<b></b>STRUCTURAL DRIVERS OF CLIMATE-RELATED FOOD SYSTEM CHANGE IN SOUTHERN ZAMBIA’S AGRO-ECOLOGICAL REGION I

Climate change is transforming food systems; however, the intricate interplay of climate-related and non-climate factors that shape their resilience remains poorly understood. This study investigates the factors driving food system change and their varied effects across crucial food system components in Agro-ecological Region I of Southern Zambia, an area experiencing climate stress marked by rising climate instability, growing market integration, and land scarcity. A mixed-methods explanatory sequential design was used to collect household survey data (n = 202) from the districts of Kazungula, Gwembe, Chirundu, and Siavonga. This study applies a multi-model analytical framework that combines penalised logistic regression to address outcome imbalance and small- sample bias, ordinal regression (for ordered severity outcomes), multinomial modelling, and principal component analysis to disentangle perceived food system change from impact severity across food availability, agricultural practices, social relationships, ecosystem services, and environment–food system linkages. The results indicate that 76% of households perceived a change in the food system over the past decade. Although climate change dominated reported narratives (66%), market dynamics and technology were also identified as drivers. Market-driven change was independently associated with a higher odds of perceived food system change relative to climate-driven change (OR = 3.30, 95% CI [1.05–10.40], p = 0.041). After controlling for household characteristics, the effects of gender and district were not statistically significant. Across food system pillars, impact severity was consistently structured by land access and household size rather than by declared drivers. Larger household size significantly increased the odds of severe impacts on food availability (p = 0.002). In contrast, moderate landholdings (6–10 ha) were protective across food availability (p = 0.023), ecosystem services (p < 0.01), and environment–food system linkage (p < 0.001). Social relationship impacts followed a distinct pattern, with older respondents (≥45 years; p = 0.010) and landholding size (6–10 ha; p < 0.001) emerging as key predictors. Principal component analysis of the system-wide food system impact index revealed that approximately 31% of the total variance was attributable to the system, suggesting that, when considered independently, cumulative food system vulnerability is primarily driven by structural factors rather than climate drivers. Findings indicate that food system change is widely perceived. However, the severity of impact across pillars is more strongly associated with structural household conditions, particularly land access and household size, than with reported causal narratives. These associations are robust to penalised estimation and alternative outcome specifications. The overall findings support a compound-risk interpretation, indicating that climate stress is influenced by market exposure and land governance; as a result, climate-resilient food-system policy must address both structural barriers and climate-smart initiatives.