Species record occurrences

The geographical coordinates of species occurrences used to build the SDM analysis for A. fraterculus, D. longicaudata, and D. areolatus were obtained using the “occ” function from the “spocc” RStudio package [49]. The occurrence records of these species were sourced from the following databases: The Global Biodiversity Information Facility (GBIF; https://www.gbif.org), The Vertebrate Network (VerNet, http://vertnet.org/index.html), iNaturalist (https://www.inaturalist.org/), and The Integrated Digitized Biocollections (iDigBio). Additional occurrences of D. areolatus were extracted from the records reported by Moreira et al. [50]. After filtering the data to remove duplicates, we obtained a total of 193 records for A. fraterculus, 21 for D. longicaudata, and 205 for D. areolatus. The georeferenced occurrences for each species were then compiled and saved in a “.csv” file for subsequent analysis.

Both occurrences and absences are usually necessary to build SDMs. Nevertheless, due to the lack of true absence data in the databases, “pseudo-absence” locations were generated randomly to characterize the environmental background of the study area. To do this, we used the function “randomPoints” from the “dismo” package [51] in R statistical computing software v4.3.2; [52]. These pseudo-absences are locations where it is assumed that there is no record or presence of the species, thus providing a point of comparison from those locations where the species has been recorded. In this study, a ratio of 2:1 pseudo-absence to presence locations was generated for each species, reducing over-fitting and maintaining the capacity of the model to discriminate between areas of presence and absence [53].

Environmental variables

SDMs use data on environmental conditions in which a species is known to exist to predict suitable habitats for that species. Environmental data for current and future climate conditions in the Americas were obtained from the WorldClim database (Global Climate Data; available at: https://www.worldclim.org/data/index.html) [54]. These data consist of 19 bioclimatic variables derived from temperature and precipitation in “raster” format (Table 1), at a spatial resolution of 2.5 arc-minutes (approximately 4.5 km).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Bioclimatic variables.

https://doi.org/10.1371/journal.pone.0325761.t001

Bioclimatic variables used for SDMs (extracted from www.worldclim.org/data/bioclim.html).

The current environmental conditions were downloaded from the WorldClim database version 2.1 and correspond to historical records from 1970 to 2000 [54]. The environmental conditions for future climate scenarios were obtained from 26 climatic projections or Global Climate Models (GCM) from the Coupled Model Intercomparison Project Phase 6 (CMIP6; available at https://www.worldclim.org/data/cmip6/cmip6climate.html, [55]) (S1 Table). These projections consist of climatic predictions made by various modeling groups based on CO₂ emissions and mitigation effort scenarios up to the year 2100, called Shared Socioeconomic Pathways (SSPs). We selected two SSPs at two time periods, from 2021 to 2040 and from 2041 to 2060: (a) SSP2–4.5, which represents a stabilization scenario that assumes the implementation of climatic policies to mitigate greenhouse gas emissions [56]; and (b) SSP5–8.5, which corresponds to high greenhouse gas emissions pathway and is considered a baseline scenario without any specific targets for climate mitigation [57].

Suppression of collinearity in predictor variables

Collinearity, also known as multicollinearity, refers to the strong interdependence among explanatory variables typically observed in regression models. This phenomenon can lead to two primary issues: inflated estimates of variable effects and uncertainty in model extrapolation [58]. The most common approach for addressing collinearity issues is a priori variable selection, which involves either excluding the most correlated variables or combining them into new explanatory terms [59]. For this purpose, each species occurrence datum was projected into the current environmental conditions database, and the values of each bioclimatic variable at those points were extracted. Subsequently, the Variance Inflation Factor (VIF) was calculated using the “vifstep” function from the “usdm” package [60]. This test excluded the highly correlated variables with VIF values >10; the remaining variables were used to run the SDMs projections.

Modeling and projections

Random Forest is a powerful machine-learning method for building SDMs and training models with small sample sizes [61–63]. This method consists of hundreds of trees constructed using random samples from the databases of occurrences/pseudo-absences and selected bioclimatic variables (predictors). Each tree is built from a bootstrap sample, which is a random sample with replacement from the original dataset. At each node of the tree, the best predictor is selected from a small, randomly chosen subset of the predictor variable pool. This process helps introduce randomness and reduce overfitting of the model [64]. The final prediction is then made by aggregating the predictions of all individual trees. This ensemble approach tends to improve the robustness and generalization performance of the model compared to the use of a single decision tree [65].

Random Forest involves several parameters that control the structure of each tree, the overall structure and size of the forest, and its randomness properties. For example, the number of randomly drawn candidate variables (mtry), the node size, and the number of trees (ntree). The optimal selection of these parameters significantly influences model prediction accuracy [66]. To determine the best values for these parameters, strategies such as “Tuning” have been developed [67]. In this context, parameter selection was performed using the “train” function from the “caret” package [68]. For all predictions made for A. fraterculus and D. areolatus, a ntree of 2000, a mtry of 3, and a node size of 2 with 25 replications were selected. However, for all predictions involving D. longicaudata, a ntree of 2500, a mtry of 3, and a node size of 2, with 10 replications were selected. Subsequently, all the models were adjusted using 75% of the occurrence data to train the model and the remaining 25% for testing purposes. An ensemble was created using all replications for the current projection. For future projections, an ensemble was formed by combining all 25 Global Climate Models for each climatic period (2021–2040 and 2041–2060) and Representative Concentration Pathway (SSP).

SDMs using Random Forest usually provide numerical predictions in the range of 0–1. However, to make the data more practical, it is essential to establish a threshold for interpreting the predictions using levels of suitability and transforming them into binary data indicating species presence or absence. Suitability scores higher than the threshold value are considered suitable areas for the establishment of a given species. In this study, threshold values were obtained for each insect species by extracting all the Receiver Operating Characteristic (ROC) curve data using the function “roc” of the “pROC” package [69]. Subsequently, the threshold was determined as the mean of all the threshold values in each replication for each species (25 for A. fraterculus and D. areolatus, and 10 for D. longicaudata). We used four suitability levels represented on a color gradient from green to red on the plotted maps: a) highly suitable areas depicted in green color correspond to suitability values close to 1; b) moderately suitable areas illustrated in yellow designate suitability values close to 0.5; c) low suitable areas depicted in red show suitability values close to the threshold for each species (0.08 for A. fraterculus, 0.31 for D. areolatus and 0.15 for D. longicaudata); and d) unsuitable areas represented by the absence of color correspond to values below the threshold and close to 0.

Lastly, the suitability values for each predicted scenario and period were converted into binary values using a species-specific threshold. This transformation was conducted for two key purposes: 1) to quantify the suitable and unsuitable areas in square kilometers for all species, facilitating the assessment of expansion areas across different countries in the Americas; 2) to generate maps illustrating the overlap of suitable areas between each parasitoid and the pest A. fraterculus. For the first one, we utilized the “expanse” function from the “terra” package [70]. The resulting data were subsequently saved in a “.csv” file. To aid interpretation, the top 10 countries with the largest suitable areas for the fruit fly A. fraterculus were selected. The same countries were also chosen for quantifying the suitability areas of the parasitoids D. areolatus and D. longicaudata, and the data were visualized using a bar plot.

All the maps generated illustrating the suitable areas for each species, as well as the overlap of suitable areas between the parasitoids and pest, were generated using the function “ne_countries” of the “rnaturalearth” RStudio package [71].

Model performance

To assess the performance of the models, we used the ROC curve analysis, specifically examining the area under the curve (AUC) [72]. The AUC value is a widely accepted metric for evaluating the accuracy of SDMs. Generally, AUC values below 0.7 indicate poor performance, those between 0.7 and 0.9 suggest moderate performance, and values exceeding 0.9 indicate good performance [73]. For each species, the relative importance of the selected bioclimatic variables was determined through analysis of multicollinearity and AUC metrics. This comprehensive approach allowed us to gauge both the predictive power of the models and the significance of individual variables in the SDMs.

The Anastrepha fraterculus SDM

Our models predict that A. fraterculus has an extensive suitable area for establishment in South America, with Brazil exhibiting the largest extent (up to 5 million km²), followed by Peru (up to 1 million km²) and Colombia (up to 0.8 million km²) (Fig 1). In Brazil, the projected suitable areas for the current time occupy 61% of the national territory, with a potential increase of up to 75% under future climate scenarios (S1 Table). These predicted suitable areas were predominant in the southern states (Rio Grande do Sul, Santa Catarina, Paraná), southeastern states (São Paulo, Minas Gerais, Rio de Janeiro, Espírito Santo), and parts of Bahia State in Brazil (Fig 2). In Peru, the projected suitable areas for A. fraterculus extend up to 98% of the country, with minimal variation in future projections (S1 Table); these suitable areas were predicted in the Andes mountains and adjacent valleys (Fig 2). Conversely, Colombia is projected to experience a significant expansion of suitable areas, from 70% to 99% of its territory under future scenarios (Fig 1), with new suitable areas emerging in the Plains (East Colombia) and Amazon (South Colombia) regions, showing moderate to low suitability for A. fraterculus (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Top 10 countries with the largest suitable areas (Km²) for Anastrepha fraterculus.

Predicted suitable areas for the parasitoids Diachasmimorpha longicaudata and Doryctobracon areolatus and the pest A. fraterculus in current and future time predictions [(SSP2-4.5, SSP5-8.5), (2021-2040, 2041-2060)].

https://doi.org/10.1371/journal.pone.0325761.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Suitable areas predicted for Anastrepha fraterculus in the Americas.

(A). Suitable areas predicted in the current time period; (B-E). Suitable areas predicted under two future climate change scenarios (SSP2-4.5 and SSP5-8.5) (periods 2021-2040 and 2041-2060); (B) scenario SSP2-4.5 time period 2021-2060; (C) scenario SSP2-4.5 time period 2041-2060; (D) scenario SSP5-8.5 time period 2021-2040; (E) scenario SSP5-8.5 time period 2041-2060. The maps show suitability values in a green to-red color where: high suitable areas depicted in green color correspond to suitability values close to 1; moderately suitable areas illustrated in yellow designate suitability values close to 0.5; low suitable areas depicted in red show suitability values close to the threshold of the species (0.08); unsuitable areas represented by the absence of color correspond to values below the threshold and close to 0.

https://doi.org/10.1371/journal.pone.0325761.g002

In North America, current time predictions indicate suitable areas for A. fraterculus along the east coast of the United States (0.9 million km²) and in Mexico (0.8 million km²) (Fig 1), particularly in the coastal states of Yucatán, Campeche, Chiapas, Oaxaca, and Tabasco. In the United States, the suitable areas for this pest are predicted to increase in future climate scenarios by 12% of its territory, concentrated in the states of Ohio, Indiana, and Kentucky (Fig 2). In contrast, Mexico was projected to experience a slight decline of suitable areas for A. fraterculus from 39% to 36% of its territory in future scenarios (S2 Table). The Mesoamerican countries such as Panama, Costa Rica, Guatemala, and Honduras, display low to high suitability across nearly all their territories in both current and future projections, with minimal temporal variation (Fig 2).

The fruit fly parasitoid, Doryctobracon areolatus SDM

Moderate to high suitable areas for D. areolatus were identified across South America, particularly along the western and coastal regions of Brazil, as well as in northern Venezuela, Paraguay, Colombia, and Bolivia (Fig 3). In contrast, areas with low suitability were predicted in the Amazon region and some areas in the Andes mountains, spanning Chile, Peru, Ecuador, and Colombia. Additionally, some areas were classified as low to moderately suitable for the establishment of D. areolatus in Central and North America (Fig 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Suitable areas predicted for Doryctobracon areolatus in the Americas.

(A) Suitable areas predicted in the current time period; (B-E). Suitable areas predicted under two future climate change scenarios (SSP2-4.5 and SSP5-8.5) and periods (2021-2040 and 2041-2060); (B) scenario SSP2-4.5 time period 2021-2060; (C) scenario SSP2-4.5 time period 2041-2060; (D) scenario SSP5-8.5 time period 2021-2040; (E) scenario SSP5-8.5 time period 2041-2060. The maps show suitability values in a green to-red color where: high suitable areas depicted in green color correspond to suitability values close to 1; moderately suitable areas illustrated in yellow designate suitability values close to 0.5; low suitable areas depicted in red show suitability values close to the threshold of the species (0.31); unsuitable areas represented by the absence of color correspond to values below the threshold and close to 0.

https://doi.org/10.1371/journal.pone.0325761.g003

In Brazil, current projections estimate that 69% of the territory is suitable for the establishment of D. areolatus, with future climate scenarios predicting an increase of up to 89% (Fig 1). In Peru, suitability is projected to increase from 16% to 36%, and in Colombia, from 26% to 56% (Fig 1), mainly in the Plains and Amazon regions (Fig 3). In Mexico, the current suitable areas predicted for D. areolatus are concentrated in the southeast, southwest, and central-south (Fig 3), with future projections suggesting a 10% increase (Fig 1; S2 Table). In the United States, suitable areas were consistently identified along the Southeast Coast, particularly in Florida, under all projections (Fig 3; S2 Table).

Overlapping suitable areas between A. fraterculus and D. areolatus were projected in the south and littoral region of Brazil, in the north of Colombia and Venezuela, and in the majority of Mesoamerica (Fig 4). In Mexico, overlapping suitable areas were more predominant in the Yucatan peninsula and in the south part of the country (Fig 4). In contrast, in the United States of America, overlapping areas were restricted to the south of Florida (Fig 4). Future projections show an increase of overlapping suitable areas between A. fraterculus and D. areolatus in almost all the Brazilian territory, except in some areas in the Amazon region and the center region of the country (Fig 4; S1 Fig). However, in Colombia, common suitable areas for A. fraterculus and D. areolatus, were predicted in the Amazon and the Plains (Fig 4). In contrast, in Mexico and the United States of America, common suitable areas for the pest and the parasitoid had no significant change in future predictions (Fig 4; S1 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Overlapping suitable areas predicted for Doryctobracon areolatus and Anastrepha fraterculus in the Americas.

(A) Overlapped suitable areas predicted in the current time period; (B) Overlapped suitable areas predicted in the future 2041-2060 scenario SSP5-8.5.

https://doi.org/10.1371/journal.pone.0325761.g004

The fruit fly parasitoid, Diachasmimorpha longicaudata SDM

Our predictions showed that suitable areas for D. longicaudata occupy 35.07% of the Americas territory (Table 2). Future projections indicate a slight expansion of suitable areas of up to 38.06% in the Americas (Table 2). Highly to moderately suitable areas for D. longicaudata were projected in the Center-West and North of Brazil, with an increase in suitable areas of 14% in its territory in future predictions (Fig 5, S2 Table). In Peru, suitable areas for the current period were predicted to cover 63% of the national territory, which could increase to 67% in future scenarios (Fig 1; S2 Table). In Colombia, future projections indicate that up to 90% of the country will be suitable for D. longicaudata (Fig 1; S2 Table), with moderate to high suitability across most regions, except for the Andes mountains, which remain unsuitable in future predictions (Fig 5.). In the United States, suitable areas are limited to central and southern Florida, with no significant expansion observed in future scenarios (Fig 5). Across Mesoamerica, future projections suggest moderate suitability for D. longicaudata along coastal areas and low suitability in the Northeast, lowlands, and central regions of Mexico (Fig 5). Suitable areas in Mexico are expected to increase 17% in future projections.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Suitable areas predicted for Diachasmimorpha longicaudata in the Americas

(A) Suitable areas predicted in the current time period; (B-E). Suitable areas predicted under two future climate change scenarios (SSP2-4.5 and SSP5-8.5) and periods (2021-2040 and 2041-2060); (B) scenario SSP2-4.5 time period 2021-2060; (C) scenario SSP2-4.5 time period 2041-2060; (D) scenario SSP5-8.5 time period 2021-2040; (E) scenario SSP5-8.5 time period 2041-2060. The maps show suitability values in a green to-red color where: high suitable areas depicted in green color correspond to suitability values close to 1; moderately suitable areas illustrated in yellow designate suitability values close to 0.5; low suitable areas depicted in red show suitability values close to the species threshold (0.15); unsuitable areas represented by the absence of color correspond to values below the threshold and close to 0.

https://doi.org/10.1371/journal.pone.0325761.g005

Notably, there is a substantial overlap of suitable areas for D. longicaudata and A. fraterculus, particularly in South America and Mesoamerica (Fig 6; S2 Fig). Current predictions showed suitable overlapping areas across the Amazon region in Brazil, Peru, and Colombia; these areas are predicted to have a significant expansion in future projections (Fig 6; S2 Fig). In Mexico, overlapping suitable areas for the establishment of D. longicaudata and A. fraterculus were predicted along the coast in both current and future predictions (Fig 6; S2 Fig). In the United States of America, overlapping suitable areas for the parasitoid and the pest were predicted only in the state of Florida (Fig 6; S2 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Predicted overlapping suitable areas for the establishment of Diachasmimorpha longicaudata and Anastrepha fraterculus in the Americas.

(A) Overlapping suitable areas predicted in the current time period; (B) Overlapping suitable areas predicted in the future 2041-2060 scenario SSP5-8.5.

https://doi.org/10.1371/journal.pone.0325761.g006

Anastrepha fraterculus is predicted to expand to new territories in the Americas under future climate change scenarios

Our models predict a reduced ability of A. fraterculus to establish in warmer, dry regions, and a high ability to establish in temperate areas (Fig 2). These predictions agree with previous laboratory bioassays that found a reduced emergence of A. fraterculus adults at temperatures exceeding 30°C, whereas the emergence rate exceeded 70% at 15 °C [74]. Similarly, Santos [75] concluded that water stress due to scarcity can cause mortality of A. fraterculus under warmer conditions. Our results are comparable with those reported for Anastrepha grandis SDM, where the precipitation of the warmest quarter (BIO10) was the most influential variable, suggesting that extreme drought events may drive the future distribution of this species [76]. Also, in our predictions, regions with temperate climates showed moderate and high suitability for the establishment of A. fraterculus, whereas low-suitable and unsuitable areas include some tropical climate regions. Based on this, we suggest that temperate climate areas with high humidity seem to be more suitable for the establishment of this species at the current time and in the future. Indeed, the spread of A. fraterculus in temperate climate regions due to climate change has already been reported in the field [77].

In Brazil, temperate climates are predominant in the southern and southeastern regions and are characterized by hot summers, no dry seasons, and cooler winters [78]. However, our study found suitable areas for the establishment of A. fraterculus in the coastal region characterized by high temperature, low altitude, and high humidity [79]. Our projections concur with previous studies reporting the presence of different A. fraterculus morphotypes in these regions. For example, the Brazilian-2 and Brazilian-3 morphotypes seem to be more adapted to coastal areas, whereas the Brazilian-1 morphotype seems to thrive in temperate areas [22,80,81]. Our model predicted areas with low or no suitability for A. fraterculus in the Brazilian Amazon region and central-western states. These regions experience tropical climates, characterized by dry winters and summers, whereas the Brazilian Amazon has predominantly tropical monsoon and tropical rainforest climates [78,82]. The low suitability predicted in these areas might be the result of not conducting separate analyses for each A. fraterculus morphotype. SDMs developed by Selivon et al. [81] for each A. fraterculus morphotype in Brazil, found suitable areas only in the southeast and south of the country. On the other hand, populations of A. fraterculus in Mesoamerica, known as the ‘Mexican morphotype’, were found at low elevations (30–1,400m) and thrive in both temperate and tropical climates [22,23]. Therefore, our results of the expansion of the suitable regions for A. fraterculus in the Brazilian Amazon are for the species without differentiation between its morphotypes. However, the continued presence of A. fraterculus in future scenarios in the Brazilian southern and southeastern regions could pose a risk for crops, such as apples and peaches in these regions [26,83].

The Peruvian mountains, predicted by our study as suitable for A. fraterculus, are characterized by two types of climates, temperate and arid [78]. A study conducted by Ramos Peña et al. [84] in the Abancay Valley in Apurimac, Peru, found an increase in fruit fly populations during the rainy season and the transitions to the dry season. The authors also suggested that low temperatures and the availability of suitable hosts increase fruit fly populations. Studies conducted in guava orchards in Oxapampa, Peru, revealed an increase in the abundance and uniform distribution of A. fraterculus at higher elevations. Moreover, these trends appear to be influenced by both seasonality and altitude [85]. In Colombia, the presence of A. fraterculus in coffee crops in the Andes mountains was previously reported by Jaramillo et al. [28], representing 99.9% of all specimens of fruit flies in coffee berries; however, its effects on fruit development and coffee quality are unknown. Additionally, at least 37 other hosts for A. fraterculus have been reported in Colombia [86]. Some of these hosts are important crops in the Plains and Amazon regions and will be at risk in the future, such as mango [87,88], soursop (Annona muricata L.), papaya (Carica papaya L.) [87], and tangerine (Citrus reticulata Blanco) [27]. Given the potential increase of suitable areas in the Plains and Amazon regions of Colombia in the coming years, efforts to monitor the dispersion of A. fraterculus are highly recommended.

Our results show suitable areas for A. fraterculus in the Yucatan peninsula in Mexico, where this pest is currently present [23]. The loss of suitable areas predicted by our models in Mexico can be explained by a reduction of temperate climate regions, transitioning to arid climates in the future [78]. This suggests that arid climates are not optimal for the establishment of the species. Indeed, states such as Chihuahua, Durango, Coahuila, Nuevo Leon, and Zacatecas, characterized by arid climates, were found unsuitable for A. fraterculus in all our predictions. The presence of A. fraterculus in the United States was previously reported in southern Texas, where the species appears to be restricted [89]. However, our future projections suggest that this pest can spread to new suitable habitats. Therefore, despite the low suitability of the new regions for the establishment of A. fraterculus, it remains crucial to monitor the potential expansion of this pest in the U.S.

Overall, our results predict an expansion in the suitable areas for A. fraterculus in temperate regions and some tropical areas across the Americas in the coming years. These predictions support our first hypothesis about the pest’s potential distribution. The models suggest this expansion could occur despite some regions becoming less suitable due to climate change. However, additional studies will be needed to determine if the predicted suitable areas for this pest overlap with compatible areas for the cultivation of susceptible crops. Our continental-scale projections provide valuable insights into the pest distribution, but we recognize that local microclimates, agricultural practices, and the presence of different pest morphotypes are essential to determine the establishment of this pest. Nevertheless, monitoring efforts should prioritize regions identified as suitable in our models, especially if valuable host crops are cultivated in those areas.

The fruit fly parasitoid, Doryctobracon areolatus as a promising control agent in temperate regions

Suitable areas for D. areolatus are predicted in arid and temperate regions, while tropical climates and high-altitude regions seem to be unsuitable or low-suitable for the establishment of this species. Similar results were reported by Núñez-Campero et al. [90], who built SDM for D. areolatus in the current period and found that this species is associated with semiarid conditions while high humidity decreases suitability for its establishment. In their projections, precipitation of the driest quarter (BIO17) and precipitation of the coldest quarter (BIO19) were the variables with the greatest contribution to the model, with values of 18% and 33%, respectively. Nonetheless, in our models, the bioclimatic variables associated with temperature and precipitation had the highest contribution values; this discrepancy could be related to different methodologies to construct the models and to different geographic areas of study. However, previous studies demonstrate that the best thermal conditions for the growth of D. areolatus are between 22 and 25 °C [91]. Also, studies conducted at different elevations have demonstrated that the parasitism rates of D. areolatus are significantly reduced at high altitudes, supporting our results.

In Brazil, certain areas classified by Beck et al. [78] as tropical rainforests in the northern regions of the country were predicted as low suitable or unsuitable for D. areolatus, particularly in the Amazonas state (Fig 3). The limited suitability of this parasitoid species in tropical rainforest regions can be attributed to the high precipitation levels in these areas, which affect the population dynamics and reduce the parasitism capacity [5,92]. Our current and future projections indicate that the temperate zones in southern and southeastern Brazil are more suitable for the establishment of D. areolatus. These regions are also suitable for the establishment of A. fraterculus, suggesting that D. areolatus could be a good biological control agent for this pest (S1 Fig). For example, in the state of Rio Grande do Sul, the parasitoid was found in association with A. fraterculus, exhibiting parasitism rates between 8% and 25% [93,94].

Moreover, overlapping suitable areas for A. fraterculus and D. areolatus were predicted in other countries across Latin America (Fig 4; S1 Fig). For instance, in Peru, Colombia, Venezuela, Ecuador, and Bolivia, the parasitoid D. areolatus has been found in association with A. fraterculus [20,95]. In Peru, some suitable areas for D. areolatus were predicted in the “la Sierra” region, which is heavily planted with fruit crops and is characterized in part by a temperate climate [78,96]. Additionally, a study conducted by Salazar-Mendoza et al. [97] in Oxapampa, Pasco (low elevation in la Sierra region), found 11.76% of A. fraterculus parasitized by D. areolatus in loquat (Eriobotrya japonica Lindl.). However, our predictions for future climate scenarios show low suitability of these regions for the establishment of the parasitoid D. areolatus. Therefore, we suggest the use of other biocontrol agents to manage A. fraterculus in these regions in the future.

Moderately and highly suitable areas in current and future projections were predicted in the valley of the Andes Colombia, specifically in the Departments of Huila, Valle del Cauca, and Tolima (Fig 3). The presence of D. areolatus in the Department of Valle del Cauca was previously reported by Carrejo and González [98] preying on Anastrepha obliqua (Macquart, 1835) in mango crops. Also, D. areolatus in Colombia seems to be more common in areas of low elevations [99], which coincides with our predictions of moderate to low suitability in the Atlantic Coast and Plains regions of Colombia. Therefore, the parasitoid D. areolatus could help mitigate the effect of A. fraterculus and other Anastrepha species in crops planted in low-altitude areas.

In Mexico, we found suitable areas for D. areolatus in regions characterized by a low elevation and a tropical savannah climate [78]. Previous studies reported that D. areolatus was the most predominant parasitoid across various host species of Anastrepha in the states of Vera Cruz and Chiapas [100]. Also, Sivinski et al. [101] found that D. areolatus was more frequent in low altitudes in the state of Veracruz, which supports our findings.

In the United States of America, the parasitoid D. areolatus was introduced in 1969 in Florida to control the Caribbean fruit fly, Anastrepha suspensa (Loew, 1862) [102]. This parasitoid was successfully established in Florida except in the southern part of the state. Eitam et al. [103] pointed out that the absence of D. areolatus in this region can be attributed to competitive displacement due to the introduction of D. longicaudata. These authors also found that D. areolatus was more common in sites with lower mean temperatures, where greater seasonality and less homogeneous host distribution favor this species.

Collectively, our results suggest that the native parasitoid D. areolatus is a promising biological control agent against Anastrepha fruit flies, particularly in temperate and arid regions, where its establishment and parasitism rates are highest. Tropical and high-altitude areas seem to be unsuitable or low-suitable due to excessive humidity and cooler temperatures, a trend reinforced by future climate projections. The suitable areas for the establishment of D. areolatus are predicted to increase between 4.72 and 8.13% in future climate change scenarios, contradicting our second hypothesis of no expansion in suitable areas for this species. This suggests that climate change could expand suitable areas also for pests’ parasitoids, reinforcing the viability of biological control programs in the future.

The fruit fly parasitoid, Diachasmimorpha longicaudata as a promising control agent in tropical regions

According to our projections, highly and moderately suitable areas for D. longicaudata have tropical climates, including areas characterized as tropical rainforests like the Amazon region. Low suitability or unsuitability is observed in areas with temperate or arid climate, such as the South of Brazil, Paraguay, along the Andes mountains, and certain states in Northwest and Northeast Mexico (Fig 5). Previous studies demonstrated that the optimum temperature for the development of immature stages and fecundity of D. longicaudata is between 25 ºC and 27 ºC, with complete development at 30 ºC in the hosts Bactrocera dorsalis (Hendel, 1912) and C. capitata [104–106]. In another study developed by Meirelles et al. [104], regions with an average temperature higher than 25 ºC were more favorable for establishing D. longicaudata as a control agent in the field. Also, according to Harbi et al. [105], temperatures between 16 ºC and 24 ºC and relative humidity between 45% and 60% are ideal for parasitism by D. longicaudata. According to the same authors, a decrease in relative humidity reduces parasitism rates, while an increase in temperature increases parasitism. These findings collectively suggest that tropical regions provide more suitable conditions for the successful establishment and performance of D. longicaudata.

In Northern Brazil, areas suitable for D. longicaudata overlap with those of the pest A. fraterculus (Fig 6; S2 Fig). However, this tropical region has low suitability for the pest, suggesting that D. longicaudata could be a good biological control agent of other tephritids, such as Bactrocera carambolae (Drew & Hancock, 1994) (Diptera: Tephritidae) in these zones [107]. Moderately suitable areas were also identified in Southeast Brazil, particularly in São Paulo, where D. longicaudata could aid in tephritid control in citrus orchards. Paranhos et al. [33] reported that summer releases enhance parasitism, while winter conditions allow high survival rates despite heavy rainfall. In contrast, a low adaptation of D. longicaudata was observed before in the state of Amapá in North Brazil [108], which aligns with our current projections for some areas classified as low suitable. However, future predictions indicate that Southeast and North Brazil may become moderately to highly suitable for the establishment of this parasitoid, suggesting that D. longicaudata could be used in forthcoming biological control programs of tephritid fruit flies. In the south of Brazil, previous studies developed in Rio Grande do Sul, demonstrated that the parasitoid has limited tolerance to winter, leading to less number of generations [104]. Therefore, other biocontrol agents that are better adapted to the environmental conditions of southern Brazil, such as D. areolatus, may be better options for controlling A. fraterculus.

In Peru, unlike D. areolatus, high and moderately suitable areas were projected across the entire region of “la Selva” and low suitable areas in almost the entire region of “la Sierra” (Fig 4 and 6). This suggests that the parasitoid D. longicaudata could be an important biocontrol agent for A. fraterculus in Peru’s fruit-growing regions. Similarly, in Colombia, D. longicaudata may be an effective biological control agent of A. fraterculus particularly in tropical regions that are predicted to be unsuitable or low suitable for the establishment of D. areolatus.

In future scenarios, suitable areas for the establishment of D. longicaudata were predicted across all Mesoamerica (Fig 5). The coastal regions of Mexico, characterized by a tropical climate, were predicted as suitable for the establishment of D. longicaudata, while the temperate and arid regions of Northeast, and Central Mexico were unsuitable. Our predictions indicate an expansion of suitable areas for D. longicaudata in future climate change scenarios, which can be attributed to a projected shift in certain Mexican regions from temperate to tropical [78]. This change could benefit the expansion of D. longicaudata populations in the region. Some suitable areas for D. longicaudata in Mexico, such as the southern part of Chiapas cultivated with mango, exhibit parasitism rates between 0.44% and 29.23% on other Anastrepha species [109]. Augmentative biological control of Anastrepha species in this region has shown that D. longicaudata can reduce pest populations without negatively affecting parasitoid local richness [110,111]. Therefore, D. longicaudata is a promising biocontrol agent for A. fraterculus and other Anastrepha species in the tropical climate areas in Mexico.

In the United States, D. longicaudata was introduced in 1988 to control C. capitata in Florida. This parasitoid reduced the fruit fly populations and is now established in the region [103,112]. The distribution of D. longicaudata proposed a decade ago aligns closely with the suitable areas predicted by our model [103]. According to our results, both D. areolatus and D. longicaudata could be used for the control of A. fraterculus in the State of Florida (Fig 4 and 6). However, if the pest spreads to other northern states in the future, where conditions are unsuitable for both parasitoids according to our predictions, the use of alternative control tactics or biocontrol agents may become necessary.

In summary, suitable habitats for D. longicaudata may expand in tropical regions under future climate scenarios, contradicting our second and third hypotheses. This suggests that non-native parasitoids like D. longicaudata could occupy a broader geographical range than native species under climate change. Importantly, D. longicaudata could serve as an effective biocontrol agent against fruit flies, particularly in areas where D. areolatus is unsuitable, thereby nearly covering the entire potential range of the pest.