A Multi-Regional Time-Varying Trivariate Weather Dependence Model for Maize Yield: Evidence from the Southern Highlands of Tanzania

Abstract This study develops a multi-regional, time-varying trivariate climate-dependence framework to examine how rainfall, temperature, and evapotranspiration jointly influence maize production in the Southern Highlands of Tanzania. Monthly observations from Iringa, Mbeya, Rukwa, and Ruvuma are analysed using copula-based dependence modelling, while rolling-window Kendall’s diagnos- tics track temporal changes in climatic interactions. Results show clear spatial heterogeneity: rainfall–temperature Kendall’s τ ranges from 0.196 (Rukwa) to 0.419 (Iringa), rainfall–evapotranspiration dependence ranges from weak (τ = −0.090) in Ruvuma to strongly negative (τ = −0.531) in Rukwa, and temperature–evapotranspiration dependence reaches 0.252 in Ruvuma. The time-varying analysis also reveals substantial temporal shifts, indicating non- stationary dependence across all regions. To represent compound climatic stress more realistically, we introduce a weighted trivariate weather stress index (WTWSI) that estimates the relative contribution of each driver to maize yield from the data. The findings show that maize productivity is controlled by a cou- pled climatic system that is both region-specific and evolving over time. Overall, the framework improves the basis for yield forecasting, agricultural risk manage- ment, and climate-adaptation planning, and provides a robust quantitative tool for regionalized climate-impact assessment in Eastern and Southern Africa.