干旱作为一种长期缺水事件，是对农业、生活生产和社会经济造成灾难性影响的自然灾害。其中，农业干旱严重影响农业生产，可能会造成农作物减产和绝收。农业干旱直接威胁粮食安全并会加剧地区贫困，因此受到越来越多的关注。基于模型对土壤湿度、干旱指数和可利用水量（降水量－蒸散发量）的模拟结果表明，气候变暖背景下未来几十年内干旱风险将会增加。高质量数据的缺失和干旱指数的差异可能会对未来的干旱预测以及对时空模式的理解产生很大的不确定性。土壤湿度是衡量地表干湿变化的一个重要变量，其与以作物可利用水量亏缺为主要特征的农业干旱密切相关，因此，土壤湿度被视为农业干旱信息最重要的表征因子。此外，政府间气候变化专门委员会（IPCC）第五次评估报告（AR5）指出，自20世纪中叶以来，由于气候内部变率较大，人为强迫对全球陆地区域干旱变化影响的检测可信度较低，且相关研究较为有限。因此，在全球变暖的背景下，土壤湿度的多时空演变和驱动因子检测与归因仍然是一个挑战。

本研究基于多源数据集提出了不同深度土壤湿度数据多时空尺度综合评估方法，评估了不同下垫面条件和气象要素对土壤湿度及其数据质量的影响；基于温度区间和滑动窗口等方法，动态探索了不同植被覆盖下土壤湿度对温升的差异化响应；基于正则最优指纹法，研究了历史时期不同强迫下土壤干旱的演变规律，量化了不同强迫对季风区和非季风区土壤干旱的影响；基于气候模式未来土壤湿度数据，探究了不同气候变化情景下未来土壤湿度的时空演变规律，并定量分析了主要气象要素对土壤湿度变化时空特征的影响。主要研究结果如下：

（1）不同土壤湿度产品的不同评估指标表现不同，应根据研究目的，采用适用的评估指标选择高质量土壤湿度数据集。土壤湿度再分析产品能反映不同深度土壤湿度随时间的变化，包括不同深度土壤湿度变化的时滞效应。ERA5-Land（第五代欧洲中期天气预报中心再分析陆面数据）和CLDAS-CMA（中国气象局陆面同化数据）能更好地反映表层土壤湿度随时间的变化，即与实测土壤湿度有更高的相关性。土壤表层（0~10cm）通常含有最低的土壤湿度。随着深度的增加，大部分地区的土壤湿度先急剧增加，然后缓慢减少，其中CLDAS-CMA很好地反映了平均土壤湿度随深度的变化。随着土壤深度的增加，土壤湿度产品的R（相关性系数）和KGE（Kling-Gupta效率）逐渐减小，ubRMSE（无偏均方根误差）逐渐增大，而BIAS（偏差）几乎无明显变化。在站点尺度上，四个指标（R、KGE、BIAS、ubRMSE）均表明CLDAS-CMA表层土壤湿度的表现最好，但其时间跨度较短，无法满足长时间序列的分析；ERA5-Land的R值和KGE值表现较好，BIAS和ubRMSE表现较差，说明它能反映表层土壤湿度的长期变化，但对土壤湿度的均值和变异性的估计较差。与空间尺度相比，不同的时间尺度对土壤湿度产品的表现影响不大。ERA5-Land和ERA-Interim（欧洲中期天气预报中心再分析数据）在不同的空间尺度上表现比较稳定。ERA5-Land在区域尺度上表现良好，而CLDAS-CMA在不同空间尺度上表现出高度不稳定性。降水对黄河流域土壤湿度产品评估结果的影响比温度的影响更大。所有四种土壤湿度产品的评估结果都受降水变化的显著影响，而只有CLDAS-CMA和ERA5-Land数据受温度显著影响。沙土含量和黏土含量对表层土壤湿度有显著影响，而有机碳含量上三层土壤湿度有显著影响。在土壤湿度较高的地区往往黏土含量较高，沙土含量较低，且各土壤湿度产品表现更好。

（2）不同植被覆盖类型下土壤湿度随气温的升高而减少，且土壤湿度变干的速率随气温的升高而加快。温度升高往往会导致土壤湿度减少，在较暖的植被覆盖类型中，表层土壤湿度随着温度升高而下降得更快，这表明在不同的植被覆盖类型上温度对缩放因子的影响比植被覆盖类型更大。同一种植被覆盖类型上，较高的温度同样会加快表层土壤湿度变干的速率，且在除稀疏草地以外的所有植被覆盖类型上都达到显著性水平。在较暖的条件下，缩放因子表现出明显的空间异质性和更大的变化范围，表明在较暖的条件下缩放因子不确定性更大，且主要受植被覆盖类型和土壤类型等局部因素的影响。在较暖的条件下，不同植被覆盖类型下的缩放因子变化幅度更大，表明在较暖的条件下不同的植被覆盖类型对蒸散发的影响更大，进而对表层土壤湿度减少速率的影响更大。当表层土壤湿度超过上临界值时，土壤湿度补给充足，所以缩放因子基本不变，且接近于0；当表层土壤湿度低于下临界值时，缩放因子会随着土壤湿度下降而急剧减小，直至为0。在这种情况下，由于表层土壤湿度的限制，蒸散发过程迅速减弱，直至完全停止。当表层土壤湿度适中时，由于蒸散发过程，缩放因子受多种因素影响，但主要是温度。

（3）全国大部分地区的干旱呈增加趋势，且在季风区人为强迫对干旱的影响更容易被检测到。在非季风区和季风区，更多区域的干旱逐渐增强，其中季风区68.94%（64.25%）的区域干旱强度（干旱极值）呈增加趋势，非季风区53.61%（57.41%）的区域干旱强度（干旱极值）呈增加趋势，而干旱历时变化较小。相较于干旱历时，各强迫对干旱强度和干旱极值的影响更容易被检测到，而且各强迫对干旱的影响在季风区比非季风区更容易被检测到。ANT强迫（人为强迫）引起的干旱强度在中国范围内变化0.16/世纪，但区域差异较大，其中ANT强迫引起季风区干旱强度变化0.59 /世纪。GHG强迫（温室气体强迫）引起全国范围干旱强度变化为0.13/世纪，占ANT强迫引起的干旱强度变化的81.25%。ANT强迫引起的干旱极值在不同区域变化差别不大：整个中国为0.32/世纪，季风区为0.42/世纪，非季风区为0.28/世纪。仅中国区域和季风区的ANT强迫对干旱历时的影响能从ALL强迫（全强迫）中检测到。不同区域的ANT强迫对干旱强度的影响均能从ALL强迫中检测到，仅季风区的GHG强迫对干旱强度的影响能从ANT强迫和AA强迫（人为气溶胶强迫）中检测到。不同区域的ANT强迫和GHG强迫对干旱极值的影响均能从ALL强迫中检测到。仅季风区的ANT强迫和GHG强迫对干旱极值的影响能从NAT强迫（自然强迫）中检测到。

（4）中国土壤湿度未来呈现“南方变干，北方变湿”的空间格局，这种空间格局主要受降水变化的空间差异性影响。气候模式能较好地模拟历史时期的表层土壤湿度变化，但无法很好地反映深层土壤湿度的变化。其中，25个CMIP5模式中有16个模式可以捕捉到1981-2005年表层土壤湿度历史趋势的方向。RCP2.6情景下，中国大部分地区的表层土壤湿度无明显变化规律；但在其他三种排放情景下，表层土壤湿度在干旱和半干旱地区总体呈减少趋势，在南方大部分地区呈增加趋势。气温上升越高，变化显著的区域越多，表层土壤湿度变化幅度越大。在RCP8.5情景下，中国56.9%的区域可以检测到表层土壤湿度逐渐减少，其中44.2%的区域表层土壤湿度显著减少（p<0.05）。而在中国43.1%的区域可以观察到表层土壤湿度呈增加趋势，其中29.4%的区域表层土壤湿度显著增加（p<0.05）。对于所有区域，降水是表层土壤湿度变化空间差异的最主要气候因素。它对中国表层土壤湿度变化空间分布的相对解释率高达43.4%。温度、风速和相对湿度对表层土壤湿度变化空间分布的相对解释率分别为22.5%、19.9%和14.2%。四个变量的相对解释率在各地区之间存在差异，相对于其他地区，温度在华北地区占有最高的比重，相对解释率为28.5%。

Drought, as a long-term water shortage event, is a natural disaster that has a catastrophic impact on agriculture, production and social economy. Agricultural drought seriously affects agricultural production, which may cause crop yield reduction and no harvest. Agricultural droughts have received increasing attention because they directly threaten food security and exacerbate regional poverty. Model-based simulations of soil moisture, drought indices, and available water (precipitation-evapotranspiration) suggest that the risk of drought will increase over the next few decades under a warming climate. The absence of high-quality data and differences in drought indices may create large uncertainties in future drought predictions and understanding of spatiotemporal patterns. Soil moisture is an important variable to measure the change of surface dryness and wetness, and it is closely related to agricultural drought, which is mainly characterized by the shortage of water available to crops. Therefore, soil moisture is regarded as the most important characterization factor for agricultural drought information. In addition, the Fifth Assessment Report (AR5) of the Intergovernmental Panel on Climate Change (IPCC) pointed out that since the middle of the 20th century, due to the large internal variability of the climate, the detection reliability of the impact of anthropogenic forcing on global terrestrial drought changes is relatively low, and related research is quite limited. Therefore, in the context of global warming, the detection and attribution of multi-temporal and spatial evolution and drivers of soil moisture remains a challenge.

Based on multi-source datasets, this study proposes a multi-scale spatiotemporal comprehensive evaluation method for soil moisture data at different depths and explores the effects of different underlying surface conditions and meteorological elements on soil moisture and its data quality. Based on the temperature interval and sliding window, this study dynamically explores the effect of temperature rise on soil moisture under different vegetation covers. Based on climate model data, this study explores the evolution of soil drought under different forcings in historical periods, and quantifies the effects of different forcings on soil moisture drought in monsoon and non-monsoon regions. Based on the climate model data of future soil moisture, the temporal and spatial evolution of soil moisture in the future was explored, and the influence of different meteorological variables on the temporal and spatial characteristics of soil moisture changes was analyzed. The main findings are as follows:

(1) Different evaluation indicators of different soil moisture products have different performances, and high-quality soil moisture datasets should be selected according to the research purposes and using applicable evaluation indicators. The soil moisture reanalysis product can reflect the changes of soil moisture with time at different depths, including the time-lag effect of soil moisture changes at different depths. ERA5-Land (the fifth-generation European Centre for Medium-Range Weather Forecasts re-analysis land surface data) and CLDAS-CMA (land surface assimilation data of China Meteorological Administration) can better reflect the change of surface soil moisture with time, that is, it has a higher correlation with the in-situ measured soil moisture. The soil surface layer (0-10cm) usually contains the lowest soil moisture. With the increase of depth, soil moisture in most areas increases sharply at first and then decreases slowly, among which CLDAS-CMA well reflectes the change of average soil moisture with depth. As soil depth increases, R (correlation coefficient) and KGE (Kling-Gupta efficiency) of soil moisture products gradually decrease, ubRMSE (unbiased root mean square error) gradually increases, while BIAS (bias) hardly changes. At the site scale, the four indicators (R, KGE, BIAS, ubRMSE) all show that the surface soil moisture of CLDAS-CMA has the best performance, but its time span is short, which cannot satisfy the analysis of long-term series; the R and KGE of ERA5-Land perform better, while BIAS and ubRMSE perform poorly, indicating that they could reflect long-term changes in surface soil moisture, but poorly estimates of the mean and variability of soil moisture. Different time scales have little effect on the performance of soil moisture products compared to spatial scales. The performance of ERA5-Land and ERA-Interim (European Centre for Medium-Range Weather Forecasts reanalysis data) is stable across different spatial scales. ERA5-Land performs well at regional scales, while CLDAS-CMA exhibits high instability at different spatial scales. The impact of precipitation on the evaluation results of soil moisture products in the Yellow River Basin is greater than that of temperature. The assessment results for all four soil moisture products are significantly affected by changes in precipitation, while only the CLDAS-CMA and ERA5-Land data are significantly affected by temperature. Sand content and clay content have a significant effect on the surface soil moisture, while the organic carbon content has a significant effect on the upper three-layer soil moisture. Areas with higher soil moisture tended to have higher clay content and lower sandy soil content, and each soil moisture product performed better.

(2) Soil moisture decreases with the increase of air temperature under different vegetation cover types, and the drying rate of soil moisture accelerates with the increase of air temperature. Soil moisture drought intensifies with the rise of air temperature in all land cover types, suggesting there is a clear negative effect of temperature on soil moisture under drought conditions.

(3) Drought is on the rise in most parts of China, and the effects of anthropogenic forcing on drought are more easily detected in the monsoon region. The drought in more regions gradually increases. 68.94% (64.25%) of the monsoon region shows an increasing trend in drought intensity (drought maximum), and 53.61% (57.41%) of the non-monsoon region showes an increasing trend in drought intensity (drought maximum), however, the drought duration changes little. Compared with the drought duration, the effects of each forcing on the drought intensity and the drought maximum were more easily detected, and the effects of each forcing on drought were more easily detected in the monsoon region than in the non-monsoon region. The drought intensity caused by ANT forcing (anthropogenic forcing) varies by 0.16/century in China, but there are large regional differences, among them, ANT forcing causes a 0.59/century change in the drought intensity in the monsoon region. GHG forcing (greenhouse gas forcing) causes a nationwide change of 0.13/century in the drought intensity, accounting for 81.25% of the drought intensity change caused by ANT forcing. The drought maximum caused by ANT forcing varies little in different regions: 0.32/century in China, 0.42/century in the monsoon region, and 0.28/century in the non-monsoon region. Only the effects of ANT forcing on the drought duration in China and the monsoon region can be detected from ALL forcing (all forcing). The effects of ANT forcing on the drought intensity in different regions can be detected from the ALL forcing, and the effects of GHG forcing in the monsoon region on the drought intensity can be detected from ANT forcing and AA forcing (anthropogenic aerosol forcing). The effects of ANT forcing and GHG forcing on drought maximum in different regions can be detected from ALL forcing. The effects of ANT forcing and GHG forcing on drought maximum can only be detected from NAT forcing (natural forcing) in the monsoon region.

(4) Surface Soil moisture in China will present a spatial pattern of "drier in the south and wetter in the north" in the future, which is mainly affected by the spatial differences in precipitation changes. Climate models can well simulate the changes in surface soil moisture in historical periods, but cannot well reflect changes in deep soil moisture. The direction of the historic trends of surface soil moisture during 1981 and 2005 can be captured by 16 out of the 25 CMIP5 GCMs. Under RCP2.6 scenario, surface soil moisture has no significant change in most regions across China. But under the other three emission scenarios, arid and semi-arid regions are dominated by wetting surface soil moisture and south of China is dominated by drying surface soil moisture. The higher the global temperature rises, there are the more regions with significant changes and larger magnitude of surface soil moisture changes. Under RCP8.5 scenario, drying SSM can be found in 56.9% of regions across China (44.2% with <0.05 P-value). As a comparison, wetting SSM trends can be observed in 43.1% of the regions across mainland China (29.4% with <0.05 P-value). For all regions considered, precipitation is the most important factor for spatial variations of surface soil moisture among the four meteorological variables considered. On average, its relative explanation rate to SSM variation can reach over 43.4%. Temperature, wind speed and relative humidity have relative explanation rates of 22.5%, 19.9% and 14.2% to SMM changes respectively. There is a small difference in the relative contributions of the four variables among the regions.