积雪覆盖时空建模分析及融雪径流模拟

发布时间：2018-06-13 06:28

本文选题：积雪覆盖 + 隐马尔科夫随机场　；参考：《华东师范大学》2016年博士论文

【摘要】：积雪是地表覆盖的重要组成部分。季节性积雪的累积和消融不仅影响了整个区域的气候和水资源平衡,而且能够进一步影响全球的能量平衡和气候变化。因此,获取准确的积雪覆盖信息、分析积雪覆盖的时空动态变化,以及进行融雪径流建模对于水资源管理和气候变化研究具有非常重要的科学意义。MODIS积雪覆盖产品被广泛地应用于区域性积雪覆盖信息提取和水文建模分析。但是作为光学传感器,MODIS受到了云层的严重影响,导致其积雪产品,特别是逐日积雪产品出现了大量的数据空缺。此外,MODIS积雪覆盖产品在森林地区、地形复杂的山区,以及积雪覆盖较薄和暂时性积雪时期精度较低。季节性积雪覆盖的时空动态变化能够揭示区域性气候变化和水资源平衡。为了监测和分析季节性积雪覆盖的时空变化,之前的研究通常采用基于像元的统计方法,从时间序列积雪遥感影像中获取积雪覆盖的空间变化和季节性时长。但是这种基于像元的方法不能反映积雪场(Snowpack)之间的时空演化关系,同时也无法正确地获取到暂时性积雪现象,会造成季节性时长信息偏差。融雪径流模型(Snowmelt RunoffModel, SRM)是目前在模拟和预测融雪径流方面应用最广泛的模型之一并且以卫星影像上获取的逐日积雪覆盖作为其中的输入参数。由于标准的MODIS积雪覆盖产品受到了云层的严重影响,之前的研究通常采用积雪消融曲线重建的方法获取逐日积雪覆盖信息。目前尚缺乏不同积雪覆盖参数对SRM模拟的径流差异影响的相关研究。针对现有研究的不足,本论文基于隐马尔科夫随机场(Hidden Markov Random Field, HMRF)时空建模技术改进了MODIS积雪覆盖产品,提出了一种面向对象的积雪累积和消融过程时空分析方法,以及基于SRM模拟的融雪径流结果分析了融雪季积雪覆盖差异导致的模拟径流差异。本研究的主要内容以及结论如下：(1)基于HMRF时空建模技术改进MODIS积雪覆盖产品。本研究基于时间序列MODIS积雪覆盖产品,通过时空建模的方法将光谱信息、时空背景信息和环境相关信息以最优的组合方式融入HMRF分析框架,生成完全没有数据空缺的逐日积雪覆盖产品。改进后的积雪覆盖产品通过与美国Rio Grande流域2006-2007积雪季的33个SNOwpack TELemetry (SNOTEL)站点的实测观测数据以及原始MODIS积雪覆盖产品对比进行精度评价。原始的Terra和Aqua MODIS合成积雪覆盖由于受到云层遮挡,在整个积雪季有高达34.1%的数据空缺。对于原MODIS标准积雪产品中的数据空缺,本研究填补以后的积雪覆盖总体精度为88.1%；对于MODIS标准积雪产品中无数据空缺的区域,改进后的积雪覆盖总体精度从原始的85.3%提高到88.6%,高于MODIS标准积雪产品3.3%。在HMRF分析框架中逐步加入时空背景信息和环境相关信息后,数据空缺的填补能力和积雪覆盖产品的总体精度都逐渐提高。原始MODIS积雪覆盖产品在积雪转化时期和森林覆盖地区具有相对较低的精度。结果表明,本研究基于HMRF的方法能够在整个积雪转化时期提高原始MODIS积雪覆盖产品4.2%的精度,其中在3月积雪消融时期能够提高5.8%。此外,原始MODIS积雪覆盖产品在常绿森林和混合森林地区的精度也得到了明显的提升。(2)提出了一种基于面向对象的积雪累积和消融过程时空分析框架。积雪场的累积和消融是一个复杂和动态的地理过程。本研究将积雪场定义为一种时空场对象(field-object),并首次提出了一种面向对象的积雪累积和消融过程时空分析框架。这个分析框架根据积雪场在时空由低到高的聚合等级,依次分为了积雪区(Snow Zone)、积雪序歹(Snow Sequences)和积雪过程(Snow Processes)三个层次。积雪区是从积雪专题遥感影像中获取得到的空间连续积雪像元的区域；积雪序列是指具有时空连续性的一系列积雪区的集合；积雪过程是具有时空相关性的一系列积雪序列的集合。本研究采用了面向对象的方法组织和存储积雪场在这三个层次的专题属性、空间属性、时间属性以及专题关系、空间关系和时间关系。本研究以美国Upper Rio Grande流域为实例样区,分析了2006-2007积雪季积雪累积和消融的过程。结果表明本研究提出的面向对象的分析方法不仅能够表达每一个积雪场在其生命周期中累积和消融的时空过程,还能够揭示积雪场内部和积雪场之间的时空演化关系。(3)基于SRM模拟的融雪径流结果分析了融雪季积雪覆盖差异导致的模拟径流差异。本研究分别采用基于HMRF时空建模技术改进后的逐日积雪覆盖和对原始MODIS积雪产品进行积雪消融曲线重建拟合得到的逐日积雪覆盖作为SRM的积雪覆盖输入参数,对Rio Grande Headwater流域2007年融雪季进行融雪径流建模,并分析了积雪覆盖差异导致的模拟径流差异。两个积雪覆盖产品产生的积雪消融曲线在整个融雪期的总体变化趋势较相似,其中利用HMRF时空建模技术改进后的逐日积雪覆盖可以捕捉到更多的积雪覆盖时空变化细节。基于两个积雪覆盖产品模拟的径流量与实测径流量之间都具有很好的相关关系,其中利用HMRF方法改进后的积雪覆盖模拟得到的径流精度明显高于MODIS积雪覆盖产品模拟的径流精度。本研究结果表明,在融雪季节的不同时期,由积雪覆盖差异导致的模拟径流差异不同。在融雪季的前期,由于空气温度仍然较低,导致低温条件下由积雪融化形成的径流量也低,因此积雪覆盖率的差异并不会对模拟的径流造成明显的差异；在融雪季的中后期,由于空气温度的回升,使得径流量对于融雪贡献径流的敏感性增强,因此积雪覆盖率的差异能够直接导致模拟径流的差异。
[Abstract]:Snow is an important part of surface cover. The accumulation and ablation of seasonal snow not only affects the climate and water balance of the whole region, but also can further influence the global energy balance and climate change. Therefore, the accurate snow cover information is obtained, the spatial and temporal changes of snow cover are analyzed, and the snowmelt diameter is analyzed. Flow modeling has a very important scientific significance for water resource management and climate change research..MODIS snow covered products are widely used for regional snow cover information extraction and hydrological modeling analysis. But as optical sensors, MODIS is seriously affected by the cloud layer, leading to its snow products, especially day by day snow products. There are a large number of data vacancies. In addition, MODIS snow cover products in forest areas, complex terrain in mountainous areas, and low snow cover and temporary snow time. Seasonal snow cover time and space dynamic changes can reveal regional climate change and water balance. In order to monitor and analyze seasonal snow cover Spatiotemporal change, previous studies usually use a pixel based statistical method to obtain the spatial variation and seasonal length of snow cover from the snow remote sensing images of time series, but this pixel based method can not reflect the spatio-temporal evolution of the snow field (Snowpack), and can not obtain the temporary snow now correctly. The snow melt runoff model (Snowmelt RunoffModel, SRM) is one of the most widely used models for simulating and predicting snowmelt runoff and the daily snow cover obtained on satellite images as input parameters. The standard MODIS snow cover products are severely affected by the cloud layer. The previous study usually uses the method of snow ablation curve reconstruction to obtain daily snow cover information. There is still a lack of research on the influence of different snow cover parameters on the runoff difference of SRM simulation. In view of the shortage of existing research, this paper is based on the construction of Hidden Markov Random Field (HMRF). Model technology improved MODIS snow cover products, proposed a spatio-temporal analysis method of object oriented snow accumulation and ablation process, and the results of snowmelt runoff based on SRM simulation analysis of simulated runoff differences caused by snow cover difference in snowmelt season. The main contents of this study are as follows: (1) based on HMRF spatio-temporal modeling techniques In this study, MODIS snow covered products were improved. Based on the time series MODIS snow cover products, the spectral information, spatio-temporal background information and environmental related information were integrated into the HMRF analysis framework by the time space modeling method to generate a day by day snow covered product without data vacancy. The improved snow cover production was made. The accuracy is evaluated by comparison with measured observations of 33 SNOwpack TELemetry (SNOTEL) sites in the 2006-2007 snow season of the Rio Grande Valley in the United States and the original MODIS snow covered products. The original Terra and Aqua MODIS synthetic snow cover is covered by cloud layer and up to 34.1% of the data vacancy in the whole snowpack season. The total precision of snow cover in the original MODIS standard snow product is 88.1%. For the area with no data vacancy in MODIS standard snow products, the improved snow cover overall precision is increased from original 85.3% to 88.6%, which is higher than that of the MODIS standard snow product 3.3%. in the HMRF analysis framework. After the time and space background information and the environment related information, the filling ability of the data vacancy and the overall precision of the snow covered products have been improved gradually. The original MODIS snow covered products have relatively low precision in the period of snow conversion and the forest covered areas. The results show that the method based on HMRF can be used in the period of the whole snow transformation. The precision of 4.2% of the MODIS snow covered products at the beginning of the plateau, which can improve the 5.8%. in the period of the snow melting in March, is also improved. The precision of the original MODIS snow covered products in the evergreen forest and the mixed forest area has also been significantly improved. (2) a spatio-temporal analysis framework based on the object oriented snow accumulation and ablation process is proposed. The accumulation and ablation is a complex and dynamic geographical process. The snow field is defined as a spatio-temporal field object (field-object), and a spatio-temporal analysis framework for the accumulation and ablation process of an object oriented snow accumulation is proposed for the first time. This framework is divided into the accumulated snow field from low to high aggregation level in the time space. The snow area (Snow Zone), the snow sequence (Snow Sequences) and the snow process (Snow Processes) are three levels. The snow area is the area of the spatial continuous snow pixel obtained from the snow thematic remote sensing images; the snow sequence is a collection of a series of snow areas with space-time continuity; the snow process is a spatio-temporal correlation. This study uses an object-oriented method to organize and store the thematic attributes of the snow field at these three levels, spatial attributes, time attributes, and thematic relationships, spatial relationships and time relations. This study takes the Upper Rio Grande basin in the United States as an example area to analyze the accumulation and elimination of snow in the 2006-2007 snow season. The results show that the object-oriented analysis method proposed by this study not only can express the spatio-temporal process of each snow field accumulated and ablation in its life cycle, but also can reveal the spatio-temporal evolution relationship between the snow field and the snow field. (3) snow melting season based on the SRM simulated snowmelt runoff results in snow melting season In this study, the daily snow cover based on HMRF space-time modeling technology improved and the snow cover of the original MODIS snow products were used as the input parameters for the snow cover of SRM, and the snow melting season of the Rio Grande Headwater Valley in 2007 was snowmelted. Modeling of runoff, and analysis of simulated runoff differences caused by snow cover differences. The overall change trend of snow melting curve produced by two snow covered products is similar in the whole snowmelt period, and more details of spatio-temporal change of snow cover can be captured by the improved snow cover of HMRF spatio-temporal modeling technology. Based on two There is a good correlation between the simulated runoff of a snow covered product and the measured runoff, and the runoff precision obtained by the improved snow cover simulation using the HMRF method is significantly higher than that of the MODIS snow covered product. The results show that the snow cover difference in the different periods of the snowmelt season In the early period of the snowmelt season, because the air temperature is still low, the flow of snow melting under low temperature is also low, so the difference of snow cover will not cause significant difference to the simulated runoff; in the middle and late period of the snowmelt season, due to the rise of air temperature, the runoff is made. The sensitivity of snowmelt to runoff is enhanced, so the difference of snow cover can directly lead to the difference of simulated runoff.
【学位授予单位】：华东师范大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：P426.635

【参考文献】