伴随着全球化进程的加快，极端气候事件不断增多，社会暴露性逐渐增强，自然灾害造成的经济损失日益严重，已经成为制约地区乃至国家社会经济可持续发展的障碍，准确评估灾害造成的经济损失对灾害风险管理决策显得尤为重要。从区域灾害系统理论的角度来看，致灾因子的强度和频次具有区域差异性，而承灾体（人口、基础设施、社会经济等方面）的空间分布也存在着差异，因此灾害造成的损失在空间上也呈现出分异特征。承灾体相关基础数据的质量直接影响灾害损失评估的结果，因此，高精度的社会经济空间网格数据对于灾害经济损失空间估计精度的提高具有重要意义。目前社会经济数据的空间化主要是利用线性回归模型来实现，集中于寻找社会经济要素与具有空间属性的自然要素的关系，而这种关系通常不明显，往往很难鉴定，所以在地理学界和经济学界尚没有明确的定量描述社会经济空间分布的方法，也没有成熟或公认的空间化数学模型。同时，可利用的社会经济空间统计数据空间分辨率低在很大程度上制约了社会经济要素空间分析的深入，因此，通过模型模拟社会经济活动空间分布是弥补空间统计数据缺乏的一个重要途径。自然灾害的经济损失可以分为直接经济损失和间接经济损失，和直接经济损失一样，灾害造成的间接经济损失也正在逐渐引起人们的重视，相比于直接经济损失，间接经济损失的影响可能会更深远。而目前间接经济损失评估模型主要是基于行政单元的经济数据进行模拟的，同时需要分部门的直接经济损失数据作为输入，而仅通过行政单元的经济数据直接去量化灾害损失的空间分布是远远不够的，空间分辨率低。因此，如果能够获得空间网格上的分部门经济数据，对于提高灾害直接经济损失评估精度，以及合理评估相应的间接经济损失具有重要意义。空间Copula地统计模型（Spatial Copula model，SCM）在空间化建模时不需要对样本数据作正态分布的假设，可以模拟在空间上具有多样性、连续性以及异质性等特征的随机变量的分布，通过Copula函数来描述空间相依性，且样本变换对Copula函数没有影响。目前该模型的应用主要集中于地质、水文、气象等自然要素方面，而在社会经济要素方面的应用甚少。因此，本研究尝试将基于地统计学理论的SCM引入到社会经济要素的空间模拟中，充分考虑社会经济要素的空间连续性、依赖性和异质性，以及随机性等特征，结合等级转换理论、空间分析理论，以及地统计相关理论，实现了社会经济要素的空间化模拟，为社会经济空间化研究提供一个新的研究思路和方法。湖南省是我国受自然灾害影响最为严重的省份之一，洪涝灾害尤为严重，因此，本文选择湖南省洪涝灾害作为案例，进行经济部门产值空间化研究，为灾害损失评估提供有效的基础数据。为了能够与灾害间接经济损失评估中的产业部门相一致，选择了部门产值作为空间化研究的对象。由于交通运输业对自然灾害的影响比较敏感，灾害引起的交通“生命线”中断，往往会造成严重的直接损失和波及效应，考虑到研究对象的典型性和数据的可获取性，本研究选择了交通运输邮电业（简称“交通业”）为代表，进行经济部门产值空间化方法的探讨。经济部门之间存在着复杂的非线性关系，为了实现经济部门产值空间化，首先采用非线性非参数的分类与回归树（CART）模型来客观确定交通业部门内部各经济要素之间的定量关系；进而鉴于产值密度空间分布的随机性、连续性、相关性、异质性等特征，采用SCM实现产值密度的空间估计，使估计值能在一定的置信区间内更加接近实际分布，从而为基于空间网格的灾害经济损失评估提供数据支撑。主要研究内容可以概括为以下几个方面：1) 针对交通业部门内部要素的非线性关系，构建了部门要素（铁路、公路、航空、水运及邮电）空间化的指标体系，利用湖南省1978-2010年部门各要素与产值相关的时间序列数据，采用CART模型确定了部门内部各要素在总产值中的比重。模型结果显示，铁路客运和公路客运对总产值的影响最大,各要素的权重在18%~22%之间波动，其中，公路运输的比重最大（21.42%），邮电业的比重相对最小（18.31%），最终模型的解释率达到95.9%。将由CART模型求算的部门要素的权重与相应规则网格内的部门数量相结合，求算网格中心点的产值密度值，并将其作为经济部门产值空间化的基础数据。2) 针对目前对社会经济空间化研究的需要和已有研究方法的不足，在空间分析理论、尺度转换理论，以及地统计学的基本理论的基础上，引入了空间Copula模型，对空间上未知点的值进行估计，实现产值的空间化。分别采用空间高斯Copula模型（Spatial Gaussian Copula Model，SGCM）和空间卡方Copula模型（Spatial Chi-square Copula Model，SCCM）对交通业产值进行了空间估计。结果显示，两模型估计的产值密度及其相应的标准差在空间上具有很好的一致性。从留一法（leave-one-out method）交叉验证的平均标准误差和平均绝对误差可以看出，SCCM的估计精度略高于SGCM。通过交叉验证计算估计值的中心中值的概率间隔覆盖范围来量化模型估计值的置信区间，同样发现SCCM的估计结果更接近于实际，不确定性相对较小。从标准残差统计值来看，在设定的97.5%置信区间内，SCCM（0.005）的估计精度要高于SGCM（-0.028）。3) 不同空间化方法与不同空间尺度下的空间化结果精度比较。本研究进一步选择了常用的地统计插值方法—普通克里格法（Ordinary Kriging，OK）进行了产值的空间化，并将所得结果与SCM结果进行比较，结果显示不同方法的估计结果在空间上的总体分布趋势是一致的，但OK模型的估计值分布相对离散，相应的估计值标准差也较大。根据空间化的产值密度，所得的交通业产值在空间上成同心圆状分布，且估计值由长沙-株洲-湘潭三市向周围逐渐递减，但又存在一定的差异，SGCM和SCCM的空间估计结果比较接近，都比较平滑，OK模型的结果则比较离散，其中，SCCM的产值空间分布最为平滑。长沙市市区、湘潭市的东北部、以及株洲市的西北部是全省最高值所在，低值出现在湖南省的边缘地区，包括岳阳市的东南角、浏阳市的东北部、湖南省西北部（湘西州、张家界）、怀化市的北部地区和西部地区、邵阳市的西部区域、永州的西南部，以及郴州的东南部。这些区域的共同特点是均为山地丘陵地区，交通运输业的发展受到不同程度的限制，因而产值相对较低。采用不同空间尺度的方法进一步比较空间估计的精度，将湖南省交通业产值密度图按市级行政区划边界作进一步的分割，在更小区域范围内比较估计值与统计值的差值。结果显示，在各地级市的估计值与统计值比较接近，但均比实际统计值偏低，其中SCCM估计的结果最接近于统计值，SGCM次之，OK模型的估计结果相对较差。说明基于省域尺度的产值空间分布结果是可信的，且SCM模型模拟社会经济空间分布的效果相对较好。4) 产值空间化结果在洪涝灾害直接经济损失评估中的应用。将空间化的湖南省交通业产值应用于洪灾损失评估中，设置不同洪水强度情景，分区模拟洪水的不同淹没范围。结果显示：常德和益阳的交通业产值受洪灾影响最大，随着淹没范围的扩大，损失增加迅速，其次是湘潭市、株洲市和永州市。地形变化较大的区域，随着淹没深度的增加，淹没面积变化不大，甚至没有显著变化，如怀化市、湘西自治州。湖南省内其他区域的淹没面积也随着水深的增加而扩大，但不是十分显著。从总体来看，交通业损失随水深增加的变化规律为：淹没深度在0-1m时，交通业损失增加迅速，损失值在0-200亿元（2011年价格）范围内；1-1.5m时，损失变化不大，基本保持不变；1.5-2m时，损失又进一步增加，除常德、益阳、湘潭外，增幅均不及0-1m时的情景。水深超过2m，除长沙市之外，损失增加不显著。根据不同的淹没情景，结合经济增长指数换算，可以快速估计出各区域在不同淹没程度下交通业的产值损失。本论文将CART模型和SCM引入经济部门产值空间化研究中，在充分考虑社会经济产值的空间连续性、依赖性和异质性，以及随机性等特征的基础上，结合等级转换理论、空间分析理论以及地统计的基础理论，实现了以交通运输邮电业为代表的经济部门产值空间化，为灾害损失评估精度的提高奠定了理论和方法基础。空间化的结果可以用作社会经济要素的空间分析，为灾害的直接和间接损失评估提供了数据支撑，同时也是灾害风险管理决策的重要参考依据。

With the rapid globalization, the extreme weather events and the social exposure continue to increase, and the economic loss caused by the natural disasters is becoming more severe simultaneously, which has become the obstacle for the sustainable development of social economy. Thus the more accurate quantification of the disaster loss is especially important for the disaster risk management. From the perspective of regional disaset system theory, there are great differences of the hazard intensity and frequency in space, and the spatial disparity of the hazard-affected body (population, infrastructure, social economy, et al.), the corresponding economic loss from disasters appears different in space. The result of the disaster loss evaluation is influenced directly by the quanlity of base data about the hazard-affected body. Thus high resolution grid data of the social-economic is necessary for the accuracy improvement of loss assessment. At present the linear regression models are the main methods for the spatilization of socioeconomic, and the relationships between social-economic elements and physical features with geographic properties are the mian concerns, but usually the relations are unapparent between them for the complexity. Therefore, there are no specific simulations or mathematical models about the spatialization of economic elements in geography and economics at present. Meanwhile, lack of available socioeconomic data is also the reason that restricts the further research about the spatial analysis of the socioeconomic. Thus the spatial simulation of the social economic activities by models is an alternative way to make up the insufficiency of spatial statistical data.The economic loss of natural disaster includes the direct and indirect economic loss, and the indirect economic loss has gradually cause people’s attention same as the direct loss for the impact maybe even far beyond. At present the simulation of the industry relatedness in the indirect loss assessment is based on the administrative unit, and the direct economic loss data with different sectors are also the input parameters. While the spatial distribution simulation based on the administrative unit data is not enough for the accuracy requirement, and the spatial resolution is low. Thus obtaining the spatial grids data about the economic sectors will be great significant for the accuracy improvement of the economic loss that with spatial disparity properties.Spatial Copula Model (SCM) can simulate the distribution of the random variable with characteristics of spatial diversity, continuity, heterogeneity without the asumption of Gaussian distribution. And the model can simulate the spatial dependence by copula fuctions, which is not affected by data transformation. The SCM has been only applied in geology, hydrology, and meteorology by now, while the application in the spatial socioeconomic elements is rare. Thus the SCM is introduced in this study, which considers the spatial properties based on the theories of rating convert, spatial analysis theory, and basic theories of geostatistics to realize the spatialization of the socioeconomic elements and provide a new ideal and approach for the socioeconomic spatialization.Hunan is one of the most serious influenced provinces by natural disasters in China. Especially floods occur frequently. For these reasons, flood in Hunan was selected as the study case. To be consistent with industry sectors of disaster indirect loss assessment, the output of the economic sector was used as the object for the spatialization. The economic sector of the transportation is one of the most sensitive industries that affected greatly by the disaster, and which was chosen as the representative for the spatialization. The transportation sector is often combined with the post and telecommunication sector, with which related closely, and considering the typicality of the research object and the availability of data, the transportation, post and telecommunication sector (abbreviated as transportation) was selected for spatialization.There are complicated nonlinear relationships between the elements of economic sectors, the classification and regression tree model was used to quantified the weight of each element occupied the total value. For the properties of output density with the randomness, continuity, heterogeneity, the SCM was introduced to implement the spatial estimation, and which enable the estimated values to close to the actual distribution. The spatialized result is the data base for the disaster economic loss assessment with the spatial grid resolution. The main contents can be summarized as follows: 1) Considered the interactions and relationships between the various elements (railway, highway, aviation, water transport, post and telecommunications) within the department of Transportation, constructed the index system of the output spatialization, and determined the weight of each element on the total industrial output based on the time series data from 1978 to 2010 of Hunan province using the nonlinear and nonparametric model of the classification and regression Trees (CART). The model results show that the output value of the railway and road passenger transport make the greatest contribution to the total output of the transportation. The weights of the five elements fluctuated from 18% to 22%, where the road transport occupied the largest proportion (21.42%) and the post and telecommunications is the relatively minimum (18.31%). The final optimal model explained 95.9%. The output density on each grid was the product of weights from the CART and the corresponding number of departments. The output density in each grid center was used as the basic data for the spatialization.2) For the needs of spatialization research of socioeconomic and against the shortages of existing methods, SCM (Spatial Gaussian Copula Model，SGCM; Spatial Chi-square Copula Model，SCCM) was introduced to estimate the output value and corresponding standard deviation to achieve the spatialization of output value based on the theories of rating convert, spatial analysis, and basic theories of geostatistics. The estimated of the two models (SGCM and SCCM) results are in good agreement with each other in space, and the output value kept a high consistency with the corresponding standard deviation in space. The average standard error and the average absolute error of cross-validation through the leave-one-out method show that the estimated accuracy of the SCCM is slightly higher than the SGCM. The center median probability interval coverage calculated by the cross-validation were used to quantified the confidence interval of the estimated value, and the results also show that the estimated values by the SCCM are more closer to the actual truth, and the uncertainty is relatively small. The standard residual errors also show the estimation accuracy of the SCCM(0.005) is higher than the SGCM (-0.028) under the 97.5% confidence interval.3) The comparisons of different approaches and spatial scales of spatialization results. In this study, the traditional geostatistical method - Ordinary Kriging was also selected to spatialize the output on the basis of the same data, and the spatialization result was compared with the SCM. The results show the overall trend of the estimated values in space is the same, yet the distribution of the estimated values is relatively more discrete, and the corresponding standard deviation is larger. According to the spatialized output density by the three models, the transportation output value is assigned to the space on the regular grids. It can be seen that the transportation output decreased gradually appeared as concentric circles from the three center cities (Changsha, Zhuzhou and Xiangtan) to the surrounding areas. While there is a certain difference that the results are similar with each other from the SGCM and the SCCM, and the interpolation results appear more smoothly, that of Ordinary Kriging method seems more discrete. The smoothest interpolation is the SCCM by compared the three models. The high values concentrated in the city center of Changsha, the northeast of Xiangtan, and the northwest of Zhuzhou. The low values scattered at the boundary of the Hunan province, including the southeast corner of Yueyang, the northeast of Liuyang, the northwest of the Xiangxi Autonomous Prefecture and Zhangjiajie, the northern and western regions of Huaihua, the western of Shaoyang, the southwest of Yongzhou, and southeast of Chenzhou. The common characteristics of these areas are mountainous and hilly areas, and the development of transportation was restricted by the terrain, thus resulted in the relatively low output.For the further comparison of the estimated accuracies of different models, the output density map transport industry in Hunan was divided by the municipal administrative boundaries, and the differences were compared between the estimated values and the statistical results within smaller areas. The results show that the estimated values of the prefectural-level city are all lower than the observed. And the estimated values by the SCCM are the closest to the statistic data, which followed by the SGCM, that of the Ordinary Kriging is relatively poor. Therefore, the estimated results in space are credible in the provincial scale.4) The application of the spatialized output of the transportation industry in Hunan for the flood damage assessment by the partition simulation with different inundated scenarios. The results show that, the transportation industry in Changde and Yiyang is influenced by the floods greatest, and the loss increased rapidly with the submerged area expansion. The transportation industry losses in Xiangtan, Zhuzhou and Yongzhou are also serious ranked only second to that in Changde and Yiyang. In the areas that terrain changes greatly, the flooded area extends little with submerged depth increases, even no significant change, such as areas of Huaihua, Xiangxi Autonomous Prefecture. The inundated areas of other regions in Hunan Province are also expanded with the increasing water depth, but it is not significant. In all, the rules for the transportation loss with the water depth increased: between 0-1 m, the transportation loss increased rapidly, and the values around 0~20 billion; between 1-1.5 m, the loss values change a little; between 1.5- 2m, the loss continue to increase except the Changde, Yiyang, and Xiangtan, but the amplification is smaller than that of 0-1 m. Over 2 m, the values increase unobvious except Changsha. According to the simulation of different inundated areas, combined with the economic growth index, the regional industrial output inundated by the flood under different water depths can be quickly evaluated.This study introduced the SCM for the spatialization of the economic sector output, which laid a significant foundation for the improvement of loss assessment accuracy. The spatialization result can be used for a spatial analysis of economic factors, but also provide a vital data base to assess the direct and indirect losses from disasters, the spatialized result is also an important reference for decision-making in disaster risk management simultaneously.