【作者】 张丽文；

【导师】 黄敬峰；

【作者基本信息】 浙江大学 ， 农业遥感与信息技术， 2013， 博士

【摘要】 当生长季内热量条件不足或在关键生育期内遭遇持续低温就会发生低温冷害从而造成作物减产。研究表明,低温冷害在今后相当长的时期内仍然是影响东北地区水稻的主要农业气象灾害之一。对水稻冷害进行及时、准确地监测与预警,对稳定粮食生产意义重大。地理信息系统(GIS)与遥感(RS)技术为宏观和动态地监测农业气象环境和农业生产过程提供了良好的技术手段,是未来构建立体化农业气象服务体系的必然发展趋势。目前低温农业气象灾害遥感研究对象以作物冻害为主,直接利用遥感数据进行作物冷害监测与评价的研究尚不多见。本文选择东北三省为研究区,运用GIS空间分析方法和卫星遥感技术,以冷害综合风险评估与区划、基于全天候气温遥感估算的冷害遥感监测和水稻冷害产量损失量遥感预测为主要研究内容,对近13年东北地区水稻冷害开展了监测与评估研究,系统建立起基于GIS和遥感的水稻冷害监测与评估技术框架,为今后建立完整的农业气象灾害服务系统奠定理论基础。本文的主要研究工作成果如下：(1)依据自然灾害风险评估理论,以日平均温、水稻生长发育期及水稻产量和面积作为基础资料,借助GIS平台,对冷害致灾因子危险性、承灾体脆弱性和承灾体损失度三大风险要素的多个单项评价指标进行了年际统计与空间分析。采用加权综合分析法和基于熵值法和层次分析法的综合赋权法构建冷害各风险要素评估模型及东北地区水稻低温冷害综合风险评估模型。依据冷害综合风险评估指标数据大小,将东北地区划分为较低、低、中等、较高和高风险五个水稻综合冷害风险分区。分别对冷害综合风险评估指标及风险分区进行定量和定性验证,结果表明冷害综合风险评估指标与典型冷害年水稻平均减产率达到0.01水平极显著相关；风险区划结果也与任意冷害类型发生频率的空间分布特征一致，说明本文提出的冷害综合风险评估与区划方法具有一定的合理性和应用价值,能客观反映各地区水稻低温冷害风险等级差异。(2)在总结国内外气温遥感估算方法研究进展的基础上,本文提出了基于多平台MODIS地表温度(LST)数据的全天候平均气温遥感估算方法。首先采用高级统计法对多平台LST数据源晴空像元对应的平均气温分别进行估算。借助多平台MODIS数据的时间互补优势,构建了两种基于时间融合和局部窗口空间插补的全天候气温遥感估算方案。通过分析非晴空像元气温估算的误差来源及大小,得出LST产品的反演误差对气温估算精度引入的不确定性明显小于空间插补算法引入的误差,确定基于全幅LST的时间融合-空间插补方案为最优全天候气温估算方法。检验结果显示,2000-2012年晴空、非晴空及全天候8天平均气温遥感估算RMSE分别为1.4-1.8℃、1.6-2.3℃和1.4-2.0℃,13年间共有12年全天候气温估算误差绝对值小于3℃的样本百分数超过90%；与台站8天平均气温时间序列对比得出,遥感估算气温在夏季有理想结果,而在初春和秋末阶段存在普遍高估。同时对日LST产品运用改进的时间融合-空间插补算法计算日平均气温,并比较了全天候日气温和8天气温合成月平均气温的精度差异,结果显示由8天LST数据源估算的月平均气温与台站观测气温相比有更高的相关性和更小的RMSE,可为后续冷害遥感监测的温度指标计算提供有效的数据支持。本文提出的基于时间融合-空间插补的全天候平均气温遥感估算方法同样适用于全天候最低、最高气温数据的遥感估算。(3)参考现有气象行业标准中的冷害温度指标,以全天候8天平均气温时间序列和植被指数时间序列为基础数据,针对像元及县级两种空间尺度,分别构建了以T5.9距平和相对累积生长度日距平为温度指标的冷害遥感监测指标。经分析,遥感估算的两种冷害温度指标均与台站估算值之间具有高度一致的年际变化趋势,能有效反应水稻生长季内研究区热量条件空间分布的实际年际差异。以地面台站气温数据辨识的冷害发生地点对2000-2012年遥感监测结果进行验证,结果显示在发生大范围延迟型冷害的年份,遥感监测结果与实际灾情的空间一致性较高,像元尺度的一般延迟型冷害监测准确率超过均70%,严重冷害监测准确率超过80%,可用于计算冷害受灾面积。分生育阶段统计相对AGDD距平指标,可对县级尺度的水稻冷害区域进行遥感动态监测。(4)东北水稻冷害灾损遥感预测方法研究以水稻生育期降雨总量、不同水稻生育阶段有效积温(AGDD)、各月月平均气温及水稻关键生育期EVI平均值为驱动因子,预测水稻单产中的气象产量及随机产量,通过累加上一年真实趋势产量,得到预测年水稻单产。结果显示,基于水稻产量水平分区的遥感估产精度好于不分区估产精度；县级及地市级单产遥感估产精度R2均大于0.7,且地市级估产精度好于县级结果。在前面章节水稻面积和水稻关键生育期遥感识别、水稻生长季热量指标遥感估算及水稻冷害受灾区遥感监测等研究成果的基础上,利用水稻冷害灾损模型对冷害年份的水稻产量灾损量进行计算,预测2009年和2011年水稻冷害灾损量分别至少达到26.61和2.17万吨。更多还原

【Abstract】 Chilling damage(also called as cold damage or chilling injury), which will occur when there is lack of accumulated temperature during the growing season or consecutive extreme low temperature below the optimum temperature at key development stage (booting, heading or flowering), is a main agro-meterological disaster for thermophilic crops in regions with low heat accumulation. Some studies have stated that Chilling damage is still a significant challenge for paddy rice in northeast China in the future. Therefore, to monitor and forecast the distribution and intensity of rice chilling damages timely and accurately is not support the local economic development but also stabilize the grain production safety. It must be the development trend for the agro-meteorological services system to using the advancemed technology such as Geographic Information System (GIS) and remote sensing (RS) to dynamically and preciously monitor the agro-meteorological environment and agricultural production at the regional scale in the future. There are several studies focusing on the freeze injury monitoring using remotely sensed data, while few application for chilling damage.In this paper, northeast China is selected to be study region since it is one of important production base for commodity rice but frequently occued low temperature weather in rice growth season.The main contents of the current study are including the risk assessment and risk zoning of rice chilling damage based on GIS, methodology to estimate the air temperature (Ta) under the all sky condition using Terra and Aqua moderate resolution imaging spectroradiometer (MODIS) Land surface temperature (LST)data, monitoring the delayed-type chilling damage of rice based on the all sky Ta data estimated from MODIS LST. and predicting the rice production losses due to chilling damage by combining the identification of planting area and key development stages.The main research achievements in this dissertation were as follows: (1) On the basis of theory of risk evaluation for natural disaster, data including daily mean air temperature, development stages, yield and planting area of rice were employed to analyze the hazard of climate factors, rice vulnerability and yield losses in chilling damage risk in Northeast China. When after determining the integrated weight for each assessment indicator by using the combination of the Entropy Method and Analytic Hierarchy Process (AHP), a comprehensive risk assessment index of rice chilling damage and its zoning were established by weighted comprehensive analysis method (WCA) method. The result of risk assessment index and risk zoning was validated quantitatively and qualitatively in two ways:1) analyzing the correlation between integrated index of chilling damage risk and average reduction rate of rice yield in chilling damage years, which showed a significant correlation at0.01level;2) comparing the risk zoning and the distribution of occurrence frequency for any typed chilling damage, which reveals a strong spatial consistency. In conclusion, our method was proved to be scientific and reasonable to support the prevention and mitigation of chilling damage in northeast China.(2) Numerous studies have developed frameworks to estimate near-surface air temperature from remote sensing for clear sky condition, which can not satisfy the need of data integrity in studies on low temperature disaster monitoring. To solve this critical problem in actual application, a novel spatio-temporal algorithm to estimate air temperature under the all sky condition from Terra and Aqua MODIS LST data is presented in current paper. Firstly, stepwise regression analysis was employed to estimate the four images of mean Ta with daytime and nighttime LST from two satellite platform and images of other factors for clear sky condition respectively.When after merging the four clear sky Ta images onto one Ta image, to estimate the cloudy sky Ta by applying a spatial interpolation algorithm based on the linear regression relationship between Ta and elevation in local sliding window. The methodology that applying the spatio-temporal algorithm in LST pixels under the all sky condition is determined to be the optimal method to estimate the mean air temperature under the all sky condition.The result reveals that. The root mean square errors (RMSE) of8-day mean MODIS_Ta under the clear, clouy and all sky condition comparing to ground-based measurements are1.4-1.8℃,1.6-2.3℃and1.4-2.0℃, respectively. There are over90%of all mean air temperature estimations under the all sky condition.were within3℃absolute bias of12years in total13years. There were big errors of MODIS_Ta estimation occurs in early spring and late autumn while quite small difference in summer days. Finally, we compared the estimation precious of monthly MODIS_Ta derived from the daily MODIS_Ta and8-day mean MODIS_Ta, and then choosed8-day mean MODIS_Ta as the optimal data source to eatimate the temperature indices of chilling damage monitoring.The spatio-temporal algorithm proposed in current paper for mean Ta is also an effective way to estimate the maximum and minimum Ta under the all sky condition.(3) According to the existing temperature indices of rice chilling damage, temperature indices with T5-9anomalies and relative accumulative growing degree day (AGDD) anomalies derived from the time series of8-day mean MODIS_Ta estimation under the all sky condition were proposed to monitor the area of chilling damage in northeast China at the pixel and county scales, respectively. The interannual variation trend of two categories of indices estimated from MODIS_Ta and ground observation are consistent.The significant sptio-temporal change of the heat accumulation in rice growing season from2000-2012are showed in forms with MODIS T5-9anomalies and MODIS rAGDD anomalies.The results reveal that it is an effective way to monitor the delayed chilling damage by using remotely sensed temperature indices when there is a wide range chilling damage occured. such as the northeast China in2003and2009. The precious of remotely sensed monitoring for moderately delayed chilling damage and seriously delayed chilling damage at the county scale are over70%and80%. respectively.(4) Taking total rainfall from May to Septermber, AGDDs over the different development stages periods and average EVI at different key development stages of rice as the predicting factors to estimate the meteorological yield by stepwise regression analysis, and then to predict the rice yield by adding the historical trend yield to the estimated meteorological yield.The result reveals that,the R2between predicting yield derived from MODIS data and actual statistics are all over the0.7at the county and prefecture levels, and the predicting precious is higher of prefecture levels than county levels.On the basis of research achievement with identification of rice key growth developments and the areas with different grades of chilling damage occurred; we used a losses evaluation model to predict the rice production losses caused by chilling damage.There are at least26.61and2.17million tons of rice production losses of northeast China in2009and2011, respectively.更多还原