A Multivariate Empirical (MVE) simulation approach to modeling sunflower yield risk and productivity under alternative farming practices in Tanzania
Year: 2026
Type: article
Abstract: Increasing domestic sunflower (Helianthus annuus L.) production is important to Tanzania's efforts to reduce edible oil imports and improve smallholder livelihoods. Yet yields remain low and vary considerably among production systems. This study used a Multivariate Empirical (MVE) model within a Monte Carlo simulation framework to examine yield distributions and production risk under six combinations of seed and fertilizer use. We obtained nationally representative household data from the 2019/20 National Sample Census of Agriculture. The systems comprised local seed with no fertilizer (LS_F0), organic fertilizer (LS_F1), or inorganic fertilizer (LS_F2), and improved seed with no fertilizer (IMS_F0), organic fertilizer (IMS_F1), or inorganic fertilizer (IMS_F2). For each system, we simulated 500 yield outcomes from the observed data using Latin Hypercube sampling. We then used empirical probability density functions and Stoplight charts to estimate the likelihood of yields falling below 1.0 t ha-1 or rising above 2.0 t ha-1. Nearly 87% of the sampled sunflower-producing households used local seed without fertilizer. Simulations for this system showed a 54% probability of yields below 1.0 t ha-1 and only a 3% probability of yields above 2.0 t ha-1. Applying inorganic fertilizer with local seed increased the probability of exceeding 2.0 t ha-1 to 7%. The combination of improved seed and inorganic fertilizer produced the highest simulated mean yield (1.14 t ha-1) and the highest probability of exceeding 2.0 t ha-1 (16%). However, this result should be treated cautiously because the system was represented by only 11 observed households. Improved seed combined with organic fertilizer performed comparatively poorly. The results show that seed and fertilizer combinations are associated with distinct yield distributions and production uncertainty. They provide a basis for designing productivity interventions that consider both expected outcomes and production risk. Further research using larger samples, multi-year observations, and explicit climatic variables is required to verify these patterns and determine their relevance to climate resilience.
Source: PLoS ONE
Institution: Mbeya University of Science and Technology
Language: English
Subfield: Plant Science
Domain: Life Sciences
SDG: Zero hunger
Open Access status: gold
Estimated APC paid: $2,382
OpenAlex ID: W7236418944
