Description succincte
(Anglais)
|
Forest gap models are often used to project forest dynamics under the impacts of anthropogenic climate change, but a systematic approach to model parameterization, validation and uncertainty analysis is lacking to date. We propose to (1) improve the formulation of growth and mortality processes in the ForClim model; (2) identify the key parameters of ForClim using a systematic sensitivity analysis, calibrate and validate the model using Bayesian methods and a wide array of unique datasets available to us; (3) apply ForClim under up-to-date climate change scenarios to assess future trajectories of forest dynamics in the European Alps and quantify the uncertainty that is inherent in such projections. For model development and calibration, we will employ methods of sensitivity analysis and recently operational-ized Bayesian statistics, using the following datasets: Swiss National Forest Inventory; tree-ring investigations; the network of Swiss Growth-And-Yield plots; Swiss as well as Northwestern German Forest Reserve data; and Holocene-long pollen records. Part of the data will be used for model calibration (focusing on the shorter time series), whereas a substantial part will be used for model validation (the longer time series). For model application, we will quantify the uncertainty induced by the different model structures, parameteriza-tions, and climatic inputs (IPCC AR4 vs. AR5) for projections of forest dynamics at a wide range of sites along climatic gradients.
|
Résumé des résultats (Abstract)
(Anglais)
|
Forest gap models are often used to project forest dynamics under the impacts of anthropogenic climate change, but a systematic approach to model parameterization, validation and uncertainty analysis is lacking to date. We propose to (1) improve the formulation of growth and mortality processes in the ForClim model; (2) identify the key parameters of ForClim using a systematic sensitivity analysis, calibrate and validate the model using Bayesian methods and a wide array of unique datasets available to us; (3) apply ForClim under up-to-date climate change scenarios to assess future trajectories of forest dynamics in the European Alps and quantify the uncertainty that is inherent in such projections. For model development and calibration, we will employ methods of sensitivity analysis and recently operationalized Bayesian statistics, using the following datasets: Swiss National Forest Inventory; tree-ring investigations; the network of Swiss Growth-And-Yield plots; Swiss as well as Northwestern German Forest Reserve data; and Holocene-long pollen records. Part of the data will be used for model calibration (focusing on the shorter time series), whereas a substantial part will be used for model validation (the longer time series). For model application, we will quantify the uncertainty induced by the different model structures, parameterizations, and climatic inputs (IPCC AR4 vs. AR5) for projections of forest dynamics at a wide range of sites along climatic gradients. In the present reporting period, two major tasks were accomplished: first, the PhD student wrote a detailed research plan for the project, which partly deepend and partly slightly modified the project plan outlined in the original proposal; and second, the work with the sensitivity analysis was begun, primarily to prove that its design is not only theoretically sound, but also technically feasible on the state-of-the-art computational infrastructure that is available at ETH Zurich. First results from the sensitivity analysis are available already, but a consolidated picture can be provided only with next year’s report.
|
