Post-establishment environmental suitability of Aedes aegypti and Aedes albopictus in Iran under SSP-based climate scenarios

Abstract

Dengue fever is a major and expanding global health threat, exacerbated by climate change. The confirmed establishment of Aedes aegypti and Aedes albopictus mosquitoes in multiple Iranian provinces raises serious concerns about potential outbreaks of dengue and other arboviral diseases. From a One Health perspective, understanding how climatic and environmental conditions shape vector establishment is essential for strengthening integrated surveillance and prevention. This underscores the urgent need for evidence-based projections of vector distribution to guide surveillance and control. This study applied the Maximum Entropy (MaxEnt) model with updated Shared Socioeconomic Pathway (SSP) scenarios to predict the current and future habitat suitability for both vectors across Iran. Occurrence records were compiled globaly and spatially thinned. Eleven uncorrelated bioclimatic and elevation variables were used for modeling, with model performance validated using multiple performance metrics, including AUC, sensitivity, specificity, TSS, and omission rate, based on independent field occurrence data collected during 2024–2025. Results indicate high suitability for Ae. aegypti in southern and southeastern provinces. Aedes albopictus suitability is highest in Northern provinces, with additional areas in the south. Approximately 12,000 rural communities currently reside in suitable zones. Future projections suggest a gradual decline in some regions by the 2050s and 2070s, though core northern and southern hotspots remain. Aedes aegypti is strongly associated with Isothermality, while Ae. albopictus is linked to the mean temperature of the driest quarter, indicating distinct climatic drivers. Model performance was robust across multiple metrics, and recent field detections aligned with predicted high-suitability areas. These findings emphasize the necessity for targeted, climate-informed vector control, particularly in persistent hotspot provinces and at ports of entry. From a One Health perspective, linking environmental suitability with vector surveillance can help anticipate changes in Aedes distribution and inform timely public health action.

Highlights

• •MaxEnt models predict hotspots for invasive Aedes species in Iran.
• •∼12,000 villages are currently at risk of exposure.
• •Future climate change may reduce at-risk areas.
• •Key drivers are temperature stability and dryness.
• •Maps guide targeted surveillance and control efforts.

1. Introduction

Dengue Fever (DF) is now recognized as the most important arboviral disease globally. According to the World Health Organization (WHO), as of 30 April 2024, over 7.6 million dengue cases have been reported globally in 2024 alone, including 3.4 million confirmed cases, over 16,000 severe cases, and more than 3000 deaths [1]. Currently, 90 countries report active dengue transmission, although this number likely underestimates the true burden due to underreporting and limited surveillance capacities. Over 4 billion people are at risk of this disease, with Asia being the most severely affected region, accounting for around 70% of the global disease burden [2].

Dengue virus is transmitted to humans through the bite of infective female mosquitoes, primarily Aedes aegypti and Aedes albopictus, which have rapidly expanded their range over the past half-century [3]. Globally, they are now reported in 167 and 126 countries, respectively [1], [4]. Historically, in Iran Ae. aegypti was documented in Khorramshahr in 1920 and Bushehr in the early 1950s, but its presence disappeared until it reemerged in Hormozgan Province in 2019. It has since been confirmed in Hormozgan, Sistan and Baluchistan, and Bushehr provinces. Ae. albopictus was first reported in 2009, later confirmed in Sistan and Baluchistan in 2013, and reappeared in 2023 in Gilan Province. It has since spread to Mazandaran, Ardabil, Qazvin, and East Azerbaijan provinces [5], [6], [7], [8]. These areas engage in land-sea trade and share borders with dengue-endemic countries, potentially leading to not only health consequences but also economic, political, and social impacts [8], [9], [10]. Since 2024, locally transmitted cases of dengue fever have been reported in Iran, particularly in Sistan and Baluchistan (Chabahar city) and Hormozgan (Bandar-e-Lengeh) provinces [8]. This represents a concerning shift from the period between 2016 and 2024, when all reported cases were imported [11]. Recently, in 2025, a fatal case due to this disease has also been reported in Iran, highlighting the increasing public health significance of dengue fever [12]. Moreover, the established presence of Ae. aegypti and Ae. albopictus in the country raises the risk of transmission of other arboviral diseases such as chikungunya and Zika, which, although currently only reported as imported cases in Iran, have the potential for local transmission if vector populations are not effectively controlled [8], [9], [13]. These developments emphasize the need for coordinated surveillance that considers vector populations, environmental conditions, human health, and the factors that connect them. Such an integrated perspective is consistent with the One Health approach, which recognizes the close interconnection between human, animal, and environmental health and promotes coordinated efforts across these domains to address complex health challenges [14].

From an ecological perspective, dengue transmission is strongly shaped by the interactions between vector biology, environmental conditions, human settlement patterns, and climate variability [15]. Aedes aegypti and Ae. albopictus are highly adaptive species with distinct ecological niches, breeding behaviors, and levels of anthropophily, which influence their capacity to establish stable populations and sustain virus transmission [16]. Changes in temperature, precipitation patterns, land use, and urban expansion directly affect mosquito survival, reproduction, and virus replication dynamics, thereby altering spatial and temporal transmission risk [17]. Epidemiologically, the co-circulation of multiple dengue virus serotypes, increasing human mobility, and the expansion of competent vectors amplify the risk of outbreaks and severe disease, particularly in regions undergoing rapid environmental and socio-economic change [15]. The One Health approach supports the integration of mosquito vector surveillance with environmental and climatic monitoring to better understand and anticipate conditions that may favor vector establishment and disease transmission. Understanding these interactions is therefore essential for integrated vector surveillance, early warning, and evidence-based disease prevention [18]. There is increasing interest in projecting the future distribution of vectors in the country with sufficient lead time to implement effective vector control and public health interventions [19], [20], [21].

Despite the growing public health relevance of dengue fever in Iran, significant knowledge gaps remain regarding the current and future ecological suitability of its primary vectors [22], [23]. In particular, there is limited understanding of how climate change, combined with local environmental conditions and socio-economic drivers, may reshape the geographic distribution of Ae. aegypti and Ae. albopictus at fine spatial scales. Existing studies have been constrained by outdated climate scenarios, limited use of locally confirmed occurrence data, or insufficient integration of ecological drivers relevant to vector establishment and persistence. Few studies have been conducted in Iran regarding the habitat suitability of these two species [13], [24]. One of the main studies conducted in 2023 in Iran used Representative Concentration Pathways (RCPs), which are now considered an outdated generation of climate scenarios. Furthermore, the species occurrence data in that modeling were not based on locally confirmed records [24]. These limitations have reduced the accuracy of previous predictions and hindered evidence-based decision-making for national health planning. Given the recent widespread reports of Ae. aegypti and Ae. albopictus across several provinces in Iran and the availability of updated SSPs [25], there is a clear need to re-evaluate and project the habitat suitability of these vectors using more recent data and advanced climate scenarios. SSPs offer standardized projections of socio-economic development and climate change, integrating drivers such as urbanization, population growth, and climate variability in risk assessments [26], [27]. Addressing these gaps is essential for improving risk assessment, strengthening early warning systems, and supporting evidence-based vector control and disease prevention strategies in Iran.

Moreover, machine learning approaches are increasingy important for modeling vector habitat suitability, as they can capture complex spatial patterns, incorporate diverse environmental variables, and improve early warning systems for targeted interventions [28]. Previous studies on Ae. aegypti and Ae. albopictus have employed models such as Genetic Algorithm for Rule Set Production (GARP), XGBoost binary classifiers, Random Forest (RF), MaxEnt and so on [11], [29], [30], [31]. The MaxEnt model is widely used in species distribution and habitat suitability studies due to its ability to produce accurate predictions from presence-only data and handle complex environmental variables [32], [33].

Unlike earlier studies conducted before the emergence of locally transmitted dengue in Iran, the present work focuses on post-establishment vector dynamics and operational village-level potential habitat suitability assessment. Specifically, this study aims to predict the baseline and future habitat suitability of Ae. aegypti and Ae. albopictus in Iran using the MaxEnt model, and to identify and estimate villages with potentially suitable habitats for these vectors. By incorporating updated local occurrence data and SSPs climatic scenarios, this research generates spatially explicit risk maps that can guide vector control programs and public health preparedness in Iran.

2. Materials and methods

2.1. Study area

Iran is located in southwestern Asia, extending approximately from 25°03′ to 39°47′ N latitude and 44°02′ to 63°18′ E longitude, and covers an area of approximately 1.65 million km². The country is characterized by substantial topographic and climatic heterogeneity, including the Caspian lowlands in the north, the Alborz and Zagros mountain ranges, extensive central plateaus, and the low-lying coastal regions along the Persian Gulf and Gulf of Oman. Climatic conditions vary considerably across the country, from relatively humid and temperate conditions along the southern Caspian coast to warm and semiarid conditions in the southwest and predominantly arid and hyperarid conditions across large parts of central, eastern, and southeastern Iran. Precipitation is also highly heterogeneous, with substantially higher rainfall in the northern and western mountainous regions and very low precipitation across much of the central and eastern plateau [34], [35]. This pronounced variation in temperature, precipitation, elevation, and moisture availability creates diverse environmental conditions across Iran and provides an appropriate geographic and climatic context for assessing the potential distribution and establishment of Aedes species.

2.2. Occurrence data

Presence records for Ae. aegypti and Ae. albopictus in Iran were obtained from peer-reviewed literature, various online scientific sources (Google Scholar, PubMed, Web of Science, SID, Irandoc, Magiran) [5], [6], [7], [36] and entomological surveillance reports for the years 2011–2023. In Iran, these species have been reported since 2019. Therefore, to supplement the limited national dataset, global occurrence records for both vector species were obtained from the Global Biodiversity Information Facility (GBIF) (www.gbif.org) (S1–2 Files). Incorporating occurrence information beyond the invaded range can be particularly useful during the early stages of invasion, when observations within the newly invaded area are sparse [37]. This approach allowed for the identification of ecological patterns and niche characteristics beyond the Iranian context, while ensuring that the resulting projections remained applicable to national risk assessments. Furthermore, updated occurrence data for the years 2024–2025 were obtained through the Ministry of Health and Medical Education, Center for Communicable Disease Management (CCDM) (S3–4 Files). These records were not included in model calibration or training; instead, they were reserved exclusively as an independent dataset for external model validation.

To ensure data quality and minimize spatial bias, we first removed duplicate entries. Although occurrence records were compiled from multiple sources with potentially heterogeneous sampling effort, spatial thinning was used to reduce the influence of uneven sampling effort and clustered occurrences, while ensuring consistency with the spatial resolution of the environmental predictors [38]. The accuracy of the environmental layers used in the study was 5 × 5 km², and to avoid spatial autocorrelation, the distance between the points of presence was reduced to at least 5 km. This ensured a single occurrence point per 5 × 5 km² grid cell. This process was carried out in the R environment using the spThin package [39]. Finally, 8572 and 6780 points were used for modeling the ecological niches of Ae. aegypti and Ae. albopictus, respectively. In addition, 1676 independent presence records for Ae. aegypti and 220 for Ae. albopictus in Iran were reserved exclusively for external testing. The database was then imported into QGIS v3.40.10 and mapped for visualization (S5–6 Figs).

2.3. Geographical variables

A set of bioclimatic variables for the historical data (1970–2000) and future periods (2020–2040, 2041–2060, and 2061–2080) were obtained from the WorldClim database v2.1 (www.worldclim.org) at a spatial resolution of 5 × 5 km². For all future periods, three Shared Socioeconomic Pathways, SSP1–2.6, SSP2–4.5, and SSP3–7.0, from the sixth version of the Model for Interdisciplinary Research on Climate (MIROC6) were selected [40]. MIROC6 is particularly valuable due to its capacity to represent complex climate processes, including monsoon behavior and atmospheric circulation, which are essential for characterizing regional climate variability in Iran. These capabilities provided the rationale for conducting decadal predictions with the MIROC6 model [41]. The selection of these scenarios was made to capture a broad range of plausible future climate and development trajectories. This range allows for a comprehensive assessment of vector distribution under varying socio-environmental conditions. Moreover, the MIROC6 climate model was chosen due to its high performance in simulating temperature and precipitation patterns in arid and semi-arid regions such as Iran, ensuring better regional accuracy for ecological modeling.

In addition, an altitude layer was also downloaded from the WorldClim v2.1 website at the same spatial resolution (5 × 5 km²). To ensure the accuracy and reliability of the analysis, we accounted for multicollinearity among geographical variables. Highly correlated climate variables can introduce biases in statistical models, leading to misleading results. Therefore, collinearity among climatic variables was assessed using the Variance Inflation Factor (VIF) method in the ‘usdm’ package in R software. Variables with a VIF > 10 were removed to minimize redundancy and enhance model performance [42]. A final set of 11 uncorrelated variables was selected from an initial pool of 20 layers (19 bioclimatic variables and one altitude layer) for modeling (Table 1).

Table 1.

Geographical variables used in species distribution modeling.

Abbreviations	Variables
Bio 2	Mean Diurnal Range (Mean of monthly (max temp - min temp))
Bio 3	Isothermality (BIO2/BIO7⁎) × 100
Bio 4	Temperature Seasonality (standard deviation ×100)
Bio 8	Mean Temperature of Wettest Quarter
Bio 9	Mean Temperature of Driest Quarter
Bio 13	Precipitation of Wettest Month
Bio 14	Precipitation of Driest Month
Bio 15	Precipitation Seasonality (Coefficient of Variation)
Bio 18	Precipitation of Warmest Quarter
Bio 19	Precipitation of Coldest Quarter
ALT	Altitude (m)

^⁎Temperature Annual Range (Max Temperature of Warmest Month-Min Temperature of Coldest Month).

2.4. Species distribution modeling

To predict the habitat suitability of the two Aedes species, we employed a widely used machine learning algorithm: MaxEnt [43], using MaxEnt v3.4.3 software. The model was run 10 times (replicates), and the average prediction across replicates was used as the final output for each method. For model training and evaluation, 80% of the occurrence data were randomly selected for training and 20% for testing. In the MaxEnt model, 10,000 background points were randomly generated to represent the available environmental space [44]. This number was selected based on established recommendations and previous applications of MaxEnt, in which 10,000 background points have been used to adequately characterize the environmental background and achieve robust model performance [45], [46]. After running the model, the output maps were clipped to the geographical boundaries of Iran to focus the analysis. In addition to generating habitat suitability maps, we also evaluated the relative importance of geographical variables in determining species distribution. To better understand species–environment relationships, response curves were generated for key variables, showing how habitat suitability changes across environmental gradients. These outputs provide ecological insights into how climatic and topographical factors shape the potential distribution of the target species. All spatial analyses and final map visualizations were conducted using QGIS v3.40.10.

2.5. Identifying villages with potentially suitable habitats for vector species

To identify villages with potentially suitable habitats for these vectors to at least one or both vector species, we first converted the continuous habitat suitability outputs of MaxEnt into binary maps, displaying regions classified as highly suitable compared with regions of no predicted suitability [47]. This conversion was performed using the maximum training sensitivity plus specificity threshold. This threshold was selected based on the training data by identifying the suitability value that maximizes the sum of sensitivity and specificity, thereby providing a balance between correctly identifying presences and absences [48]. The resulting threshold values were 0.36 for Ae. aegypti and 0.38 for Ae. albopictus. Pixels with suitability values equal to or above the corresponding species-specific threshold were classified as suitable, whereas those below the threshold were classified as unsuitable.

Subsequently, the national village layer was spatially overlaid onto the binary suitability maps to determine the number and location of villages intersecting with areas predicted to be suitable for either Ae. aegypti, Ae. albopictus, or both species simultaneously. This spatial analysis was conducted in QGIS v3.40.10, where villages falling within the predicted suitable zones were classified as being at potential risk of vector presence. The procedure was repeated under different climate scenarios and across multiple temporal periods, allowing for the identification of villages with potentially suitable habitats for these vectors under both baseline and projected future conditions. The resulting maps were generated and visualized in QGIS, providing a spatially explicit framework for assessing vector exposure risks.

2.6. Model performance

Several evaluation measures have been developed to assess the predictive performance of habitat suitability models [49], yet the choice of an optimal metric remains context-dependent, and no universally accepted measure exists. Therefore, we employed widely used metrics, including Area Under the Receiver Operating Characteristic Curve (AUC), Sensitivity, Specificity, Omission Rate, and True Skill Statistic (TSS) to provide a comprehensive assessment of predictive accuracy. AUC values closer to 1 indicated stronger discrimination, whereas values near 0.5 reflected random performance [50], [51]. TSS was calculated as Sensitivity + Specificity −1 and used as a complementary performance metric because it accounts for both omission and commission errors and is less affected by prevalence than AUC [47], [52]. All metrics were calculated in R v4.5.3.

For external validation, an independent occurrence dataset from 2024 to 2025 was employed. These records, which had been excluded from the calibration and training phases, were overlaid on the baseline habitat suitability maps generated by the MaxEnt model. Moreover, predicted environmental suitability values were extracted at the locations of independent occurrence records. Records with missing raster values were excluded from calculations of mean suitability and the proportion of occurrences located in suitable areas. Suitable areas were defined using maximum training sensitivity plus specificity threshold, which was 0.36 for Ae. aegypti and 0.38 for Ae. albopictus. This procedure, carried out in QGIS v3.40.10, allowed a spatial comparison between the predicted suitability patterns and the independently observed occurrence locations. By mapping these points against the modeled suitability surfaces, the capacity of the models to anticipate suitable habitats beyond the calibration period was objectively assessed.

3. Results

3.1. Prediction of environmental suitability under climate scenarios in Iran

Our predictions show suitability levels ranging from low (blue) to high (red), highlighting regions where environmental conditions are favorable for the presence of these vectors species. The MaxEnt model under baseline climatic conditions (using historical bioclimatic data) predicted high habitat suitability for Ae. aegypti in the southwestern, southern, and southeastern regions of Iran (Fig. 1). The southern parts of Khuzestan, Bushehr, Hormozgan, Fars, Kerman, and Sistan and Baluchestan provinces are likely to provide favorable climatic conditions for this vector both at present and in the future conditions. Under future climate scenarios, the geographic pattern of environmental suitability for Ae. aegypti is expected to remain largely consistent with baseline conditions. An decrease in environmental suitability is expected in the southwestern region, particularly in Khuzestan and Bushehr province in the 2030s and across SSP2–4.5 and SSP3–7.0 scenarios (Fig. 1).

Fig. 1.

Predicted environmental suitability of Aedes aegypti under baseline and future climate change scenarios using MaxEnt model in Iran. The map was generated using QGIS v3.40.10.

On the other hand, baseline projections indicate that environmental suitability for Ae. albopictus is concentrated in both the northern and southern parts of the country. In the north, suitable areas extend from the northwest through the north and into the northeast, particularly across the provinces of Gilan, Mazandaran, Ardabil, East Azerbaijan, West Azerbaijan, Zanjan, North Khorasan, Golestan, and a small portion of northern Semnan. In the south, high suitability is mainly associated with Fars, Hormozgan, and Kerman, while only limited suitable patches are predicted in Bushehr and Sistan and Baluchestan (Fig. 2). Compared with Ae. aegypti, the magnitude of future changes in habitat suitability is more pronounced. Under projected climate scenarios, a decline in suitability is expected in Golestan, North Khorasan, Semnan, Ardabil, and East Azerbaijan, particularly under SSP3–7.0, as well as in southern provinces during the 2050s and 2070s. Other regions are projected to remain relatively stable, and Gilan and Mazandaran consistently retain high suitability across all climatic scenarios (Fig. 2).

Fig. 2.

Predicted environmental suitability of Aedes albopictus under baseline and future climate change scenarios using MaxEnt model in Iran. The map was generated using QGIS v3.40.10.

3.2. Variable importance and response curves

The most influential geographical variables varied between the two species and models. For Ae. aegypti, Bio 3 (Isothermality), which reflects the relative stability of temperature throughout the year by comparing the mean diurnal temperature range to the annual temperature range was identified as the most important predictors in the MaxEnt model. In the case of Ae. albopictus, Bio 9 (Mean temperature of driest quarter) was most important in the MaxEnt model (Fig. 3).

Fig. 3.

Importance of variables on the distribution of Aedes aegypti and Aedes albopictus based on MaxEnt model. The plot was generated using R-4.4.2v and the ggplot2 package.

Fig. 4 presents the response curves illustrating the relationship between predicted habitat suitability and the most influential variables for each species, as estimated by the MaxEnt model (Fig. 4). According to the MaxEnt response curves, for Ae. aegypti, the variable Bio3 (Isothermality) had a significant influence within the range of 30 to 100. The species showed a peak response between 65 and 75, indicating a preference for regions with moderate to high temperature stability. For Ae. albopictus, the most influential variable was Bio9 (Mean Temperature of the Driest Quarter, in °C). The response to Bio9 ranged from approximately 3 °C to 28 °C, suggesting that moderately warm conditions during the driest season are optimal (Fig. 4).

Fig. 4.

Response curves of Aedes aegypti and Aedes albopictus to the most influential variables, as identified by the MaxEnt model. Bio3: Isothermality (the degree of temperature consistency throughout the year, particularly the relationship between daily (diurnal) and annual temperature variations); Bio9: Mean Temperature of the Driest Quarter, in °C.

3.3. Villages with potentially suitable habitats for dengue vectors

Table 2 presents the spatial distribution of villages in Iran with potentially suitable habitats for dengue vectors under baseline and future climate scenarios projected to the 2080s. During the baseline period (1970–2000), 12,131 villages were identified as having potentially suitable habitats for these vectors. In the near future (2021–2040), the number of villages is projected to decline to 9655 under SSP1–2.6, 9446 under SSP2–4.5, and 9348 under SSP3–7.0. A further decrease is expected in the mid-century period (2041–2060), with the number ranging between 8327 and 8426 villages across the three scenarios. By the late century (2061–2080), the projected number of villages shows the lowest values, reaching 8131 under SSP1–2.6, 7442 under SSP2–4.5, and 7562 under SSP3–7.0 (Table 2).

Table 2.

Projected number of villages in Iran with potentially suitable habitats for dengue vectors under baseline and future climate scenarios.

Time period	Scenario	No. Villages
1970–2000	Baseline	12,131
2021–2040	SSP1–2.6	9655
	SSP2–4.5	9446
	SSP3–7.0	9348
2041–2060	SSP1–2.6	8333
	SSP2–4.5	8327
	SSP3–7.0	8426
2061–2080	SSP1–2.6	8131
	SSP2–4.5	7442
	SSP3–7.0	7562

The spatial distribution analysis revealed that villages with potentially suitable habitats for dengue vectors are located across approximately 13 provinces of Iran. However, the level of suitability for vector presence varies considerably among these regions. Provinces along the northern part, particularly Gilan, Mazandaran, and Golestan, as well as the southern coastal provinces of Hormozgan and Bushehr, host the largest numbers of vulnerable villages due to favorable climatic and environmental conditions. Other provinces, including East Azerbaijan, Ardabil, Zanjan, and North Khorasan in the north, together with Fars, Khuzestan, Kerman, and Sistan and Baluchestan in the south, also contain areas of moderate suitability where villages are distributed. Under future scenarios, the projections indicate a declining trend in the number of villages over time, with more pronounced reductions by the 2050s and 2070s. This decrease is especially evident in provinces such as Golestan, North Khorasan, and Bushehr, where the suitability for dengue vector occurrence is expected to diminish in the future (Fig. 5).

Fig. 5.

Spatial distribution of villages with potentially suitable habitats for dengue vectors in Iran based on habitat suitability predicted by MaxEnt model under baseline and future climate scenarios by 2080s. The map was generated using QGIS v3.40.10.

3.4. Model evaluation and significant geographical variables

The MaxEnt models showed good predictive performance for both Aedes species. For Ae. aegypti, the model yielded an AUC of 0.796, with a Sensitivity of 0.868, Specificity of 0.828, Omission Rate of 0.132, and TSS of 0.695. For Ae. albopictus model performed an AUC of 0.781, Sensitivity of 0.903, Specificity of 0.819, Omission Rate of 0.097, and TSS of 0.722. Overall, the results indicate good agreement between predicted habitat suitability and the observed occurrence records for both species (Table 3).

Table 3.

Performance metrics of the MaxEnt models for Aedes aegypti and Aedes albopictus.

Species	AUC	Sensitivity	Specificity	Omission Rate	TSS
Ae. aegypti	0.796	0.868	0.828	0.132	0.695
Ae. albopictus	0.781	0.903	0.819	0.097	0.722

To further evaluate model performance, a set of independent occurrence records not used in model training, was mapped onto the predicted suitability layers to evaluate the accuracy of the models. For both Ae. albopictus and Ae. aegypti, mapping the test points onto predicted habitat suitability revealed that the majority of these independent test points coincided with areas predicted as highly suitable (Figs. 6), confirming the model's accuracy in identifying favorable habitats. In particular, occurrence data of Ae. albopictus in East Azerbaijan and Zanjan fell squarely within zones of high predicted suitability, despite being excluded from the training dataset (Fig. 6 A, see circled occurrence data). A few exceptions were observed where test points of Ae. aegypti occurred in areas predicted as unsuitable. Notably, these occurrences were singular reports with no subsequent observations, which will be further addressed in the Discussion. (Fig. 6 B, see circled occurrence data).

Fig. 6.

MaxEnt model performance illustrated by independent test occurrence data for Aedes albopictus (A) and Aedes aegypti (B) under baseline condition in Iran. The map was generated using QGIS v3.40.10.

We quantitatively evaluated the predicted environmental suitability at the locations of these independent records. It showed different levels of agreement between observed occurrences and predicted environmental suitability for the two species. For Aedes aegypti, 1470 of 1676 independent records had valid predicted suitability values, with a mean suitability of 0.310 ± 0.086 (range: 0.064–0.448). Of these records, 644 (43.8%) occurred in areas classified as suitable according to the predefined suitability threshold. For Aedes albopictus, 195 of 221 independent records had valid predicted suitability values, with a substantially higher mean suitability of 0.578 ± 0.058 (range: 0.217–0.632). Among these records, 192 (98.5%) occurred within areas classified as suitable (Table 4).

Table 4.

Quantitative external validation of model predictions using independent occurrence records.

Species	Independent records (n)	Valid records (n)	Missing values (n)	Mean suitability ± SD	Suitability range	Records in suitable areas, n (%)
Ae. aegypti	1676	1470	206	0.310 ± 0.086	0.064–0.448	644 (43.8%)
Ae. albopictus	221	195	25	0.578 ± 0.058	0.217–0.632	192 (98.5%)

4. Discussion

In recent years, the introduction of invasive Ae. aegypti and Ae. albopictus mosquitoes, the main vectors of dengue viruse, has raised significant public health concerns in Iran [8], [9], [53]. Iran has substantial climatic and environmental heterogeneity due to its diverse topography and geographic extent. The country ranges from cool and relatively humid conditions in the north and west to warm, semiarid climates in the southwest and predominantly arid to hyperarid conditions in central, eastern, and southeastern regions [34], [35]. These spatial gradients, together with pronounced variability in temperature and precipitation and recent warming trends, shape water availability and habitat suitability across the country. Such variation is particularly relevant to Aedes establishment, as temperature, precipitation, humidity, elevation, and the availability of suitable aquatic habitats strongly influence mosquito survival and population persistence [54]. This study offers a comprehensive evaluation of how climate change is likely altering environmental suitability for both vectors across Iran and how such changes are likely to influence the number and spatial distribution of rural communities at risk of exposure. Our modeling results highlight that under both baseline and projected climate conditions, persistent areas of high suitability exist for Ae. aegypti in the southern and southeastern provinces, and for Ae. albopictus in both the northern and southern parts of the country. The stability of these suitable habitats across different climate scenarios suggests that these vectors will continue to pose long-term risks in regions where environmental conditions remain favorable. Specifically, for Ae. aegypti, the southern parts of Khuzestan, Bushehr, Hormozgan, Fars, Kerman, and Sistan and Baluchestan consistently provide favorable conditions. For Ae. albopictus, highly suitable areas were identified along the northern provinces, including Gilan, Mazandaran, and Golestan, extending into Ardabil, East Azerbaijan, and North Khorasan, as well as in selected southern provinces. These findings align with previous research in Iran and neighboring countries, which has also indicated high suitability for Aedes species in coastal regions [24], [55], [56], [57]. Predicted suitable areas from this study show substantial overlap with those reported by Sedaghat et al. [24], indicating convergence on the core regions where both vectors are likely to be present. Nonetheless, some differences are apparent. For Ae. aegypti, suitability in northern provinces (Gilan, Mazandaran, Golestan) was indicated in Sedaghat et al., while these areas were less pronounced in the present predictions. Similarly, for Ae. albopictus, western provinces such as Kurdistan and Kermanshah were highlighted in their study, whereas these regions appeared less suitable in the current analysis. These variations may reflect differences in methodological approaches, including the use of RCP scenarios in Sedaghat et al. versus SSP scenarios here. In recent years, SSP scenarios have been used as the latest climate models, and not only the greenhouse gas concentration but also how climate change will change in response to socio-economic indicators such as population, economy, land use, and energy change will be considered [27]. Moreover, the reference study did not consider the existence of multicollinearity between climatic layers. It is essential to avoid redundancy and potential biases in species distribution models. Apart from climate, other factors such as using locally validated occurrence records plays a crucial rule in predictions. Modeling new invaders is challenging due to data limitations, disequilibrium with the environment, and potential shifts in ecological niches between native and invaded ranges [58]. While some studies advocate using native range data [59], others favor invaded range occurrences [60] or a combination of both [61], [62]. In some regions of Iran, occurrences of the species have been reported only once; incorporating both global and native range data, as suggested by studies on invasive species [37], allows for a more comprehensive representation of the species' environmental niche, improving predictions even when local observations are sparse. Additionally, model performance is influenced by the choice of environmental predictors, spatial resolution, and data processing steps. Understanding the strengths and limitations of each modeling approach, particularly across different stages of invasion, is essential for producing reliable predictions that can inform monitoring and control strategies [58]. Generally speaking, such factors can influence model outputs and may account for the observed differences in predicted habitat suitability [63].

Beyond methodological performance, it is important to clarify how the present study advances previous investigations of Aedes aegypti and Aedes albopictus distribution in Iran. While earlier studies, including our own, explored the potential environmental suitability of these vectors, they were conducted prior to the recent establishment, geographic expansion, and onset of locally transmitted dengue cases in the country. The present study is explicitly situated in a post-establishment and post-emergence epidemiological context, incorporating newly confirmed national occurrence records and independently validated surveillance data from 2024 to 2025, which were unavailable at the time of previous analyses. Methodologically, this work represents a substantive update through the adoption of SSPs and the MIROC6 climate model, replacing the earlier generation of Representative Concentration Pathways (RCPs). This shift enables a more realistic integration of climate change with socio-economic drivers such as urbanization and population dynamics, which are particularly relevant for dengue vector expansion. In addition, the use of a large, quality-controlled occurrence dataset, spatial thinning aligned with environmental resolution, and external validation using independent post-calibration records strengthens model robustness and predictive credibility. Most importantly, the present study moves beyond suitability mapping as an end point and translates model outputs into operationally relevant village-level potential habitat suitability assessments. By explicitly identifying villages intersecting with predicted suitable habitats under baseline and future scenarios, this work provides actionable insights for targeted surveillance, vector control, and early warning systems. This applied risk-oriented framework, combined with updated data and scenarios, distinguishes the present study as a new generation of dengue vector modeling in Iran rather than a repetitive analysis of earlier work.

Our modeling demonstrates practical value in identifying potential hotspots ahead of confirmed detections. For instance, recent reports of Ae. albopictus in Zanjan and East Azerbaijan were consistent with areas predicted as suitable in our study, underscoring the utility of such models for guiding early surveillance. Provinces such as Khuzestan, Golestan, North Khorasan, and Kerman, where no occurrences of these vectors have yet been reported, were also highlighted as climatically suitable. These regions should therefore be prioritized for intensified monitoring, as early detection is critical to preventing vector establishment. While the presence of vectors in locations predicted as unsuitable may appear to reflect model limitations, such events are not necessarily failures of the modeling approach. Species distribution models are inherently probabilistic, and occasional detections may represent transient or accidental introductions rather than stable populations. For example, Ae. aegypti was once recorded in Sistan and Baluchestan at an airport, but likely failed to persist due to unfavorable local conditions. These cases highlight both the strengths and boundaries of predictive models, which should be interpreted as tools to optimize surveillance and resource allocation rather than as deterministic forecasts. Integrating modeling results with field-based entomological monitoring provides the most effective strategy for anticipating and managing the risk of vector-borne diseases in Iran [64]. However, the external validation using the 2024–2025 occurrence records should be interpreted with caution because the dataset covers a relatively short period and may not fully represent the geographic distribution of the species in Iran. These records provide an independent test of how well the model predicts more recent occurrences, but they cannot fully assess the long-term performance of the models. Further validation using longer-term and more geographically representative surveillance data will be needed to confirm the robustness of our predictions.

The evaluation metrics indicate that the MaxEnt models provided reliable predictions for both Aedes albopictus and Aedes aegypti. The relatively high AUC and TSS values suggest that the models captured meaningful environmental associations rather than random patterns in the occurrence data. The low omission rates, together with satisfactory sensitivity and specificity, further indicate that the predicted suitability patterns were broadly consistent with the observed occurrence records. The model performance was comparable to that reported in previous studies, indicating good predictive accuracy and supporting the reliability and robustness of the SDMs for assessing Aedes habitat suitability [24], [31], [65], [66], [67]. Nevertheless, model performance should be interpreted in light of potential sampling bias, spatial clustering of occurrence records, and the inherent uncertainty associated with presence-only modeling.

Species distribution, particularly on extensive spatial and temporal scales, is fundamentally constrained by abiotic factors [68]. Crucially, our analysis indicates that the two vector species, Ae. aegypti and Ae. albopictus, exhibit distinct responses to key climatic variables. These findings confirm that the climatic limiting factors for different vectors are indeed distinct, necessitating that each species be treated as a unique ecological entity in vector surveillance and public health policy [69]. The distribution of Ae. aegypti was strongly correlated with Isothermality. This bioclimatic variable captures the relative stability of temperature throughout the year by comparing the mean diurnal temperature range to the annual temperature range. It showed a peak response between 65 and 75. This suggests that low thermal variation, less extreme fluctuation in temperature both daily and seasonally, provides optimal conditions for this species' survival, larval development, and reproductive success. For Ae. albopictus was more influenced by the mean temperature of driest quarter. The response to Bio 9 ranged from approximately 3 °C to 28 °C. This suggests a considerable level of thermal plasticity that allows the mosquito to persist under both cooler and warmer conditions. Such flexibility may facilitate its establishment in new environments, particularly in temperate or seasonally dry regions, and aligns with previous evidence of epigenetically driven cold hardiness adaptation in this species [70]. Surprisingly, rainfall was not a major factor in predicting the presence of both vectors. in our study. Our finding aligns with a study in Kenya [31], which also reported precipitation is one of the least important variables for Ae. aegypti, despite some research, like studies in Tanzania [71] and China [72], finding that precipitation during the driest period was highly important. The low importance of precipitation is likely a reflection of the vectors' container-breeding behavior, where human water storage in urban environments provides abundant, buffered larval habitat, making the mosquitoes less reliant on natural rainfall patterns [73].

These ecological and climatic insights have important implications for regional risk assessments. In Iran, the predicted sustained or growing suitability in provinces such as Gilan, Mazandaran, and East Azerbaijan is concerning, given recent detections of Ae. albopictus there. Observed warming trends in northern and northwestern Iran, combined with drought-induced water storage practices, are likely to enhance oviposition opportunities for these vectors, facilitating their expansion [74]. Plus, Iran's geographic position, with its northern Caspian provinces and southern coastal ports, places it at the crossroads of multiple potential introduction routes. This is consistent with broader evidence from Europe and Asia showing that climate change can shift the geographic distribution of Aedes mosquitoes, often expanding their range into previously unsuitable regions [75], [76]. Iran's eastern borders, shared climate characteristics with neighboring Pakistan and Afghanistan, and major international ports expose it to transboundary arboviral threats [2], [24], [77]. Past flooding in Pakistan has coincided with dengue outbreaks, increasing risk in adjacent Iranian provinces [74], [77], [78]. Consequently, systematic vector surveillance at ports and border crossings is critical. Without a strategic, climate-informed vector control framework focusing on high-risk provinces and transit points, Iran risks shifting from potential exposure to active disease transmission. Effective mosquito control requires collaboration across sectors and community awareness on water management and personal protection to curb breeding sites and minimize human-mosquito contact [24], [79], [80]. In line with the One Health framework proposed for vectors surveillance and integrated vector management, effective vector control should consider not only mosquito occurrence and climatic suitability but also the broader environmental and territorial context in which transmission risk emerges. This requires coordination among entomological surveillance, environmental and climate monitoring, public health services, local authorities, and affected communities. In Iran, integrating our village-level suitability maps with these complementary surveillance and management activities could help identify gaps, prioritize interventions, and adapt vector control strategies as climatic and epidemiological conditions change [18]. The projections based on village-level potential habitat suitability analysis revealed that approximately 12,000 rural communities are currently located in areas suitable for vector presence. However, this number is projected to decline steadily across future time horizons, particularly by the 2050s and 2070s. The projected decline refers to the total number of rural villages located within areas predicted to be suitable for at least one of the two Aedes species or for both species simultaneously. Thus, this represents an overall village-level measure of potential habitat suitability and should not be interpreted as a direct decline in the geographical distribution of species. The number of suitable villages may decrease overall because climatic conditions can become less suitable in some areas while remaining or becoming suitable in others. This variation depends on the environmental requirements and climatic responses of each species. Aedes species require specific environmental conditions for survival and development, particularly within a suitable temperature range. Temperatures above or below this range may reduce habitat suitability. Therefore, future climate change may result in both expansion and contraction of suitable habitats across different regions [81], [82]. Similar approaches have been adopted in other studies to identify local transmission foci and improve the precision of vector control strategies [83], [84].

Despite utilizing advanced modeling technique and diverse climate scenarios, this study also has several limitations that future research could address, and it should be considered when interpreting the results. Species-climate relationships were assumed to remain constant over time, despite the possibility of behavioral shifts in response to climate change. Key variables such as human behavior, which may significantly influence on presence of both vectors. Future work should include additional local factors, such as proximity to roads, water sources, livestock density, and land use, to improve prediction accuracy. It is important to note that the village-level estimates in this study are based primarily on climatic suitability and therefore do not represent actual vector abundance or disease transmission risk. Factors aforementioned were not included in the present models and may influence the presence, establishment, and spread of Aedes mosquitoes. For example, urban areas and water-storage practices may provide additional breeding opportunities, while human population density and movement can influence mosquito dispersal and the potential for disease transmission [85], [86]. Therefore, the estimated number of suitable villages should be interpreted as an indicator of potential climatic suitability, rather than a direct measure of public-health risk. In reality, mosquito responses may change over time through adaptation or other ecological processes. In addition, the models do not account for mosquito dispersal or interactions between Ae. aegypti and Ae. albopictus, which may influence their ability to reach and establish in climatically suitable areas [87]. Therefore, these projections represent potential climatic suitability rather than definitive future distributions. Incorporating additional environmental and human-related factors and species dispersal, ecological interactions, and changing species–climate relationships in future studies could reduce these sources of uncertainty.

Plus, land use changes, were not included due to the unavailability of reliable long-term data. Moreover, the lack of absence data meant models relied exclusively on presence-only data, inherently increasing uncertainty. Although the MaxEnt model achieved good predictive accuracy, it has drawbacks. Its exponential probability function can overestimate conditions outside the observed area, and the model acts like a “black box,” assuming species prevalence and sometimes misclassifying absences. It also does not account for species competition or dispersal. Despite these issues, the model provides a useful basis for identifying high-risk areas. An important limitation of this study is the use of a single global climate model (MIROC6) for future projections. It does not capture the uncertainty arising from differences among climate models. The use of multiple GCMs and ensemble-based projections could provide a more comprehensive assessment of climate-related uncertainty and potentially improve the robustness of future species distribution estimates. However, a formal evaluation of inter-model uncertainty was beyond the scope of the present study and is recommended as an important direction for future research.

5. Conclusion

This study underscores the critical importance of integrating updated ecological data and advanced climate scenarios to accurately assess the potential distribution of invasive dengue vectors. Beyond mapping suitable habitats, it highlights the utility of predictive modeling for proactive public health planning, enabling early identification of at-risk communities and optimization of surveillance and control strategies. The work demonstrates that combining validated occurrence data with robust modeling frameworks provides actionable insights, bridging the gap between ecological predictions and practical disease prevention. Ultimately, this approach offers a roadmap for evidence-based vector management in regions facing emerging arboviral threats.

The following are the supplementary data related to this article.

Supplementary material 1

Global and national geographic coordinates of occurrence records of Aedes aegypti used for model training.

Supplementary material 2

Global and national geographic coordinates of occurrence records of Aedes albopictus used for model training.

Supplementary material 3

Geographic coordinates of Aedes aegypti occurrence records in Iran used for model testing.

Supplementary material 4

Geographic coordinates of Aedes albopictus occurrence records in Iran used for model testing.

Supplementary Fig. S5.

Distribution maps of Aedes aegypti and Aedes albopictus in the world. The map was generated using QGIS v3.40.10.

Supplementary Fig. S6.

Distribution maps of Aedes aegypti and Aedes albopictus in Iran. The map was generated using QGIS v3.40.10.

Supplementary Fig. S7.

Global environmental suitability for Aedes aegypti and Aedes albopictus. The map was generated using QGIS v3.40.10

Declaration of generative AI and AI assisted technologies in the writing process

During the preparation of this work, the author used Grammarly to check for spelling and grammar errors. After using this tool, the author reviewed and edited the content as needed and takes full responsibility for the content of the published article.

CRediT authorship contribution statement

Faramarz Bozorg-Omid: Writing – review & editing, Writing – original draft, Investigation, Data curation. Mohammad Rahimi: Writing – review & editing, Methodology. Anooshe Kafash: Writing – review & editing, Methodology. Abbas Rahimi Foroushani: Writing – review & editing, Methodology. Amir Ahmad Akhavan: Writing – review & editing. Robert W. Snow: Writing – review & editing. Samuel K. Muchiri: Writing – review & editing. Abbas Ostadtaghizadeh: Writing – review & editing, Supervision, Methodology, Conceptualization. Ahmad Ali Hanafi-Bojd: Writing – review & editing, Supervision, Project administration, Methodology, Funding acquisition, Conceptualization.

Funding

The present study was carried out with the financial support of the Research Deputy, Tehran University of Medical Sciences, Grant No. 53127.

Declaration of competing interest

The authors declare that they have no known competing interests.

Data availability

All data used for modeling presented in the appendix.