气候变化背景下植被动态响应及其对陆地生态系统水循环的影响是当前全球变化生态水文学的研究热点。滦河流域位于中国北方半干旱半湿润过渡区，是京津冀重要水源涵养区和生态安全的绿色屏障，水资源和生态系统对气候变化的响应尤为敏感。20世纪80年代以来，气候变暖促进了北半球中高纬度地区植被变绿和生长周期显著延长。植被物候响应气候的动态变化如何影响陆地生态系统水循环过程？相关研究，特别是流域尺度的研究十分缺乏。以流域为单元探究植被物候波动及其对流域水循环的影响，对于理解陆地生态系统对气候变化的响应，全面评价植被动态与流域水量平衡的关系，进而保障流域水资源和生态系统健康稳定，具有重要的研究价值和现实意义。

本文以滦河流域为研究区，以揭示气候变化-植被动态-流域水文要素关系为主要研究目标，基于遥感和地面观测数据，首先分析了1982-2015年间滦河流域气候因子、植被春秋季物候及地表径流的变化特征，在流域尺度上解析了水热因子对春秋季物候的影响机制，在此基础上构建了耦合水热因子的植被物候模型；其次，明确了植被生长动态及物候变化对地表径流的影响，并将植被物候动态耦合到SWAT模型中，提高了模型模拟精度；最后，利用CMIP6不同气候模式数据与物候模型、SWAT改进模型相结合的方法，预估了未来不同气候变化情景下滦河流域植被物候和水文要素的演变趋势，并为流域应对气候变化和实施水资源与生态管理提供了参考建议。

本文得到的主要研究结论如下：

（1）1982-2015年，滦河流域整体呈现暖干化趋势，降水变化的空间差异显著，中下游降水减少量是上游的两倍。流域内植被变绿，生长季延长约20天，具体表现为春季物候（SOS）显著提前（约14天）而秋季物候（EOS）呈延迟趋势（约6天），由春季物候提前导致生长季延长的区域占比约70%。以草地覆盖为主的流域上游地区SOS普遍延迟，而以森林覆盖为主的中下游地区SOS提前趋势显著。受降水、植被类型和植被生长动态空间差异的共同影响，流域中下游地区径流减少趋势是上游的4倍。

（2）植被春秋季物候受到温度和降水协同控制，其中SOS响应水热因子变化的时段更短，敏感性更高，流域整体SOS温度和降水敏感性分别为-3.0 ± 0.3天/℃和-4.1 ± 0.2天/10mm。森林春季物候的温度敏感性约为草地的两倍，降水敏感性低于草地，秋季物候温度和降水敏感性之间无显著差异。基于此，构建了耦合水热因子的春季物候过程模型和机器学习模型，运用模拟退火法在网格尺度上实现模型参数优化，模型对研究区春季物候模拟效果良好，过程模型内外部检验相对误差约为7天，支持向量机模型小于5天。

（3）降水是影响流域径流的最主要因素，与地表径流的灰度关联系数达到0.7以上；植被变绿和物候波动改变了蒸散发和土壤水含量，但并未显著改变流域径流。由植被生长周期和物候波动对径流影响关系入手，耦合植被物候动态到SWAT模型，优化了模型植被生长过程曲线，并在水文要素模拟中表现出较好的区域适用性。研究发现，SWAT原始模型对森林SOS模拟值较实际SOS晚8天，而森林和草地EOS模拟值较实际EOS晚近60天，因此原始模型高估了生长季LAI而低估了非生长季LAI，导致非生长季蒸散发和土壤水被低估约13%。

（4）在未来不同气候情景下，滦河流域整体呈现暖湿化趋势。水热耦合过程模型在物候期预测中较机器学习模型具有更好的适用性。2016-2100年滦河流域SOS总体呈现显著提前趋势，SSP126、SSP245、SSP585情景下2022-2055年较1982-2015年分别提前14天、15天和15天，2062-2095年较2022-2055年分别提前2天、4天和6天。由于森林植被物候的温度敏感性更强，而草地植被物候的降水敏感性更强，在SSP126情景下，流域中下游森林植被可能出现SOS提前量减小甚至延迟的现象，而草地SOS呈现随降水增加持续提前的趋势。

（5）未来不同气候情景下，SOS提前将促进蒸散发量增加，对流域产流量减少贡献约为10%，而对地表径流的影响存在较大不确定性。在流域规划和区域生态工程建设中应统筹生态环境保护与水资源承载能力，充分发挥植被在保持水土、涵养水源、自然固碳等方面的功能优势，同时也不能忽略植被变化导致的水分耗散量增加而产流量减少对流域水资源储量的负面效应。

The dynamic response of vegetation to climate change and its impact on the terrestrial ecosystem water cycle has been the focus of the current research of the global change ecohydrology. Luanhe River basin is located in the semi-arid and semi-humid transition area of northern China, which is a green barrier and an important water conservation area for the ecological security of Beijing-Tianjin-Hebei region. The response of water resources and ecosystem to climate change are particularly sensitive. Since the 1980s, climate warming has significantly prolonged the growing season and promoted the growth of vegetation in the middle and high latitudes of the Northern Hemisphere. However, how does the response of plant phenology variation affect the terrestrial hydrological processes? Relevant studies, especially at the watershed scale, are very lacking. Therefore, it is of great research value and practical significance to investigate the variation of plant phenology and its impact on river runoff, in order to better understand the response of terrestrial ecosystem to climate change, to comprehensively evaluate the relationship between vegetation dynamics and regional water balance, as well as to help ensure the health and stability of water resources and ecosystems in the basin.

Taken Luanhe River basin as the study area, this study aims to reveal the relationship between climate change, vegetation dynamics and watershed runoff. Based on both remote sensing data and ground observations, we first analyzed the characteristics and changes of climate change, spring and autumn phenology, river runoff in Luanhe River basin from 1982 to 2015, explored the influence mechanism of evironmental factors (i.e. temperature and precipitation) on spring and autumn phenology at the watershed scale, and on this basis, proposed an improved phenology model coupling hydrothermal factors. Secondly, we identified the grey correlations between growing season length, spring phenology and river runoff, and coupled the dynamic parameters of vegetation phenology into SWAT model, which improved the simulation accuracy. Finally, the CMIP6 dataset derived from different climate models was used to drive the improved phenology model and SWAT model, and predict the evolution trend of spring phenology and hydrological elements in Luanhe River basin under different climate change scenarios in future. The results also provided suggestions for the watershed to cope with climate change and implement better water resources and ecological management.

The main conclusions of this study are as follows:

(1) From 1982 to 2015, the Luanhe River basin showed a trend of warming and drying, and the spatial pattern of precipitation change was significant. The precipitation reduction in the middle and lower reaches was twice that in the upper reaches. The basin was getting greening with the growth season extended by about 20 days. Specifically, the spring phenology (SOS) is significantly advanced about 14 days, and the autumn phenology (EOS) was delayed by about 6 days. About 70% of the resiongs with prolonged grow season were affected by earlier spring. In the upper reaches of the basin where grassland is the dominant land cover type, the SOS was generally delayed. Howerver, in the middle and lower reaches of the basin where forest is dominant, SOS tended to advance significantly. Affected by the spatial differences of precipitation, vegetation cover and their growth dynamics, the runoff in the middle and lower reaches decreased four times as much as that in the upper reaches.

(2) Both the spring and autumn phenology is corrdinated by temperature and precipitation. By comparison with EOS, the response periods of SOS to the changes of temperature and precipitaion are shorter and the sensitivities are higher. The average sensitivity of SOS to temperature and precipitation is -3.0 ± 0.3 days / ℃ and -4.1 ± 0.2 days / 10mm, respectively. The temperature sensitivity of SOS for forest is about twice that of grassland, but the precipitation sensitivity of SOS for forest is lower than that of grassland. There is no significant difference in the sensitivity of EOS to temperature and precipitation. Thus, the process-based spring phenology model and the machine learning models were built by coupling hydrothermal factors, and the model parameters are optimized using simulated annealing algorithm on the grid scale. The improved models showed better adaptation in spring phenology simulation in the study area. The internal and external RMSE of the process-based model is about 7 days, and the SVM model is less than 5 days.

(3) Precipitation is the most important factor affecting river runoff, and the grey correlation coefficient is larger than 0.7. Vegetation greening and the phenology dynamics directly affected evapotraspiration and soil moisture, but not significantly changed the river runoff. By coupling the dynamic parameters of SOS and EOS to SWAT model, the vegetation growth process curve was optimized, and the hydrological elements were simulated more accurately. The results showed that the simulated forest SOS by original SWAT model was 8 days later than the remotes sensed value, and the EOS of forest and grassland was nearly 60 days later than the remotes sensed value. Thus, the LAI during growing season was overestimated, but the LAI of non-growing season was underestimated by the original SWAT model. Such inaccuracy in modeling vegetation processes resulted in serious underestimation of evapotranspiration and soil water content in non-growing season, with a ratio of 13% around.

(4) Under different scenarios of climate change in future, Luanhe River basin was predicted to be warmer and wetter. The process-based model has better applicability than machine learning model in spring phenology prediction. From 2016 to 2100, the overall SOS in Luanhe River basin will become much earlier. Compared with 1982-2015, the SOS during 2022-2055 under the scenario of SSP126, SSP245 and SSP585 would be advanced by 14 days, 15 days and 15 days, respectively. And continues to be 2 days, 4 days and 6 days earlier than 2022-2055 during 2062-2095, respectively. Because the forest phenology is more sensitive to temperature and the grassland phenology is more sensitive to precipitation, the advance of SOS of forest vegetation may decrease or even delay in the middle and lower reaches under SSP126, but the SOS trend of grassland would continues to advance with the increase of precipitation.

(5) Under different scenarios of climate change in future, SOS is predicted to contribute to 10% of water yield reduction by promoting evapotranspiration, and its impact on surface runoff is uncertain. In watershed planning and regional ecological engineering construction, it is necessary to balance the ecosystem protection and water resources carrying capacity. On one hand, it should give full play to the functional advantages of vegetation in soil and water conservation, and natural carbon fixation. On the other, it is of great importance to avoid the negative effects of the increase of water dissipation and the decrease of water yeild caused by phenology dynamics on the water resources reserves of the watershed.