~在全球气候变化的背景下，干旱事件发生的频率、持续时间和严重程度都呈增加趋势，严重威胁陆地生态系统的平衡及造成巨大的经济损失。在干旱事件中，植被会受到严重的水分胁迫作用，导致植被生理状态变化和生产力下降，对粮食安全造成巨大威胁，还会导致森林大面积死亡，也影响碳循环等问题。因此，干旱准确监测及预警意义重大。目前，随着卫星遥感技术发展，可在区域和全球尺度上较好的监测干旱事件。近年来，卫星遥感反演的太阳诱导叶绿素荧光（SIF）与植被的光合作用直接相关，可以更好地反映植被生理状态，逐渐成为监测区域植被生长条件和环境胁迫的新型遥感数据源。
本文利用SIF数据、MODIS产品数据（NDVI、EVI、GPP和FPAR等）、气象水文数据（降水、温度、土壤湿度、比湿度、蒸散发和地表径流等）、气象站和农业气象站数据和遥感干旱指数数据等监测中国地区（东部、中部、南部、东北及西部部分区域）不同植被类型的干旱事件、探究时间序列上干湿的时空变化与格局及与土壤湿度（SM）时滞性。本研究围绕如何提高旱情监测精度及准确评估时空分布特征这一核心问题开展研究，主要的研究内容和结论如下：
（1）利用SIF、植被指数（VIs）、气象水文数据及遥感干旱指数等研究了SIF监测旱情的敏感性，SIF与农业、气象指标的时空分布关系及不同地表下SIF与各个指标的相关性。研究结果表明SIF比VIs对干旱胁迫更敏感。在2009年8月至10月森林发生干旱期间，SIF normalized by absorbed photosynthetically active radiation（SIFyield）分别减少了23.10%、34.88%和30.52%，其减少幅度明显大于SIF（3.25%、9.43%和3.88%）。而且，在2014年7月和8月作物干旱事件中，SIFyield（17.73%、20.66%）减少幅度也大于SIF（10.96%和11.85%）。但在2016年7月和8月草原干旱事件中，Total Emitted SIF（SIFte）（29.71%、33.63%）减少幅度大于SIF（28.91%、32.53%）和SIFyield（19.26%、19.60%）。因此，SIF指标对不同植被类型干旱的敏感性存在差异，SIFyield对森林和作物类型干旱最敏感；SIFte对草原干旱最敏感，其敏感性仅稍微强于SIF和SIFyield。由于SIFyield消除了太阳辐射的影响，对干旱变得更加敏感。同时，小波相干分析表明SIF指标比NDVI可以更超前反应SM的变化情况。
SIF和SIFyield与农业、气象干旱指标的空间分布基本一致。SIF、SIFyield与农业干旱指标（VHI和SM）的空间一致准确率分别为89.08%和68.00%（VHI）、96.44%和78.97%（SM），且相关系数分别为0.760和0.532（P<0.001，VHI）、0.734和0.705（P<0.001，SM）。SIF、SIFyield与气象干旱指标（气温（Air_Ts）、降水距平百分比（Pa）和比湿度（SH））的空间准确率分别为96.58%和66.70%（Air_Ts）、83.36%和82.44%（Pa）、97.19%和72.45%（SH），且相关系数分别为0.482和0.511（P<0.001，Air_Ts）、0.746和0.706（P<0.001，Pa）、0.744和0.716（P<0.001，SH）。因此，在长时间序列上SIF与农业干旱、气象干旱的空间一致性更好。最后，探究SIF指标与不同植被类型下各指标（GPP、ET、LST、ATs、PPT、GSM、ESM、SH和WS）的相关性的差异，结果表明，SIF与各指标存在较好的相关性，大部分情况高于NDVI。总体上，SIFyield最适用于森林和作物植被类型旱情监测，SIFte对草原旱情最敏感；在时间序列上，SIF与农业/气象干旱空间分布一致性较好，且在不同植被类型下SIF与各指标均具有良好的相关性。
（2）本文将地表温度（LST）、SIF和水量平衡（降水、蒸散发和地表径流）利用标准化异常、主成分分析和欧氏距离法构建了一个新的三维空间干旱指数——温度太阳诱导叶绿素荧光水量平衡干旱指数（TSWDI）。利用农业气象站的相对土壤湿度（RSM）、气象站计算的标准化降水指数（SPI）/标准化蒸散指数（SPEI）、基于遥感数据的SM、SPEI、干旱指数（VCI、TCI和PCI）和GPP的相关性及空间分布验证TSWDI的准确性；进一步利用中国和美国典型干旱事件验证TSWDI的适用性。结果表明，TSWDI比上述干旱指数能更好地监测区域干旱情况。进一步基于TSWDI时间序列数据探究了研究区干湿状态的空间变化及格局。结果表明，研究区平均以0.0065/年的速率变湿润，其中达到显著性变湿润的像元比例为52.89%，不显著变湿润的面积为26.07%，且重心呈湿润趋势且向西北方向偏移大约184.49km。研究区干湿状态的Moran’s I在月、季节和年尺度介于0.743~0.930之间，表明空间格局具有强的聚集性，破碎程度较小；局部空间自相关表明以高-高聚集和低-低聚集的比例在23.85%~35.74%之间，低-高聚集和高-低聚集的比例小于1%，表明空间呈湿湿和干干分布。最后小波相干分析表明，在主要显著周期的时间段内，不同植被类型下和气候带的TSWDI超前SM变化的比例分别为83.33%和84.09%。因此，TSWDI可以更准确的监测区域干湿状态及超前反应SM的变化情况。
 

In the context of global climate change, the frequency, duration, and severity of drought events are increasing, which seriously affects the balance of terrestrial ecosystems and causes huge economic losses. During drought events, vegetation is subjected to severe water stress, resulting in changes in vegetation physiology and reduced productivity, which poses a great threat to food security and also leads to extensive forest mortality and affects the carbon cycle, among other issues. Therefore, accurate drought monitoring and early warning are of great significance. Currently, with the development of satellite remote sensing technology, drought events can be better monitored on a regional and global scale. In recent years, the sun-induced chlorophyll fluorescence (SIF) retrieved by satellite remote sensing is directly related to the photosynthesis of vegetation, which can better reflect the physiological state of vegetation. SIF has gradually become a new remote sensing data source for monitoring regional vegetation growth conditions and environmental stress. This paper uses SIF data, MODIS product data (NDVI, EVI, GPP, FPAR, etc.), meteorological and hydrological data (precipitation, temperature, soil moisture, specific humidity, evapotranspiration and surface runoff, etc.), meteorological stations and Agro-meteorological stations data and drought indices data, etc. to monitor drought events in different vegetation types in China region (eastern, central, southern, northeastern and western parts), and explore the spatial and temporal variability and patterns of dryness and wetness on time series and the time lag with soil moisture (SM). This paper focuses on the core issue of how to improve the accuracy of drought monitoring and accurately assess the characteristics of temporal and spatial distribution. The main research contents and conclusions are as follows: (1) Using SIF, vegetation indices (VIs), meteorological and hydrological data and remote sensing drought indices to study the sensitivity of SIF to drought, the temporal and spatial distribution of agricultural and meteorological indicators, and the correlations between SIF and various indicators under different vegetation types. The results show that SIF is more sensitive to drought stress than VIs. During the forest drought from August to October 2009, SIF normalized by absorbed photosynthetically active radiation (SIFyield) was reduced by 23.10%, 34.88% and 30.52%, respectively, which was significantly greater than SIF (3.25%, 9.43% and 3.88%). Moreover, the reduction in SIFyield (17.73% and 20.66%) was also greater than that of SIF (10.96% and 11.85%) in the July and August 2014 crop drought events. However, in the grassland drought events in July and August 2016, total emitted SIF (SIFte) (29.71% and 33.63%) decreased more than SIF (28.91% and 32.53%) and SIFyield (19.26% and 19.60%).Thus, SIF indicators are different in the sensitivity of different vegetation types to drought. SIFyield is the most sensitive to forest and crop drought. SIFte is the most sensitive to grassland drought, and its sensitivity is only slightly higher than SIF and SIFyield. As SIFyield eliminates the effects of solar radiation, it becomes more sensitive to drought. Simultaneously, wavelet coherence analysis shows that the SIF indicators can over-respond to SM changes more than NDVI. In addition, research shows that SIF and SIFyield are basically consistent with the spatial distribution of agricultural drought indicators (VHI and SM) and meteorological drought indicators (SH, Pa and Air_Ts). The spatially consistent accuracy rates of SIF, SIFyield with VHI, SM are 89.08% and 68.00% (VHI), 96.44% and 78.97% (SM), respectively. And the correlation coefficients between SIF, SIFyield and them are 0.760 and 0.532 (P<0.001, VHI), 0.734 and 0.705 (P<0.001, SM), respectively. The spatial accuracy of SIF, SIFyield and meteorological drought indicators are 96.58% and 66.70% (Air_Ts), 83.36% and 82.44% (Pa), 97.19% and 72.45% (SH), respectively. And the correlation coefficients between SIF, SIFyield and them are 0.482 and 0.511 (P<0.001, Air_Ts), 0.746 and 0.706 (P<0.001, Pa), 0.744 and 0.716 (P<0.001, SH), respectively. Hence, the spatial consistency between SIF and agricultural drought and meteorological drought is better in time series. Finally, this article explores the differences in the correlation between SIF indicators and various indicators (GPP, ET, LST, ATs, PPT, GSM, ESM, SH and WS) under different vegetation types. The results show that SIF has a good correlation between with various indicators, most of which are higher than NDVI. The conclusions are that in typical drought events, SIFyield is the most applicable to forest and crop vegetation types, and SIFte is the most sensitive to grassland drought. On the time series, the spatial distribution of SIF and agricultural/meteorological drought is the most consistent, and has good correlation with various indicators under different vegetation types. (2) This paper combines land surface temperature (LST), SIF and water balance (precipitation, evapotranspiration and surface runoff) data using standardized anomalies, principal component analysis and Euclidean distance method to construct a new three-dimensional drought index—temperature sun-induced chlorophyll fluorescence water balance drought index (TSWDI). First, the accuracy of TSWDI is verified using relative soil moisture (RSM) from agrometeorological stations, standardized precipitation index (SPI)/standardized evapotranspiration index (SPEI) calculated from meteorological stations, SM, SPEI, drought indices (VCI, TCI and PCI) and GPP based on remote sensing data; and then the applicability of TSWDI is verified using typical drought events in China and the United States.The results show that TSWDI can better monitor regional drought conditions. In addition, the spatial variation and pattern of dry and wet states in the study area are further explored based on TSWDI time series data. The results showed that the study area became wet at an average rate of 0.0065/year. Among them, the proportion of pixels that reached P<0.05 significantly wetting is 52.89%, and the area that dose not significantly wetter is 26.07%. And the wetting area is shifted to the northwest by about 184.49 km. The Moran’s I of the dry and wet state in the study area ranges from 0.743 to 0.930 on the monthly, seasonal, and annual scales, indicating that the spatial pattern has strong agglomeration and less fragmentation. The local spatial autocorrelation shows that the ratio of high-high aggregation and low-low aggregation is between 23.85% and 35.74%, andthe proportion of low-high aggregation and high-low aggregation is less than 1%, indicating that the space is distributed wet-wet and dry-dry. Finally, wavelet coherence analysis shows that the proportions of TSWDI leading SM changes under different vegetation types and climatic zones are 83.33% and 84.09%, respectively. Therefore, TSWDI can more accurately monitor the regional dry and wet state and the change of over-response SM.