EGI OpenIR
地质水文灾害易发性及社区风险认知分析-以卢旺达的洪水和滑坡例
Alternative TitleSusceptibility analysis for geo-hydrological hazards and Community Risk perception – a case of flood and landslide in Rwanda
Richard MIND’JE
Subtype硕士
Thesis Advisor李兰海
2020-06-30
Degree Grantor中国科学院大学
Place of Conferral北京
Degree Discipline理学硕士
Keyword洪水 滑坡 易发性 影响因素 卢旺达 Flood Susceptibility Influencing factors Landslide Rwanda
Abstract在气候变化的背景下, 近年来洪水和山地滑坡等地质灾害频发,导致大量人员伤亡,严重影响发展中国家的社会经济发展。 人类很难阻止这些自然灾害的发生, 但是以居民作为灾害易发区的关键关注点,通过大力普及与宣传社区防灾减灾知识为切入点,制定洪水和山体滑坡方面的防灾减灾策略, 最终能够有效避免或减轻自然灾害给人类社会带来的损失。 本论文通过系统分析不同地区洪水和滑坡的影响因素, 综合确定洪水和滑坡灾害高风险区。 研究结果将有助于人们更好地了解灾害风险, 进而制定合理有效的风险管控措施, 为灾害风险的可持续管理提供理论依据。东非地处热带, 地形条件复杂。 洪水和滑坡频发对该区域社会经济和人类生命财产造成严重影响, 并且这种影响呈现持续增加趋势。 卢旺达地区人口密集、 丘陵众多, 气候因素驱动产生的洪水和滑坡已成为阻碍该区域人类活动和社会经济发展的主要自然灾害。 目前主要针对洪水或滑坡开展的单一研究较多, 但是对于卢旺达地区,滑坡常常与洪水事件相伴而生。 因此, 本研究将洪水和滑坡灾害相结合, 确定灾害同时发生的条件。 目前主要采用辅助性数据来对灾害风险管理进行研究, 通过使用定性和描述性的方法来分析灾害风险意识。 在进行与灾害风险管理相关的研究时, 必须对洪水与滑坡灾害相关的主要变量进行筛选, 并对其进行易发性分析, 以揭示洪水与滑坡灾害的危害程度以及与潜在影响因素之间的空间相关性和区域的不确定性。 然而对影响洪水与滑坡危害程度的因素仍未得到充分的分析和理解, 人们对灾害有关风险的认识缺乏重视, 人们的灾害意识在防灾减灾中的作用被忽略。 针对卢旺达的地质水文灾害发生的频率和强度不断增加, 有必要加强洪水和滑坡易发性评估, 以减少灾害的风险。综上所述, 本研究将解决以下科学问题: (1) 卢旺达洪水和滑坡发生的主要致灾因素有哪些? (2) 洪水和滑坡同时发生的共同影响因素和条件是什么? (3) 社区灾害风险意识对其所在区域的洪水和山体滑坡灾害易发性产生多大影响? 本论文的研究目标: (1) 分析洪水和滑坡致灾因素与危害之间的空间关系, 确定造成洪水和滑坡易发性的主要因素; (2) 确定安全和易受灾害危害的区域, 并绘制区域灾害易发性图;(3) 确定两种害条同时发生的因素和灾件; (4) 评估社区对减少洪灾和滑坡风险的最佳预防和缓解措施。为了达到以上目标, 本研究基于历史灾害事件的信息, 通过认知历史灾害事件造成危害发生的条件, 据此选择历史灾害事件的影响因素及其相关的基础数据。 应用遥感(RS) 、 地理信息系统(GIS) 和统计和概率方法分析灾害易发性的时空变化。 通过频率比(FR) 模型的统计方法确定因果关系和过去滑坡事件之间的关系; 应用Logistic 回归(LR) 模型生成有助于预测该地区洪水的系数来评估洪水事件的发生概率及与其相关因素之间的关系; 综合文献, 结合卢旺达洪水和滑坡影响因素的实际情况, 本研究选择 13 个预测因素进行洪水和滑坡灾害易发性分析。 这 13 个预测因素包括: 海拔、 坡度、 坡向、 剖面曲率、 到河流的距离、 到道路的距离、 归一化植被指数(NDVI) 、 归一化土壤指数(NDSI) 、 地形湿度指数(TWI) 、 河流动力指数(SPI) 、降雨量、 土地覆盖与土地利用(LCLU) 和土壤质地。 本研究一共收集过去发生的 153个洪水和 423 个滑坡事件, 应用地统计分析工具及其在 GIS 中再分区的子集功能, 将这些定位点的事件随机抽取 75%的训练点用于模型构建, 剩余 25%训练点用于模型验证。 采用问卷调查的方法完成社区有关灾害危险意识看法的实地调查。 根据洪水与滑坡灾害的发生次数确定调查问卷区域, 选取的调查区包括西部省(Nyabihu, Rubavu,Rusizi, Karongi, Rutsiro, Ngororero 区) , 北部省(Musanze 和 Burera 区) , 南部省(Nyaruguru 区) , 东部省(Rwamagana 区) 和基加利市(加索博区) 等 11 个地区。 在每个选定的地区, 使用随机抽样和滚雪球抽样方式选择 50 名参与者, 最终共有550 名来自所选地区的不同领域的受访者接受了问卷调查。 本研究的主要结论如下:(1) 卢旺达西部、 北部和南部省份是最容易发生洪水和滑坡的地区, 不同区域洪水和滑坡的影响因素存在较大差异。 西部省主要受山脊和高原地形的影响, 包括具有特殊气候和地形特征的刚果尼罗河; 北部省主要因在陡坡上的农业活动所引起的土壤侵蚀, 土壤进入河道, 从而产生沉积物淤积, 排水能力减弱; 南部省主要受地形地貌影响; 东部省由于降雨开始较晚, 最容易受到滑坡和洪水灾害的共同影响。 此外,快速城市化进程与极端天气条件的共同作用, 增加了基加利市滑坡和洪水灾害易发性空间分布的不确定性。(2) 洪水易发性的 Logistic 回归系数结果表明, NDVI、 NDSI、 降雨、 海拔、剖面曲率和长宽比与洪水发生率呈正相关, 而 TWI、 距道路的距离、 距河的距离和坡度则呈负相关。 其中, NDVI 和降雨是最重要的直接影响因素(回归系数分别为 0.5938 和 0.5159) , 而距道路的距离的回归系数最低(-7.6265) , 这与历史记录的洪水事件相符。(3) 坡度、 LCLU、 降雨和高程是研究区滑坡发生的主要影响因素。 滑坡极高和高等级易发区域分别占到全国的 13.2%和 14.5%, 主要分布在穆山(Musanze) 的恩戈罗雷罗(Ngororero) 、 加肯克(Gakenke) 、 尼亚比亚(Nyabihu) 和纳马加贝(Nyamagabe) 等地区。 此外, 中等强度的滑坡区占全国的 24.3%, 主要分布在卡莫尼(Kamonyi) 、 鲁汉戈(Ruhango) 、 尼亚鲁古(Nyaruguru) 和尼扬扎(Nyanza) 等地区, 而处于低和极低等级强度的滑坡区占比分别为 40.1%(占比最高) 和 7.9%。降雨和植被覆盖度是引起洪水和滑坡事件的主因。(4) 灾害的暴露度因人口密度而异, 各个地区存在差异。 近期人口密度变化表明, 人口快速增长将使大多数家庭(尤其是农村地区) 继续居住在洪水和滑坡易发区域。 这些区域内的居民主要依靠农业为生, 生活极端贫困, 致使社区进一步将林地和草地转变为农业用地, 导致洪水和滑坡更容易发生。(5) 基于曲线下面积(AUC) 方法对生成的易发性图的验证结果表明, 洪水易发性模型的预测和成功率分别为 79.8%和 80.4%, 滑坡易发模型的预测和成功率分别为 84.6%和 81.2%。 尽管输入数据的准确性存在局限性, 但这结果仍是令人满意的,预测模型在研究区内具有良好的适应性和预测洪水和滑坡发生的能力。(6) 许多社区成员, 特别是在农村地区缺乏有关减少洪灾和滑坡风险中的作用和责任方面的认知。 这主要由于教育和社会保护保障的有关机构在传播、 制定和实施减灾措施相关责任的信息方面投入太少, 导致社区周边环境更容易遭受洪水和山体滑坡的侵袭; 居民通常自己不采取措施来减少这些危害, 而是希望政府和地方领导人代表他们采取行动。 因此在制定防灾减灾计划中, 必须加强社区, 尤其在农村地区的减灾和预防知识方面的普及和教育。正如本研究的结果所言, 降低洪水与滑坡灾害最佳的解决措施就是将灾害易发区的居民迁出并重新安置于安全区中。 因此, 在政府制定防灾减灾计划时该措施应当被广泛实施。 另外, 本研究所展示的信息可作为辅助性决策工具为卢旺达地区针对洪水和滑坡的区域规划、 未来灾害研究以及风险管理提供科学的理论参考。
Other AbstractThe incidences of geo-hydrological hazards such as floods and landslides have beenon the increase in the recent decades due to climate change especially in developing countries.Flood and landslide have caused significant societal and economic damage and large numberof fatalities worldwide. It may not be feasible to control nature and stop the occurrence ofthese natural phenomena, but the efforts could be made to avoid turning into disasters andalleviate their negative effects on human lives, infrastructure and property by also taking intoaccount community perceptions on the hazards risk as a key factor when seeking to developsystems, practices and policies to protect local populations based on their susceptibility to ahazard. Therefore, it is of importance to analyzing the factors influencing floods andlandslides in different areas, and identifying disaster high risk zones on floods and landslides,which can significantly lead to a better understanding of disaster risk and put in placemeasures for risk reduction and contribute a lot to the process of sustainable management ofdisaster risks.In East African regions where there exists complex topography with tropic weather,floods and landslides occurred frequently and have caused serious and diverse impacts.Existing figures have confirmed their rise in damages and losses. In Rwanda, given theclimatic, topographic, demographic, and geologic conditions in combination with internalfluctuations in the climate system, flood and landslide incidences constitute the normal geohydrological phenomena with which society must live. Landslide incidence is oftenassociated with flood event, and they occur in a cycling system. Existing studies that havebeen carried out in Rwanda focused on flood or landslide separately. Therefore, this study hascombined both flood and landslide together to identify under which conditions or factorsthese hazards occur at the same time. Flood and landslide events in Rwanda are constantlyincreasing with a scarcity of related scientific research as there have been studies aboutdisaster risk management with significant emphasis on only describing the hazards andmeasuring awareness using qualitative, social, descriptive approaches and secondary datasources. Dissimilar to this, expert in the related field explained that studies in relation todisaster risk management have to be highlighted by a susceptibility analysis involvingimportant predicting variables to reveal the spatial correlation between the hazards and theirpossible influencing factors which are still not well understood and uncertain in the studyarea. On the top of this, the perception of community on their risk related to the hazards hasarguably not received enough attention and seldom used as an influencing factor in the scientific literature. Given the increase in frequency and intensity of geo-hydrological hazardsin Rwanda, limited efforts have been made to predict floods and landslides to prevent orreduce the risks from these hazards.Therefore, this study is conducted to address and answer to the following mainresearch questions: (1) What are factors mostly influencing the occurrence of flood andlandslide in Rwanda?, (2) What are common factors and conditions under which flood andlandslide occur at the same time?; (3) Are community perceptions on flood and landsliderisks play a role on the susceptibility of an area in Rwanda?. And similarly, the objectives ofthis study are: (1) To analyze the spatial correlation between predicting factors and thehazards for determining and providing a scientifically driven explanation of the main factorsmostly contributing to flood and landslide susceptibility; (2) To produce susceptibility mapsdisplaying safe and prone areas to the hazards; (3) To identify the factors and conditionsunder which both hazards occur at the same time; (4) To link the community ’s perception offlood and landslide risk with the resulting susceptibility index to assess the community ’sability to reduce the impacts of the hazards; and (5) To assess community perception on thebest prevention and mitigation measures for flood and landslide risk reduction.With the genesis of remote sensing (RS) and development of geographicalinformation system (GIS), different studies on hazards susceptibility analysis have useddifferent factors with the potential of predicting the occurrence of the hazards in specificareas. Remote sensing was significant as it provided several data and information needed forthe study while GIS was used as a tool to analyze those data and all the models used wereGIS-based. To achieve the objectives of this study, statistical and probabilistic methodswere used. Statistical method through Frequency ratio (FR) model was used to determinethe relationship between causal factors and past landslide events given by the ratio of areawhere landslide happened to area that has not been affected by the hazard. Whereas, theprobabilistic method through Logistic Regression (LR) model was used to assess the linkagebetween the presence and lack of flood events with related factor by generating coefficientsthat help to predict floods in the area. The choice of the model to be applied in susceptibilityanalysis always goes hand in hand with the availability of historical events which serves asthe basic data source for understanding conditions contributing to the occurrence of hazards,and the selection of influencing factors. Owing to the above, the revision of the literatureenables the listing of various factors that have been said to have the potential of causing bothflood and landslide occurrence. Once listed,each factor was placed in the context of its contribution in predicting and understandingsusceptibility to the hazards. Therefore, thirteen predicting factors including: the elevation,slope, aspect, profile curvature, distance to rivers, distance to roads, the NormalizedDifferent Vegetation Index (NDVI), the Normalized Difference Soil Index (NDSI), theTopographic Wetness Index (TWI), the Stream Power Index (SPI), rainfall, Land CoverLand Use (LCLU) and Soil texture were used. Furthermore, a total of 153 past flood and423 past landslide events were collected, mapped, and used. These events represented aslocation points were randomly split into a proportion of training (75%) for model buildingand testing (25%) points for validation steps using the geostatistical analyst tool and itssubset feature of repartition in GIS. In regard to community perceptions, a field survey wasconducted whereby questionnaires were distributed in different districts with frequent floodand landslide events based on the occurrence rather than population size. Based on this,eleven districts were, therefore selected such that in Western province (Nyabihu, Rubavu,Rusizi, Karongi, Rutsiro, Ngororero districts), Northern Province (Musanze and Bureradistrict), southern province (Nyaruguru district), Eastern province (Rwamagana district) andfinally in Kigali city (Gasabo district) were selected. The questionnaires were administered insuch way that in each selected district, 50 participants were chosen using random andsnowball sampling. The questionnaires were administered to 550 respondents in thosedifferent areas. The main findings of the study are as follows:(1) The results indicated the western, northern, and southern provinces as the mostsusceptible areas to both flood and landslide occurrence. This distribution has been attributedto different reasons containing the used factors. The status in the western province wasattributed to the dominant ridges and plateaus including the Congo Nile with certain climaticand topographic features. The situation in the northern province was attributed to the erodedsoil from agricultural activities practiced on steep slopes which eventually ends up in waterchannels, thus reducing the drainage capacity to accommodate peak runoff and sedimentsaccumulation while the condition in southern province was attributed to the geomorphologiccharacteristics of the area. In contrast, the eastern province was the least prone to bothhazards due to its high rainfall deficit and late rainfall onsets. Furthermore, the spatialdistribution of susceptibility in Kigali city for both hazards was exacerbated by rapidurbanization in combination with extreme weather conditions.(2) From the calculated coefficients of logistic regression for flood susceptibility,NDVI, NDSI, rainfall, elevation, profile curvature, and aspect factors showed a positiverelationship with flood incidence while TWI, distance from roads, distance from rivers andthe slope have exhibited negative relationships. Out of all these factors, NDVI and rainfallwere the most influencing variables as they have shown the highest coefficients (0.5938 and0.5159 respectively) while the distance from roads showed the lowest (-7.6265). Theseresults were testified by the collected past flood locations through the inventory map whereheavy rain zones and areas with positive NDVI values accounted for a big number of floodevents. Hence, flooding and their impacts should be mostly expected in areas susceptible toflood and strong mitigation efforts should be mostly deployed in areas modeled as high andvery highly susceptible. This should be based on factors that have proved to be positivelycorrelated to flood occurrence.(3) For landslide, the slope, LCLU, rainfall and elevation were the main and highlyinfluencing factors for landslide occurrence in the study area. This result implied thatvegetation and topographical related factor (LULC) stimulated by rainfall are conditionscausing landslide to occur. District wise, Ngororero, a part of Musanze, Gakenke, Nyabihu,and Nyamagabe fell into very high and high susceptible classes covering 13.2% and 14.5%respectively in the entire country. Besides, moderate susceptible areas covering 24.3% weredistributed throughout the country in many districts such as Kamonyi, Ruhango, Nyaruguruand Nyanza while at the contrary, 40.1% (the highest ratio) and 7.9 % fell into the low andvery low susceptible classes, respectively. Thus, it was depicted that both hazards will occurin areas receiving much amount of rainfall with less vegetation coverage to a certainelevation.(4) The levels of population’s exposures are different based on the population density,and each area presents its own particularities especially in terms of influencing factors. Hence,compared to the recent population density of the country, the results revealed that for bothhazards, households especially in rural areas keep on residing into susceptible zones due tothe growing population. These people are known to be living in extreme poverty and mainlyrelying on agriculture for their livelihood. This situation pushed the community in convertingforestlands and grasslands into agricultural areas; consequently, resulting into deforestationwhich increases susceptibility and easily become a trigger for flood and landslide occurrence.(5) The generated susceptibility maps were validated using the area under curve(AUC) method which portrayed the prediction and success rate of 79.8% and 80.4% for floodsusceptibility modeling and 84.6% and 81.2% for landslide susceptibility modeling,respectively. These percentages were considered satisfactory despite the input data limitationand accuracy; and explained how well LR and FR models along with the factors performed inpredicting flood and landslide occurrence in the study area.(6) The results have been linked with community perception on the hazards risk.The findings revealed that many community members especially in rural areas are unawareand lack enough information and knowledge on their roles and responsibilities in flood andlandslide risk reduction. This finding revealed that the relevant institutions in charge ofcommunity awareness, education and social protection guarantee do not go great length indisseminating information related to the responsibility of developing and implementingdisaster mitigation measures. Consequently, this condition makes community’s immediateenvironment more susceptible to flood and landslide since people do not take measures tolessen the occurrence of these hazards but instead, expect the government and local leaders totake actions on their behalf. This finding implicated the need of establishing programs toinculcate a sense of mitigation and prevention in communities, especially in rural areas.In conclusion, mitigation measures to reduce the impacts of the hazards are needed.As disclosed by the results of this study, relocation, and resettlement from prone to safe zonesis one of the best measures to be adopted. The information obtained from this whole studycan be used as a baseline for planners, future researchers, disaster risk managers and beused as a supplementary decision-making tool in the country as far as flood and landsliderisk mainstreaming is concerned.
Subject Area环境科学
Language英语
Document Type学位论文
Identifierhttp://ir.xjlas.org/handle/365004/15397
Collection中国科学院新疆生态与地理研究所
研究系统
Affiliation中国科学院新疆生态与地理研究所
First Author Affilication中国科学院新疆生态与地理研究所
Recommended Citation
GB/T 7714
Richard MIND’JE. 地质水文灾害易发性及社区风险认知分析-以卢旺达的洪水和滑坡例[D]. 北京. 中国科学院大学,2020.
Files in This Item:
There are no files associated with this item.
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[Richard MIND’JE]'s Articles
Baidu academic
Similar articles in Baidu academic
[Richard MIND’JE]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[Richard MIND’JE]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.