Aedes aegypti ranks among the most significant arboviral vectors worldwide, transmitting pathogens responsible for dengue, yellow fever, chikungunya, and Zika virus infections (Powell and Tabachnick, 2013). Native to sub-Saharan Africa, this species has colonized tropical and subtropical zones globally, predominantly utilizing artificial water-holding containers for larval development in close association with human habitation (Kraemer et al., 2015). Climate change is expected to alter the geographic distribution of Ae. aegypti by modifying thermal and precipitation regimes that govern vector survival and reproduction (Mordecai et al., 2019), potentially facilitating poleward range expansion into previously unsuitable temperate territories (Liu-Helmersson et al., 2016).

Species distribution models (SDMs) correlate species occurrence records with environmental variables to estimate habitat suitability across geographic space and time (Elith and Leathwick, 2009). Among SDM algorithms, Maximum Entropy (MaxEnt) is widely employed for modeling invasive species distributions due to its robust performance with presence-only data (Phillips et al., 2006;Merow et al., 2013).

The Republic of Korea (ROK) maintains no confirmed records of established Ae. aegypti populations through national vector surveillance networks. However, several broad-scale modeling investigations have portrayed portions of the Korean Peninsula as climatically compatible for this species (Kraemer et al., 2018; Kamel et al., 2019). These global-scale assessments often employ occurrence data aggregated from multiple sources without systematic verification of record authenticity, potentially introducing spatial errors (Beck et al., 2014). Investigation of GBIF reveals a solitary Ae. aegypti record attributed to the ROK, embedded within a global occurrence compilation assembled in 2015 (Kraemer et al., 2015). This entry lacks fundamental collection details—specifically collector identification, sampling date, and specimen repository—rendering its authenticity questionable. Incorporation of potentially spurious locality data into predictive frameworks may artificially elevate suitability estimates for regions lacking genuine occurrence documentation (Elith et al., 2010).

This investigation aimed to: (1) construct a species distribution model for Ae. aegypti utilizing verified regional occurrence data while excluding the unverified ROK record; (2) evaluate present and anticipated habitat suitability across the ROK under divergent emission trajectories; and (3) discuss implications of occurrence data quality for invasion risk assessment.

Materials and Methods

The analytical framework comprised four stages (Fig. 1): (1) compilation and spatial filtering of occurrence records from GBIF; (2) acquisition and selection of bioclimatic variables from WorldClim; (3) model optimization through ENMeval; and (4) MaxEnt model construction and projection across current and future climate scenarios. Model performance was evaluated using 10-fold cross-validation, which randomly partitions occurrence data into 10 subsets, iteratively using nine subsets for model training and one for testing (Muscarella et al., 2014).

Occurrence data compilation

Georeferenced Ae. aegypti records were retrieved from the Global compendium of Aedes aegypti occurrence dataset (Page et al., 2016) spanning 14°S to 39°N latitude and 90°E to 160°E longitude. Data cleaning included: (1) removal of records with coordinate precision below 1 km; (2) exclusion of duplicates; (3) elimination of records in oceanic areas; and (4) spatial rarefaction at 1 km resolution (Boria et al., 2014). The singular ROK occurrence was omitted given absent verification of collection provenance.

Environmental predictors

Nineteen bioclimatic parameters from WorldClim version 2.1 at 2.5 arc-minute resolution served as candidate predictors (Fick and Hijmans, 2017; Table 1). Variables demonstrating absolute Pearson correlation coefficients exceeding 0.85 were identified using the "corrplot" package in R (Wei and Simko, 2021), retaining only the highest-contributing member of correlated pairs (Dormann et al., 2013). Final predictors comprised seven variables: annual mean temperature (BIO1), temperature seasonality (BIO4), minimum temperature of coldest month (BIO6), mean temperature of wettest quarter (BIO8), mean temperature of warmest quarter (BIO10), annual precipitation (BIO12), and precipitation of driest quarter (BIO17). Nonclimatic factors were excluded as bioclimatic variables predominantly determine broad-scale distributional limits for this species (Kraemer et al., 2015).

Model development

MaxEnt estimates species distributions by finding the probability distribution of maximum entropy subject to constraints representing incomplete information about species’ environmental requirements (Phillips et al., 2006). The algorithm was executed via ‘dismo’ version 1.3-9 (Hijmans et al., 2022) and ‘ENMeval’ version 2.0.4 (Kass et al., 2021) in R version 4.3.1, interfacing with MaxEnt version 3.4.4.

MaxEnt employs regularization multipliers (RM) to control model complexity and feature classes (FC) to transform environmental variables (linear, quadratic, product, threshold, hinge; Phillips and Dudík, 2008). Model optimization evaluated RM from 0.5 to 5.0 in 0.5 increments combined with feature class combinations (L, LQ, H, LQH, LQHP, LQHPT), yielding 60 candidate models. Optimal parameterization followed corrected Akaike Information Criterion (AICc) minimization (Warren and Seifert, 2011). The selected parameters were RM = 1.0 with feature classes L.

MaxEnt output was generated using the logistic format, producing habitat suitability indices ranging from 0 to 1 (Phillips and Dudík, 2008). In this study, we did not apply a binary threshold to classify suitable versus unsuitable areas, as maximum suitability values across all scenarios remained below 0.5, indicating consistently low habitat suitability throughout the ROK. Instead, we focused on relative changes in suitability values across climate scenarios to assess potential future trends.

Climate projections

Prospective suitability was modeled under three Shared Socioeconomic Pathway (SSP) scenarios: SSP1-2.6 (low emissions), SSP2-4.5 (intermediate), and SSP5-8.5 (high emissions, Riahi et al., 2017). Bioclimatic projections from the MIROC6 (Tatebe et al., 2019) were obtained from WorldClim for periods 2021–2040, 2041–2060, and 2061–2080.

Results

Current climate conditions yielded uniformly low Ae. aegypti suitability throughout the ROK (Fig. 2). Maximum suitability reached 0.157, with a territorial mean of 0.033 (Table 2). Marginally elevated values occurred on Jeju Island and southern coastal zones, though remaining substantially below levels associated with established populations elsewhere.

Variable contribution analysis indicated that BIO1 (annual mean temperature) was the highest contributor (48.1%), followed by BIO8 (20.9%) and BIO6 (16.5%) (Table 3). However, permutation importance revealed that BIO6 (minimum temperature of coldest month) exerted the strongest influence on model predictions (45.9%), suggesting that winter temperatures are the primary limiting factor for Ae. aegypti distribution. Current coldest-month minimum temperatures across the ROK (-12°C to -2°C) fall substantially below the thermal tolerances of Ae. aegypti, which cannot endure prolonged exposure below 10°C (Tun-Lin et al., 2000).

Progressive suitability enhancement emerged across all emission pathways (Table 2). SSP1-2.6 elevated maximum suitability to 0.214 by 2061–2080 (+64.8% from baseline). SSP2-4.5 produced 0.278 peak suitability (+137.3%). Most pronounced shifts materialized under SSP5-8.5, where maximum values climbed to 0.454 by 2061–2080 (+283.8%).

Geographic patterns demonstrated northward displacement of elevated-suitability zones under warming scenarios (Fig. 2). By 2061–2080 under SSP5-8.5, enhanced suitability expanded from Jeju toward the Seoul metropolitan vicinity. Nonetheless, even under the most extreme scenario, maximum suitability remained below 0.5.

Discussion

Our findings indicate that Ae. aegypti establishment prospects in the ROK are negligible under current conditions and remain improbable through the late 21st century. These results contrast with previous global-scale assessments that projected moderate-to-high suitability for portions of the Korean Peninsula (Kraemer et al., 2019;Kamel et al., 2018). Our analysis, which excluded the unverified ROK occurrence record and employed regionally calibrated models, produced substantially lower suitability estimates. This discrepancy likely stems from inclusion of the spurious ROK record in global models, which artificially trained algorithms to recognize Korean environmental conditions as suitable (Elith et al., 2010).

Minimum winter temperatures emerged as the decisive impediment to colonization. Ae. aegypti demonstrates pronounced cold sensitivity, with developmental arrest below 10°C and mortality accompanying sustained cold exposure (Tun-Lin et al., 2000). Consistent with this, BIO6 exerted predominant control over predicted suitability in our model. Current coldestmonth minima (-6°C to -12°C) throughout most ROK territory substantially exceed Ae. aegypti thermal tolerances. Although climate trajectories forecast gradual thermal amelioration, even aggressive emission scenarios fail to elevate suitability to levels observed in regions with established populations.

Our conclusions do not preclude episodic Ae. aegypti introductions. Metropolitan heat island phenomena can generate microenvironments 2–5°C warmer than peripheral areas (Murdock et al., 2017), conceivably permitting transient warm-season persistence in localized urban settings. International transit hubs maintaining regular connections with endemic territories constitute plausible entry points warranting vigilant monitoring (Im et al., 2021).

The discrepancy between our results and previous global assessments highlights the importance of occurrence data quality in invasion risk evaluation. Biodiversity repositories aggregate observations across institutions and methodologies exhibiting variable reliability. For vectors of public health significance, we recommend critical appraisal of occurrence record authenticity, particularly for coordinates near distributional margins or in regions lacking corroborating surveillance records.

Certain constraints merit acknowledgment. Model training drew exclusively from tropical and subtropical populations, potentially underrepresenting responses to temperate thermal regimes. Reliance upon single climate model projections introduces uncertainty addressable through ensemble approaches. Biological processes including dispersal mechanics, interspecific dynamics, and evolutionary adaptation remained unconsidered.

In summary, our assessment excluding the unverified ROK record indicates that Ae. aegypti establishment is unlikely through 2080 across examined scenarios. Progressive climate warming will incrementally enhance habitat compatibility, yet stable colonization appears improbable given persistent thermal constraints. Sustained vigilance at international entry points remains prudent, though present evidence does not substantiate Ae. aegypti as an immediate establishment threat to the ROK.