Species aggregation models resolve essential foraging habitat: Implications for conservation and management

Abstract

Species aggregations are a culmination of behavioral events arising from an array of biophysical interactions, dynamically shifting in space and time. Prediction of species' aggregation dynamics remains a challenge in studies of their distribution patterns. Species distribution models (SDMs) are statistical tools for understanding spatial patterns of marine biodiversity, ranging from essential species habitat, assessing fisheries bycatch, and projecting future distribution changes. SDMs involving pelagic species abundance generally do not typically resolve aggregation patterns. We use a 20-year observation record of seabird species aggregations, with seabirds being the most easily quantified "pelagic" species, to develop SDMs and a regional ocean modeling system to identify physical drivers and changes in aggregation location and intensity over time. We apply a conceptual ecosystem model to organize environmental covariates according to habitat production within coastal upwelling systems. The SDM used a 2-step modeling approach: a presence/absence model and a binary aggregation model. Thus, we aim to predict factors that characterize baseline ocean habitat for a species (presence/absence) and that aggregate large numbers of the species. Prediction of seabird aggregation results in realistic spatial distribution patterns that reflect known species habitat associations. Temporally, aggregation indices indicate mixed responses both within and between resident and migrant species, reflecting interannual effects of warm/cool ocean years and mesoscale structure supporting enhanced or decreased productive foraging habitat. The most abundant species were more likely to form aggregations during warmer years, indicating a response to a decrease in productive foraging habitat. The occurrence of species aggregations in spring is predictable by examining ocean-climate conditions in the preceding winter, thus providing a potential early warning system of anticipated ecosystem shifts. We contend that the aggregation occurrence model may improve the realism of pelagic SDMs and their utility for assessing spatial and temporal variability of trophic interactions. We discuss the utility of species aggregation models for quantifying the variability in critical pelagic habitats, the ecology and response of seabird species as indicators, advancement of ecosystem modeling and monitoring, and conservation applications (e.g., bycatch, wind energy, and oil spills).

Keywords: aggregation; conservation; distribution; ecosystem oceanography; foraging habitat; patchiness; realism; spatial organization; species distribution model.

FIGURE 1

Conceptual overview of the species aggregation model and environmental scope of seabird aggregations as a function of the Triad (enrichment, concentration, and retention processes) as applied to spring/summer ocean conditions within the California Current upwelling ecosystem. The species aggregation model applies mostly physical oceanographic variables derived from a Regional Ocean Modeling System and is organized according to physical drivers that generate productive foraging habitat for seabirds. Although seabird aggregations are likely driven by prey availability and distribution, physical proxies are used in order to make predictions. The spring/summer California Current Ecosystem is influenced by ocean basin‐scale conditions (dashed box) and is characterized by high inter‐annual variability that follows warm and cool ocean conditions that is dependent on conditions during winter and early spring that typically result in early or delayed spring upwelling transition resulting in stanzas of cooler and strong upwelling versus warm and weaker upwelling. Expanded (reduced) cool thermal habitat area and greater (weaker) persistence of fronts occurs during cooler/stronger (warm/weaker) upwelling periods. We examined seabird species aggregation response (e.g., intensity, spatial variance) for a selection of resident and migrant species, and we hypothesized the influence for each variable used in the formation of aggregation models. Model covariates derived and modified from Suca et al. (2022). See Table 1 for generalized species life history information. Taxa silhouettes for common murre, Cassin's auklet were drawn by the author (Jarrod Santora). Silhouettes for the albatross (Alexandre Vong) and the gull and shearwaters (Juan Carlos Jerí) are derived from www.phylopic.org under a CC0 1.0 Universal Public Domain Dedication license (https://creativecommons.org/publicdomain/zero/1.0/).

FIGURE 2

Spatial and temporal observations of total aggregations observed off central California used to train the species aggregation distribution model. Spatial maps represent total number of aggregations per grid cell, reflecting generalized habitat affinities per species. (a–c) Resident breeding species: CAAU, Cassin's auklet; COMU, common murre; WEGU, western Gull; (d–f) seasonal migrant species: SOSH, sooty shearwater; PFSH, pink‐footed shearwater; BFAL, black‐footed albatross. Gaps in in time series are missing years (2011, 2020). Taxa silhouettes for common murre and Cassin's auklet were drawn by the author (Jarrod Santora). Silhouettes for the albatross (Alexandre Vong) and the gull and shearwaters (Juan Carlos Jerí) are derived from www.phylopic.org under a CC0 1.0 Universal Public Domain Dedication license (https://creativecommons.org/publicdomain/zero/1.0/).

FIGURE 3

Summarized model output representing spatial mean probability of seabird aggregation, (left) spatial variance pattern of first component of Empirical Orthogonal Function (EOF1), and (right) time series of normalized aggregation index (calculated from threshold analysis of climatology) and the corresponding first principal component (PC1) that represents the temporal signal of the spatial variance of EOF1. (a–c) CAAU, Cassin's auklet; COMU, common murre; WEGU, western Gull; (d–f) SOSH, sooty shearwater; PFSH, pink‐footed shearwater; BFAL, black‐footed albatross. Spearman rank correlation coefficients (ρ) between the normalized aggregation index and PC1 are shown (all significant at p < 0.01). Taxa silhouettes for common murre and Cassin's auklet were drawn by the author (Jarrod Santora). Silhouettes for the albatross (Alexandre Vong) and the gull and shearwaters (Juan Carlos Jerí) are derived from www.phylopic.org under a CC0 1.0 Universal Public Domain Dedication license (https://creativecommons.org/publicdomain/zero/1.0/).

FIGURE 4

Summarized model output representing spatial mean probability of seabird aggregation, (left) spatial variance pattern of first component of Empirical Orthogonal Function (EOF1), and (right) time series of normalized aggregation index (calculated from threshold analysis of climatology) and the corresponding first principal component (PC1) that represents the temporal signal of the spatial variance of EOF1. (a–c) SOSH, sooty shearwater; PFSH, pink‐footed shearwater; BFAL, black‐footed albatross. Spearman rank correlation coefficients (ρ) between the normalized aggregation index and PC1 are shown (all significant at p < 0.01). Taxa silhouettes for the albatross (Alexandre Vong) and the gull and shearwaters (Juan Carlos Jerí) are derived from www.phylopic.org under a CC0 1.0 Universal Public Domain Dedication license (https://creativecommons.org/publicdomain/zero/1.0/).

FIGURE 5

Radar plots representing the relative importance of model covariates for predicting the probability of occurrence of seabird species aggregations: (a) Cassin's auklet, (b) common murre, (c) western gull, (d) sooty shearwater, (e) pink‐footed shearwater, and (f) black‐footed albatross. See Figure 1 for model covariate descriptions. Radar plots for presence/absence models shown in Appendix S1: Figures S2–S7. Taxa silhouettes for common murre and Cassin's auklet were drawn by the author (Jarrod Santora). Silhouettes for the albatross (Alexandre Vong) and the gull and shearwaters (Juan Carlos Jerí) are derived from www.phylopic.org under a CC0 1.0 Universal Public Domain Dedication license (https://creativecommons.org/publicdomain/zero/1.0/).

FIGURE 6

Environmental variables used for independent model evaluation and their spatial and temporal variance described using EOFs: sea‐surface temperature (SST) and sea level anomaly (SLA). These physical variables are not used in the model and derived from satellite observations. SLA is the difference between actual sea surface height (SSH) and the long‐term mean SSH, and it reflects the regional extent of anomalous water in the coastal ocean. (Top) Spatial variance patterns of PC1 and (bottom) normalized time series of PC1 for SST and SLA over January–June.

FIGURE 7

(Top) Spearman rank correlations between the first principal component (PC1) time series for May of the normalized seabird species aggregation index and independent sea surface temperature (SST) not included in the model. All rank correlations are significant at p < 0.05. (Bottom) Lagged spatial correlation maps between the May seabird aggregation time series and January–May observed SST. Bold contour lines highlight the area and indicate where correlations are significant at p < 0.05. Note that lagged spatial correlation maps indicate broad scale connections with both warm and cool conditions in the north Pacific (except for Cassin's auklet which reflect localized conditions). CAAU, Cassin's auklet; COMU, common murre; WEGU, western Gull; (d–f) SOSH, sooty shearwater; PFSH, pink‐footed shearwater; BFAL, black‐footed Albatross. Taxa silhouettes for common murre and Cassin's auklet were drawn by the author (Jarrod Santora). Silhouettes for the albatross (Alexandre Vong) and the gull and shearwaters (Juan Carlos Jerí) are derived from www.phylopic.org under a CC0 1.0 Universal Public Domain Dedication license (https://creativecommons.org/publicdomain/zero/1.0/).

FIGURE 8

(Top) Spearman rank correlations between first principal component (PC1) time series for May of the normalized seabird species aggregation index and independently observed sea level anomaly (SLA) that was not included in the model. All rank correlations are significant at p < 0.05. (Bottom) Lagged spatial correlation maps between the May seabird aggregation time series and January–May observed SLA. Bold contour lines indicate significance (p < 0.05) and highlights the area where correlations are significant at p < 0.05. Note that lagged spatial correlation maps indicate regional mesoscale connections with sea level anomaly conditions within the California Current. CAAU, Cassin's auklet; COMU, common murre; WEGU, western gull; (d–f) SOSH, sooty shearwater; PFSH, pink‐footed shearwater; BFAL, black‐footed albatross. Taxa silhouettes for common murre and Cassin's auklet were drawn by the author (Jarrod Santora). Silhouettes for the albatross (Alexandre Vong) and the gull and shearwaters (Juan Carlos Jerí) are derived from www.phylopic.org under a CC0 1.0 Universal Public Domain Dedication license (https://creativecommons.org/publicdomain/zero/1.0/).