Modelling the terrestrial carbon cycle – drivers, benchmarks, and model-data fusion

Abstract: The terrestrial ecosystem sequesters about one-third of anthropogenic emissions each year, thereby providing a critical ecosystem service that slows the rate of increase of atmospheric carbon dioxide and helps mitigate climate change. Observed atmospheric carbon dioxide concentrations exhibit a large inter-annual variability which is considered to be caused primarily by the response of the terrestrial ecosystem to climate change and anthropogenic activity. A better understanding of the functioning of the terrestrial ecosystem is therefore required to improve our ability to predict the global carbon cycle and climate change.Ecosystem models integrate and apply knowledge of ecological processes (e.g. photosynthesis, respiration, allocation, and other plant physiological and microbial processes) to simulate net primary production, biomass accumulation, litterfall and soil carbon amongst others, in terrestrial ecosystems worldwide. These models are widely applied to explore, analyze and further our understanding of the complex interactions among biomes as well as the flows of carbon, nutrients and water through ecosystems over time in response to climate change and disturbances. Ecosystem models also allow the projection of the evolution of the carbon cycle under different scenarios of future possible carbon dioxide concentrations. However, current studies have demonstrated large uncertainties in predictions of past and present terrestrial carbon dynamics which limits our confidence in projections of future changes. These uncertainties, originating from model structure, parameters and data that drives the model, greatly limits our ability to accurately assess the performance of ecosystem models as well as our understanding of the response of ecosystems to environmental changes.This thesis aims to analyze these caveats by disentangling the causes of uncertainties in modeling terrestrial carbon dynamics to inform future model improvement. A state-of-the-art ecosystem model LPJ-GUESS is employed as the model platform for this study. Climate data induced uncertainty in model-based estimations of terrestrial primary productivity are analyzed and quantified for different ecosystems. Also, different climate variables are identified as the main contributors to total climate induced uncertainty in different regions. In addition, this thesis assesses the suitability of contemporary climate datasets with respect to a given research purpose and study area, and quantifies the effect of land use and land cover changes on the terrestrial carbon sink. Moreover, a matrix approach, which reorganizes the carbon balance equations of the ecosystem models into one matrix equation while preserving dynamically modeled carbon cycle processes and mechanisms, is applied to identify which ecological processes contribute most strongly to model-data disagreement in term of terrestrial carbon storage and flux.Identifying and reducing uncertainty in estimations of the terrestrial carbon cycle via a modeling approach enables us better understand, quantify, and forecast the effects of climate change and anthropogenic activity on the terrestrial ecosystem, but is also of increasing relevance in the context of climate change mitigation policies.