关键词: 模型验证框架, 时空代替法, 多源数据, 森林数据匮乏区

Abstract: The forest landscape model (FLM) has been recognized as a crucial tool for simulating the dynamics of forest landscapes. The efficacy of predictions of the model will determine its width and depth of application in forest management. The limited availability of survey data is a main constraint on model validation. Choosing appropriate methods to compensate for data gaps and addressing the challenge of model validation in data-scarce regions can promote the wider application of FLM. In this study, a framework for model validation was established based on multi-source data and spatiotemporal substitution in regions with limited forest survey data. We used the LANDIS PRO forest landscape model to simulate the future forest landscape dynamics of the Qinghai-Tibet Plateau. The specific validation framework was as follows. First, the model parameters were calibrated using forest survey data. Then, the short-term model results were validated using multiple sources of data (forest survey data, remote sensing products, and existing research results). Finally, both spatiotemporal substitution and previous research results were used to validate the long-term results. The validation results showed that the initial simulation results accurately represented the real forest attributes. The short-term simulation results of LANDIS PRO based on multiple indicators were relatively consistent with the validated data at both plot and landscape scales (RMSE<80 t·hm^-2, SD<80 t·hm^-2, R²=0.81). The long-term dynamic characteristics of forests were consistent with previous research, with the model predictions of forest composition, structure, and succession being consistent with the growth trajectory of native forest in the region. This suggests that the established validation framework of forest landscape model under the conditions of the lack of forest resource data can effectively validate the simulation results of forest landscapes.

Key words: model validation framework, spatiotemporal substitution, multi-source data, forest survey data-poor region