基于中国水稻种植区划的水稻模型优化研究

A variety of rice models have been developed in the last three decades. However, simulations results from these models are largely different, especially in yields and phenology events. A rice model may provide good estimations for a specific region but it may have a poor performance for other regions, because it may be developed based on observations obtained in a region with similar climate conditions. Rice is widely planted in many regions with different climate conditions in China. Typically, it is difficult to get a good estimation of rice phenology and yield by using single one rice model even it has been calibrated. In this study, a map for rice planting zones in China will be used as a reference. Several different rice development temperature response functions, which are the key functions for simulating phenology in rice models, will be evaluated. At the same time, several rice models will also be compared against observations at the plant zone level. The function and model with the best performance in estimating phenology and yields will be selected for a specific plant zone. Then, the parameters used in the selected model will be calibrated at each plant zone. Finally, the model which are optimized at plant zones will be used to simulate rice phenology and yields in China.

经过近30年的发展，水稻模型已具备模拟水稻生长发育和产量的能力，但水稻模型采用的水稻发育速率温度响应函数和产量模拟方案各不相同，直接导致水稻模型的物候期和产量模拟结果差异大。尤其是我国水稻种植区域分布广，不同种植区气候差异显著，更增加了我国水稻模拟研究的难度。大量研究表明，针对某一特定区域设计和优化的模型，比通用模型具有更高的模拟精度。因此，在中国针对不同水稻种植区的气候特点，选择适当的水稻模型，开展模型优化研究是减少我国水稻模拟研究不确定性的有效途径。本研究拟以在中国水稻种植区划数据的基础上，开展水稻发育速率温度响应函数对比研究，以水稻物候期为指标，选择各水稻种植亚区最适的函数形式；将最适的水稻发育速率温度响应函数耦合进入多个水稻模型，以水稻产量为指标选择各水稻种植亚区最适的水稻模型，开展水稻模型参数优化研究，提高中国水稻产量的模拟精度，最后对水稻模型的模拟精度及不确定性进行评估。

水稻是世界上重要的粮食作物之一。在气候变化背景下,作物模型常被用来研究气候变化对水稻生产的影响。在作物模型中，物候模拟是一个重要部分。本研究收集了1981-2015年中国气象局农业气象观测站点的气象数据、田间管理数据和水稻物候期及产量数据,选择了4种有代表性的温度响应函数进行对比，结果表明每个温度响应函数在结构上存在着差异，由此导致模型参数的初始化上也存在着差异。我们把W响应函数和J响应函数耦合到MCWLA-Rice模型的物候模块中，在发生了高温热害的站点上，改进后的模型对水稻物候期模拟的精度高于其它没有发生热害的站点。在水稻模型优化方面，利用贝叶斯模型平均方法估算4种函数形式的权重，选取权重最大的作为该种植亚区最适的函数形式。研究结果表明华南双季稻稻作区的最适模型为J函数，华中双单季稻稻作区的最适模型为W函数，西南高原单双季稻稻作区的最适模型为J函数，华北单季稻稻区的最适模型为M函数，东北早熟单季稻稻作区的最适模型为M函数，西北干燥区单季稻稻作区的最适模型为J函数。