基于多源数据的土壤水分估算及森林火灾风险评估应用

发布时间：2018-03-22 02:08

本文选题：土壤水分　切入点：异质性地表　出处：《中国科学院大学(中国科学院遥感与数字地球研究所)》2017年博士论文　论文类型：学位论文

【摘要】：土壤水分是全球水循环中最为活跃的变量,影响着地表和大气之间的物质和能量交换,在全球水、碳循环、气候变化预测中扮演着重要角色。鉴于土壤水分重要的科研及实用价值,如何准确获取长时间的土壤水分信息成为相关研究必须面临的首要问题。微波遥感技术同时具备空间覆盖范围广、时间分辨率高、对表层土壤水分变化敏感的特点已成为大尺度土壤水分观测中的主要手段,并已有多个微波遥感卫星持续提供土壤水分产品。虽然土壤水分产品能够应用于大尺度地表水文过程研究,但同时,较低的空间分辨率(=25km)由于无法捕捉精细尺度下土壤水分的空间异质性,已成为流域及小尺度下生态水文模型的主要不确定性之一。而且近年来随着水文和农业研究的开展,土壤水分产品代表的0-5cm土壤水分分布信息已经无法满足生态水文相关预测模型对0-30cm甚至更深层的0-200cm的根区土壤水分信息的需求,获取深层土壤水分信息成为农业生态相关的领域的迫切需要。值得注意的是,不论是遥感观测的近地表、深层土壤水分产品,还是其他遥感观测技术,其最终的目的是实现社会和经济效益,服务于社会发展。虽然近地表、深层土壤水分产品以及微波观测技术已经广泛应用于多个领域,但是尚未涉及到森林灾害监测领域。如能将近地表、深层土壤水分及相关微波遥感技术应用于森林火灾监测,发挥其不易受云雨干扰,覆盖范围广的优势,对于提高森林预测模型精度,减小森林火灾造成的社会经济损失,实现遥感技术的社会经济价值均具有重要意义。随着多源地面-遥感观测技术的发展,多源观测数据为解决上述已有土壤水分产品存在的分辨率低、代表深度有限、应用潜力挖掘有限的问题提供了可能的手段,而综合有效利用多源数据的优势则是解决上述问题的关键。因此,本文基于多源多尺度对地观测数据,开展以下三方面的研究:(1)发展了考虑异质性地表的高分辨率近地表土壤水分反演算法。针对光学遥感反演高分辨率近地表土壤水分算法应用于异质性地表存在的反演精度低的问题,借助更高分辨率更小像元内部异质性的机载光学-热红外数据,对广泛应用于光学反演土壤水分的温度植被特征空间方法在异质性地表存在的植被指数饱和和干扰像元(无法表现土壤水分信息的地表类型像元)进行有效去除,实现了温度植被特征空间的合理构建,提高了高分辨率近地表土壤水分的反演精度。在改进光学遥感异质性地表土壤水分反演算法的基础上,基于贝叶斯最大熵融合框架,融合与土壤水分信息相关的多源数据,包括地面观测的灌溉统计数据、无线传感器网络观测的土壤水分数据、星载光学数据(ASTER)和机载被动微波(PLMR)土壤水分产品等,获取了更加准确的异质性地表高分辨率土壤水分分布信息。(2)发展了长时间序列深层土壤水分反演算法。基于长时间序列(1978-2014)的近地表土壤水分产品(ECV_SM),通过迭代指数滤波算法,确定算法核心参数T=15天,反演了长时间序列的深层土壤水分产品。验证结果表明反演的深层土壤水分产品与地面观测数据具有较好的一致性,可以捕捉到深层土壤水分的时空变化。(3)近地表、深层土壤水分及相关微波观测在森林火灾监测的应用潜力挖掘。开展近地表、深层土壤水分、微波极化因子(MPDI)和植被光学厚度产品(VOD)与森林火灾敏感因子(可燃物含水率,LFMC)的相关性分析。结果表明上述微波因子与可燃物含水率的敏感性依次为:近地表土壤水分深层土壤水分MPDIVOD。特别是X波段的植被光学厚度与LFMC相关性最好,对于LFMC监测具有很大潜力,为光学监测LFMC提供了重要的数据补充,也为下一步森林火灾预测模型提供了重要的数据源参考。本研究基于多源观测数据,对目前微波土壤水分产品存在的分辨率低、代表深度有限、应用领域有限等部分问题进行了初步探讨,其理论研究以及实际应用发展还有待更进一步的探索。
[Abstract]:Soil moisture is the most active variable in the global water cycle, affecting between the surface and the atmosphere exchange of material and energy in the global carbon cycle, water, and plays an important role in the prediction of climate change. In view of scientific research and practical soil moisture important value, how to obtain accurate information of soil water for long time has become the most important problem related study on microwave remote sensing technology must face. At the same time with space coverage, high temporal resolution, the characteristics of surface soil moisture variation has become a major means of large scale soil moisture observation, and has a plurality of microwave remote sensing satellite continues to provide soil moisture products. Although soil moisture products can be used in large scale surface hydrological process study, but at the same time, low spatial resolution (=25km) due to the inability to capture the spatial heterogeneity of soil moisture in fine scale, and has become a watershed One of the main uncertainty of eco hydrological model under small scale. But in recent years with the hydrological and agricultural research, 0-5cm soil moisture distribution information of soil moisture products has been unable to meet the relevant prediction of eco hydrological model and even deeper 0-200cm to the 0-30cm root zone soil moisture information needs, obtain the deep soil moisture information has become the urgent the relevant requirements of the agricultural ecological field. It is worth noting that both remote sensing observations of near surface, deep soil moisture products, or other remote sensing technology, the ultimate goal is to achieve social and economic benefits, service to the society development. Although the near surface soil moisture products and technology has been widely used in microwave observation a number of areas, but is not related to forest disaster monitoring field. Such as the near surface soil moisture, and microwave remote sensing technology Operation applied to forest fire monitoring, the less susceptible to cloud interference, covering a wide range of advantages, to improve the prediction accuracy of model forest, social and economic loss caused by forest fires, are of great significance to realize the remote sensing technology social economic value. With the development of multi - ground remote sensing technology, multi-source data to solve the soil moisture products already exist low resolution, on behalf of the limited depth of potential Mining Limited provides possible means, and effectively use of multi-source data superiority is the key to solve the above problems. Therefore, the multi-source and multi-scale earth observation data based on research carried out in the following three aspects: (1) considering the development of high resolution heterogeneous surface near surface soil moisture inversion algorithm for high resolution remote sensing inversion of near surface soil moisture algorithm applied to optical The heterogeneity of the surface inversion precision is low, with a higher resolution picture element heterogeneity of airborne optical and thermal infrared data, has been widely applied to the inversion of soil moisture optical temperature vegetation feature space heterogeneity of vegetation index method in surface saturation and interference (the type of surface pixel to pixel information of soil water the) can effectively remove the reasonable construction, realize the temperature vegetation feature space, improve the accuracy of inversion of soil moisture near the surface of high resolution optical remote sensing. Based on improving the heterogeneity of soil moisture inversion algorithm, Bayesian maximum entropy fusion framework based on multi-source data fusion and soil moisture information, including irrigation statistics the ground observation data, soil moisture observation data in wireless sensor networks, satellite optical data (ASTER) and airborne passive microwave soil (PLMR) Water products, to obtain high resolution surface heterogeneity of soil moisture distribution information more accurate. (2) the development of a long time series of deep soil moisture inversion algorithm. Based on the long time series (1978-2014) of the near surface soil moisture products (ECV_SM), through the iterative filtering algorithm to determine the core index, algorithm parameters inversion of T=15 days. The long time series of deep soil moisture products. The verification results show that the inversion of deep soil moisture products and ground observation data have good consistency, can capture the temporal and spatial variation of soil moisture. (3) the near surface, mining potential of deep soil moisture and related microwave observations in forest fire monitoring. In the near surface deep, soil water, microwave polarization factor (MPDI) and vegetation optical depth (VOD) and the forest fire sensitive factor (fuel moisture, LFMC). The correlation analysis results table The sensitivity of the microwave water content factor and fuel rate is as follows: the near surface soil moisture in deep soil moisture MPDIVOD. especially X has the best correlation band vegetation optical thickness and LFMC, has great potential for LFMC monitoring, provides an important supplement for the optical monitoring of LFMC data, also provides an important reference for the data source prediction model forest fires in the next step. This research is based on multi-source data, the microwave soil moisture products such as low resolution, on behalf of the limited depth of field limited part of the problem is discussed, its theoretical research and practical application development has to be further explored.

【学位授予单位】：中国科学院大学(中国科学院遥感与数字地球研究所)
【学位级别】：博士
【学位授予年份】：2017
【分类号】：S762;S714

【相似文献】