Study area

We conducted the study in the municipalities of La Huerta, Cihuatlan, and Tomatlán in the southwestern coast of Jalisco, Mexico (Fig 1). Natural vegetation in this region is dominated by Tropical Dry Forest (56.1%), while agricultural and pasture areas cover 25.8% [65]. The study sites at La Huerta are located in the Chamela-Cuixmala Biosphere Reserve, a reserve created in 1993 that protects 1311.42 Km² of well-preserved TDF and wetlands [66]. The climate of the study area is predominantly warm subhumid with summer rains, with mean annual temperature ranging from 20 to 28°C, and mean annual precipitation ranging from 600 to 2000 mm; the dry season goes from November to May [65]. Several pollinator-dependent crops are cultivated during this season including watermelon (Citrullus lanatus), green tomato (Physalis ixocarpa), chili (Capsicum annuum), cucumber (Cucumis sativus), chayote (Sechium edule), papaya (Carica papaya), squash (Cucurbita moschata and C. pepo) and tomato (Solanum lycopersicum). These crops are usually located along river basins. Cultivation period is during the dry season as it allows farmers to reduce crop losses due to flooding and herbivory during the wet season [67,68]. According to the Agri-Food and Fisheries Information Service of Mexico (http://infosiap.siap.gob.mx/gobmx/datosAbiertos_a.php [69]), during 2017, and for the three municipalities of the study site, the cultivated area for watermelon and green tomato was 1,258 ha, the total value production was 8,233,714 US dollars. In the same period and at the study site, watermelon production accounted for 28.7% of the total production value of all crops, utilizing 7.3% of the cultivated area, whereas green tomato production comprised on average 6% of the total production value and 7% of the cultivated area.

Download:

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

Fig 1. Study site.

Study site and municipalities of La Huerta, Cihuatlan and Tomatlán where watermelon (Citrullus lanatus) and green tomato (Physalis ixocarpa) crops were sampled during the dry seasons of 2008, and 2014 to 2017. Letters A and B are examples of the land use/cover classes; A shows sampling sites distant from the Chamela-Cuixmala Reserve, and B shows sampling sites near the Chamela-Cuixmala Reserve.

https://doi.org/10.1371/journal.pone.0301402.g001

Study species crops

Citrullus lanatus (Watermelon): is native to Northeastern Africa and has been domesticated at least 4000 years ago [70]. Flowers are solitary, 2–3 cm in diameter, with five light yellow petals. Watermelon is a self-compatible monoecious plant, with pistillate and staminate flowers (Fig 2), although some cultivars have hermaphrodite flowers [71]. The major pollinators of C. lanatus are bees, especially Apis mellifera, bumblebees and some species of solitary bees [1,72]. In the study region, farmers use seedless watermelon as recipient plants and seeded watermelon as pollen donors in a 3:1 proportion, respectively. Seedles watermelon do not produce viable staminate flowers, they require to be pollinated with viable pollen from seeded plants [73–75]. Apis mellifera is widely used as a pollinator of watermelon in the study region, where farmers have contracts with beekeepers to use A. mellifera hives to supplement watermelon pollination (at a cost of around $16 dollars per hive for the season during the study years).

Download:

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

Fig 2. Flowers of Watermelon (Citrullus lanatus) from Jalisco, Mexico.

(A) pistillate flowers and (B) staminate flowers.

https://doi.org/10.1371/journal.pone.0301402.g002

Physalis ixocarpa (Green tomato): Physalis fruits have been used as a food source since the pre-Columbian era. There are about 90 species of Physalis, and of these, 70 species are found in Mexico [76]. The main species of Physalis cultivated in Mexico is green tomato, the fifth economically most important vegetable. The flowers are hermaphrodite, and solitary (Fig 3) and they open before anther dehiscence. The fruit is a fleshy green to yellow berry of variable size that is wrapped by a broad and persistent calyx [77]. The species has been classified as self-incompatible in greenhouse studies [78–80].

Download:

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

Fig 3. Hermaphrodite flowers of green tomato (Physalis ixocarpa) from Jalisco, Mexico.

https://doi.org/10.1371/journal.pone.0301402.g003

These crops offer different perspectives to farmers. On one hand, watermelon is primarily an export crop to the United States, making it highly profitable economically. However, it requires many inputs, which not all farmers can afford. On the other hand, green tomato is a crop for the domestic market, and although it is not as profitable as watermelon, it requires fewer inputs, making it accessible to both large and small producers. As a result, farmers rely heavily on the productivity of both crops. Studying these two crops is crucial to understanding the different dynamics and challenges faced by farmers, and the importance of pollinators in the sustainability and profitability of their agricultural practices.

Study design

To evaluate pollinator effectiveness and the influence of the landscape on pollinator richness and abundance, over five years we sampled 39 plots of watermelon in 2008 and 2014–2017 (5, 11, 9, 6, 8 plots respectively), and 51 plots of green tomato in 2008, 2014–2017 (16, 14, 10, 7, 4 plots respectively). The number of crops varied in time because crop locations changed from year to year. To select the study plots for each crop, we searched the river basins of the study region in the early dry season each year. We selected plots where the study crops were at peak flowering (i.e., individual plants had a high proportion of flowers to fruits, about >0.8 of proportion). We georeferenced each surveyed plot, positioned at the center of each sampled plot, where we took GPS coordinates. Pollinator community composition and visitation patterns, pollen load on pollinators bodies and pollinator dependence and pollinator efficiency experiments were recorded in years 2015–2017. Pollinator abundance data for the assessment of landscape and mapping crop pollination services using InVEST, were recorded in 2008 and 2014–2017.

Pollinator community and visitation patterns

To characterize the community of floral visitors that are potential pollinators of the study crops, we surveyed flowers in seven watermelon plots for three years (2015, 2016, and 2017), and four green tomato plots for two years (2015 and 2016). We conducted video camera recordings in two to 10 plants per plot in one flower per plant from 900 to 1200 hrs.; each flower was recorded for 180 minutes. In watermelon plants, we surveyed both pistillate and staminate flowers. For both crops, we only registered floral visitors that contacted the reproductive organs of the flower. We recorded pollinator species or taxonomic group, time of arrival to the flower, and duration of the visit. We identified the pollinators visually when they arrived at the flower; when this was not possible, we captured the visitor with an insect net for further identification. We identified insects to the lowest possible taxonomic category with the help of taxonomic keys and field guides. We calculated visitation rates as the number of visits per flower per hour for each pollinator taxa. Time of arrival to the flower was used to obtain mean visitation rates by time of the day.

We conducted generalized linear mixed models with the GLIMMIX procedure in SAS version 9.4 [81] to evaluate differences among pollinator species (fixed effect) in pollinator visitation rates, and duration of individual pollinator visits (response variables). The GLIMMIX procedure extends generalized linear models by incorporating random effects, which are particularly useful for modeling data with multiple levels of variability, such as repeated measures or clustered data. In the case of watermelon crops, we also evaluated how flower gender, pollinator species (fixed effects), and their interaction influenced the same response variables. We included year and field plot as random effects in both analyses. The Poisson distribution and log link function for both response variables were chosen based on theoretical considerations, given the count nature of the data, and confirmed by model diagnostics which indicated that these choices were appropriate. We specified the ILINK option of the LS-MEANS statement to obtain back-transformed least-square means and used a Tukey adjustment for multiple comparisons.

Pollen loads on pollinators’ bodies

To determine the capacity of different floral visitors to carry pollen of watermelon (Citrullus lanatus) and green tomato (Physalis ixocarpa), we captured insects visiting flowers (3–15 pollinator species) in the same study plots selected to document visitation frequencies in 2015. To count pollen loads on pollinator body, we followed the same methodology used by Delgado-Carrillo on Cucurbita moschata crops [19]. Insects were captured along the transect where counts for richness and abundance were conducted; only insects that contacted the reproductive organs of the flower were captured. In the case of watermelon, we only sampled pollinators in staminate flowers. We placed insects in separate vials and later removed pollen from each individual, dabbing one piece of fuchsin gel over four different parts of the body: back, head, ventral abdomen, and ventral torso. We did not remove pollen from specialized structures for pollen transport (i.e. corbiculae, scopae). We used one slide for each body part and counted the number of pollen grains across the four samples collected per specimen, and we used a stereoscopic microscope and the Zen program V 1.1.2 [82]. Because of the difficulty to see green tomato pollen grains under a stereoscopic microscope, we modified the methodology for pollen counts, and counted grains in five regions of the slide under a 40X magnification using an optical microscope. We averaged the number of pollen grains across counts and slides to obtain a number of pollen grains per specimen. To evaluate the capacity of different floral visitors to carry pollen of green tomato and watermelon, we performed a generalized linear mixed model with GLIMMIX procedure in SAS version 9.4 [81]. The model included pollinator species as a fixed effect and the total pollen counts across the body or slide regions as a response variable. This analysis used a Poisson distribution and a log link function, the ILINK option of the LS-MEANS statement was used to obtain back transformed least square means, and field plot was included as a random effect in the model.

Pollinator dependence and pollinator efficiency experiments

To determine the degree of dependence of watermelon (Citrullus lanatus) and green tomato (Physallis ixocarpa) on pollination by animals, in 2015 we conducted field experiments on 10 plants per plot on 11 plots for watermelon and 10 plots for green tomato. Within the plot, we selected plants that were more than 10 meters from the edge and had a similar number of flowers. The plants were selected haphazardly within the transects used to measure pollinator diversity and abundance. We marked two virgin flowers per plant that were assigned to one of two treatments: (1) open pollination, we marked flowers and left them exposed to visits by all potential pollinators; (2) pollinator exclusion, we covered flowers with meshnet bags one day before anthesis to exclude all floral visitors. On the third day after anthesis, we removed the bags and marked the flowers. For both treatments, we followed fruit development until maturation.

To determine the effectiveness of the most frequent floral visitors known to carry pollen of watermelon and green tomato, we added one more treatment in 2016 and 2017: (3) pollinator exclusion with one pollinator visit. We used 10 plants and one flower per plant and bagged the flowers one day before anthesis. We removed the bag the day after anthesis and allowed for a single pollinator visit. After the visit, we bagged the flower again and removed it three days after anthesis. We obtained data for individual visits of Apis mellifera and Trigona fulviventris, which were the most frequent pollinators.

In the first sampled green tomato plots, we observed that the plants produced fruits in the pollinator exclusion treatments; therefore, we tested for apomixis adding a fourth emasculation treatment to the green tomato plants evaluated in 2016 and 2017. One day before anthesis, one flower per plant was emasculated through the removal of all stamens using fine tweezers. The flowers were bagged until three days after anthesis, and fruits were followed until maturity.

For both crop species, we compared fruit set among pollination treatments using the GENMOD procedure in SAS version 9.4 [81]. The GENMOD procedure fits generalized linear models, which extend traditional linear models to accommodate response variables that follow distributions other than the normal distribution. In our analysis, the model used pollination treatment as the independent variable and the proportion of flowers that developed into fruit as the response variable. We specified a binomial distribution and a logit link function for fruit set, as this response variable represents a binary outcome (fruit set vs. no fruit set). Tukey-adjusted P-values were used for multiple comparisons.

We also tested the relationship between pollinator abundance (predictor variable) and fruit set (response variable) with a simple linear regression with the REG procedure in SAS version 9.4 [81]. We used the insect abundance data from the transect surveys (see study design section) and the fruit set data from the open pollination treatments. The analysis used a binomial distribution and a logit link function for the variable fruit set.

Influence of the landscape

To evaluate floral visitor abundance and richness at the landscape scale, we established 50 m random transects 10 meters away from the edge (one within each plot) and conducted flora visitor surveys in 2008 and 2014–2017. We sampled each plot only during sunny or partly cloudy days (when insects are more active), and at least after three days of any application of agrochemicals. We walked along each transect for 10 minutes registering all flower visitors that contacted the reproductive organs of the flowers. The insects were identified to the lowest possible taxonomic level in the laboratory with the help of taxonomic keys. To correlate the richness and abundance of pollinators of watermelon and green tomato to landscape configuration, we generated land-use and land-cover (LULC) maps for years 2007, 2012, and 2017, which provided the best quality maps to conduct analyses in close temporal proximity to the study years. LULC maps were elaborated through visual interpretation of SPOT5 satellite images at 5 and 10 m resolution from 2007, Rapideye satellite images at 5 m resolution from 2012, and Planet satellite images at 3 m resolution from 2017. Acquisition dates of satellite images were between February and May (the dry season when leaves fall off the trees in the tropical dry forest), corresponding closely to the sampling period of the bees. For the SPOT5 images, a fusion process was carried out between the 5-meter resolution panchromatic band and the multispectral bands to increase spatial resolution to 5 meters. To avoid spurious changes in LULC maps, we conducted an interdependent classification process, starting in 2007 and updated after 2010 and 2018. The satellite images were classified into nine LULC classes: (1) secondary forest (e.i., including pastures and forests in early successional stages), (2) deciduous tropical forest of advanced successional age (including intermediate and late successional stages), (3) riparian forest (including both gallery forest along large rivers and vegetation along the stream banks), (4) mangroves, (5) seasonal crops (e.g., corn, green tomato, tomato, chili, melon, watermelon, cucumbers, squash), (6) orchards (e.g., mango, papaya, coconut or citrus), (7) bare soil, (8) water, and (9) urban settlements. Image processing was done in ArcGIS 10.2 [83] and Qgis [84] at a scale of 1: 15 000. To validate LULC maps, we followed the methodology proposed by [85]. We selected 553 random points from each LULC map, and then carried out verifications for each point using Google Earth images. For this procedure, we designated visual interpreters who had not been involved in the construction of the LULC maps. The overall accuracy of the classifications for 2007 and 2012 is 94%, while for 2017 is 96%. We obtained this calculation using the AccurAssess plugin for Qgis [86].

Around each polygon (i.e. plot) centroid, where the richness and abundance of pollinators were measured, we calculated circular buffer areas of 500m, 1000m, 1500, 2000m, and 2500m. Using the classified LULC maps for 2007, 2012, and 2017, we estimated the following parameters: (1) percentage of forest cover -including mangroves, tropical deciduous forest (secondary and late forest), and riparian vegetation- for each buffer area surrounding watermelon and green tomato crops, and (2) distance to the different forest types. Then, we assessed the correlation between the percent of forest cover and the distance to the different classes of forests vs. the following variables: (1) abundance and richness of Apis mellifera, (2) abundance and richness of native wild pollinator species, (3) total abundance. For each survey, we selected the LULC map closest to the field sampling year to estimate forest cover for each buffer area and to calculate Pearson’s correlation coefficient between forest areas and pollinator species abundance. Normality of residuals was assessed with the Shapiro-Wilk test. We conducted the spatial and correlation analysis using the free software environment R [87].

Mapping crop pollination services using InVEST

To model crop pollination services provided by bee pollinators to watermelon (Citrullus lanatus) and green tomato (P. ixocarpa) crops, we used the crop pollination model implemented in InVEST [88] The model estimates an index of bee abundance across a landscape based on the availability of nesting sites (the different substrates and availability where bees can nest, in this study was cavity or ground substrates) and floral resources (plants with flowers that bee uses) within pollinator flight ranges as described below [89,90]. The index of abundance produced by the model depicts the abundance of bees for each cell. The inputs required by the model are (a) a land use and land cover map, (b) land cover table describing LULC classes, and (c) a table describing guilds or species of pollinators to be modeled, their nesting biology and flight ranges [89,90]. We used LULC map for 2017 (see “influence of the landscape” section) as the LULC input map. To estimate bee floral resources available in each LULC class, we conducted a literature review searching for plant species flowering from January to March in the TDF of the Chamela-Cuixmala Biosphere Reserve. We obtained the percentage of flowering plants from January to March found in each land cover class (S1 Table). We used the list of pollinator species obtained from video recordings and through direct observation in sampled transects. We estimated each bee foraging distance from intertegular distance in mm using individuals captured for pollen counts. We estimated each bee’s foraging distance based on the intertegular distance (ITD) measured in millimeters from individuals captured for pollen counts. The intertegular distance, which is the distance between the bases of the wings, is a reliable predictor of flight capability and foraging range in bees [90]. For each species, we recorded 2 to 20 ITD measurements. Using these measurements, we calculated the typical foraging range (in social bees, the distance at which 50% of foraging bees return to the nest) in meters (S2 Table). For this calculation, we used the regression formula proposed by Greenleaf et al. [90], which can also be applied to solitary bees. The regression formula provides a robust estimate of foraging distance based on ITD, enabling us to predict the foraging behavior of different bee species. We also reviewed the literature to determine nesting suitability (cavity or ground) for each bee species (S3 Table). To compare INVEST predictions against the observed abundance of pollinators in watermelon and green tomato crops, we performed a simple linear regression using the REG procedure in SAS version 9.4. We aimed to test the hypothesis that the INVEST model predictions would significantly correlate with the actual observed pollinator abundances. By doing so, we sought to validate the accuracy and reliability of the INVEST model in predicting pollinator abundance in these agricultural settings.

Contribution of the main pollinators: Frequency and duration of visits

Watermelon (Citrullus lanatus): we observed eight species of bees visiting watermelon flowers. We found significant differences in pollinator visitation rates among species (F_(7,40) = 17, P ≥ 0.0001), and A. mellifera was the main visitor, accounting for 93.3% and 94.8% of total visits to pistillate and staminate flowers, respectively (Fig 4). Peak visitation occurred between 9:00 am and 11:00 am (Fig 5A). There were no significant differences in pollinator visitation rates by flower gender (F_(1,40) = 0.56, P = 0.45), nor for the interaction between flower gender and pollinator species (F_(5,40) = 1.27, P = 0.29). The average duration of a pollinator visit in flowers was 0.15±0.16 minutes. We did not find significant differences in the duration of a pollinator visit by flower gender (F_(1,1171) = 0.09, P = 0.77, Fig 6A), pollinator species (F_(5,1171) = 0.42, P = 0.8) or the interaction between flower gender and pollinator species (F_(4,1171) = 0.1, P = 0.98).

Download:

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

Fig 4. Composition and relative abundance of bee species visiting the male and female flowers of watermelon (Citrullus lanatus), and the bisexual flowers of green tomato (Physalis ixocarpa) in Chamela, Jalisco, Mexico across years 2015–2017.

https://doi.org/10.1371/journal.pone.0301402.g004

Download:

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

Fig 5. Pollination visitation rates (mean ± SE number of pollinator visits flower ^-1 h ^-1) to watermelon and green tomato flowers in Jalisco, Mexico during years 2015 to 2017.

(A) Visits to watermelon male and female flowers, (B) visits to green tomato bisexual flowers. Different colors indicate different bee species and flower gender for the monoecious species watermelon (Citrullus lanatus).

https://doi.org/10.1371/journal.pone.0301402.g005

Download:

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

Fig 6. Pollinator visitation rates by bee species.

Pollinator visitation rates (mean ± SE number of pollinator visits flower -1 h -1) for: (A) male (M) and female (F) flowers of watermelon (We did not find significant differences), and (B) bisexual flowers of green tomato (We find significant differences) in Jalisco, Mexico during years 2015 to 2017. For Fig 6B, different letters indicate significant differences between groups (P < 0.05) with Tukey´s ad hoc test.

https://doi.org/10.1371/journal.pone.0301402.g006

Green tomato (Physalis ixocarpa): we observed five bee species visiting green tomato flowers. Apis mellifera was also the main visitor, accounting for 95% of total visits (Fig 4). Peak visitation rate occurred between 10:00 and 12:00 hrs (Fig 5B). We found significant differences in pollinator visitation rates by pollinator taxa (F_(7,33) = 156.42, P ≤ 0.0001; Fig 6B). The average duration of pollinators in flowers was 0.15 ±0.12 minutes. We did not find significant differences in the duration of individual pollinator visits by pollinator taxa (F_(4,1008) = 0.22, P = 0.9).

Contribution of the main pollinators: Pollen loads on pollinators’ bodies

Watermelon (Citrullus lanatus): We captured 40 individuals of seven bee species in watermelon male flowers. All captured bees had pollen on their bodies. We found significant differences in pollen count by pollinator taxa (F_{(6, 21)} = 592, P ≤ 0.0001; Fig 7A). Trigona fulviventris (n = 4) and Exomalopsis sp (n = 4) had the largest pollen grain loads, while halictid bees Augochlora pura (n = 6) and Agapostemon sp. (n = 4) had the smallest loads; Trigona nigra (n = 5), Apis mellifera (n = 13) and Lassioglosum sp. (n = 4) carried intermediate pollen loads.

Download:

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

Fig 7. Pollen grains on the bodies of pollinators.

Mean (± SE) number of pollen grains on the bodies of pollinator taxa collected in (A) Watermelon (Citrullus lanatus) and (B) Green Tomato (Physalis ixocarpa), in the region of Chamela, Jalisco, Mexico during years 2015 to 2017. Different letters indicate significant differences between groups (P < 0.05) with Tukey´s ad hoc test.

https://doi.org/10.1371/journal.pone.0301402.g007

Green tomato (Physalis ixocarpa): We captured 45 individuals of eight bee species that had pollen on their bodies. We found significant differences in the number of pollen grains per load by pollinator taxa (F_{(7, 33)} = 156.42, P ≤ 0.0001; Fig 7B). Colletes sp (n = 3) had the largest pollen loads, Apis mellifera (n = 13), Anthophora sp (n = 3), Lassioglosum sp (n = 4), and Exomalopsis sp (n = 12) carried intermediate loads, and bees of the Augochlorini group (n = 3), Trigona nigra (n = 3) and Scaptotrigona hellwegeri (n = 4) carried the smallest pollen loads.

Contribution of the main pollinators: Pollinator efficiency experiments on watermelon and green tomato

Watermelon (Citrullus lanatus): For the open pollination treatment 41 out of 142 marked flowers produced fruits, and fruit set was zero for the pollinator exclusion treatment. For the one pollinator visit treatment, out of 20 flowers with one visit of Apis mellifera, no fruits were produced. The regression between pollinator abundance and open fruit set was not significant; less than one percent of the variation in fruit set was explained by pollinator abundance (F_{(1, 9)} = 0.07, P = 0.8, R² = 0.007, Fig 8A).

Download:

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

Fig 8. Pollination efficiency treatments.

Mean (± SE) fruit set of pollination efficiency treatments conducted to determine the potential for autonomous fruit production (in the absence of pollinators) and the pollination efficiency of Apis mellifera (one pollinator visit). (A) Watermelon (Citrullus lanatus) (B) Green tomato (Physalis ixocarpa) plants. Different letters indicate significant differences between treatments (p < 0.05) with Tukey ad hoc test.

https://doi.org/10.1371/journal.pone.0301402.g008

Green tomato (Physalis ixocarpa): For the open pollination treatment, 68 out of 169 flowers produced fruit. For the pollinator exclusion treatment, we obtained 19 fruits out of 167 flowers. For the one pollinator visit treatment, we only obtained two fruits from 21 flowers. For the emasculation treatment conducted to assess for apomixis, we only obtained one fruit from 74 flowers. We found significant differences between treatments, where the fruit set of the open pollination treatment was higher than the pollinator exclusion, one pollination visit and emasculated flowers treatments (χ² = 65.94, df = 3, P ≤ 0.0001; Fig 8B). For instance, in the absence of pollinators, green tomato crops reduced their fruit production by 65%. We found a significant positive association between pollinator abundance and the fruit set of open-pollinated flowers (R² = 0.68; F_{(1, 8)} = 17.63, P = 0.003). This regression indicates that 68% of the variability in fruit set can be explained by the abundance of pollinators.

Influence of the landscape

For watermelon (Citrullus lanatus) crops, in five years we observed 15 pollinator species in 39 plots (for year 2008; 5 plots, 2014; 11 plots, 2015; 9 plots, 2016; 6 plots, and 2017; 8 plots) (S4 Table). For green tomato (Physalis ixocarpa) crops, in five years we observed 19 species of pollinators in 51 plots (for year 2008; 16 plots, 2014; 14 plots, 2015; 10 plots, 2016; 7 plots, and 2017; 4 plots) (S4 Table). In crops of green tomato for year 2008, we found significant positive regressions where forest cover predicted the total abundance of pollinators at the following distances to the focal crop: 500m (R² = 0.5, P = 0.003), 1000m (R² = 0.5, P = 0.001), 1500m (R² = 0.55, P = 0.001), and 2500m (R² of 0.5, P = 0.003) (Fig 9). We did not find significant differences with distance to the forest and forest cover in 2014 or 2016. For watermelon, we did not find significant effects of landscape configuration on pollinator diversity.

Download:

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

Fig 9. Correlation between the abundance of pollinators in green tomato (Physalis ixocarpa) crops and forest cover in 16 crops for the year 2008.

https://doi.org/10.1371/journal.pone.0301402.g009

Use of InVEST natural capital project program

We obtained the per-pixel total abundance of 12 pollinators species for three years. We found a significant positive regression between the pollinator abundance predicted by the model for 2008 and: (1) the observed total pollinator abundance (F_{(1, 19)} = 5.41, R² = 0.22, P = 0.03), (2) the observed abundance of social bees (F_{(1, 8)} = 5.37, R² = 0.22, P = 0.03), which validates the model’s predictions for these groups. We did not find a strong regression between the observed abundance and the predicted abundance of INVEST for 2014 and 2016 pollinator abundance.

Pollination of watermelon and green tomato crops in western Mexico

Watermelon (Citrullus lanatus) and green tomato (Physalis ixocarpa) are two economically important crops for Mexico, the first one native to Africa and the second one native to Mesoamerica. Both crops depend on pollinators. The results of this study showed that, despite their different geographic origins and floral morphologies, the honeybee A. mellifera is the main floral visitor, accounting for more than 90% of the visits to flowers of both species. Honeybees visited flowers throughout floral anthesis and in terms of pollen carrying capacity, they were preceded only by a native bee species in the genus Colletes, a rare floral visitor in both crops. Honeybees are key pollinators of many crops worldwide [8,21,91], and part of their success is related to their social behavior, their high reproductive rates, and their ability to establish large populations and feed on diverse floral resources [21,92,93]. Moreover, africanized honeybees have successfully established populations across diverse habitats in the New World, including natural ecosystems, agricultural land and urban areas [21,22,94]. In the Chamela-Cuixmala Reserve, africanized honeybees have established feral populations interacting with native plants and neighboring crops [19,22]. The high visitation frequency and pollination effectiveness of A. mellifera make this bee species the most important provider of pollination services for watermelon and green tomato in the study region.

However, the dominance of A. mellifera as the main floral visitor raises potential concerns. Dependence on a single pollinator species can increase risks associated with pollinator decline due to diseases, pests, or environmental changes [3,29,31]. Additionally, the predominance of honeybees may impact the genetic diversity of pollinated plants by limiting the range of pollen donors [95]. Therefore, while A. mellifera provides significant pollination services, it is crucial to consider strategies to support and maintain diverse pollinator communities to ensure the long-term sustainability of crop pollination.

A key trait of honeybees as pollinators of watermelon and green tomato is their ability to forage throughout the year, even in seasonal environments, such as the Chamela-Cuixmala region [19,22]. Moreover, africanized honeybees can locate and exploit floral resources in times of limited floral availability, and they may travel large distances to find food [93,96,97]. These behaviors enable them to sustain colonies during the dry season. In contrast, native bees include both social and primitively solitary species that have smaller populations and foraging ranges [97] particularly during the dry season. Eight species of native bees were observed visiting the flowers of both crops at low frequencies in flowers, but some of them were able to carry high pollen loads on their bodies. These included social Trigona fulviventris and solitary Exomalopsis sp for watermelon, and solitary Colletes and Anthophora sp for green tomato crops (Fig 4). Native bees can be effective pollinators of watermelon crops in other regions, as was demonstrated for the sweat bee Lassioglosum sp. in Pakistan, and for Bombus impatiens in USA [91,98]. However, in our study system, native bees were found in low abundance, possibly due to the time of cultivation of watermelon and green tomato. It has been demonstrated that the abundance and diversity of flowers and floral visitors experience a notable reduction in the dry season in the Chamela-Cuixmala region [22]. During the rainy season in this region or in less seasonal ecosystems, native bees and other pollinators may play an important role in the pollination of commercial crops, even in places where honeybees are dominant [8,21,99,100].

The results of this study suggest that, in order to reach the natural fruit set observed in watermelon (23%) and green tomato (40%), both plant species require multiple pollinator visits. In this context, native bees likely play a complementary role in the pollination of these crops. Research has shown that there is often complementarity between A. mellifera and native pollinators, which can enhance crop pollination; this complementary effect has been observed in various crops, including almonds and sunflowers [24,25]. A previous study found that honey bees were the most abundant and important pollinators of squash crops (C. moschata) during the dry season, when native pollinators are not present; however, during the rainy season, native bees of the genus Peponapis were the most important pollinators of squash crops due to their abundance and greater effectiveness [19]. Similarly, the native pollinators of wild Physalis species are only active during the rainy season and they are oligolectic, meaning that they feed on a small range of plant species. For example, the main pollinator of Physalis viscosa is the social stingless bee Perdita halictoides, which feeds primarily from species of the genus Physalis [101]. Therefore, the study of wild crop relatives and their interaction with native bees is crucial to the understanding of crop pollination dynamics.

In addition to assessing the interactions between crops and their native and managed pollinators, it is essential to evaluate the extent to which a crop relies on pollinators for reproduction. Our results showed a 65% decrease in green tomato yield in the absence of pollinators when contrasted with open pollination (Fig 8), suggesting the species is highly dependent on pollinators for reproduction. However, despite being a self-incompatible species, fruit set is possible in the absence of pollinators in P. ixocarpa, suggesting a possible breakdown of the self- incompatibility reproductive system towards the last days of anthesis in the absence of cross-pollination. Our study revealed that green tomato flowers have a lifespan of three days; however, the anthers become dehiscent on the second and third day, enabling delayed self-pollen deposition. This process has been described for other Solanaceae species with plastic self- incompatibility systems. In Nicotiana alata and Solanum carolinense, the decay of the RNAses involved in the self-incompatibility system occurs before the end of anthesis if the flowers are not previously cross-pollinated. This phenomenon suggests the potential for reproductive assurance through self-pollination [102,103].

Assessment of landscape influence on pollinator diversity and abundance

The conservation condition and distance to natural forests are determinant factors that ensure the richness and abundance of insects; therefore, the proximity of crops to continuous forests is expected to increase pollination service to agricultural systems [38,48]. Moreover, landscapes that incorporate large expanses of natural habitat exhibit a greater diversity and abundance of pollinating insects and a corresponding increase in pollination services compared to landscapes with scant natural habitat [104]. In the case green tomato (Physalis ixocarpa) crops, which are pollinated by honeybees present in the wild-agricultural interface, we observed a positive relationship between pollinator abundance and forest cover for year 2008. This indicates that as forest cover increases, the abundance of pollinators in green tomato crops also increases. Other studies have demonstrated that proximity to continuous natural habitats enhances pollination services to agricultural fields by increasing pollinator abundance, visitation and diversity, pollen deposition, fruit set, and crop productivity [39,46–48,104,105]. However, the relationship between forest cover and pollinator abundance and diversity was not significant for green tomato during 2014–2017. We also found a similar pattern for watermelon crops of all sampled years. This lack of relationship between the forest landscape and pollinator diversity may be due to a series of factors that may act in synergy. The first factor to consider is the use of Apis mellifera hives in watermelon crops to supplement the pollination of seedless watermelon. A. mellifera hives have been used to guarantee the pollination service of seedless watermelon crops. In this study, all watermelon plots supplemented A. mellifera hives during the flowering period. Farmers hire bee hives during the watermelon flowering season to ensure crop production. Consequently, the presence of these managed hives likely influenced the visits of native pollinators during the sampling of these crops. This could mask the effects of distance from the forest on pollinator abundance within the sampled agroforestry landscape, as A. mellifera are known for their extensive foraging ranges and large populations in agricultural fields and natural habitats [106].

In second place, we have a record of the impact of two hurricanes in the region of Chamela, which occurred after the sampling carried out in 2008, Jova in 2011 (category 3) and Patricia in 2015 (category 5). The impacts of hurricanes were evidenced by dramatic changes in vegetation greenness [107] and across extensive areas of fallen trees and branches across the Chamela-Cuixmala Biosphere Reserve [107,108]. Although a previous study demonstrated positive effects of Hurricane Patricia on herbivorous and predatory insects [109], it has been shown that bees can be susceptible to natural catastrophic events such as hurricanes, which can significantly reduce bee diversity [110]. The impact of hurricanes includes damage to nesting sites and depletion of floral resources, both of which are crucial for the seasonal activity and survival of bee species [111].

Hurricane disturbances can have similar negative effects with anthropogenic disturbances on vegetation, which has been demonstrated in a tropical dry forest for epiphytic Tillandsia species [112]. After a hurricane, the vegetation structure may be simplified to the point of resembling early stages of secondary succession, [112], where it has been shown that floral visitors are less diverse compared to advanced stages of succession [22]. In addition to natural disturbances, habitat loss due to cropland expansion and intensification can further exacerbate the decline in bee populations. Deforestation could have reduced the habitat complexity, making it more difficult for pollinators to find nesting sites and forage, while herbicide use might have limited the diversity and abundance of wildflowers, which are essential for pollinator nutrition [113,114]. These factors together, the use of A. mellifera hives during watermelon flowering season, the entry of two hurricanes during the sampling after 2008 and the simplification of the agroforestry landscape may affect the diversity and abundance. of pollinators in relation to cover and distance to the forest. Therefore, more studies are necessary to determine the seasonal patterns of bees and to evaluate the combined effects of agricultural practices and natural events on native bee populations. Understanding these interactions is crucial for developing effective conservation strategies to support pollinator communities and ensure crop productivity.