moisture inversion algorithm, with direct access to the satellite brightness temperature observations as the data source, the brightness temperature observations derived parameters and can depict surface cover characteristics of spatial heterogeneity index based on the prediction analysis and periodic error may exist in the region. (4) hydrologic data assimilation by hydrological observation and model with physical mechanism combining multi-source and heterogeneous data sources can alleviate the uncertainty, get out Physical consistency, continuous spatio-temporal land surface hydrological process state estimation and prediction. Although Calman set filtering algorithm and its variants have been demonstrated fully in the field of data assimilation, some problems still exist in the specific application. This paper selects eight Chinese upstream of Heihe River Basin as the study area, through the organization model of geography and drive input data, calibration of hydrological model sensitivity parameters, based on the data assimilation model. Data driven by different sources, including site observation and weather forecast model reanalysis data, build observation system simulation experiment framework. According to the data assimilation problem sets Calman smoothing algorithm and the complex hydrological model, the semi distributed hydrological model for model based on the introduction of improved method of local factors and expansion, improve the background error covariance matrix estimation, improve collection card Coleman smoothing algorithm for state estimation of utility. Through data assimilation of surface soil moisture observations, improved estimates of deep soil moisture and root zone and key hydrological variable. The improved algorithm combined with analysis of spatial heterogeneity of input data and parameters and data assimilation, uncovering the influence of multiple factors, shown on the effect of data assimilation including precipitation. Soil types and land use data.

【学位授予单位】：武汉大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：S152.71;S127

【参考文献】