Predicting climate-driven habitat shifts of Varroa destructor using MaxEnt and CMIP6 data

Abstract

Climate change is reshaping global temperature, humidity, and precipitation patterns, key factors influencing habitat suitability and facilitating the spread of invasive species. Varroa destructor, a parasitic mite that shifted hosts from Apis cerana to Apis mellifera, has become one of the most destructive threats to apiculture worldwide. Despite existing control measures, the mite continues to expand its range. In this study, we applied the Maximum Entropy (MaxEnt) model to predict current and future global distribution of climatically suitable habitats for V. destructor. Occurrence records were compiled from the Global Biodiversity Information Facility (GBIF), the Center for Agriculture and Bioscience International (CABI), and published literature. The model achieved acceptable predictive accuracy (mean AUC values = 0.87, TSS = 0.60, CBI = 0.86), demonstrating strong robustness and discriminatory power. We found that isothermality (38.3%), precipitation of the driest month (19.8%), maximum temperature of the warmest month (16.4%), precipitation seasonality (14%), and precipitation of the wettest month (7.1%) were the most influential variables. Projections under three future climate scenarios (SSP245, SSP370, and SSP585) indicate a decline in suitable habitats by the 2050s and 2070s, particularly under SSP585, with an estimated 18.38% reduction in suitable areas. However, parts of the Northern Hemisphere, including North America and Europe, are likely to remain highly suitable. These findings highlight the need for region-specific, sustainable management strategies to prevent further spread of V. destructor in high-risk areas.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-46817-y.

Introduction

Invasive species pose a growing threat to global biodiversity, ecosystem function, and economic stability^1–3. Climate change further accelerates their spread by altering environmental variables such as temperature, humidity, and precipitation, which in turn influence habitat suitability^4,5. These changes enable invasive species to colonize previously unsuitable areas through both natural and anthropogenic dispersal. Ectothermic and parasitic organisms are especially affected due to their physiological dependence on external environmental conditions^6,7. Thus, identifying areas where invasive species are likely to be established under current and future climate conditions is critical for informing surveillance and control efforts.

Honey bees (Apis mellifera) play a vital role in global agriculture through pollination services and honey production^8–10. However, bee populations worldwide are declining due to a complex interplay of factors, among which the parasitic mite Varroa destructor¹¹(Mesostigmata: Varroidae)¹². Originally parasitizing Apis cerana, V. destructor shifted hosts to A. mellifera in the 20th century¹³, leading to widespread infestations across continents. The mite feeds on bee fat bodies and hemolymph¹⁴ and vectors a range of debilitating viruses, contributing significantly to colony collapse¹⁵.

Varroa destructor is now recognized as the primary cause of colony losses in A. Mellifera globally^15–17. Loss rates vary widely, from < 15% in countries like China, Saudi Arabia, and New Zealand, to > 40% in the United States and Latin America^18–21. In Europe, colony losses in Poland increased from 8% to 22% between 2012 and 2016^22,23, while similar trends were reported elsewhere between 2006 and 2008. In South Korea, colony numbers declined from 2.69 million in 2021 to 2.51 million in 2022, largely due to V. destructor-associated viral outbreaks²⁴. These losses threaten the sustainability of both commercial and local beekeeping operations^25,26. Although management approaches, such as acaricide use and selective breeding for resistant bee exist, the mite’s rapid reproduction and adaptability hinder eradication^27,28. Understanding the environmental conditions that support its survival and spread remains essential for regionalized control strategies. Species Distribution Models (SDMs) provide a framework for predicting the geographic range of invasive species based on environmental variables. Among these, the Maximum Entropy (MaxEnt) model is widely used for species with presence-only data and limited occurrence records^29–31. MaxEnt offers high predictive accuracy even with small datasets and complex ecological requirements. It has proven effective in forecasting the distribution of agricultural pests and disease vectors under both current and future climate scenarios³². Given the cryptic and progressive spread of V. destructor, MaxEnt is particularly well suited to identify climatically favorable regions for its establishment and expansion^29,30,33.

While some A. mellifera populations have shown natural resistance to V. destructor in Northeastern United States³⁴, most untreated colonies fail to survive infestations³⁵. The mite continues to expand into new territories, from Asia¹¹, Americas³⁶, Europe³⁷, Africa³⁸, with recent detection in Australia’s Port of Newcastle in 2022³⁹. Despite its global impact, there is paucity of information on spatial distribution of V. destructor on a global scale. In this study, we modeled the global habitat suitability of V. destructor using the MaxEnt model under both current and projected future climate scenarios. Our study seeks to answer these questions; (i) what are the major drivers of V. destructor distribution? (ii) will suitable areas of V. destructor expand or contract in the current time? and (iii) will habitat suitability for V. destructor expand or contract in the future? The specific objectives were: (1) to identify the key environmental variables associated with the mite’s occurrence; (2) to map areas of current global suitability; and (3) to forecast future shifts in habitat suitability. The findings aim to support evidence-based management, inform policy, and guide monitoring efforts to mitigate the spread and impact of V. destructor on apiculture and ecosystem services.

Materials and methods

Species data

Occurrence records of V. destructor were compiled using targeted searches for species name, location information, and geographical coordinates (latitudes and longitudes). Data sources included the Global Biodiversity Information Facility (GBIF), the Center for Agriculture and Bioscience International (CABI), and relevant literature^11,39–45. This search procedure yielded 273 records. To ensure data quality, we excluded spatial outliers, including records located in capital cities, centroids, oceans, and institutional coordinates. To minimize sampling bias and spatial autocorrelation, particularly in regions with high record density, a spatial thinning approach was applied using a 15 km distance filter reducing the records to 263 for the final simulation (Appendix A, Supplementary Figure S1). Data on honeybee colony density or hive-level data were not considered because it provides country-level summaries depending on management techniques, which are too coarse for predictive modeling. Therefore, we used Varroa occurrence records compiled from GBIF, CABI, and published literature, which inherently represent areas where managed honeybee colonies are present, as Varroa may not exist independently of its host.

Environmental variables

Nineteen bioclimatic variables (Appendix A, Supplementary Table S1) for the baseline period (1970–2000) were downloaded from the WorldClim v2.1 database^46–48. Of these, four variables; Mean Temperature of Wettest Quarter, Mean Temperature of Driest Quarter, Precipitation of Warmest Quarter, and Precipitation of Coldest Quarter were excluded due to reported spatial artifacts that may cause artificial climate discontinuities⁴⁹. To address multicollinearity among the remaining 15 variables, we conducted a Variance Inflation Factor (VIF) with correlation analysis, excluding variables with VIF values > 0.7⁵⁰. The final six variables retained for modeling were: Mean Diurnal Range, Isothermality, Maximum Temperature of the Warmest Month, Precipitation of the Wettest Month, Precipitation of the Driest Month, and Precipitation Seasonality. For future projections, we used climate data from three Shared Socioeconomic Pathways (SSPs)—SSP245 (intermediate), SSP370 (regional rivalry), and SSP585 (fossil-fueled development)—sourced from Coupled Model Intercomparison Project Phase 6 (CMIP6)—downscaled via WorldClim v2.1. Projections were made for mid-century (2050s: 2041–2060) and late-century (2070s: 2061–2080) periods. All layers maintained a spatial resolution of 2.5 arc-minutes (~ 5 km²). Climate simulations were generated by the Coupled Model Intercomparison Project Phase 6⁵¹ and analyzed in R v4.4.1⁵². All future climate projections were generated using the Model for Interdisciplinary Research on Climate version 6 (MIROC6) general circulation model from CMIP6. We selected MIROC6 because it provides balanced simulations of global temperature and precipitation, has been widely validated for biodiversity and ecological-climate studies, and offers complete coverage across the SSP245, SSP370, and SSP585 scenarios. The percentage change from present was calculated as: .

Model development

We used the Maximum Entropy (MaxEnt) algorithm to estimate habitat suitability for V. destructor under current and future climate scenarios (Fig. 1)^45,53,54. Out of the 263 occurrence records compiled, 236 were used for training and 27 for testing. The calibration area was delineated using a minimum convex polygon buffered by 8 km around the species’ presence points, within which 10,000 background points were randomly generated. To optimize MaxEnt model complexity, we considered feature classes (Linear [L], Quadratic [Q], Hinge [H], Product [P]) and regularization multipliers (RM = 1–5) which generated 25 candidate models (Appendix A, Supplementary Table S2). The best-performing model, selected based on the minimum corrected Akaike Information Criterion (AICc), used feature classes LQHP and RM = 1. The model parameter optimization was performed using the Automated Tuning and Evaluations of Ecological Niche Models (ENMeval) package in R⁵⁵ and WorldClim v2.1 dataset⁴⁸. Ten-fold cross-validation was performed with 10 replicates. We checked the response curves and Jackknife. Model performance was evaluated using Area Under the Receiver Operating Characteristic Curve (AUC), True Skill Statistic (TSS), and Continuous Boyce Index (CBI)^56–58. AUC, TSS, and CBI values close to 1 indicate excellent performance (Appendix A, Supplementary Table S3). Environmental variable importance was assessed using the Jackknife test in the MaxEnt algorithm⁵⁹. A threshold of maximum test sensitivity plus specificity (≥ 0.3868) was applied to convert continuous probability maps into binary presence–absence maps^60,61. Suitable habitat areas were calculated in square kilometers (km²). All analyses were conducted using MaxEnt.jar in R v4.4.1^52,62.

Fig. 1.

ROC curve of the MaxEnt model based on validation data.

Results

Model performance and key environmental predictors

The MaxEnt model demonstrated strong predictive performance, with a mean Area Under the Curve (AUC_mean), True Skilled Statistics (TSS), and Continuous Boyce Index of 0.9, 0.6, and 0.9, respectively, indicating a high discriminatory power in predicting the global distribution of V. destructor (Fig. 1). This performance suggests the model reliably differentiates suitability from unsuitable habitats based on the selected environmental variables. Of the initial 19 bioclimatic variables evaluated, six were retained after multicollinearity filtering and selection based on ecological relevance (Table 1). These were: mean diurnal range, isothermality, maximum temperature of the warmest month, precipitation of the wettest month, precipitation of the driest month, and precipitation seasonality. Among these, five variables contributed more than 5% to the model’s performance and were identified as the principal predictors of V. destructor distribution: Isothermality (38.3%), Precipitation of the Driest Month (19.8%), Maximum Temperature of the Warmest Month (16.4%), Precipitation Seasonality (14%), and Precipitation of the Wettest Month (7.1%). Mean Diurnal Range was the least influential, contributing only 4.5%. The relative importance of each variable was further validated using a Jackknife test of regularized training gain (Fig. 2). This analysis confirmed that Isothermality, Precipitation Seasonality, Maximum Temperature of the Warmest Month, Precipitation of the Driest Month, and Precipitation of the Wettest Month were the most informative variables for accurately modeling the current and potential distribution of V. destructor, collectively accounting for 95.6% of the model’s explanatory power.

Table 1.

Relative contribution of bioclimatic variables retained in the MaxEnt model for predicting the global distribution of Varroa destructor.

BIO Variable	Description	Percent Contribution
BIO2	Mean Diurnal Range (Mean of monthly (max temp - min temp))	4.5
BIO3	Isothermality (BIO2/BIO7) × 100	38.3
BIO5	Max Temperature of Warmest Month	16.4
BIO13	Precipitation of Wettest Month	7.1
BIO14	Precipitation of Driest Month	19.8
BIO15	Precipitation Seasonality (Coefficient of Variation)	14

Fig. 2.

Jackknife Test for bioclimatic variables. bio2 = Mean Diurnal Range (Mean of monthly (max temp - min temp)), bio3 = Isothermality, bio5 = Maximum Temperature of the Warmest Month, bio13 = Precipitation of the Wettest Month, bio14 = Precipitation of the Driest Month, and bio15 = Precipitation Seasonality.

Influence of environmental variables on the occurrence of Varroa destructor

Response curves were generated to assess how each environmental predictor influences the predicted habitat suitability of V. destructor, while holding all other variables constant (Fig. 3). These curves help isolate the marginal effect of individual bioclimatic variables on species’ occurrence probability.

Fig. 3.

Response curves of environmental variables used in the MaxEnt simulation. Each curve represents a model built with a single variable, illustrating how predicted habitat suitability for Varroa destructor changes as the value of that variable varies, while all others are held constant. A = Mean Diurnal Range (Mean of monthly (max temp - min temp)), B = Isothermality, C = Maximum Temperature of the Warmest Month, D = Precipitation of the Wettest Month, E = Precipitation of the Driest Month, and F = Precipitation Seasonality.

The occurrence probability of V. destructor remained steady with increasing mean diurnal temperature range, peaking around 6 °C, after which it gradually declined with wider temperature fluctuations (Fig. 3A). Similarly, habitat suitability increased with isothermality up to approximately 35%, beyond which it steadily decreased (Fig. 3B). A comparable unimodal response was observed for the maximum temperature of the warmest month, with suitability peaking near 25 °C and sharply declining at higher temperatures (Fig. 3C).

For precipitation-related variables, the predicted probability of V. destructor occurrence increased with the precipitation of the wettest month, reaching an optimum at approximately 150 mm and remained steady to 1200 mm before declining sharply (Fig. 3D). A similar trend was seen for precipitation of the driest month, with optimal suitability at around 50 mm (Fig. 3E). In contrast, the probability of occurrence decreased sharply with increasing precipitation seasonality to about 30 mm with a gradual increase peaking at approximately 125 mm, suggesting that V. destructor prefers more stable precipitation regimes (Fig. 3F).

Overall, these results indicate that V. destructor favors moderate and relatively stable climatic conditions, with well-defined optimal thresholds for both temperature and precipitation. Suitability declines when these environmental parameters exceed critical limits, highlighting the species’ ecological sensitivity to climate variability.

Predicted climate suitability for the global distribution of Varroa destructor

The current potential climatic suitability for V. destructor was modeled based on species occurrence data (Appendix A, Supplementary Figure S1). Under present climate conditions, V. destructor is predicted to potentially establish across all continents except Antarctica (Appendix A, Supplementary Figure S1). However, the probability of occurrence varies geographically. Of the Earth’s terrestrial surface, representing approximately 29.2% of the planet’s total area, about 20% (~ 29.8 million km²) is projected to be climatically suitable for V. destructor establishment (Fig. 4A). The model predicts the highest habitat suitability in the Northern Hemisphere, particularly across Europe and North America.

Fig. 4.

Predicted global climatic suitability for Varroa destructor. A Current climate suitability overlaid with species occurrence records, B Binary map of suitable habitats based on Maximum Test Sensitivity Plus Specificity (MTSPS) threshold (> 0.3868).

Using the Maximum Test Sensitivity Plus Specificity (MTSPS) threshold value of > 0.3868, binary habitat suitability was classified to delineate areas with favorable conditions for V. destructor presence (Fig. 4B). When suitability zones were further categorized, highly suitable areas were predominantly concentrated in Europe and North America. In contrast, moderately suitable habitats were more widely distributed, spanning parts of South America, Asia, Africa, and Australia (Fig. 4B).

Predicted future climatic suitability for the global distribution of Varroa destructor

To assess how climate change may affect the future distribution of Varroa destructor, we modeled habitat suitability under three shared socioeconomic pathway (SSP) scenarios: intermediate development (SSP245), regional rivalry (SSP370), and fossil-fueled development (SSP585) for two future periods—the 2050 s (2041–2060) and 2070 s (2061–2080) (Fig. 5; Table 2; Figure S2). Across all scenarios, the total area predicted to be climatically suitable for V. destructor is projected to decline over time. In the 2050 s, climatically suitable areas are expected to decrease by 10.23%, 9.64%, and 13.54% under SSP245, SSP370, and SSP585, respectively, when compared to current predictions (Table 2). This decline is projected to intensify in the 2070 s, with reductions of 14.11%, 14.31%, and 18.38% under the same respective scenarios. These findings suggest a progressive loss of suitable habitats, with the steepest reduction occurring under the high-emission SSP585 scenario in the 2070s.

Fig. 5.

Predicted global climatic suitability for Varroa destructor. A Current climate suitability overlaid with species occurrence records, B Binary map of suitable habitats based on Maximum Test Sensitivity Plus Specificity (MTSPS) threshold (> 0.3868). Generated using: https://www.R-project.org/ R v4.4.1.

Table 2.

Projected suitable areas (km²) of Varroa destructor under current and future climate scenarios.

Time period	Scenario	*No threshold (km²)	*Threshold (km²)	*Percentage decline from present (%)
Baseline (Current)	–	3.29 × 10⁷	2.95 × 10⁷	–
2050s	SSP245	3.05 × 10⁷	2.65 × 10⁷	10.23
2050s	SSP370	3.03 × 10⁷	2.67 × 10⁷	9.64
2050s	SSP585	2.95 × 10⁷	2.55 × 10⁷	13.54
2070s	SSP245	2.94 × 10⁷	2.54 × 10⁷	14.11
2070s	SSP370	2.89 × 10⁷	2.53 × 10⁷	14.31
2070s	SSP585	2.75 × 10⁷	2.41 × 10⁷	18.38

*Values represent both continuous habitat suitability and binary suitability (threshold ≥ 0.3868), along with the percentage change from present-day suitable areas.

Despite the overall decline in global suitability, several regions, including parts of North America and Europe, are projected to remain highly suitable for V. destructor through both future time periods, based on the Maximum Test Sensitivity Plus Specificity (MTSPS) threshold (> 0.3868) (Appendix A, Supplementary Figures S2 and S3). Additional patches of persistent suitability are predicted in parts of South America, Asia, and Australia. However, Africa is projected to have consistently low climatic suitability under all scenarios for both time periods.

Discussion

As the impact of V. destructor on the apiculture industry continues to intensify, it is increasingly important to develop strategies aimed at limiting its spread and mitigating its effects. Environmental stressors, particularly those associated with climate change, are known to influence the abundance and distribution of honey bee populations^63,64. However, the extent to which these factors also affect the distribution of their parasites remains less clearly defined. In this study, we employed the MaxEnt model to predict the current and future global distribution of V. destructor, based on key bioclimatic variables. Model performance was acceptable, with an AUC, TSS, and CBI values of 0.9, 0.6, and 0.9, respectively, confirming the model’s strong discriminatory power.

Among the environmental variables tested, isothermality, precipitation of the driest month, precipitation of the wettest months, maximum temperature of the warmest month, and precipitation seasonality were the most influential predictors, with optimal occurrence predicted at temperatures below 30 °C. These findings are consistent with earlier studies suggesting that climatic factors are primary drivers of V. destructor occurrence^65,66. This aligns with laboratory observations indicating higher infestation rates and reproductive success of V. destructor within this range^37,67. In support of the idea that V. destructor prefers cooler conditions, Kablau et al.⁶⁸ demonstrated increasing hive temperatures could be used to manage mite infestations. Therefore, seasonal patterns in temperature and precipitation directly shape suitable habitats and the distribution of species⁶⁹.

Our results indicate optimal precipitation thresholds around 150 mm for the wettest month and approximately 45–50 mm for the driest. These findings support previous reports that link the parasite’s occurrence to regions with moderate rainfall and low climatic variability⁴⁵. Additionally, temperature and humidity regimes influence Varroa reproduction, survival, and dispersal⁷⁰. A recent study by García-Figueroa et al.⁶⁷ further reported a significant negative correlation between annual mean temperature and V. destructor infestation. On a local scale, significantly higher infestation rates of V. destructor have been recorded in cooler regions of Ethiopia compared to warmer areas⁷¹. In line with other studies, the precipitation of the wettest month and precipitation of driest months were also identified as influential factors, with optimal values of around 150 mm and near 50 mm, respectively⁴⁵.

Spatially, the current model predicts highest climate suitability for V. destructor across the Northern Hemisphere, particularly in Europe and North America, with smaller patches in the Southern Hemisphere. These areas not only support large populations of A. mellifera but also provide climatic conditions favorable to mite establishment and reproduction. In contrast, the parasite is absent from Antarctica, likely due to extreme temperatures and the lack of host populations⁷². This pattern highlights the importance of environmental conditions in shaping the parasite’s distribution. The severity of the parasite in the Northern hemisphere may be attributed to climatic conditions that foster favorable agroecological environments. These areas often experience optimal temperatures and precipitation for mite establishment⁶⁷. In contrast, areas like Antarctica characterized by extreme cold, lack both suitable climatic conditions and the presence of the host (A. mellifera), explaining the parasite’s absence in that region⁷².

Projections under three shared socioeconomic pathway (SSP) scenarios indicate a consistent decline in climatically suitable areas for V. destructor by the 2050 s and 2070s. The most severe reductions, up to 18.38%, were observed under the SSP585 scenario by the 2070s. This decline aligns with earlier studies predicting future habitat contraction for the species⁴⁵. Conversely, future risk projections from Tanzania indicated mixed responses of the potential spread of V. destructor, highlighting both increase and decrease in the mid-century 2055 and late-century 2085 on different sites⁴¹. The marginal variation in both studies could be attributed to local and global models divergence in predictions due to differences in input data and calibration extent, but both remain valid within their respective spatial contexts. They further indicated that there is a general decline of highly suitable areas of V. destructor in mid and late century across all scenarios⁴¹. However, our findings also suggest that highly suitable areas, particularly in parts of Europe and North America, are likely to remain favorable despite overall reductions. As a result, infestation levels may become more intense in these remaining hotspots.

In this study, our prediction shows suitability in all continents except for Antarctica where A. mellifera is absent⁷². The prevalence of V. destructor in regions where the host is abundant has major implications for honey production, as approximately 85% of global honey is produced in the Northern hemisphere⁷³. According to FAO⁷³, Asia, Europe, and the Americas ranked the world’s leading natural honey producers. All climate change scenarios predicted greater suitability for V. destructor in the Northern Hemisphere compared to the Southern Hemisphere. These areas are expected to worsen in the future due to shrinking suitable habitats and the resulting concentration of probability of presence in remaining favorable zones. These findings suggest that future intensification of V. destructor infestation in these regions could have serious implications for pollination services and food security. The economic impact is substantial: the loss of 1.6 million colonies is estimated to cost the industry approximately $600 million, including losses in honey production, pollination services, and colony replacement⁷⁴. Broader agricultural and economic consequences could affect up to $18 billion annually through reduced crop yields, higher food prices, and disrupted supply chains.

While our model assessed climate suitability with acceptable performance metrics, we did not account for microclimate variations which may influence the establishment of the mite, particularly in regions predicted to be unsuitable. Other factors such as propagule pressure, natural enemies, dispersal abilities, beekeeping practices and host-parasite interactions should be taken into consideration when analyzing the results. Future studies should incorporate higher-resolution data and additional ecological variables to improve prediction accuracy. Although abiotic factors such as colony management practices may slightly influence parasite spread⁷⁵, these were not considered in this study. Despite the model’s strong predictive performance, more occurrence records should be incorporated in the future studies to improve model accuracy and reliability. Also, future models integrating honeybee colony density or hive-level data could improve ecological realism and interpretability. Considering the fact that V. destructor is a parasitic mite of managed honeybee colonies, its occurrence records are limited by the availability and reporting of apiary data rather than by true absence in unsampled regions. Therefore, data scarcity and regional imbalance are inherent limitations of global parasite datasets and future global efforts to develop harmonized Varroa surveillance databases are being encouraged. Also, incorporating global-scale agroecological layers such as land-cover or apiculture-intensity proxies would be an important next step to enhance ecological realism in future models. Nevertheless, our predictive models can guide the use of control agents in high-risk zones⁷⁶. Therefore, we recommend early-warning and surveillance systems in high-risk regions (Europe, North America); stronger biosecurity and trade-inspection measures; and adaptive management and beekeeper-training initiatives in emerging risk zones.

Conclusion

This study identifies key bioclimatic variables shaping the global distribution of V. destructor, with isothermality, precipitation of the driest month, maximum temperature of the warmest month, and precipitation of the wettest months being the most influential. The MaxEnt model demonstrated high predictive power, and projections under three climate change scenarios suggest a decline in suitable habitat for V. destructor by the mid and late 21 st century. Despite this projected contraction, highly suitable areas are expected to persist, particularly in the Northern Hemisphere. These zones may experience more concentrated and severe infestations in the future. Therefore, region-specific, sustainable control strategies must be prioritized, especially in high-risk areas, to protect managed bee colonies and ensure the stability of pollination services and agricultural productivity under changing climatic conditions. Also, practical applications such as informing international bee-movement quarantine, regional risk zoning, and targeted surveillance programs should be implemented by stakeholders. Future research can integrate host (honeybee) distribution data, use ensemble SDMs to reduce uncertainty, and incorporate socio-economic drivers of apiculture expansion.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Peter Njukang Akongte, Owusu Fordjour Aidoo, Dongwon Kim and Yong-Soo Choi: Conceptualization. Peter Njukang Akongte and Owusu Fordjour Aidoo: Methodology. Peter Njukang Akongte and Owusu Fordjour Aidoo: Software; Formal analysis; writing—review and editing. Daegeun Oh, Chang-hoon Lee and Yong-Soo Choi: Validation; Resources; Visualization. Peter Njukang Akongte, Owusu Fordjour Aidoo, Daegeun Oh, and Jin-Myung Kim: Investigation. Peter Njukang Akongte, Owusu Fordjour Aidoo, Daegeun Oh, Chang-hoon Lee, and Jin-Myung Kim. Data curation. Dongwon Kim: Supervision. Dongwon Kim and Yong-Soo Choi: Project administration. Dongwon Kim: Funding acquisition. All authors have read and agreed to the published version of the manuscript.

Funding

This study was supported by the 2024/2025 RDA Fellowship Program of the National Institute of Agricultural Sciences, Rural Development Administration (RDA), Republic of Korea (project number PJ01779802).

Data availability

Data are within the manuscript and available on request from the corresponding author.

Declarations

Competing interests

The authors declare no competing interests.