Robust sensitivity analysis to uncertainties in environmental and socio-economic scenarios: A perspective from a global socio-ecological system model

Abstract

Uncertainties exist in modelling and simulating complex socio-ecological systems due to unknown, unmeasurable, and uncontrollable occurrence probabilities of future conditions, which are often expressed through multiple representative future scenarios. The sensitivities of uncertain scenario parameters that comprise these scenarios may be different under different scenarios. To largely reduce unnecessary efforts of refining insensitive parameters, it is helpful to identify scenario parameters that remain sensitive in all plausible future scenarios. Based on a well-established global socio-ecological system model—Functional Enviro-economic Linkages Integrated neXus (FeliX), this study conducted robust sensitivity analyses of the scenario parameters that defined five representative environmental and socio-economic scenarios. The study explored their sensitivities on eco-environmental carrying capacity represented by six planetary boundary (PB) indicators. A three-step method was proposed for these analyses. First, an uncertainty analysis was adopted to explore the impacts of uncertain scenario parameters on PB indicators in each scenario using the Latin hypercube sampling and nonparametric Kruskal-Wallis test. Second, a sensitivity analysis was utilized to quantify the global sensitivity of each parameter using the Extended Fourier Transform Sensitivity Test (eFAST). Third, we conducted a robust sensitivity analysis and identified 12 common sensitive parameters that remained sensitive under all five scenarios by combining the eFAST's results according to four decision criteria representing different decision preferences. This study provides a better understanding of the sensitivity of scenario parameters on PB indicators influenced by future uncertainties across five representative scenarios in a complex socio-ecological system modelled by FeliX. The findings make it easier to improve and apply FeliX or other socio-ecological system models by focusing on key parameters and understanding important linkages of sectors in socio-ecological systems.