Global sensitivity analysis of an agroforestry system

Abstract

Agroforestry systems are naturally complex with multiple input and output factors. Understanding how these inputs interact and influence the system can be significant in helping researchers and practitioners make more informed decisions and design more effective systems. Statistical methods such as global sensitivity analysis (GSA) investigate the effect of uncertainty in the inputs of a model on the uncertainty in the model’s output. This can provide informative insights into parameter relationships. However, there is little interdisciplinary work applying GSA in agroforestry thus far. Integrating GSA into agroforestry research has the potential to gain a deeper understanding of these systems and significantly impact the conclusions that can be drawn from the data. In this thesis, this is demonstrated by developing a tool combining the agroforestry simulation Hi-sAFe, developed by the French National Research Institute for Agriculture Food and the Environment (INRAE), and the uncertainty quantification framework UQLabs developed by the Federal Institute of Technology Zurich (ETH Zurich). The tool is tested on an experiment originally run by INRAE investigating the influence of latitude on light availability for intercrops in an alley cropping system. Using GSA, it was possible to identify an input factor not accounted for in this simulation proving an assumption of the original work could be erroneous. Furthermore, the next experiment was focused on agroforestry’s role on carbon farming. Due to the recent European Parliament agreement to certify carbon farming through agroforestry, there will likely be a rise in interest for quantifying and optimising carbon sequestration in agroforestry systems. Thus, GSA was again applied to a Hi-sAFe simulation to understand how the carbon-to-nitrogen ratio of soil organic matter, initial soil organic nitrogen content, and denitrification properties including depth and rate interact to influence total carbon humus stock in agro forestry systems. Results demonstrate that the influence of soil organic nitrogen is most influential. Overall, the main contribution of this work is introducing GSA to agroforestry research. The development of the GSA tool combining UQLabs and Hi-sAFe makes it accessible for agroforestry researchers to use GSA enabling them to gain deeper insights from their experiments.