Who knows about climate change? Determinants of climate awareness among rural farmers in Zambia using supervised machine learning

Climate-related shocks increasingly threaten smallholder agriculture in sub-Saharan Africa, highlighting the need to strengthen farmers’ climate change awareness for effective adaptation. Using survey data from 1499 rural farmers in Zambia, we apply supervised machine learning models – logistic regression, ridge, and Least Absolute Shrinkage and Selection Operator (LASSO) to identify key predictors of climate change awareness, and a random forest classifier to capture nonlinear effects and variable importance. This approach generates robust empirical evidence on the determinants of climate change awareness among rural households, which constitutes the study's core contribution. We then compare model performance using accuracy, sensitivity, specificity, and area under the curve, which are standard evaluation metrics in supervised classification. We find that 85.2% of farmers are aware of climate change and its impacts on agriculture, with the models yielding consistent and high classification accuracy. Key predictors include socioeconomic characteristics (household headship, education, household size, cattle ownership), farm practices (use of crop residues, manure application, agroforestry, adjusting planting dates, yield changes), institutional access (cooperative participation, mobile phone and internet access, farming training), and recent exposure to climate shocks (drought and unusual rainfall). These findings underscore the need to decentralise climate information through cooperatives, agricultural hubs, and strengthened training services to enhance awareness and resilience, supporting efforts to improve adaptive capacity and food security in climate-vulnerable regions.