The Pacific coast population of Western Snowy Plovers (Charadrius alexandrinus nivosus) is a federally listed threatened species, having experienced significant and pervasive population declines within its range in California, Oregon, and Washington.
Recovery of the species depends on the effective use of management resources because human- associated disturbance is a key factor in reducing or eliminating nesting habitat. Within the large extensive tracts of potential habitat, managers must decide where to pursue conservation actions and weigh the benefits of those actions against the costs of implementation. Geographic Information Systems (GIS)-based predictive habitat modeling is an approach that can identify and rank nesting habitats for Western Snowy Plover. We implemented a series of habitat models for the central coast of California (Recovery Unit 5 in the U.S. Fish and Wildlife Service Recovery Plan for the species).
Recovery Unit 5 (RU5) encompasses roughly 700 km of coastline in San Luis Obispo, Santa Bar- bara, and Ventura counties. It has remarkable geographic diversity, with predominantly west- facing beaches in San Luis Obispo County and south-facing beaches in much of Santa Barbara and Ventura Counties. The recovery unit also includes the northern California Channel Islands of San Miguel, Santa Rosa, Santa Cruz, Anacapa, Santa Barbara, and San Nicolas. An average of 1,000 Western Snowy Plovers nest in RU5, which is approximately half of the population of the Pacific Coast. A broad array of management actions is taken for Western Snowy Plover in RU5, ranging from nothing to intensive management.
We took two approaches to develop habitat suitability models for Western Snowy Plover.
The first is a deductive method, where environmental variables to predict habitat were selected based on the existing scientific literature and its description of the habitat preferences of the species. Such preferred habitat includes sand spits, dune-backed beaches, estuary and lagoon salt pans, beaches at creek and river mouths, bluff- backed beaches, dry salt ponds, and sand and gravel bars in rivers.
We consulted the literature to identify variables that could best describe optimal habitats, with our final models including elevation; slope gradient; distance from the coast; distance from streams and estuaries; distance from major rivers; landward boundary (e.g. dune, bluff); beach substrate (e.g. sand, gravel); beach width plus adjacent sand dunes/river sand bars; wave height; wind speed; and air/sea temperature. We intentionally excluded vegetation, presence of beach wrack, and other factors that could be influenced by management. Data were obtained or calculated from existing sources or developed for this purpose from remotely sensed information (e.g., description of landward boundary, beach width, and beach substrate from aerial photography). The continuous variables were standardized using fuzzy-logic linear or nonlinear functions with inflection points based on recorded species’ preferences before being used in the deductive models. Habitat suitability models were calculated from the continuous and categorical variables using Idrisi multi-criteria evaluation procedures.
Inductive models are built from species occurrence data and the modeling procedure selects those environmental variables and their values that best predict existing occurrences to extrapolate to potential habitat. We used nest site data obtained from beach managers to run the Maxent model with the same environmental variables as our deductive models, while keeping some nest site data aside to check the accuracy of the predictions.
Our deductive habitat models were also tested using nest site data and performed well. Beaches identified in the Recovery Plan as recovery sites encompassed high habitat suitability values as would be expected. For most beaches, our habitat suitability values for nest sites were statistically greater than at non-nest sites within recovery beaches. In instances where this was not true, a lack of management at high suitability value areas was the obvious and overwhelming explanation. For example, at Coal Oil Point in Santa Barbara County, which is intensively managed for Western Snowy Plovers, nest sites are concentrated in areas with high suitability values. By contrast, nest sites on the Morro Bay Strand in San Luis Obispo County are also found at some locations with high habitat suitability values, but are conspicuously absent from others. This reflects the different patterns of ownership and associated level of management.
With uneven management effort, nest sites are located in high suitability sites within managed areas, but also within lower habitat suitability sites that are intensively managed. Sites without management, despite having high suitability values defined by physical variables in the model, are not used for nesting.
Our inductive models were not as useful. They returned high suitability values for areas with nest site data, but did not return generalized rules that were able to extrapolate such high values to areas with known high habitat value but no nest point data.
We conclude a deductive approach provides a number of advantages for conservation planning in a heavily human-dominated landscape, even though it does depend on the existence of a well-developed natural history for the species in question. With this information, however, it can extrapolate the ideal conditions to locations where the species is no longer present and indeed is useful for identifying locations that would be excellent habitat if appropriate management were undertaken. Such sites are simply not ranked highly by inductive models when the training data are geographically clustered and the actual potential range is large.
Our model results should be useful for identifying, accurately delimiting, and assessing critical habitat for the species within RU5. Our deductive model gives gradations of habitat suitability within (and outside) existing recovery sites that might be used both to concentrate efforts within those sites and to reconfigure them during future recovery planning efforts.
By comparing the models that we developed with existing survey data we provide convincing evidence that many sites are indeed nesting habitat for western snowy plover, but it is only ongoing beach disturbance that consistently and chronically interferes with nesting. The model results, validated by the nest site and historic data, provide the basis for strong argument that “take,” as defined under the Endangered Species Act, is regularly occurring at high habitat suitability value sites that have wintering populations of Western Snowy Plovers but are not managed for the species.