台风是影响我国沿海地区最严重的自然灾害之一。尤其是在气候变化的背景下，我国沿海地区的台风灾害损失严重。而降低人类社会对台风灾害的脆弱性是适应全球环境变化、减少气候变化不利影响的关键所在。然而，脆弱性受多方面因素的影响，具有一定的复杂性。

因此，本文围绕脆弱性及其变化的影响因素展开，首先基于灾情数据分析了我国台风灾害经济脆弱性和人口脆弱性的时间变化、省间差异等特点，然后基于地理探测器量化了致灾因子危险性、孕灾环境敏感性、社会适应能力三个维度不同指标对台风灾害脆弱性的解释率和指标间的交互作用，最后结合事件对的分析方法，揭示了台风灾害脆弱性变化的影响因素。研究得到的主要结论如下：

（1）2009-2018年台风灾害县级平均直接经济损失率9.3‰，平均人口受灾率为101.8‰。其中平均直接经济损失率和平均人口受灾率最高的是2013年，分别为23.0‰和223.0‰；样本数量最多的前六个省中，海南省的平均直接经济损失率和平均人口受灾率最高，分别为18.94‰和257.56‰。

（2）对2009-2018年台风灾害经济脆弱性和人口脆弱性解释率最高的指标分别是历史台风频次和小时最大雨强，解释率最高的指标组合分别是小时最大雨强和第二产业占GDP的比例、小时最大雨强和历史台风频次。就不同省份而言，经济脆弱性方面，浙江省和广东省主要受到社会适应能力维度中的产业结构相关指标的影响，云南省主要受到社会适应能力维度的收入水平的影响，福建省和广西壮族自治区分别主要受到致灾因子危险性维度的大风强度和降水强度的影响，海南省则主要受到致灾因子危险性和社会适应能力两个维度的影响。人口脆弱性方面，福建省和云南省的主要驱动因素分别为社会适应能力的历史台风频次和非钢混结构房屋占比，对广东省人口脆弱性贡献最大的也是社会适应能力维度的历史台风频次，但一些孕灾环境敏感性指标，如NDVI和地貌类型，对广东省人口脆弱性的贡献也较大。浙江省、广西壮族自治区和海南省的人口脆弱性主要受致灾因子危险性指标的驱动。对于不同路径的台风来说，第一类登陆型和西进型的台风灾害，其脆弱性受到风雨致灾因子解释率的差异通常是小于第二类转向型台风的，而孕灾环境敏感性方面，前者受三个孕灾环境敏感性指标的影响高于第二类路径的台风，在社会适应能力方面，历史台风频率对两类台风脆弱性的影响表现为第一类大于第二类，而房屋质量这一指标对两类台风人口脆弱性的影响表现则为第二类大于第一类。

（3）台风灾害脆弱性的变化是多方面因素综合作用的结果。台风风雨强度的差异对脆弱性变化的影响是不容忽略的。对于时间跨度大和季节不同的两场台风，与致灾因子危险性指标和社会适应能力指标相比，孕灾环境敏感性维度的植被覆盖的变化能够更好的解释两场台风间脆弱性的变化，如对1003号台风“灿都”和1522号台风“彩虹”之间脆弱性的差异解释率最高的指标是NDVI。对于致灾因子强度相近，发生季节相同的两场台风来说，社会适应能力指标的变化所导致的脆弱性的差异也是很明显的。

总体而言，本文建立了2009-2018年我国部分台风灾害脆弱性数据集，针对台风灾害脆弱性的复杂性，基于地理探测器量化了台风灾害脆弱性指标对脆弱性的解释程度，并在此基础上探索了因子间的交互作用。同时，本文根据台风路径选取了台风事件对，通过对重复受灾县脆弱性指标变化和脆弱性变化的分析，给出了前后两次台风脆弱性差异的驱动因素。本研究为定量分析自然社会因素对台风灾害脆弱性的影响以及台风灾害脆弱性评估模型指标优选提出了一种新思路，有利于管理者具有针对性的采取措施来降低台风灾害脆弱性，适应气候变化。

Typhoon is one of the most serious natural disasters affecting coastal areas of China. The loss of typhoon disasters in coastal areas of China is high, especially under global warming. Reducing the vulnerability of human society to typhoon disasters is the key to adapting to global environmental changes and reducing the adverse effects of climate change. However, vulnerability is affected by many factors and is complicated.

Therefore, this paper focuses on the vulnerability and the influencing factors of vulnerability changes. First, based on the disaster data, the temporal changes and inter-provincial differences of typhoon economic vulnerability and population vulnerability were analyzed, and then the explanation of vulnerability indicators from hazard intensity dimensions and environment sensitivity dimensions, and social adaptability dimensions were quantified based on geographic detectors, the interpretation rate and interaction between indicators of different indicator group were also analyzed. Finally, combined with the ‘paired event’ method, the influencing factors of typhoon vulnerability change are revealed. The main conclusions of the study are as follows:

(1) From 2009 to 2018, the county-level average direct economic loss rate of typhoon disasters was 9.3‰, and the average affected population rate was 101.8‰. Among them, the average direct economic loss rate and the average affected population rate were the highest in 2013, which are 23.0‰ and 223.0‰ respectively; among the top six provinces with the most samples, Hainan Province had the highest average direct economic loss rate and average affected population rate, which are 18.94‰ and 257.56‰ respectively.

(2) The indicators with the highest interpretation rate for typhoon economic vulnerability and population vulnerability from 2009 to 2018 are the historical typhoon frequency and the hourly maximum rainfall intensity, respectively, and the combination of indicators with the highest interpretation rate is the group of hourly maximum rainfall intensity and the proportion of secondary industry and the group of hourly maximum rainfall intensity and historical typhoon frequency. As far as different provinces are concerned, in terms of economic vulnerability, Zhejiang Province and Guangdong Province are mainly affected by the industrial structure-related indicators in the dimension of social adaptability, Yunnan Province is mainly affected by the income level of the dimension of social adaptability, while Fujian Province and Guangxi Province are mainly affected by the gale intensity and precipitation intensity of the hazard intensity dimension, and Hainan Province is mainly affected by both the hazard intensity and social adaptability. In terms of population vulnerability, the main driving factors in Fujian and Yunnan provinces are the historical typhoon frequency and the proportion of non-steel-concrete structures of social adaptability dimension, respectively. The greatest contribution to population vulnerability in Guangdong Province is also the historical typhoon frequency of social adaptability, but some environmental sensitivity indicators, such as NDVI and landform type, also contributed significantly to the population vulnerability of Guangdong Province. Population vulnerability in Zhejiang Province, Guangxi Province, and Hainan Province is mainly driven by the hazard intensity indicator. For typhoons with different tracks, the first type, which include landfall type and westward type typhoon disasters are usually have a great difference in the impact of wind and rain more than the second type of typhoons. However, the second type of typhoons are more influenced by the indicators in environment sensitivity dimension than the first type of typhoons. In terms of social adaptability dimension, the impact of historical typhoon frequency on the vulnerability of the two types of typhoons is that the first type is greater than the second type of typhoons, and the impact of housing quality on the vulnerability of the two types of typhoon population is that the second type of typhoons is greater than the first type of typhoons.

(3) The change of vulnerability to typhoon disaster is the result of the effect of many factors. The impact of the difference in rain intensity on vulnerability cannot be ignored. For the two typhoons with large time spans and different seasons, compared with the indicators of hazard intensity and social adaptability, the changes in vegetation cover can better explain the changes in vulnerability between the two typhoons. For example, the indicator with the highest interpretation rate for the difference in vulnerability between the 1003 Typhoon Chanthu and the 1522 Typhoon Mujigae is NDVI. For two typhoons with similar hazard intensity and occurred in the same season, the differences in vulnerability caused by changes of social adaptability indicators are also obvious.

In general, this paper establishes some typhoon disaster vulnerability data sets in China from 2009 to 2018. In view of the complexity of typhoon disaster vulnerability, based on GeoDetector, the degree of interpretation of typhoon vulnerability indicators for vulnerability is quantified, and the interaction between factors was explored. At the same time, this paper selects four pairs of paired typhoon events according to the typhoon track. Through the analysis of the change of vulnerability indicators and vulnerability in the repeatedly affected counties, the driving factors of the difference in vulnerability between the two typhoons in each paired typhoon event are given. This study puts forward a new idea for quantitatively analyzing the impact of natural and social factors on the vulnerability of typhoon disaster and selection of indicators in vulnerability assessment model, which is helpful for managers to take targeted measures to reduce typhoon disaster vulnerability and adapt to climate change.