Proposal of model-based methodologies for assessing climate change vulnerability in agriculture: Focusing on nutrient discharge assessment in agricultural land using a linked APEX and regression model

Scientific, evidence-based vulnerability assessments are vital for predicting and mitigating the climate change risks poses to national systems. This study proposes a climate change vulnerability assessment method by combining a process-based model and machine learning, using nutrient runoff vulnerability in agricultural land as an example. A training dataset was generated using the process-based model APEX (Agricultural Policy and Environmental eXtender) to train a nonlinear regression machine learning model. The trained nonlinear regression model accurately replicated the APEX results, with an R² value of 0.53 for nitrogen in paddy, and 0.72 for nitrogen and 0.71 for phosphorus in upland. This model was then applied to quantify nitrogen and phosphorus discharge at the si (city) and gun (county) levels. The results for each indicator were normalized, and weights were assigned using the Analytic Hierarchy Process (AHP). Vulnerability assessments were then conducted for mid-term (2041 ~ 2070) and long-term (2071 ~ 2100) future scenarios, and compared with the baseline period (1981 ~ 2010). The results showed that the average nutrient discharge vulnerability score for si and gun units was 31.4 in the baseline period, with a minimum of 0.6 and a maximum of 70.0. Under the SSP5-8.5 scenario, the average score increased to 43.6 in the mid-term future and 54.7 in the long-term future. However, when scenarios of compliance with standard fertilization rates for paddy and a 20% reduction in fertilization rates for upland were applied, the average score dropped to 33.3 in the mid-term future and 44.5 in the long-term future. These results highlight potential policy measures to mitigate the increasing vulnerability due to climate change in the agricultural sector and provide a rapid reference for policymakers when designing effective adaptation strategies.