农业生产是干旱灾害主要影响领域， 严重干旱会对作物产量造成巨大的损失，导致区域粮食产量与消耗的不平衡，是影响世界粮食安全的决定性因素。 研究严重干旱对农业的影响，将有助于采取措施保障农业产量损失最小化， 为干旱减灾管理策略提供科学依据。 本文从全球严重干旱时空变化规律及其对农业产量影响研究角度出发， 首先，基于SPEI（ Standardized Precipitation-Evapotranspiration Index） 指数， 开展了全球主要作物的易旱区域的识别， 并将此作为进一步开展干旱影响评估研究区的选择依据； 在此基础上，从作物生长机理出发， 以环境政策综合气象（ Environmental Policy Integrated Climate ， EPIC）为工具， 模拟了中国黄淮海平原和美国堪萨斯州县域单元上1981-2010 作物产量。 其次，以改进的日值 SPEI 为干旱识别指标， 探索了生长季干旱持续时间、 严重性、 频数及生长季尺度 SPEI 值四个干旱特征指标与作物产量之间的关系， 甄选能表征作物产量的最优干旱特征指标。 最后， 基于选定的干旱特征指标， 通过非干旱条件年份的识别及非干旱条件下作物参考产量的计算， 构建作物产量旱灾损失评估方法。 并将此方法扩展应用至美国堪萨斯州， 评估了该地区 5 个典型干旱年的作物因旱减产量。 主要的结论如下。1. 1902-2008 年全球干旱影响面积呈上升趋势， 1950 年后干旱易发区面积也在增加， 美国中东部、 亚洲东部、 欧洲南部和南美洲是主要的干旱易发区。 以 SPEI 为工具， 通过分析全球干旱影响面积和发生频率， 研究 100 余年的干旱时空变化规律。 1902-2008 年间全球、 北半球和南半球干旱影响面积呈上升趋势， 大洲尺度上仅大洋洲呈下降趋势， 但自 1990 年以来全球干旱影响面积呈下降的趋势。 全球干旱易发区呈簇状集中分布， 且分布面积逐步扩大， 1902-1949 年间全球有 16.19%的区域处于长期干旱易发区， 1950-2008 年间扩展到 41.09%， 美国中东部、亚洲东部、 欧洲南部和南美洲是主要作物的干旱易发区。2. 通过探索四个生长季干旱特征指标与作物产量之间的关系， 甄别出表征作物生长季干旱的最佳特征指标。 基于改进的日值 SPEI， 提取了黄淮海平原 28 个站点生长季干旱持续时间、 生长季干旱严重性、 生长季干旱频数及生长季 SPEI 值四个表征干旱特征指标， 并逐站点分析四个干旱特征指标与作物多年产量数据的关系。 研究发现， 在四个干旱特征指标中， 生长季尺度 SPEI 与作物产量关系最好。 在此基础上， 通过分析不同尺度的 SPEI 与作物产量的关系，发现冬小麦播种前 90 天（ 约 3 个月 ） 加上生长季时长尺度下的SPEI， 即 SPEIs+90 与产量关系最好， 除 2 个站点相关性显著性水平 p<0.05 外， 其他站点相关性显著性水平均 p<0.01，SPEIs+90 可以作为准确甄别生长季干旱的特征指标。3. 以作物模型模拟的作物产量为基础， 综合构造的无干旱条件下的作物参考产量， 构建了作物产量旱灾损失评估方法。 本文选择 SPEIs+90 为冬小麦生长季干旱识别指数， 在正常水分条件年份识别的基础上，通过分析上述年份下作物产量， 构建无旱条件下的作物参考产量， 并通过干旱年份下作物模型模拟的作物产量与作物参考产量的对比， 构建旱灾损失评估方法， 评估作物因旱减产量和减产率。 基于黄淮海平原 5 省（ 直辖市） 作物产量统计数据， 将构建的产量旱灾损失方法与基于统计数据构建的旱灾损失评估方法进行对比验证， 可靠性指标值达到 0.97， 评估重度及重度以上干旱对作物产量影响的可靠性指标值达到 1， 本文构建的方法可以较好地应用于旱灾损失评估。4. 扩展应用构建的作物产量旱灾损失评估方法至美国堪萨斯州， 定量评估了该地区县域尺度典型干旱年份干旱对冬小麦产量的影响。 在堪萨斯州县域水平上， 评估了五个典型干旱年份下因旱导致的作物减产率的空间分布特征，1981 年干旱主要发生在堪萨斯东南部，造成冬小麦减产 45 万吨， 1989 年干旱导致堪萨斯东部减产 126.8 吨， 2002、 2004 和 2006 年干旱发生在堪萨斯西部，各年因旱导致作物减产分别为 75.1 万吨、 27.1 万吨和 84.2万吨。

Agricultural production is one of main sectors mostly suffered from drought disaster. Severe drought usually led to huge yield losses, and brought imbalance between grain production and consumption within certain region. Thus, severe drought is one of the decisive factors affecting food security around the world. Therefore, the study on the impact of drought on agriculture is helpful to take measures to minimize agricultural production losses, and would provide scientific basis for strategy on drought disaster reduction management.From the spatial and temporal variation of severe drought and its impact on global agricultural production research perspective, firstly, drought prone area for main crops was identified with SPEI (Standardized Precipitation-Evapotranspiration Index) index, which was the basis for further selection of drought impact assessment study area. Then, from the perspective of cropgrowing mechanism, crop model EPIC (Environmental Policy Integrated Climate) was applied in Huang-Huai-Hai Plain of China and the United States Kansas ,to simulate crop yield from 1981 to 2010, with county as the basic unit. Taken improved daily SPEI as the drought indicator,this research extracted drought duration, drought severity, drought frequency in growing season,and the growing season SPEI value, and analyzed the relationships between crop yields and thesedrought characteristics. Optimal drought indicator for yield condition was identified. By identifying the non-drought condition year and calculating the crop yield of non-drought condition, optimal drought indicator was used to build the crops drought damage assessmentmethod. The method was applied in Kansas, USA and drought induced crop yield losses in five typical drought years were assessed. The main conclusions are as follows：1. Between 1902 and 2008, drought areas over the world are generally on the rise, and drought prone area increased after 1950, mid-east United States, eastern Asia, southern Europe and South America is the main drought-prone area.Based on SPEI index, the drought spatio-temporal variation characteristics during about 100 years were explored by analyzed the drought areas and drought frequency over the world. Between 1902 and 2008, drought areas over the globa;、 northern hemisphere and the southern hemisphere are generally on the rise, among continents only Oceania continent is on the decline.Since 1990, global drought area showed a downward trend. Identified global drought prone area distributed as clusters, and the distribution area was gradually expanding; during 1902-1949, 16.19% of the world was identified as long-term drought prone area, the percentage expanded to 41.09% between 1950 and 2008, the mid-east United States, eastern Asia, southern Europe and South America were the drought prone area for main crops.2. This research explored the relationship between characteristics of drought in growing season and crop yield, and identified optimal indicator for characterizing the drought in growing season for crops.Based on the improved daily SPEI, growing season drought duration, drought severity, growing season drought frequency and growing season SPEI value were extracted from 28 sites in Huang-Huai-Hai plain, analyses of four characteristics of drought and crop yield data were conducted at each site. Study found that in four indexes of drought characteristics, SPEI at growing season scale had best relationship with crop yield. By analyzed the relationship between SPEI of different level and crop yield, further studies found that the SPEI for period beginning from 90 days before sowing and ending with harvest, namely SPEIs + 90， showed best relationship with yield, with the exception of two site reached 1% significance level, other sites were all in very significance level( p<0.01),therefore, SPEIs + 90 can be used as an indicator for identifying drought in growing season accurately.3. Based on simulated crop yield and constructed crop reference yield under non-drought condition, crops drought loss assessment method was build and verified.This study choose the SPEIs + 90 as the drought index in winter wheat growing period, and based on the identification of the year with normal water condition, crop reference yield under non-drought condition was built by analyzing the crop yield of the above years. Then, crop reference yield during drought years was combined with the yield simulated by crop model todesign drought loss assessment method, and assess yield loss due to drought and crop failure rate. For 5 provinces in Huang-Huai-Hai plain, assessment results from above method were compared with those based on statistical data. Verification results showed that the reliability index is 0.90. Further analyses found that the above assessment method got higher reliability (close to 1) when applied to severe and beyond severe drought. So the method built in this study could work well with drought loss assessment.4. The drought induced crops loss assessment method was applied to the United States Kansas,and quantitatively estimated the impact of drought in typical drought year on winter wheat yield with county as the basic unit.At county level in Kansas, space distribution of crop failure rate in five typical drought years was evaluated. Drought mainly happened in southeast Kansas in 1981, the winter wheat yield loss was 450 thousand tons. In 1989, eastern Kansas lost 1268 thousand tons due to the drought. In 2002, 2004, and 2006, drought occurred in the western Kansas, crop yield losses caused by drought are 751 thousand tons, 271 thousand tons and 842 thousand tons respectively