Habitat suitability of cetaceans in the Gulf of Mexico using an ecological niche modeling approach

Introduction

Understanding species’ geographic distribution patterns and related environmental factors is a central topic of population ecology (Guisan & Zimmermann, 2000). Environmental factors include both abiotic conditions that influence the physiological response (e.g., temperature) determined in turn by the species’ adaptive responses and the interspecific interactions (e.g., prey availability) (Soberón & Peterson, 2005; Peterson et al., 2011). Cetaceans are a group of fully aquatic mammals whose anatomical, morphological and physiological adaptations have allowed them to colonize a wide variety of aquatic habitats (Katona & Whitehead, 1988). Nevertheless, their distribution is usually explained in terms of the abundance of prey, primarily controlled by dynamic oceanographic conditions (e.g., sea surface temperature and mesoscale processes), as well as by physiographic features (e.g., bottoms depth and slope) (Kenney et al., 1997; Forcada, 2018).

The Gulf of Mexico (GOM) is a semi-enclosed sea connected to the Atlantic Ocean and Caribbean Sea. It has a 1.6 million km² surface and includes the Exclusive Economic Zones (EEZ) of Mexico (which represent ∼55% of the total surface of the GOM), the United States (∼40%), and Cuba (∼5%) (De Lanza Espino & Gómez-Rojas, 2004). There are at least 21 species of cetaceans that inhabit the GOM, including one mysticete, the Bryde’s whale (Balaenoptera edeni), which is distributed exclusively in the northeastern GOM (Soldevilla et al., 2017), and 20 odontocetes (Würsig, 2017).

The distribution of cetaceans in the northern GOM (i.e., the U.S. EEZ) has been extensively studied. Based on sighting records, Maze-Foley & Mullin (2006) divided the cetaceans into two communities: (1) the continental shelf community, which includes the bottlenose dolphin (Tursiops truncatus), the Atlantic spotted dolphin (Stenella frontalis), and the Bryde’s whale, and (2) the continental slope community, which comprises the remaining species, although the rough-toothed dolphin (Steno bredanensis) can be found in both. More recently, Roberts et al. (2016) used density surface models to describe the spatial distribution of cetaceans in the northern GOM. Overall, their results are consistent with those of Maze-Foley & Mullin (2006), although they highlight the importance of the continental slope and submarine canyons, such as the Mississippi Canyon, as areas of high density of cetaceans. In contrast, the distribution of cetaceans in the southern GOM (i.e., the EEZs of Mexico and Cuba) is poorly understood. In fact, only one study has covered this region, but it was conducted by extrapolating data from the northern GOM (Mannocci et al., 2017).

Besides the ecological relevance of the GOM, it is an important economic area where fishing, tourism, and the hydrocarbon industry generate billions of dollars annually (Karnauskas et al., 2013), and it is a key transportation region (Shepard et al., 2013). Given its economic importance, the GOM ecosystem is under increasing anthropogenic pressure, threatening cetacean populations (Roberts et al., 2016). To determine the extent and impact of these hazards and to optimize threat mitigation and conservation measures, it is necessary to have accurate predictions of their distribution on a broader scale; that is, at the ecosystem level. However, the latter is complicated given the limited data on cetaceans in the southern GOM (Ramírez-León et al., 2020), but one way to achieve this is to use ecological niche models (ENM).

ENM are statistical tools that define the distribution of suitable habitats of a species based on its ecological requirements (Peterson et al., 2011). The rationale is that the records, which are discontinuous in nature, are related to environmental and/or spatial characteristics (environmental predictors) to predict the suitable areas of the species in unsampled locations; therefore, the maps produced are spatially continuous, showing the regions where greater aggregation is expected (Franklin, 2010). Thus, areas of high habitat suitability are defined as those sites were ideal (or favorable) conditions exist for a species’ long-term subsistence (Peterson & Soberón, 2012).

Our objective was to estimate the habitat suitability of the odontocetes of the GOM to identify those geographical regions that could support a high diversity of these cetaceans. The analysis included historical datasets of georeferenced sightings (presence-only data) recorded in both the south and north of the GOM. We used the maximum entropy (MaxEnt) modeling approach (Phillips, Anderson & Schapire, 2006), and five environmental predictors were selected based on their documented importance in determining the occurrence of cetaceans, either directly or indirectly: sea surface temperature, chlorophyll-a concentration, bottom depth and slope, and distance to the 200-m isobath (e.g.,  Praca et al., 2009; Fernandez et al., 2018; Pace et al., 2018).

Materials and Methods

Study and modeling area

The modeling area was not restricted to the GOM, due to the high movement capacity of cetaceans, and because there are no physicals barriers in the marine environment for them. The area was extended to include the warm-temperate and tropical oceanic provinces of the northwest Atlantic Ocean (Fig. 1A; Spalding et al., 2007).

The physiography of the GOM is complex and consists of 13 physiographic sub-provinces (Fig. 1B). The continental shelf (≤200 m deep) can be very narrow, as the Tamaulipas-Veracruz shelf, or extensive, like the Yucatan and Florida shelves. The continental slope extends from the 200 m continental shelf break to 2,800 m depth, and there are vast canyons, such as the Mississippi Canyon. The oceanic zone extends beyond the slope up to the abyssal plain, where depths >3,500 m are reached (Bouma & Roberts, 1990; Monreal-Gómez, Salas-de León & Velasco-Méndoza, 2004). The GOM oceanic waters have oligotrophic conditions that contrast with the eutrophic coastal regions, which receive a high nutrient input by river discharges, mainly in the northern GOM (Biggs, 1992; Lohrenz et al., 1999; Muller-Karger et al., 2015).

Results

Habitat suitability was modeled for only twelve species (Table 1), which were those that after filtering had the minimum sample required (≥ 30 presence records; Wisz et al., 2008). The pygmy sperm whale (Kogia breviceps), pygmy killer whale (Feresa attenuata), false killer whale (Pseudorca crassidens), and melon-head whale (Peponocephala electra) were excluded due to small sample size (Table 1).

The models showed a good degree of discriminatory ability based on the AUC scores, which ranged from 0.74 (the pantropical spotted dolphin, Stenella attenuata, model) to 0.91 (the bottlenose dolphin model) (Table 2). On the other hand, the OR₁₀ value was close to the expected value in some models, such as the rough-toothed dolphin, but in others it was higher, as in the spinner dolphin (Stenella longirostris) model (Table 2), suggesting some degree of overfitting.

The environmental predictors used in each model differ (Table 2; Supplemental Information 4). The bottlenose dolphin models had the fewest predictors, while the Risso’s dolphin (Grampus griseus), Atlantic spotted dolphin, pantropical spotted dolphin and spinner dolphin models had the most. Slope was included in 11 models, while both depth and minimum-SST in 10 (Table 2). However, in terms of contribution, depth was the most important environmental predictor, with a contribution of >25% in seven models, followed by the minimum-SST, which had an important contribution in five models.

High suitability areas for 10 species were located on the continental slope (Figs. 2–4). The pantropical spotted dolphin seems to be the species with the widest distribution, potentially occupying the entire continental slope. High suitability areas for the sperm whale (Physeter macrocephalus), short-finned pilot whale (Globicephala macrorhynchus), and spinner dolphin were found on the inner continental slope, while for the rough-toothed dolphin they were located both on the outer continental shelf and on the slope. Dwarf sperm whale (Kogia sima), Risso’s dolphin, and Cuvier’s beaked whale (Ziphius cavirostris) models show almost continuous high suitability along the northern continental slope, from Florida to Louisiana-Texas, even extending to the Rio Grande and Tamaulipas-Veracruz slopes. High suitability areas for the Clymene dolphin (Stenella clymene) were found on the northern slope, with small patches on the Tamaulipas-Veracruz slope. Striped dolphin (Stenella coeruleoalba) model identified as high suitability areas some patches on the Tamaulipas-Veracruz slope and on the bay of Campeche. The Atlantic spotted dolphin model indicates that high suitability areas are located from the continental shelf to the inner slope of the entire GOM, whereas the model of the bottlenose dolphin points to the continental shelf, from Florida to the Tamaulipas-Veracruz.

The main region of high diversity of odontocetes was located between the Mississippi Canyon and the Louisiana-Texas slope (Fig. 5). Other suitable regions were identified on the west Florida slope and on the western continental slope, between the Rio Grande and Tamaulipas-Veracruz slopes.

Discussion

ENM are powerful tools for generating spatially explicit maps of species’ habitat suitability. We used the MaxEnt approach to model the habitat suitability of the GOM odontocetes, using data from both the north and south, and managed to identify regions where high diversity can be expected. We decided to use this approach because it allows the development of reliable models of the potential distribution based on presence-only data, although it is important to emphasize that these models do not represent the probability of the presence of a species. However, due to the paucity of data, we were able to model only 12 of the 20 species present in the GOM. Furthermore, because we use historical sighting records, the resulting maps are integrated images that show no temporal variations, and no biotic interaction were considered. Biotic interactions might improve habitat suitability models, however, require abundance demographic data at population over time (Anderson, 2017).

All models had a good discrimination power, with AUC values >0.70, indicating that the results are reliable, and can be used in planning management and conservation measures (Elith et al., 2006; Raes & Aguirre-Gutiérrez, 2018). On the other hand, the OR₁₀ values were higher than the expected value; however, they are within the range reported in other studies (e.g., Kramer-Schadt et al., 2013; Arthur, Morrison III & Morey, 2019). High OR₁₀ values suggest overfitting, which could be due to sampling bias and/or noise in the presence data (Anderson & Gonzalez, 2011; Merow et al., 2014). In our study area, the south of the GOM is under-sampled compared to the north. We attempted to reduce this bias using the spatial filtering to minimize the omission error, but it may not have been completely successful, especially in the case of the spinner dolphin model.

Among the used environmental predictors, depth was the most important, followed by minimum SST and bottom slope. These results are not surprising since it is well known that these variables influence the occurrence of cetaceans directly; for example, some species display relatively persistent bathymetric associations (Yen, Sydeman & Hyrenbach, 2004; Harvey et al., 2017), but mainly indirectly by playing a determining role in the availability, distribution, and abundance of their prey (Davis et al., 2002; MacLeod, 2009; Forcada, 2018). In fact, previous studies have shown that the distribution of several species of cetaceans of the GOM is strongly related to depth (e.g., Baumgartner, 1997; Davis et al., 1998; Baumgartner et al., 2001).

Our results are consistent with the segregated distribution of cetaceans proposed by Maze-Foley & Mullin (2006) for the northern GOM. The two dolphin species of the shelf community, the Atlantic spotted dolphins and the bottlenose dolphins, use different habitats. High suitability areas of the Atlantic spotted dolphins were located on the outer continental shelf and the inner slope, while the bottlenose dolphin has coastal habitats, occupying shallower waters; actually, it is the only species that inhabits lagoons, estuaries, and bays (e.g., Mullin et al., 1990; Griffin & Griffin, 2003; Martínez-Serrano et al., 2011). The continental slope community is composed of the remaining species (Maze-Foley & Mullin, 2006; present study), although densities of these can vary seasonally, at least in the northern GOM (Roberts et al., 2016; Mannocci et al., 2017). The aggregation of multiple species reveals important biological regions capable of supporting a high cetacean diversity (Harvey et al., 2017). Ten of the 12 species modeled showed habitat suitability areas along the continental slope, consistent with the relatively high diversity of cetaceans observed on the northern GOM continental slope (Davis et al., 2002; Maze-Foley & Mullin, 2006; Roberts et al., 2016). The exploitation of different types of habitat and prey allows this co-existence (Bearzi, 2005; Schick et al., 2011). The continental slope of the GOM covers a large area and presents underwater canyons (Bouma & Roberts, 1990), where the main prey (e.g., cephalopods) of the deep-diving species can accumulate (Biggs, Leben & Ortega-Ortiz, 2000; O’Hern & Biggs, 2009; Moors-Murphy, 2014). On the other hand, the species that primarily feed on epipelagic prey preferentially use the upper layers of the water column, where mesoscale structures occur (Davis et al., 1998; Davis et al., 2002).

We identify four suitable regions. The most notable was located in the north, encompassing the Mississippi Canyon and the Louisiana-Texas slopes, consistent with that previously reported for the north of the GOM (e.g., Mullin & Fulling, 2006; (Roberts et al., 2016). The other regions were located on the west Florida slope (east-northeast of the GOM), the Rio Grande slope (west-northwestern of the GOM), and the Tamaulipas-Veracruz slope (west-southwestern of the GOM). All these regions are characterized by their high primary productivity. In the north, productivity is directly influenced by the input of nutrients from the Mississippi and Atchafalaya rivers (Lohrenz et al., 1999), while in the west by the Grande and Pánuco rivers (Salmerón-García et al., 2011). The plumes of nutrient-rich waters are transported through the continental shelf (Del Castillo et al., 2001; Morey et al., 2003; Zavala-Hidalgo, Morey & O’Brien, 2003), reaching the slope by the interactions of anticyclonic-cyclonic eddies (Toner, 2003; Martínez-López & Zavala-Hidalgo, 2009). However, the largest region suitable for cetaceans was located in the north for two possible reasons. First, the large nutrient input from the Mississippi River into the shelf ecosystem favors huge phytoplankton blooms on spatial scales of tens to hundreds of kilometers (Lohrenz et al., 1997). Second, the continental slope in this region is extremely wide, which could favor the convergence of a greater number of cetacean species.

Conclusions

We identified areas of high suitability for 12 species of odontocetes in the GOM through the implementation of an ENM. Unfortunately, the paucity of data did not allow modeling all the species, which highlights the importance to establish transboundary research and monitoring programs between the U.S., Cuba, and Mexico to improve knowledge on the cetaceans of the GOM. Even so, we were able to detect four geographic regions where a high diversity of odontocetes is expected, all located on the continental slope. These suitable regions were identified using a spatial overlay, which although it is a very conservative approach (Harvey et al., 2017), it can be useful to detect areas where to focus conservation efforts (Tolimieri et al., 2015).

Supplemental Information