Feedback Loops Between National Climate Policies and Urban GHG Mitigation: A System Dynamics and Machine Learning Analysis

Monday, July 20, 2026, 11:00 AM

Session: SOC Poster Session (In-person)

The policy-emissions relationship is often analysed at the same spatial scale, for example, country-level policies and country-level emissions. Less attention is paid to assessing the link between national climate policies (NCPs) and city-level greenhouse gas (GHG) emissions. This study addresses that gap by using a hybrid system dynamics and machine learning approach. Following a literature review, a conceptual model was developed in VenSim to map feedback loops, model boundaries, and delays. Supervised machine learning was then used to rank hypothesised variables and examine non-linear relationships empirically with a dataset of nearly 7,000 cities across 66 countries. Preliminary results support the strongest model for CO2 (N=6,967, R²=0.578, RMSE=3,075,615) and total GHG emissions (N=6,932, R²=0.516, RMSE=3,560,281). Built-up volume, NCPs, HDI, green and cropland areas are most strongly associated with emissions outcomes, while NCPs’ effects vary by emission type, time since implementation, and levels of industrialisation. The study contributes a hybrid SD–machine learning approach that explores and empirically supports how multi-level climate governance shapes urban GHG mitigation over time.