Tackling challenges in process-based forest modelling: from concept to uncertainty
19 December 2017
Campus Drie Eiken, Promotiezaal Q0.02 - Universiteitsplein 1 - 2610 Antwerpen-Wilrijk (route: UAntwerpen, Campus Drie Eiken
Organization / co-organization:
Department of Biology
Reinhart Ceulemans & Gaby Deckmyn
Public defence of the PhD thesis of Mrs. Joanna Horemans - Faculty of Science - Department of Biology
Based upon the close relationship between forests and the atmosphere, and within the framework of a changing world, both climatologically and socially, the prediction of forest change is essential for sustainable forest management. Therefore, process-based forest models (FMs) are needed for predictions. Through the incorporation of physical and mechanistic processes they provide excellent tools for system understanding and hypothesis formulation. However, they are constrained by knowledge of the system, and their results are considered unreliable, i.e. too uncertain, to be used as stand-alone tools for forest management decisions. The goal of this dissertation was to objectively contribute to the further improvement of FMs. Therefore, different case studies were performed to assess challenging aspects of forest modelling, i.e. (i) system understanding; (ii) parameter determination; (iii) sensitivity assessment; (iv) validation and (v) output uncertainty. Different models were used, selected depending on the objectives of each study.
We have quantified the relative importance of biotic and abiotic forest drivers of net ecosystem exchange of carbon (NEE) for a mixed forest in the Campine region of Belgium.
We have reparameterised the AquaCrop model, initially developed for annual crops, to simulate the productivity and water balance of a short-rotation coppice plantation in East Flanders.
The sensitivity of the AquaCrop model was assessed with regard to changes in parameter values.
The strengths and weaknesses of the current model validation methods, were assessed in a case study combining the results of three FMs. The models were used to simulate the NEE of three European beech forests for which at least 16 years of eddy covariance NEE measurements were available. Model performance was studied on different temporal scales.
We have quantified the relative importance of selected sources of uncertainty based on model results of one empirical model and one process-based model. The models were used to make stem biomass predictions for a Slovakian beech forest until 2100.
The main conclusions of this dissertation with regard to the use of FMs are: (i) system understanding should be a main focus, and empirical and process-based theory should be linked more closely; (ii) FM parameter determination and evaluation, i.e. validation and sensitivity assessment, should be comprehensively performed; (iii) better risk assessment can be obtained by running model ensembles and by focusing on the analysis and the communication of output uncertainty; (iv) FM results should be interpreted in a more holistic and probabilistic perspective.