Regional-Scale Data Assimilation of a Terrestrial Ecosystem Model: Leaf Phenology Parameters Are Dependent on Local Climatic Conditions

Abstract

Model optimization using data assimilation is an effective tool for reliable projections of environmental changes. To date, however, data assimilation has not been widely applied for terrestrial ecosystem models, especially in large-scale studies, owing to specific difficulties including heterogeneity and abruptness in terrestrial processes. To overcome the difficulties arising from the complex and abrupt behaviour of the terrestrial ecosystem model, the data assimilation by particle filter, a non-parametric and computationally intensive parameter optimization method, was applied in this study. We simultaneously optimized nine model parameters of a terrestrial ecosystem model with a satellite-based leaf area index. The optimized model successfully reproduced the leaf onset and offset phenology of temperate deciduous forests in mainland Japan. We formulated the relationship between local climate and leaf onset and offset timings which indicates that warmer temperatures were required for leaf onset in the warmer southern parts of Japan, and the northern forests retained their leaves under much colder temperatures, relative to southern forests. Unlike the findings of conventional phenology models using crude estimation with limited local data, the results of this study were based on regional big data and objective optimization. This research thus shows that data assimilation can be used to optimize complex terrestrial ecosystem models.