Environmental suitability for cultivation of sweetleaf, Stevia rebaudiana (Asteraceae, Eupatorieae), in South America: an ensemble species distribution modelling approach using bioclimatic and soil variables

Stevia rebaudiana (Bertoni) Bertoni is a South American crop of increasing agronomic relevance because its leaves accumulate steviol glycosides, high-intensity natural sweeteners used as low-calorie alternatives to sucrose. Although the species is native to Paraguay and adjacent regions, its cultivation has expanded beyond its native range, making the identification of suitable growing areas relevant for agricultural planning under climate change. Here, we modelled the current and future environmental suitability for S. rebaudiana cultivation in South America using cultivation records, bioclimatic variables, and soil predictors. We hypothesized that the main suitability core in southeastern South America would persist under mid-century climate change, with gains and losses concentrated primarily along its margins. A total of 142 georeferenced outdoor cultivation records were compiled from GBIF, iNaturalist, speciesLink, herbarium specimens, and specialized literature. After correlation filtering and exploratory variable selection, 13 predictors were retained, including eight bioclimatic and five edaphic variables. Models were calibrated under current environmental conditions using GLM, MaxEnt, and Random Forest within an ensemble species distribution modelling framework. Because GLM and MaxEnt showed the best balance between AUC and TSS, final current–future projections were based on these algorithms. Future suitability was projected for 2041-2060 under SSP245 and SSP585 using HadGEM3-GC31-LL, GISS-E2-1-G, and MIROC6 climate models combined with present-day soil predictors. Current suitability was highest in Paraguay, northeastern Argentina, Uruguay, and southern to southeastern Brazil. The main predictors were related to seasonal precipitation, annual temperature, and soil conditions. Future projections were consistent with this hypothesis, showing broad persistence of the main suitability core under both SSP245 and SSP585, with gains and losses concentrated mainly in marginal and transitional areas. Suitability gains exceeded losses under both scenarios, resulting in net increases in predicted suitable area of 14.15% and 10.92%, respectively. These results provide a spatial framework for identifying candidate regions for stevia cultivation and for prioritizing field validation under current and future environmental conditions.