华北平原冬小麦—夏玉米轮作区DLM与CLM5模型模拟对比研究

作物生长与气候的互馈是当前气候变化研究的热点之一。陆面模型作为一项重要的研究工具,其模型框架、算法设计及参数化方案的不同会直接导致模拟结果的不确定性。为探究陆面模型DLM(dynamic land model)和CLM5(community land model)在作物生长及农田热通量模拟方面的差异及原因,评估2个模型在华北平原作物研究中的适用程度,论文开展了冬小麦—夏玉米轮作站点的模拟对比研究。结果显示,DLM的夏玉米叶面积指数和生态系统总初级生产力的模拟值更高,与观测值更为接近;CLM5模型则在冬小麦模拟中略优。DLM的潜热模拟值与观测值的相关性普遍更高,可能反映了DLM采用的彭曼公式、双叶策略比CLM5采用基于水势梯度质量守恒、大叶策略的潜热计算方法更具优势。对于产量,模型当前的估测能力并不理想。总的来说,在默认设定下,2个模型的模拟结果能基本反映研究区农田站点内夏玉米和冬小麦的生长规律,但与实测值存在一定偏差。模型在该区域的适用性可能需要通过添加农田管理措施、算法优化和参数本地化等方式进一步提高。

Comparative study of DLM and CLM5 model simulations at winter wheat-summer maize rotation stations in the North China Plain

The interaction between crop growth and climate is one of the key issues in climate change studies. As an important research tool, land surface models can be used for distinct simulations with different model frameworks, algorithms, or parameterization schemes. In order to investigate the differences of the Dynamic Land Model (DLM) and Community Land Model (CLM5) in estimating crop growth and farmland heat flux and the causes of the differences, model comparisons were conducted at the agricultural stations with summer maize and winter wheat rotation in the North China Plain (NCP). The results show that the estimated leaf area index and gross primary production of DLM for summer maize were better, while for winter wheat, the opposite results were obtained. Nonetheless, DLM performed better in simulating latent heat flux, which may reflect that the Penman formula and two-leaf strategy in DLM is better than the mass conservation algorithm based on water potential gradient and big-leaf strategy in CLM5. However, the capacity of yield prediction was poor. In conclusion, with the default settings, the simulations of the two models can basically reflect the growth characteristics of summer maize and winter wheat in the study area, but there remains certain deviation from the observations. The applicability of the models for the NCP may need to be further improved through the addition of farmland management measures, algorithm optimization, and parameter localization.