生态过程模型敏感参数最优取值的时空异质性分析——以BIOME-BGC模型为例

生态过程模型是当前研究陆地生态系统水循环、碳循环有力的工具,但此类模型参数众多,参数的合理取值对模型模拟结果有重要影响.以往研究对模型参数的敏感性以及参数的优化取值有诸多的分析和讨论,但有关参数最优取值的时空异质性关注较少.本文以BIOME-BGC模型为例,在常绿阔叶林、落叶阔叶林、C3草地3种植被类型下,通过构建敏感性判别指数,筛选出模型的敏感参数,并在每种植被类型下选取两个试验站点,使用模拟退火算法结合实测通量数据构建目标函数,获取各站点敏感参数逐月的最优取值,然后构建时间异质性判别指数、空间异质性判别指数和时空异质性判别指数对模型敏感参数最优取值的时空异质性进行定量分析.结果表明:BIOME-BGC模型在3种植被类型下遴选出的敏感参数大部分一致,少数有差异,但参数的敏感性强弱在不同植被类型下的表现不尽相同;BIOME-BGC模型敏感参数的最优取值,大都具有不同程度的时空异质性,但不同植被类型下,敏感参数最优取值的时空异质性表现各异;敏感参数中与植被生理、生态相关的参数,其时空异质性相对较小,而与环境、物候相关的参数,其时空异质性普遍较大;在3种植被类型下,模型敏感参数最优取值的时间异质性与空间异质性表现出显著的线性相关性;依据其最优取值的时空异质性,可对BIOME-BGC模型敏感参数进行类型划分,以便在实践应用中采取不同的参数率定策略.本研究结论有助于加深对生态过程模型参数特性及最优取值的理解,可为实践应用中模型参数的合理取值提供一种思路和参考.

Temporal and spatial heterogeneity analysis of optimal value of sensitive parameters in ecological process model: The BIOME-BGC model as an example.

The ecological process models are powerful tools for studying terrestrial ecosystem water and carbon cycle at present. However, there are many parameters for these models, and weather the reasonable values of these parameters were taken, have important impact on the models simulation results. In the past, the sensitivity and the optimization of model parameters were analyzed and discussed in many researches. But the temporal and spatial heterogeneity of the optimal parameters is less concerned. In this paper, the BIOME-BGC model was used as an example. In the evergreen broad-leaved forest, deciduous broad-leaved forest and C3 grassland, the sensitive parameters of the model were selected by constructing the sensitivity judgment index with two experimental sites selected under each vegetation type. The objective function was constructed by using the simulated annealing algorithm combined with the flux data to obtain the monthly optimal values of the sensitive parameters at each site. Then we constructed the temporal heterogeneity judgment index, the spatial heterogeneity judgment index and the temporal and spatial heterogeneity judgment index to quantitatively analyze the temporal and spatial heterogeneity of the optimal values of the model sensitive parameters. The results showed that the sensitivity of BIOME-BGC model parameters was different under different vegetation types, but the selected sensitive parameters were mostly consistent. The optimal values of the sensitive parameters of BIOME-BGC model mostly presented time-space heterogeneity to different degrees which varied with vegetation types. The sensitive parameters related to vegetation physiology and ecology had relatively little temporal and spatial heterogeneity while those related to environment and phenology had generally larger temporal and spatial heterogeneity. In addition, the temporal heterogeneity of the optimal values of the model sensitive parameters showed a significant linear correlation with the spatial heterogeneity under the three vegetation types. According to the temporal and spatial heterogeneity of the optimal values, the parameters of the BIOME-BGC model could be classified in order to adopt different parameter strategies in practical application. The conclusion could help to deeply understand the parameters and the optimal values of the ecological process models, and provide a way or reference for obtaining the reasonable values of parameters in models application.