干旱区天然草地退化-恢复过程的生态水文耦合模拟方法研究

Due to the impact of climate change and human activities, the grassland cover in China has changed significantly during recent years. Traditional process-based integrated eco-hydrological models commonly cannot simulate the grassland degradation and restoration processes. The middle and downstream of the Heihe River Basin (HRB) is selected as the study area. Vegetation coverage and biomass are used to represent the degree of grassland degradation and restoration. This study will conduct research on the integrated eco-hydrological modelling of the degradation and restoration of natural grassland in arid region. The main research contents include: (1) Establish a data-driven model (i.e., Deep Neural Networks, DNN) to simulate the degradation and restoration of natural grassland based on machine learning, and reveal the influence of water resources conditions on grassland cover change; (2) Improve the Hydrological-Ecological Integrated watershed-scale FLOW (HEIFLOW) model by enhancing its capacity in modeling the growth of natural grass in arid region, and couple it with the DNN model to develop a model which can simulate the key eco-hydrological processes along with the degradation and restoration processes of natural grassland in arid region; (3) Investigate the responses of natural grassland to climate change and human activities in the HRB, and look for appropriate practices to control the degradation and restoration of natural grassland. The proposed project can help us understand the reason of grassland degradation in arid region, and also can provide us scientific basis for the protection and management of grassland ecosystem.

近年来，受气候变化和人类活动影响，我国草地覆被发生明显变化。传统的基于过程机理的流域生态水文耦合模型很难模拟草地的退化和恢复过程。本项目以黑河流域中下游为研究区，采用植被盖度和生物量表征草地的退化和恢复程度，研究干旱区天然草地退化-恢复过程的生态水文耦合模拟方法。主要研究内容包括：（1）基于机器学习，发展可模拟天然草地盖度变化的数据驱动模型——深度神经网络模型，揭示水资源条件对草地覆被变化的影响机理；（2）对流域生态水文耦合模型HEIFLOW进行扩展改进，增强对干旱区草地系统的模拟能力，将其与数据驱动模型耦合，发展一个可综合模拟流域生态水文过程、草地退化和草地恢复的耦合模型；（3）研究黑河流域中下游天然草地对气候变化和人类活动的响应规律，以及水资源限制条件下天然草地退化-恢复过程的调控机制。本项目将有助于理解干旱区天然草地退化的原因，为干旱区草地资源的保护与综合管理提供科学依据。

干旱区草地生态系统脆弱，易受气候变化和人类活动影响。过程机理模型很难定量模拟草地的退化和恢复过程。本项目以位于我国西北干旱区的黑河流域为研究区，围绕生态水文模拟、草地退化与恢复过程模拟、气候变化和人类活动影响、灌溉模拟与分析等内容展开研究，取得了一系列研究成果。主要包括：（1）发展了有自主知识产权的三维分布式生态水文耦合模型HEIFLOW；（2）开发了可有效预测干旱区草地盖度和归一化植被指数（NDVI）的机器学习模型；（3）通过耦合HEIFLOW和机器学习模型得到了一个可模拟草地退化-恢复过程的集成模型HEIFLOW-Grass；（4）在灌溉模拟与灌溉效率分析方面实现方法和理论创新。本项目的意义主要有：产出的HEIFLOW模型补充了我国的生态水文模型库，提出的模拟方法可为同类模型提供技术参考；探索了耦合过程机理模型和机器学习模型的范式，为同类研究提供借鉴经验；黑河流域研究成果可以为世界范围内其他内陆河流域的研究和管理提供参考。