【作者】 卓义；

【导师】 刘桂香；

【作者基本信息】 中国农业科学院 ， 草业科学， 2011， 博士

【摘要】 干旱灾害一直是影响草原牧区畜牧业生产最主要的气象灾害。在信息化飞速发展的时代背景下,利用新的遥感与地理信息系统(GIS)技术深入研究草原地区干旱灾害,建立有效的草原地区干旱监测、评估体系,可为抗灾决策部门提供更为丰富、有力的信息支持。本文以MODIS1B数据为遥感信息源,结合地面样方调查数据、长序列的气象站点数据以及社会经济统计资料,利用遥感、GIS技术手段,通过对干旱对草原植被的影响、草原干旱灾害监测预警、草原干旱灾害评估以及草原干旱灾害风险评价与区划四个方面的研究,实现了对草原地区干旱灾害的监测与评估。利用近40年的长时期序列降水资料,以降水在时间序列上的GAMMA分布概率函数为模型,使用标准化降水距平指数(SPI)作为气象干旱指标,针对不同干旱程度与不同生长时段,分析了不同类型草原植被的生长速度、产草量和建群种优势度等指标对干旱的响应。结果表明:草甸草原、典型草原和荒漠草原三种不同的草原类型中,草甸草原的生长速度变化对干旱的响应最为敏感,荒漠草原的响应最为迟钝。对草甸草原来说4、5月份的降水最为关键;对典型草原来说4、5月份和8、9月份的降水对植被的生长最为关键;对荒漠草原植被来说8、9月份的降水最为关键。本文利用LST-NDVI特征空间原理深入挖掘特征空间中相对稳定的角度信息,建立了有效的土壤湿度遥感反演模型。所建立的模型克服了传统LST-NDVI特征空间法时间可比性差的缺点,能够满足不同年份不同时段的监测需求。通过对土壤湿度、植被生长情况等干旱指标要素的逐旬遥感监测,建立了复合预警指标体系。在2009年的夏季干旱的案例中,使用复合预警指标体系进行逐旬监测。监测结果表明复合预警指数与各旬的SPI-1M指数变化相近,既可反映当前干旱状况又可预警下一时段干旱发展,监测预警效果良好。通过对草原干旱灾害的影响范围、持续时间、灾害强度以及灾情损失的评估建立了草原地区干旱灾害评估标准,采用BP神经网络法对干旱灾害类型进行了划分。与一般的统计聚类方法相比,BP神经网络分类法克服了灾害类型判读经验的限制和不同程度干旱灾害类型下各指标之间的关系的假设限制,具有一定的技术优势。利用本文建立的草原地区干旱灾害评估方法,对研究区发生的春旱、夏旱、夏秋连旱的不同年份的干旱灾害进行了分县评估分析。评估结果既能反映该年度的灾情程度,又可反映该年度干旱灾害的类型。利用气象灾害风险评价方法对研究区草原干旱灾害进行风险评价与区划。内蒙古东部草原地区干旱灾害风险等级评价结果表明:低风险区地区包括新巴尔虎左旗、扎鲁特旗等6个旗县;中风险区的地区包括苏尼特右旗、林西县等7个旗县;高风险区的地区包括新巴尔虎右旗、西乌珠穆沁旗等7个旗县。草原干旱灾害风险类型划分结果表明:阿鲁科尔沁旗草原干旱灾害风险类型为低危险型;阿巴嘎旗、镶黄旗等10个旗县草原干旱灾害风险类型为低防灾抗灾能力型;东乌珠穆沁旗、鄂温克族自治旗等9个旗县草原干旱灾害风险类型为易损型。更多还原

【Abstract】 Drought has always been the main meteorological disaster which influences livestock productionof the prairie areas. Under the context of rapid development of information, using the new remotesensing and geographic information system (GIS) to profoundly study draught in grassland areas andestablishing effective monitoring and evaluation system for disaster decision-making departments canprovide richer and more powerful information support for decision-making departments. This paperadopted remote sensing and GIS techniques to monitor and assess the draught in the eastern grasslandsof Inner Mongolia. Taking MODIS 1B data as the remote sensing information sources and at the sametime combining with the ground quadrat investigation data, a long series data of meteorological stationsand economic and social statistic data, the paper took a deep analysis of the impacts of draught on thegrassland vegetation. Furthermore, it set up stable inversion model of soil moisture through thequantitative excavation of remote sensing information. On the basis of this model, the author carried outresearch in the following four aspects: The influence of drought to the prairie grassland vegetation; thedrought monitoring and early warning of grassland; the drought assessment of grassland and the droughtrisk and zoning of grassland.By employing the nearly 40 years data on the index precipitation, using the precipitation in timeseries GAMMA distribution probability function as a model, and applying the standardized rainfallindex (SPI) as the meteorological drought index, the paper analyzed the different types of grasslands onthe aspects of growth speed, plant yield and the building of the advantage degree index to the responseof the drought by the different dry degree and different growth periods. The result indicated that in thethree different types of grasslands: meadow grassland, typical grassland and desert grassland, thegrowth rate of the meadow grassland is more sensitive to the response of the drought, while the desertgrassland is slower. Different types of grassland vegetation in different growth period responsedifferently to drought. As for meadow grassland, rainfall in April and May is the most critical; fortypical grassland, rainfall in April, May, August and September is the most critical to vegetation growth;for desert grassland vegetation the precipitation in August and September is the most critical.This paper, by using LST-NDVI character space principle to dig in the feature space relativelystable Angle information, established the effective soil humidity remote sensing data model. The modelovercame the disadvantage of traditional NDVI feature space method’s poor time comparability. It cansatisfy the monitoring requirements at the different years and periods.Through a ten-day remote sensing monitoring of drought index such as soil moisture, vegetationgrowth, a composite index system can be established. In the 2009 summer drought cases, the compositeindex system was used by the ten-day monitoring, which can not only reflect the current droughtconditions but also warn the drought development of the next phase is established. The result indicatedthat the compound warning index and the last-1 M index SPI similarly changed, which can not onlyreflect the current drought in the early but also warn the next time development and the affect of thedrought monitoring and warning is good. The paper established the drought disaster evaluation standard on the basis of evaluating of scope,duration and intensity of the disasters and classified the drought disaster types by applying the BP neuralnetwork. Compared to the general statistic clustering method, the BP neural network overcame therestrictions on interpretation experience of different types of drought disasters and the assumptionrestrictions on the relationship of each index of different degree of drought disaster types. It has acertain technical superiority. By using the evaluation method of prairie areas, the paper evaluated thedifferent years of droughts happened in spring, summer and autumn of the study areas. The result notonly can reflect the annual disaster degree, but also the type of drought disaster.By using meteorological disaster risk evaluation method to research area grassland drought disasterrisk assessment and divisions. The evaluation results of Eastern Inner Mongolia grassland area droughtdisaster risk level showed that: the low-risk areas include six Banners and Counties, such asXinBaErHuZuoQi, ZaLuTeQi and so on. The medium-risk areas include seven Banners and Counties,such as SuNiTeYouQi, linxi county and so on; The high- risk areas include XinBaErHuYouQi,XiWuZhuMuQinQi seven Banners and Counties.更多还原