Cybianthus anthuriophyllus (Primulaceae) was previously known from eastern Ecuador and northern Peru. Here we document the first confirmed occurrences of this species in Colombia and develop a distribution model based on collections from the Andean–Amazonian foothill of Caquetá, Cauca, and Putumayo departments, in order to clarify its conservation. Specifically, to refine its potential distribution and inform conservation efforts, we developed a species distribution model (SDM) using MaxEnt and assessed its extinction risk following IUCN Red List Criteria. The model identified high suitability in the northwestern Amazon, particularly along the Andean foothill, and response curves indicated that annual precipitation and isothermality were the primary predictors of habitat suitability. The species qualifies as Endangered (EN B2ab(iii)) due to its limited area of occupancy, few known locations, and ongoing habitat loss. These findings extend the distribution known range of Cybianthus anthuriophyllus, close the floristic gap in the subgenus Comomyrsine, highlight the conservation importance of the Andean–Amazon foothill corridor, and provide evidence for the continued need to prioritize biological inventories in this area.

Occurrence data collection

During botanical explorations in the municipalities of Florencia (Caquetá department), Piamonte (Cauca department), and Villagarzón (Putumayo department) from Colombia, we recorded the presence of Cybianthus anthuriophyllus. Phenological observations were noted in the field with the support of a local collaborator (see Acknowledgements). To compile the full known distribution of the species, we reviewed the original protologue (Pipoly 1998), consulted specimens housed at COL, CUVC, HEAA, HUA, HUAZ, LAMUA, and UIS (following Thiers 2023), and gathered georeferenced records from the Tropicos database (www.tropicos.org) and GBIF (2026). After removing duplicate entries, records without specimens’ images available before 2023, and records with geographic uncertainty, we retained a set of 14 presence points used for subsequent analyses (S1).

Species distribution modeling (SDM)

We estimated the potential distribution of Cybianthus anthuriophyllus using the MaxEnt algorithm (version 3.3.3k), which is particularly suitable for presence-only data (Phillips et al. 2006). Climatic predictors were sourced from the WorldClim database (www.worldclim.org) at 30 arc-second resolutions (~1 km²). To avoid overfitting under a small sample size (n = 14), we implemented a structured variable selection procedure. Initially, Pairwise Pearson Correlations were calculated, and highly correlated variables (r ≥ 0.8) were excluded to reduce multicollinearity and inflation of variable importance.  Consequently, four bioclimatic predictors with known ecological relevance to premontane humid forest were evaluated: isothermality (Bio3), temperature seasonality (Bio4), annual temperature range (Bio7), and annual precipitation (Bio12).

Model calibration used the following settings: 20 replicate runs, 10,000 background points (after testing lower background values proportional to presence points according to Rausell-Moreno et al. 2025), and a regularization multiplier of 1, which were selected after testing alternative values (0.5, 1.5, and 2) based on response curve evaluation. The accessible area M of the species (Barve et al. 2011) was defined as the Napo Province sensu Morrone (2014), and a polygon shapefile provided by Löwenberg-Neto (2014) was used as a mask layer to constrain the model geographically.

Given the small number of occurrence records, we evaluated model performance using the jackknife validation approach described by Pearson et al. (2007). This leave-one-out method assesses predictive ability by iteratively excluding each point and calculating whether it falls within the predicted suitable area of the model. We used the strictest omission threshold—Lowest Presence Threshold (LPT)—and calculated Pi (proportion of predicted area) and Xi (model success/failure) accordingly. Models including additional correlated predictors produced unstable response curves and did not improve predictive performance under jackknife validation. Therefore, the final model retained two ecologically interpretable and statistically stable variables: isothermality (Bio3) and annual precipitation (Bio12). Bio 3 captures the proportional oscillation between diurnal and annual temperature variation, which is particularly relevant in premontane environments where thermal stability may influence physiological tolerance and niche specialization (O’Donnell and Ignizio 2012). Whereas Bio 12 reflects overall water availability, a key driver in any terrestrial ecosystem (Gutiérrez-Hernández and García 2021). The combination of these two variables captures both thermal buffering and moisture availability, fundamental axes structuring in plant distribution.

Conservation assessment

We assessed the preliminary conservation status of Cybianthus anthuriophyllus based on the IUCN Red List Categories and Criteria proposed by the IUCN Standards and Petitions Committee (2024). We focused on Criterion B, which includes the Extent of Occurrence (EOO, subcriterion B1) and the Area of Occupancy (AOO, subcriterion B2). These metrics were calculated using the GeoCAT tool (www.geocat.kew.org), applying a 2 km² grid size for AOO, as recommended by the IUCN guidelines.

To evaluate the additional subcriteria (a), (b), and (c) required under Criterion B, we used spatial data layers on forest cover and land use obtained from national and global sources of DIVA-GIS (https://diva-gis.org), SIAC (https://www.siac.gov.co), MapBiomas project with the initiatives of Colombia (https://colombia.mapbiomas.org), Ecuador (https://ecuador.mapbiomas.org), and Peru (https://peru.mapbiomas.org). Fragmentation was inferred by overlaying known localities and the modeled distribution on recent forest/non-forest maps to assess habitat continuity and the degree of isolation between populations. Habitat degradation and ongoing decline were assessed based on recent deforestation trends and land use change across the Andean–Amazonian foothills. The number of locations was estimated using the precautionary principle, considering the spatial clustering of records and potential shared threats.