![Map of California Gnatcatcher habitat in Ventura, Los Angeles, San Bernardino, Orange, Riverside, and San Diego counties as predicted by a habitat suitability model
Sources: [21, 29, 30].](https://www.researchgate.net/profile/Barbara-Kus/publication/382023290/figure/fig1/AS:11431281258948130@1720204065075/Map-of-California-Gnatcatcher-habitat-in-Ventura-Los-Angeles-San-Bernardino-Orange_Q320.jpg)
![Distribution of California Gnatcatcher survey points at the regional scale in Ventura, Los Angeles, San Bernardino, Orange, Riverside, and San Diego counties, CA in 2016
Sources: [29–32].](https://www.researchgate.net/profile/Barbara-Kus/publication/382023290/figure/fig2/AS:11431281258889134@1720204065918/Distribution-of-California-Gnatcatcher-survey-points-at-the-regional-scale-in-Ventura_Q320.jpg)
![Distribution of California Gnatcatcher survey points at the subregional scale in San Diego and Orange counties, CA in 2016
Sources: [29–32].](https://www.researchgate.net/profile/Barbara-Kus/publication/382023290/figure/fig3/AS:11431281258948131@1720204066319/Distribution-of-California-Gnatcatcher-survey-points-at-the-subregional-scale-in-San_Q320.jpg)
![Distribution of California Gnatcatcher survey points by year of last fire in San Diego County, CA in 2016
Sources: [29–33].](https://www.researchgate.net/profile/Barbara-Kus/publication/382023290/figure/fig4/AS:11431281258898344@1720204066785/Distribution-of-California-Gnatcatcher-survey-points-by-year-of-last-fire-in-San-Diego_Q320.jpg)
Access to this full-text is provided by PLOS.
Content available from PLOS ONE
This content is subject to copyright.
RESEARCH ARTICLE
Rangewide occupancy of a flagship species,
the Coastal California Gnatcatcher (Polioptila
californica californica) in southern California:
Habitat associations and recovery from
wildfire
Barbara E. KusID*, Kristine L. Preston ID, Alexandra Houston
U.S. Geological Survey, Western Ecological Research Center, San Diego, California, United States of
America
*barbara_kus@usgs.gov
Abstract
The Coastal California Gnatcatcher (Polioptila californica californica), a federally threatened
species, is a flagship species for regional conservation planning in southern California
(USA). An inhabitant of coastal sage scrub vegetation, the gnatcatcher has declined in
response to habitat loss and fragmentation, exacerbated by catastrophic wildfires. We docu-
mented the status of gnatcatchers throughout their California range and examined post-fire
recovery of gnatcatchers and their habitat. We used GIS to develop a habitat suitability
model for Coastal California Gnatcatchers using climate and topography covariates and
selected over 700 sampling points in a spatially balanced manner. Bird and vegetation data
were collected at each point between March and May in 2015 and 2016. Presence/absence
of gnatcatchers was determined during three visits to points, using area searches within 150
x 150 m plots. We used an occupancy framework to generate Percent Area Occupied (PAO)
by gnatcatchers, and analyzed PAO as a function of time since fire. At the regional scale in
2016, 23% of the points surveyed were occupied by gnatcatchers, reflecting the effect of
massive wildfires in the last 15 years. Similarly, PAO in the post-fire subset of points was
24%, with the highest occupancy in unburned (last fire <2002) habitat. Positive predictors of
occupancy included percent cover of California sagebrush (Artemisia californica), California
buckwheat (Eriogonom fasciculatum), and sunflowers (Encelia spp., Bahiopsis laciniata),
while negative predictors included laurel sumac (Malosma laurina) and total herbaceous
cover; in particular, non-native grasses. Our findings indicate that recovery from wildfire may
take decades, and provide information to speed up recovery through habitat restoration.
Introduction
Endangered species have long been threatened by habitat loss, degradation, and fragmentation
associated with anthropogenic land use. Increasingly, wildfire is exacerbating these processes
PLOS ONE
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 1 / 35
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Kus BE, Preston KL, Houston A (2024)
Rangewide occupancy of a flagship species, the
Coastal California Gnatcatcher (Polioptila
californica californica) in southern California:
Habitat associations and recovery from wildfire.
PLoS ONE 19(7): e0306267. https://doi.org/
10.1371/journal.pone.0306267
Editor: Yang Yang, Western Carolina University,
UNITED STATES
Received: February 9, 2024
Accepted: June 12, 2024
Published: July 5, 2024
Copyright: This is an open access article, free of all
copyright, and may be freely reproduced,
distributed, transmitted, modified, built upon, or
otherwise used by anyone for any lawful purpose.
The work is made available under the Creative
Commons CC0 public domain dedication.
Data Availability Statement: Data analyzed in this
paper are available at: Kus, B.E., and Houston, A.,
2021, Rangewide occupancy and post-fire
recovery of California Gnatcatchers in southern
California: U.S. Geological Survey data release,
http://doi.org/10.5066/F7PC30JX.
Funding: This research was funded by the
following grants to BEK: Grant Agreement
#Q1950402 from the California Department of Fish
and Wildlife Cooperative Endangered Species
and poses a major threat to biodiversity worldwide [1]. In California (USA), a global biodiver-
sity hotspot [2], wildfire has emerged in the last quarter century as a leading threat to coastal
sage scrub habitat [3–6]. Among the inhabitants of sage scrub are several endemic plants and
animals, including the Coastal California Gnatcatcher (Polioptila californica; “gnatcatcher”), a
federally threatened species restricted to coastal sage scrub in southern California [7].
Although fire is a natural part of the sage scrub ecosystem, changes to the fire regime, such as
increased frequency and intensity, challenge the persistence of coastal sage scrub habitat and
associated species [8–10]. Catastrophic fires such as the Cedar fire in 2003, the Harris fire in
2007, and a complex of fires in May 2014 burned hundreds of thousands of hectares in San
Diego County alone, destroying and degrading habitat required by California Gnatcatchers.
The impacts of wildfire are exacerbated by postfire invasion of non-native grasses and other
herbaceous vegetation [5,11,12] which reduces habitat suitability for gnatcatchers [13], and
promotes future fires through its high flammability [4,14].
The Coastal California Gnatcatcher (hereafter, California Gnatcatcher) has become estab-
lished as a “flagship” species for conservation to protect coastal sage scrub habitat and its
inhabitants, and is a covered species under several regional habitat conservation plans (HCPs),
including the Multiple Species Conservation Program [15] and the Multiple Habitat Conserva-
tion Program in San Diego County [16], the Orange County Central and Coastal Subregion
Conservation Plan [17], and the Western Riverside County Multiple Species Habitat Conser-
vation Plan [18]. Effective management of gnatcatchers under these plans relies on periodic
monitoring to determine population trends and evaluate the plans’ success in protecting the
species. Historically, gnatcatcher monitoring has been limited in geographic scope, often at the
level of individual preserves or local jurisdictions, and has employed differing methodologies;
limiting the extent to which findings can be generalized across larger spatial scales and multi-
ple populations. While satisfying the monitoring requirements of individual conservation
plans, local monitoring cannot capture the full effects of landscape-scale phenomena like wild-
fire which often span jurisdictional boundaries that birds do not recognize, nor can it assess
features like connectivity among conserved lands that facilitates dispersal, colonization, and
gene flow. Gnatcatcher monitoring was expanded in the early 2000’s with a series of surveys by
Winchell and Doherty [13,19] who investigated occupancy in a study area spanning Orange
and San Diego Counties. In further work, Winchell and Doherty [20] examined vegetation
characteristics as predictors of gnatcatcher occupancy, colonization, and extinction in
unburned habitat. These studies significantly advanced our understanding of gnatcatcher dis-
tribution and habitat requirements, yielding data to better protect and restore habitat to pro-
mote gnatcatcher viability. Our goal was to build on and expand these studies to encompass
the entire range of the Coastal California Gnatcatcher, and to take advantage of the opportu-
nity provided by several recent large wildfires to improve our understanding of the effects of
fire on gnatcatchers and their habitat.
We designed and implemented a standardized protocol to address two related objectives.
First, we sought to determine gnatcatcher occupancy at the regional scale, including habitat
from throughout the species’ range in southern California, as well as in two subregions:
Orange County and San Diego County, to address specific management objectives within
those jurisdictions associated with the requirements of their HCPs. We did not examine other
subregions because gnatcatcher abundance was insufficient for our analytical approach. Sec-
ond, we collected vegetation data to better understand gnatcatcher-habitat associations that
influence occupancy. In a parallel objective, we evaluated the effect of fire on gnatcatchers and
their habitat in two ways. First, we compared occupancy and vegetation characteristics across
regional and subregional sites varying in the length of time since the last fire. Second, we cre-
ated a separate postfire dataset that categorized sites according to time since last fire and
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 2 / 35
Conservation Fund (Section 6) (www.wildlife.ca.
gov); San Diego Association of Governments
Collaborative Agreement #5004597 (www.sandag.
org). The funders played no role in the study
design, data collection and analysis, decision to
publish, or preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
compared occupancy and vegetation characteristics across categories. Together, the results of
these investigations create a baseline for future rangewide monitoring to track changes in habi-
tat condition that affect gnatcatcher occupancy, and inform management to protect important
ecological conditions and processes required for species persistence.
Methods
Study area
We surveyed for California Gnatcatchers in coastal sage scrub habitat within the U.S. portion
of the species’ range in Ventura, Los Angeles, San Bernardino, Riverside, Orange and San
Diego counties, California. To establish a sampling frame for southern California, we devel-
oped a habitat suitability model [21] to use in place of a model (Technology Associates Inter-
national Corporation [22] used in previous California Gnatcatcher surveys of coastal regions
[13,19], but not designed for inland locations which differ considerably from the coast in cli-
matic and topographic conditions. We used Geographic Information System (GIS) software to
create a grid of points oriented north to south that encompassed the entire southern California
study area including developed lands and open space, with each point falling within the center
of a 150-m (meter) x 150-m grid cell. We used ArcGIS and digital data layers to calculate vari-
ous climatic, topographic, land use and vegetation variables at each point in the landscape
grid. California Gnatcatchers are often associated with California sagebrush (Artemisia califor-
nica; e.g. [19]); however, our vegetation layers did not identify coastal sage scrub supporting
California sagebrush for the entire study area. Thus, we modelled California sagebrush habitat
suitability [23] and included sagebrush model output among the environmental variables cal-
culated for each grid point.
We used a partitioned Mahalanobis D
2
approach [23–27] to construct alternative models of
habitat suitability for gnatcatchers in southern California. Mahalanobis D
2
represents a stan-
dardized distance between the multivariate mean for environmental variables at locations
where a species occurs and values calculated for the same set of environmental variables at
each grid point in the landscape being modelled [25,26]. The more similar environmental
characteristics are at a point in the landscape to the species’ multivariate mean, the more suit-
able the habitat is for the species. Habitat suitability for each 150-m x 150-m grid cell in the
study area is indicated by a Habitat Similarity Index (HSI) value that ranges from 0 (least simi-
lar to occupied habitat and considered least suitable) to 1 (most similar to occupied habitat
and most suitable). We categorized habitat suitability for gnatcatchers based on HSI values as:
Very High: 0.75–1.00, High: 0.50–0.74, Moderate: 0.25–0.49, and Low: 0–0.24.
We compiled California Gnatcatcher location records for 2000 to 2013 from a variety of
sources including the U.S. Geological Survey, California Department of Fish and Wildlife’s
California Natural Diversity Database, U.S. Fish and Wildlife Service (USFWS) Carlsbad
Office, County of San Diego (SanBIOS), Center for Natural Lands Management, Marine
Corps Air Station Miramar, Marine Corps Base Camp Pendleton, Naval Weapons Station Fall-
brook, Natural Communities Coalition (formerly the Nature Reserve of Orange County), and
Western Riverside County Multiple Species Habitat Conservation Program, and used these to
develop and evaluate the performance of alternative habitat models. Records with spatial accu-
racy lower than 80 m radius were excluded to avoid characterizing habitat in cells where the
birds did not occur. To characterize the environment used by gnatcatchers, we used ArcGIS to
spatially join each gnatcatcher location to the center point of the closest cell in the landscape
grid. We screened for spatial redundancy (location records that were assigned the same grid
cell) and used only one gnatcatcher record per cell. We used 1,063 location records from mul-
tiple datasets to construct the models and 3,205 records from the USFWS database to
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 3 / 35
independently evaluate and compare the performance of alternative habitat suitability models.
To avoid spatially biased sampling, we employed a subsampling strategy to balance gnatcatcher
locations used in constructing the models [28]. We divided the region up into 5 sampling
units: Los Angeles/Ventura, Riverside/San Bernardino, Orange, San Diego Coastal, and San
Diego Inland. We randomly subsampled 50 gnatcatcher locations from each area (i.e., a total
of 250 gnatcatcher locations) and constructed a model. We repeated this subsampling for
1,000 iterations and then averaged the results to develop a final model from which the Mahala-
nobis D
2
values were calculated across the landscape.
Eighteen models with different combinations of variables were developed and their perfor-
mance in predicting suitable habitat for California Gnatcatchers evaluated. The top-perform-
ing model included average minimum January and maximum July temperatures, annual
precipitation, elevation, northness, eastness, slope, topographic heterogeneity, the percent of
urban, coastal sage scrub and chaparral land cover within the 150-m x 150-m grid cell, and
predicted habitat suitability for California sagebrush. We used this model to generate HSI val-
ues for each cell in the landscape grid, and used cells with HSI values 0.5 (Very High or
High suitability; hereafter, “suitable”) to define our sampling frame (Fig 1).
We selected points for the regional survey from suitable habitat on conserved and military
lands throughout southern California (Fig 2), and for the subregional surveys from suitable
habitat in Orange and San Diego counties (Fig 3). Postfire points were restricted to San Diego
County, and were selected from within the footprints of wildfires that burned in 2003–06,
2007–10, and 2011–14, periods that bracketed the three major recent fires of 2003, 2007, and
2014, respectively. A fourth category in the postfire study, that of “unburned”, was comprised
of points selected from suitable habitat in San Diego County that had last burned between
2002 and 1878, the earliest year for which digital fire records are available [33;Fig 4]. The
median year of the last fire in the unburned category was 1878.
Sampling point selection
Sampling points for the regional, subregional, and postfire analyses were selected in the same
manner unless otherwise noted.
Number of points. We ran simulations in Program MARK [34] to estimate the number
of points needed to detect specific changes in California Gnatcatcher occupancy between suc-
cessive surveys. For these simulations, we set gnatcatcher detection probability at 0.6, a conser-
vative estimate based on a prior gnatcatcher survey with similar methodology [35]. Occupancy
was set at 0.30, the mean occupancy in “high” and “very high” quality habitat surveyed in 2004,
2007, and 2009 (range = 0.24–0.39; [13]). We specified detection of a 30 percent change in
gnatcatcher occupancy as the goal for the regional monitoring, and a 40 percent change for the
subregional monitoring. Lower power to detect change at the subregional level was considered
an acceptable way to manage cost and practicalities of subregional surveys given that they were
conducted within the context of regional monitoring designed to detect regionally significant
changes in occupancy that would trigger management response [18,36]. Based on the simula-
tions, we established sample sizes of 330 points for the regional survey and 180 for each subre-
gional survey.
In a similar manner, we determined sample sizes for each of the four categories in the post-
fire study. Guided by previous surveys for gnatcatchers at burned sites [13], we estimated occu-
pancy at 0.35 for unburned habitat, 0.30 for habitat burned in 2003–06, 0.15 for habitat
burned in 2007–10, and 0.05 for habitat burned in 2011–14. We then evaluated simulations for
sample sizes needed to provide power to distinguish among the four postfire categories with
90 and 95 percent confidence. Limited by the total number of points possible given the size of
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 4 / 35
Fig 1. Map of California Gnatcatcher habitat in Ventura, Los Angeles, San Bernardino, Orange, Riverside, and San Diego counties as predicted by a
habitat suitability model. Sources: [21,29,30].
https://doi.org/10.1371/journal.pone.0306267.g001
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 5 / 35
the burned areas, we opted for sample sizes of 110 points in each category. These sample sizes
provided a power of 1.0 to detect a difference between occupancies of 0.05 (burned in 2011–
14) and 0.35 (unburned) with 90 percent confidence. Power of other comparisons between
burned and unburned points was 0.67 for occupancies of 0.15 (burned in 2007–10) and 0.35,
and 0.09 for occupancies of 0.30 (burned in 2003–06) and 0.35. Among burned categories,
power was 0.38 and 0.48 to detect differences between occupancies of 0.05 and 0.15, and 0.15
and 0.30, respectively.
Distribution of points. Survey points for the regional, subregional, and postfire compo-
nents were selected separately in ArcGIS using the Spatially Balanced Sampling tool [37]. Spa-
tially balanced sampling creates a more flexible and efficient design than random sampling for
population trend analysis, and is robust to unanticipated events requiring adjustment to the
design, such as loss of access to points, need to replace points, etc. [37,38]. Points in the
regional and subregional surveys were spaced at least 600 m apart to avoid double-counting
birds [19]. Points in the postfire study were separated by a minimum of 450 m in order to
achieve our desired samples sizes within the fire footprints.
Number and timing of visits. With an estimated occupancy probability of 0.3 and detect-
ability of 0.6, 2 visits per point are sufficient to estimate gnatcatcher occupancy (Table 6.1 in
MacKenzie et al. [39]). We conservatively chose to make 3 visits to each point during the first
year of surveys in case detectability and occupancy differed substantially from our estimates.
Points were surveyed at 2-week intervals between March 15 and April 30, 2016, coinciding
Fig 2. Distribution of California Gnatcatcher survey points at the regional scale in Ventura, Los Angeles, San
Bernardino, Orange, Riverside, and San Diego counties, CA in 2016. Sources: [29–32].
https://doi.org/10.1371/journal.pone.0306267.g002
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 6 / 35
with the pre-breeding and early breeding season when gnatcatcher vocal detectability is at a
seasonal high [40]. Points burned in 2003–06, 2007–10, and 2011–14 were also surveyed dur-
ing the same timeframe in 2015 in a pilot study of postfire plots.
Data collection
Bird surveys. Gnatcatcher detection-nondetection surveys were conducted within 150-m
x 150-m (2.25 ha) plots centered on each sampling point [35]. Surveys were performed
between dawn and late morning/early afternoon depending on weather, avoiding conditions
of excessive wind (>20 km/h), temperatures below 4.5 degrees C or above 32 degrees C, or
precipitation greater than a drizzle. Upon arriving at the plot, the surveyor recorded date and
start time, and accessed imagery on hand-held devices displaying their location and the plot
boundaries in ArcGIS Collector. After allowing 1–2 minutes for activity at the plot to settle,
the surveyor walked slowly and methodically through the entire plot, looking and listening for
gnatcatchers. Playbacks of gnatcatcher “mew” vocalizations were broadcast in a standardized
manner [35] from each of the 4 cardinal quadrants of the plot (northwest, northeast, southeast,
southwest). A single bout of songs lasting approximately 20 seconds was played in each quad-
rant, with no more than 4 bouts per plot. Broadcasts were directed towards the center of the
plot to avoid attracting birds from outside the plot. Surveys ended after 45 minutes, or when
an adult California Gnatcatcher was detected within the plot, at which time the surveyor
Fig 3. Distribution of California Gnatcatcher survey points at the subregional scale in San Diego and Orange
counties, CA in 2016. Sources: [29–32].
https://doi.org/10.1371/journal.pone.0306267.g003
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 7 / 35
recorded the location of any birds detected, the number of song broadcasts (0–4), and the sur-
vey end time.
Vegetation sampling. Vegetation data were collected at each survey plot between May 1
and August 2, 2016. We used a modified point intercept method to record data at 2-m intervals
along 2 perpendicular (north-south and east-west) 30-m transects centered within each of the
Fig 4. Distribution of California Gnatcatcher survey points by year of last fire in San Diego County, CA in 2016. Sources: [29–33].
https://doi.org/10.1371/journal.pone.0306267.g004
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 8 / 35
4 cardinal quadrants of the plot, for a total of 128 vegetation sampling points per plot. Field
surveyors used high resolution digital aerial imagery and a compass to navigate to transects
and estimate sampling locations along their length. At each sampling point, surveyors placed a
2-m long measuring pole perpendicular to the ground and recorded “hits” of shrubs, trees,
and herbs. Shrubs and trees were defined by their height at the location of the measuring pole,
with shrubs being <2 m and trees 2 m tall. In addition to hits of shrubs and trees, surveyors
recorded the associated height of the tallest shrub and tree at the point. Based on preliminary
analysis of vegetation data collected previously in suitable gnatcatcher habitat [20], we
recorded species for hits of 20 shrub, tree and herbaceous species thought to influence gnat-
catcher occupancy and habitat suitability (Table 1); all other species were combined into
“other shrub/tree” or “other herbaceous” categories and the species for woody vegetation
noted under comments. We recorded “Dead” for hits of woody vegetation where the entire
Table 1. Vegetation species, heights, and unvegetated substrates recorded at California Gnatcatcher survey plots in 2016.
Species/Substrate Taxonomic Name/Comments Code
Shrub/Tree:
Oak Quercus spp. QUER
Laurel sumac Malosma laurina MALA
Elderberry Sambucus mexicana SAMX
Lemonadeberry Rhus integrifolia RHIN
Lilac Ceanothus spp. CEAN
California sagebrush Artemisia californica ARCA
California buckwheat Eriogonum fasciculatum ERFA
Bush sunflower Encelia californica ENCA
Brittlebush Encelia farinosa ENFA
San Diego sunflower Bahiopsis laciniata BALA
White sage Salvia apiana SAAP
Black sage Salvia mellifera SAME
Coyote bush Baccharis pilularis BAPI
Deerweed Acmispon glaber ACGL
Yucca Hesperoyucca whipplei or Yucca sp. YUCC
Dead Entire plant dead DEAD
Other shrub/tree OTHSHRTRE
Shrub height Height of tallest shrub SHRBHT
Tree height Height of tallest tree TREEHT
Herbaceous:
Mustard Brassica nigra,B.tournefortii BRAS
Star thistle Centaurea melitensis CEME
Artichoke thistle Cynara cardunculus CYCA
Fennel Foeniculum vulgare FOVU
Non-native grasses Bromus spp., Avena spp., others GRASS
Other herbaceous OTHHRB
Substrate:
Bare ground BARE
Boulder BOULDER
Pavement PAVEMENT
Trees: height 2m; shrubs: height <2m. Substrate only recorded when no vegetation present at sampling point.
https://doi.org/10.1371/journal.pone.0306267.t001
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 9 / 35
plant was dead. At points where no vegetation occurred, substrate was recorded as bare
ground, boulder (large rock too heavy to lift), or pavement.
Data analysis
We calculated percent cover for each vegetation variable (Table 1) as the percent of the 128
sampling points at which the species or substrate occurred, and then averaged these over the
four cardinal quadrants to obtain an overall average for each gnatcatcher survey plot. Similarly,
we calculated average shrub and tree heights for each gnatcatcher survey plot. In 12 plots for
which shrub height was missing, we used the average shrub height calculated for the rest of the
plots in that dataset as the value for shrub height. In addition to the species and substrates
recorded in the field, we analyzed five species extracted from “Comments” for woody vegeta-
tion recorded as “other shrub” or “other tree” after input from field investigators and data
exploration revealed these species to be more common in the northern part of the gnatcatch-
er’s range than in San Diego County from which the vegetation list (Table 1) was derived.
These species included chamise (Adenostoma fasciculatum, “ADFA”), chaparral bushmallow
(Malacothamnus fasciculatus, “MAFA”), Menzie’s goldenbush (Isocoma menziesii, “ISME”),
purple sage (Salvia leucophylla, “SALE”), and cactus (Cactaceae, primarily Opuntia sp. and
Cylindropuntia sp., “CACTUS”).
We created two composite variables combining structurally similar species. “Sunflowers”
(SUNFL) included bush sunflower, brittlebush, San Diego sunflower, and Menzie’s golden-
bush, while “Sage” (SAGE) combined black and purple sages. These composite variables
allowed us to analyze taxonomically or structurally similar species that differed in distribution
across the study area. We created a variable called “Total shrub/tree” (TOTSHRTRE) that rep-
resented cover of all species (i.e., the 25 individual species plus “other” species). In calculating
total shrub/tree cover, each point received a maximum of 1 “hit”, regardless of how many spe-
cies of shrubs or trees occurred there. We present means and standard deviations (SD) to dis-
play variability among plots, and present means and standard errors (SE) in comparisons
across datasets where non-overlapping SEs indicate significant differences.
In addition to analyzing vegetation variables, we used ArcGIS to extract physical variables
for each gnatcatcher survey plot in the regional and subregional datasets, including distance to
the Pacific coast (DISTCOAST), elevation (ELEV), and slope (SLOPE) of the plot location. We
did not analyze physical variables for the postfire dataset because points were distributed non-
randomly with regard to these variables; recent fires were concentrated at lower elevations
near the coast while older fires were inland at higher elevations. We also calculated an index of
time since the last fire at each regional and subregional plot (LASTFIRE) relative to 1878, the
earliest record in the fire perimeter dataset [33]. LASTFIRE was calculated by subtracting 1877
from the year of the most recent fire at each plot to generate values that ranged from 1 (last fire
in or prior to 1878) to 137 (last fire in 2014). Structuring the index as increasing with recency
of fire rather than decreasing eliminated the need to recalculate LASTFIRE in future survey
datasets while retaining the ability to update the variable for any plots that burned subsequent
to the 2016 survey.
We modelled single season occupancy in Program MARK [34] to estimate detection and
occupancy probabilities (pand psi, respectively) separately for 2015 and 2016. The analysis for
2015 included estimates of pand psi for the postfire dataset, while those for 2016 included esti-
mates for the regional, 2 subregional, and postfire datasets. We modelled occupancy as a func-
tion of covariates (see below), and evaluated support for models using Akaike’s Information
Criterion corrected for small sample size (AIC
c
; [41]). We tested for goodness-of-fit using the
parametric bootstrapping approach of MacKenzie and Bailey [42] in Package Unmarked [43],
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 10 / 35
and present adjusted estimates of the overdispersion parameter (^c) and QAIC
c
values for over-
dispersed datasets. We modelled detection probability as constant among surveys.
We identified covariates for inclusion in models of occupancy for each dataset using a 3-step
screening process. First, we performed a Principal Components Analysis (PCA) using function
Principal with varimax rotation to identify patterns of variability in vegetation composition and
structure across the survey plots. The PCA for postfire models was performed on the unburned
plots to provide a reference condition for burned plots. To evaluate whether habitat variability
among plots was related to occupancy by gnatcatchers, we used the scores of individual plots on
each principal component as input for two inferential analyses. We ran t-tests to compare
means of plots with and without gnatcatcher detections (hereafter, “occurrence” or “presence”),
and used GLM (generalized linear model; binomial family) to evaluate the principal compo-
nents as significant predictors of gnatcatcher occurrence. Principal components significantly
(P 0.05) related to gnatcatcher occurrence in either t-tests or GLMs were evaluated further.
We assessed the contribution of variables to each significant principal component and extracted
those with loadings 0.35 to create a set of covariates for inclusion in models of occupancy.
Models were built hierarchically by first creating models based on each significant compo-
nent from the PCA and comparing them using AIC. The top-ranked model at this step was
refined and evaluated for support by adding covariates from the best-supported models for the
other principal components. Finally, the physical covariates were added to the top vegetation
model to produce a final model set with ΔAIC
c
or ΔQAIC
c
4. We removed models with
uninformative parameters (i.e., where addition or removal of a covariate had a negligible effect
on deviance; [44]).
We present the final model set for each of the datasets (regional, subregional, and postfire),
and considered models within 2 AIC
c
or QAIC
c
units of the top-ranked model to be the best
supported [41]. We estimated occupancy for each dataset by averaging over the best-supported
models. We used the “group” function in Program MARK [34] to estimate occupancy for each
of four postfire categories: unburned, 2003–2006 burned, 2007–2010 burned, and 2011–2014
burned. Relationships between covariates and estimated occupancy are plotted based on the top
model for each dataset, or the best-supported model in which that particular covariate appears.
Data analyzed in this paper are available in a U.S. Geological Survey data release [45]. This
research was observational in nature. The field study was approved by the U.S. Geological Sur-
vey Western Ecological Research Center Animal Care and Use Committee.
Results
We surveyed for gnatcatchers at 334 regional points (Table 2). Six points were dropped follow-
ing the first survey, 3 points were excluded from analysis because none of the surveys of these
Table 2. Sample sizes for points surveyed and analyzed in 2015 and 2016 regional, subregional and postfire datasets.
Dataset # points surveyed # points excluded # points analyzed # points per survey frequency
3 2 1
Regional 334 11 323 285 32 6
OC Subregional 180 1 179 172 7 0
SD Subregional 182 4 178 174 2 2
Postfire:
Unburned 103 1 102 101 0 1
2003–06 burned 107 1 106 105 1 0
2007–10 burned 111 2 109 106 2 1
2011–14 burned 106 0 106 103 3 0
https://doi.org/10.1371/journal.pone.0306267.t002
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 11 / 35
points met the 45-minute protocol in the absence of gnatcatcher detections, and 2 points
lacked vegetation data. One hundred and eighty Orange County subregional points were sur-
veyed; a landslide at 1 point prevented collection of vegetation data, so 179 points were ana-
lyzed. The San Diego subregional dataset included 178 points after excluding 2 points dropped
after the first visit and 2 points failing to meet the survey time requirement. Among the postfire
datasets, the number of points surveyed for gnatcatchers ranged from 103 to 111. Of 103
unburned points, 1 was dropped after the first survey. Of 111 points burned in 2007–10, 2
points failed to meet the survey time requirement and were excluded from analysis. Eighty-
seven to ninety-nine percent of points in each dataset were surveyed 3 times (Table 2).
Vegetation composition
Regional and subregional points. We analyzed 17 species and cover types that averaged
at least 1 percent of the cover at either regional or subregional points (Fig 5). At regional
points, grass and other herbaceous vegetation made up most of the cover. Among woody
shrubs and trees, California sagebrush and California buckwheat contributed the most cover,
followed by black/purple sage and laurel sumac. Bare ground averaged 14 percent (±13 per-
cent SD) of cover at regional points.
Gnatcatcher occurrence at points was non-random with regard to vegetation composition
(Fig 6). Cover of California sagebrush, California buckwheat, black/purple sage, sunflowers,
and bare ground was higher on average at points with gnatcatcher detections than at points
without detections. Points without gnatcatcher detections supported higher cover of grass,
mustard, and oaks than points where gnatcatchers were detected. These differences were
observed in both the Orange County and San Diego subregions, with the exception that oak
cover did not differ between points with and without gnatcatcher detections in San Diego.
Of the species making up the most cover across regional and subregional points, average
cover of grass and other herbaceous vegetation, as well as dead woody vegetation, was lower in
the Orange County subregion than in the San Diego sub-region and the region as a whole (Fig
7), while cover of California sagebrush, California buckwheat, and black/purple sage tended to
Fig 5. Average percent cover (+ SD) of vegetation species and cover types at regional points, 2016. See Table 2 for
sample sizes.
https://doi.org/10.1371/journal.pone.0306267.g005
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 12 / 35
be higher. Total shrub and tree cover, including species making up <1 percent of cover, was
higher in Orange County, and total herbaceous cover lower, than at the regional and San
Diego subregional points (Fig 8).
Postfire points. Vegetation at the unburned points was similar to that in the region as a
whole (Fig 9). Grass and other herbaceous species made up most of the vegetation, although
average cover of grass at unburned points (24 percent) was roughly two-thirds that at regional
points (33 percent, Fig 5). California sagebrush and California buckwheat dominated the
shrub cover at unburned points (Fig 9), as at regional and subregional points (Figs 7and 8).
Fig 6. Average percent cover (±SE) of vegetation and cover types at regional points occupied and not occupied by
California Gnatcatchers, 2016. See Table 2 for sample sizes.
https://doi.org/10.1371/journal.pone.0306267.g006
Fig 7. Average percent cover (+ SD) of vegetation species and cover types at regional, San Diego (SD) subregional,
and Orange County (OC) subregional points, 2016. See Table 2 for sample sizes.
https://doi.org/10.1371/journal.pone.0306267.g007
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 13 / 35
Fire altered the vegetation, which differed across points as a function of time since burn
(Fig 9). At the most recently burned points (2011–14), cover of bare ground, grass and other
herbaceous species, including fire-followers deerweed and black mustard, was higher than at
unburned points. Among shrubs, cover of California sagebrush, California buckwheat, and
black/purple sage was lower than at unburned points; in contrast, cover of laurel sumac was
slightly higher.
At sites burned in 2007–10, grass cover was even higher than at the recently burned points,
and approximately twice as high as in unburned habitat, while cover of bare ground was
reduced relative to recently burned sites (Fig 9). Cover of other herbaceous species was lower
than at the recently burned points, but still substantially higher than in unburned habitat. Cali-
fornia sagebrush cover was twice as high as at recently burned points, but still lower than at
unburned points. California buckwheat cover was over 5 times higher than at recently burned
Fig 8. Average percent cover (±SE) of total shrubs and trees, and total herbaceous vegetation, across regional, San
Diego (SD) subregional, and Orange County (OC) subregional points, 2016. See Table 2 for sample sizes.
https://doi.org/10.1371/journal.pone.0306267.g008
Fig 9. Average percent cover (+ SD) in 2016 of vegetation species and cover types at unburned points and points
burned in 2003–06, 2007–10, and 2011–14. See Table 2 for sample sizes.
https://doi.org/10.1371/journal.pone.0306267.g009
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 14 / 35
points, and exceeded the average in unburned habitat. Cover of laurel sumac was 60 percent
higher, and double that in unburned habitat.
Differences in vegetation structure relative to unburned habitat persisted even into the old-
est fire category (2003–06). Grass and other herbaceous cover remained elevated at approxi-
mately twice that in unburned habitat, as was the case for laurel sumac (Fig 9). California
buckwheat cover continued to increase and exceeded that in unburned habitat. Cover of Cali-
fornia sagebrush changed little relative to that at points last burned in 2007–10, and remained
at about 60 percent of that at unburned points. In contrast, cover of black/purple sage was
roughly 3 times higher than in the two younger fire categories, comparable to that at unburned
sites.
Overall, recently burned sites supported much more total herbaceous cover (including
grass) than unburned sites, a difference that intensified and persisted over all three burn cate-
gories (Fig 10). In contrast, total shrub and tree cover was reduced to about two-thirds of that
in unburned habitat immediately following fire, but recovered relatively quickly, although the
recovery included species such as laurel sumac not necessarily associated with gnatcatcher
occurrence.
California Gnatcatcher occupancy
Habitat covariates. A Principal Components Analysis of vegetation data collected at
regional points produced five Principal Components (PC) collectively explaining 37 percent of
the variability across points (Table 3). PC1, which accounted for 10 percent of the variability,
differentiated between sites with high cover of total shrub/tree and sage (especially black sage),
as well as relatively tall shrubs, and low cover of grass, other herbaceous, and total herbaceous
vegetation. PC2, which accounted for 7 percent of the variability, described habitat character-
ized by high cover of oaks, Mexican elderberry, other shrub/tree, total shrub/tree, and tall tree
and shrub heights, with little bare ground and low cover of brittlebush. PC3 (7 percent of vari-
ability) described habitat suggestive of disturbed coastal sage scrub, including high cover of
California sagebrush, white sage, lilac, laurel sumac, grass and star thistle, and low cover of
brittlebush and bare ground. PC4 (7 percent of variability) was suggestive of less disturbed
coastal sage scrub supporting high cover of California sagebrush, California buckwheat,
Fig 10. Average percent cover (±SE) in 2016 of total shrubs and trees, and total herbaceous vegetation, across
unburned points and points burned in 2003–06, 2007–10, and 2011–14.
https://doi.org/10.1371/journal.pone.0306267.g010
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 15 / 35
Table 3. Variables with loadings >0.35 on factors extracted by Principal Components Analysis of the regional, San Diego and Orange County subregional, and
postfire (unburned) datasets for 2016.
Regional PC1 PC2 PC3 PC4 PC5
+ SAME + SAMX + ARCA + ARCA + ADFA
+ SAGE + QUER + CEAN + ERFA + ACGL
+ TOTSHRTRE + OTHSHRTRE + CEME + MALA + YUCC
- GRASS + TOTSHRTRE + MALA + BALA + BARE
- OTHHRB - ENFA + SAAP + SUNFL - GRASS
- TOTHRB - BARE + GRASS + DEAD - TOTHRB
+ SHRBHT + SHRBHT - ENFA + TOTSHRTRE
+ TREEHT - BARE
Var. explained 0.10 0.07 0.07 0.07 0.06
Effect in GLM +++ ++
P value t-test <0.001 0.03
San Diego Subregional PC1 PC2 PC3 PC4 PC5
+ ARCA + FOVU +RHIN + SAME + ARCA
+ ERFA + GRASS +SAMX + SAGE + CEAN
+ MALA + OTHHRB +QUER - TOTHRB + CEME
+ BALA + TOTHRB + OTHSHRTRE + MALA
+ SUNFL - BARE + TOTSHRTRE + SAAP
+ DEAD - TOTSHRTRE + SHRBHT + TOTSHRTRE
+ TOTSHRTRE + TREEHT
+ PAVEMENT
- BARE
Var. explained 0.09 0.08 0.08 0.08 0.07
Effect in GLM + - + +
P value t-test 0.14 0.01 0.12 0.08
Orange County Subregional PC1 PC2 PC3 PC4 PC5
+ ARCA + BAPI + SAMX + ADFA + ACGL
+ SAME + ENCA + QUER + MAFA + CEME
+ SAGE + ISME + OTHSHRTRE + SAME + SAAP
+ ERFA + SUNFL - SAME + SAGE + OTHHRB
+ MALA + SAMX - SAGE + YUCC + TOTHRB
+ RHIN + TREEHT + OTHSHRTRE - BARE
+ TOTSHRTRE + BOULDER
- BRAS
- CYCA
- GRASS
- OTHHRB
- TOTHRB
+ SHRBHT
Var. explained 0.15 0.07 0.07 0.06 0.06
Effect in GLM ++ ++ - -.
P value t-test 0.01 0.001 0.01
Postfire PC1 PC2 PC3 PC4 PC5
+ RHIN + ARCA + ERFA + SAME + FOVU
+ SAMX + CEME + ENCA + SAGE - ACGL
+ OTHSHRTRE + MALA + BALA - ARCA - ISME
+ TOTSHRTRE + SAAP + SUNFL - CEME - BARE
(Continued)
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 16 / 35
sunflowers (especially San Diego sunflower), laurel sumac, dead, and total shrub/tree. PC5 (6
percent of variability) described a more limited type of sage scrub characterized by high cover
of chamise, deerweed, yucca, and bare ground, and low cover of grass and other herbaceous
vegetation.
Analysis of the San Diego subregion, Orange County subregion, and postfire datasets pro-
duced PCs similar to those for the regional dataset, although not always ordered the same in
terms of the amount of variability explained (Table 3). GLMs regressing gnatcatcher presence
against scores on each factor for points within the four datasets consistently identified factors
with low grass and other herbaceous cover, and high sage (SAME and SAGE) cover (PC1, PC4,
and PC1 in the regional, San Diego subregional, and Orange County subregional datasets,
respectively) as significant positive predictors of gnatcatcher occurrence. In the postfire dataset,
the conditions of low grass and herbaceous cover, and high sage (SAME and SAGE) cover, were
split across two PCs (PC1 and PC4, respectively); only PC1 was found to be a significant predic-
tor of gnatcatcher presence, but average scores on PC4 differed between points with and without
gnatcatcher detections in t-tests. Cover of California sagebrush, California buckwheat, and sun-
flowers were also positively associated with gnatcatcher occurrence at points (PC4 in the
regional dataset, PC1 in the San Diego subregional dataset, PC1 and PC2 in the Orange County
subregional dataset, PC3 in the postfire dataset). Negative predictors were found in analyses of
the San Diego and Orange County subregional datasets, and included factors describing sites
with little bare ground and high cover of herbaceous vegetation, in particular exotic species such
as fennel (San Diego subregional, PC2) and star thistle (Orange County subregional, PC5). An
additional factor in the Orange County subregional dataset, PC4, emerged as a negative predic-
tor of gnatcatcher presence; this factor is suggestive of rocky chapparal habitat with high cover
of chamise, sage (SAME and SAGE), chaparral bushmallow, and other shrubs/trees.
California Gnatcatcher occupancy models
Regional. Two models predicting gnatcatcher occupancy at the regional scale were well-
supported (ΔAIC
c
2, Tables 4and 8). Cover of California sagebrush and California
Table 3. (Continued)
- GRASS + DEAD - SAAP + SHRBHT
- OTHHRB + TOTSHRTRE - TOTHRB
- TOTHRB - BARE
+ SHRBHT + SHRBHT
Var. explained 0.10 0.08 0.07 0.07 0.05
Effect in GLM ++ ++
P value t-test 0.01 <<0.001 0.05
A “+” preceding variable name indicates a positive loading, while “-”indicates a negative loading. Effect in GLM (generalized linear model): +++ = positive effect on
gnatcatcher presence, P0.001; ++ = positive effect, P0.01; + = positive effect, P0.05;— = negative effect, P0.05; -. = negative effect, P0.10. Vegetation codes (in
alphabetical order): ACGL = Acmispon glaber (deerweed), ADFA = Adenostoma fasciculatum (chamise), ARCA = Artemisia californica (California sagebrush), BALA =
Bahiopsis laciniata (San Diego sunflower), BARE = bare ground, SAGE = black/purple sage, BOULDER = boulder, BRAS = Brassica spp. (mustard), CEAN = Ceanothus
spp. (lilac), CEME = Centaurea melitensis (star thistle), DEAD = dead, ENCA = Encelia californica (California sunflower), ENFA = E.farinosa (brittlebush), ERFA =
Eriogonum fasciculatum (California buckwheat), FOVU = Foeniculum vulgare (fennel), ISME = Isocoma menziiesii (Menzie’s goldenbush), GRASS = non-native grass,
MAFA = Malacothamnus fasciculatus (chaparral bushmallow), MALA = Malosma laurina (laurel sumac), OTHHRB = other herbaceous, OTHSHRTRE = other shrub/
tree, PAVEMENT = pavement, QUER = Quercus spp. (oak), RHIN = Rhus integrifolia (lemonadeberry), SAAP = Salvia apiana (white sage), SAME = S.mellifera (black
sage), SAMX = Sambucus mexicana (Mexican elderberry), SHRBHT = shrub height, SUNFL = California sunflower, San Diego sunflower, brittlebush, Menzie’s
goldenbush), TOTHRB = total herbaceous, TOTSHRTRE = total shrub/tree, TREEHT = tree height, YUCC = Hesperoyucca whipplei or Yucca sp.
https://doi.org/10.1371/journal.pone.0306267.t003
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 17 / 35
buckwheat appeared in both models, and both covariates were positive predictors of occu-
pancy (Fig 11). Total shrub and tree cover also appeared in both models, with the results sug-
gesting a threshold response whereby occupancy increased with increasing cover of shrubs
and trees up to about 30–40 percent, then declined thereafter (Fig 11). There was some support
for shrub height as an additional although weak positive predictor of gnatcatcher occupancy
(Fig 11). In contrast, total herbaceous cover, which appeared in all models, was negatively
related to occupancy (Fig 11). Occupancy increased with time since the last fire over a period
of about 80 years before plateauing, based on the top model (Fig 11).
San Diego subregion. Three vegetation covariates contributed to well-supported models
predicting gnatcatcher occupancy within the San Diego subregion, and a fourth received some
support (ΔQAIC
c
close to 2; Tables 5and 8). Cover of California sagebrush and time since last
fire were the strongest predictors of occupancy, with LASTFIRE appearing in every model as
either a linear or quadratic variable (Fig 12). Three covariates appeared in one model each,
including California buckwheat and sunflowers, both of which had a positive effect on occu-
pancy, and laurel sumac, which had a negative effect (Fig 12). Occupancy increased slightly
with elevation (Fig 13).
Orange County subregion. Gnatcatcher occupancy in the Orange County subregion was
best described by a model (Table 6) that included cover of California sagebrush, California
buckwheat, sunflowers, and bare ground (Fig 14, Table 8), all of which had positive effects on
occupancy. Also included in the best supported models was elevation, which unlike in San
Diego County was a negative, and stronger, predictor of occupancy (Fig 14).
Postfire. Four models describing occupancy in postfire habitat received strong support
(Tables 7and 8). The postfire analysis included four groups (unburned, 2003–06 burned,
2007–10 burned, and 2011–14 burned; for illustrative purposes, results for the unburned
group are displayed in figures). The positive effects of California sagebrush and sunflowers, as
well as the negative effect of laurel sumac, were in every model (Fig 15). Strong support was
found for models with two additional covariates, cover of grass and cover of other herbaceous
vegetation (Fig 16), both of which had a negative effect on occupancy. Sage (all black sage in
this dataset) cover emerged as a new predictor not seen in analyses at the regional and subre-
gional scales, with a positive effect on occupancy (Fig 15). Relationships between individual
covariate predictors and occupancy were similar across groups within the postfire analysis
(example shown for California sagebrush, Fig 17), and were stronger than those found for the
San Diego subregion as a whole.
California Gnatcatcher occupancy. Gnatcatcher occupancy differed across regional and
subregional scales in 2016 and was higher in the Orange County subregion than in the San
Table 4. Logistic regression models of California Gnatcatcher occupancy in 2016 as a function of habitat covariates at the regional scale.
Model ΔAIC
c
AIC
c
Weight logLik Number Parameters Deviance
1ERFA+ARCA+TOTSHRTRE
2
+TOTHRB+LASTFIRE
2
0.00 0.66 1.00 9 531.91
2ERFA+ARCA+SHRBHT+TOTSHRTRE
2
+TOTHRB+LASTFIRE
2
2.07 0.24 0.36 10 531.85
3ERFA+ARCA+SHRBHT+TOTSHRBTRE
2
+TOTHRB+LASTFIRE 3.77 0.10 0.15 9 535.67
Detection probability (p) constant across models. Models are ranked from best to worst based on Akaike’s Information Criterion for small samples (AIC
c
), the
difference between the model’s AIC
c
and the highest-ranked model’s AIC
c
(ΔAIC
c
), and AIC
c
weights. AIC
c
is based on -2 x log
e
likelihood (logLik) and the number of
parameters in the model. Minimum AIC
c
= 552.55. Only models with ΔAIC
c
4 presented. Bold means 95% confidence interval of the beta estimate for covariate does
not include 0. ERFA = California buckwheat, ARCA = California sagebrush, TOTSHRTRE = total shrub/tree cover, TOTHRB = total herbaceous cover,
LASTFIRE = time since last fire, SHRBHT = shrub height. A superscript 2 in a covariate code represents the quadratic (squared) function of that covariate.
https://doi.org/10.1371/journal.pone.0306267.t004
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 18 / 35
Fig 11. California Gnatcatcher occupancy in relation to cover of California sagebrush, California buckwheat, total
shrub/tree (quadratic function), total herbaceous, time since last fire (quadratic function; plotted as year of last
fire), and shrub height at regional points in 2016. Gray lines represent 95% confidence interval. All covariates except
shrub height from model 1; shrub height from model 2; Table 4).
https://doi.org/10.1371/journal.pone.0306267.g011
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 19 / 35
Diego subregion and the region as a whole (Fig 18). Detectability (p) ranged from 0.65 at the
regional points to 0.69 at the Orange County subregional points.
In burned habitat, Gnatcatcher occupancy varied with time since fire (Fig 19). In 2015,
occupancy was highest in habitat burned the longest ago (2003–06), where gnatcatchers occu-
pied 24 percent of points. Occupancy of points burned in 2007–10 was lower, at 15 percent,
while just 2 percent of points last burned in 2011–14 were occupied. Detectability was high,
ranging from 0.62 at points burned in 2003–06 to 0.74 at points burned in 2007–10.
By 2016, occupancy had doubled in the two most recent burn categories, but was
unchanged in the oldest burn category. Occupancy of unburned habitat, at 43 percent,
approached twice that in the older burned habitat. Detectability in 2016 was estimated at 0.76,
0.67, and 0.38 for points burned in 2003–06, 2007–10, and 2011–14, respectively, and at 0.73
for unburned points.
Discussion
This study represents the first effort to characterize habitat condition and gnatcatcher occu-
pancy throughout the entire species’ range in southern California. Our investigation docu-
mented variability in structure and composition of coastal sage scrub vegetation that
influenced occupancy by gnatcatchers. Our findings provide a rangewide perspective that will
help guide gnatcatcher management and recovery at multiple spatial scales.
Rangewide, California sagebrush dominated the woody vegetation at our plots, with Cali-
fornia buckwheat, black and purple sage, and laurel sumac making up most of the remaining
shrub cover. However, most of the cover at plots was made up of non-native grasses and other
herbaceous vegetation, reflecting a history of disturbance associated with fire and possibly
other factors such as grazing and pollution [4,11,12,46,47]. Grass and herbaceous cover at
regional plots was over 40 percent higher than at unburned plots, suggesting that recent fires
are a primary driver of habitat disturbance in our study area. We found vegetation differences
at the subregional scale indicative of differences in recent fire history: in Orange County,
Table 5. Logistic regression models of California Gnatcatcher occupancy in 2016 as a function of habitat covariates in the San Diego subregion.
Model ΔQAIC
c
AIC
c
Weight logLik Number Parameters QDeviance
1 ARCA+LASTFIRE 0.00 0.21 1.00 4 118.38
2 ARCA+LASTFIRE
2
0.96 0.13 0.62 5 117.23
3 LASTFIRE
2
1.25 0.11 0.54 4 119.63
4 ARCA+SUNFL+LASTFIRE 1.29 0.11 0.53 5 117.55
5 ARCA+ERFA+LASTFIRE 1.47 0.10 0.48 5 117.74
6 ARCA+LASTFIRE+ELEV 1.92 0.08 0.38 5 118.18
7 ARCA+MALA+LASTFIRE 2.02 0.08 0.36 5 118.29
8 ARCA+LASTFIRE+DISTCOAST 2.12 0.07 0.35 5 118.38
9 ARCA+ERFA+SUNFL+LASTFIRE 3.27 0.04 0.20 6 117.39
10 ARCA+ERFA+LASTFIRE+DISTCOAST 3.49 0.04 0.18 6 117.61
11 ARCA+MALA+TOTSHRTRE+LASTFIRE 3.80 0.03 0.15 6 117.92
Detection probability (p) constant across models. Variance inflation factor (c^
c) = 2.3. Models are ranked from best to worst based on Akaike’s Information Criterion for
small samples (QAIC
c
), the difference between the model’s QAIC
c
and the highest-ranked model’s QAIC
c
(ΔQAIC
c
), and AIC
c
weights. QAIC
c
is based on -2 x log
e
likelihood (logLik) and the number of parameters in the model. Minimum QAIC
c
= 126.61. Only models with ΔQAIC
c
4 presented. Bold means 95% confidence
interval of the beta estimate for covariate does not include 0. ARCA = California sagebrush, LASTFIRE = time since last fire, MALA = laurel sumac,
SUNFL = sunflowers, ERFA = California buckwheat, DISTCOAST = distance to coast, ELEV = elevation, TOTSHRTRE = total shrub/tree cover. A superscript 2 in a
covariate code represents the quadratic (squared) function of that covariate.
https://doi.org/10.1371/journal.pone.0306267.t005
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 20 / 35
Fig 12. California Gnatcatcher occupancy in relation to cover of California sagebrush, time since last fire (linear
and quadratic functions; plotted as year of last fire), sunflowers, California buckwheat, and laurel sumac at San
Diego subregional points in 2016. Gray lines represent 95% confidence interval. California sagebrush from model 1,
time since last fire from model 2, sunflowers from model 4, California buckwheat from model 5, and laurel sumac from
model 7; Table 5).
https://doi.org/10.1371/journal.pone.0306267.g012
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 21 / 35
where less than 4 thousand ha encompassing 19 percent of suitable (HSI 0.50) gnatcatcher
habitat had burned within 13 years of this study, overall shrub/tree cover was higher, and her-
baceous cover lower, than in San Diego, where three large fires burned 35,585 ha of suitable
habitat (56%) during the same period. Grass and herbaceous cover in the San Diego subregion
was twice as high as in unburned habitat, reflecting the extent of the impact of fire on gnat-
catcher habitat in this part of its range.
Fire altered the landscape in ways that were associated with time since burn. Within the
first 5 years postfire, sites were dominated by grass and herbaceous vegetation and supported
Fig 13. California Gnatcatcher occupancy in relation to elevation at San Diego subregional points in 2016. Gray
lines represent 95% confidence interval. From model 6; Table 5).
https://doi.org/10.1371/journal.pone.0306267.g013
Table 6. Logistic regression models of California Gnatcatcher occupancy in 2016 as a function of habitat covariates in the Orange County subregion.
Model ΔAIC
c
AIC
c
Weight logLik Number Parameters Deviance
1ARCA+ERFA+SUNFL+BARE+ELEV 0.00 0.44 1.00 7 339.58
2ARCA+ERFA+SUNFL+ELEV 1.70 0.19 0.43 6 343.45
3ARCA+ERFA+BARE+ELEV 2.19 0.15 0.33 6 343.94
4ARCA+ERFA+SUNFL+TOTHRB+ELEV 3.08 0.09 0.21 7 342.66
5ARCA+ERFA+SUNFL+MALA+TOTHRB+ELEV 3.60 0.07 0.17 8 340.99
6ARCA+ERFA+SUNFL+ELEV+LASTFIRE 3.83 0.06 0.15 7 343.41
Detection probability (p) constant across models. Models are ranked from best to worst based on Akaike’s Information Criterion for small samples (AIC
c
), the
difference between the model’s AIC
c
and the highest-ranked model’s AIC
c
(ΔAIC
c
), and AIC
c
weights. AIC
c
is based on -2 x log
e
likelihood (logLik) and the number of
parameters in the model. Minimum AIC
c
= 354.24. Only models with ΔAIC
c
4 presented. Bold means 95% confidence interval of the beta estimate for covariate does
not include 0. ARCA = California sagebrush, ERFA = California buckwheat, SUNFL = sunflowers, BARE = bare ground, ELEV = elevation, TOTHRB = total
herbaceous, MALA = laurel sumac, LASTFIRE = time since last fire.
https://doi.org/10.1371/journal.pone.0306267.t006
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 22 / 35
little shrub cover. Once established, grass cover increased over time, and even at points 13
years postfire was still twice as high as that in unburned habitat. Recovery of woody vegetation
was underway by 9 years postfire, but it was not until 13 years postfire that total shrub/tree
cover was comparable to that in unburned habitat, although species composition differed
Fig 14. California Gnatcatcher occupancy in relation to cover of California sagebrush, California buckwheat,
sunflowers, and bare ground, and to elevation, at Orange County subregional points in 2016. Gray lines represent
95% confidence interval. From model 1; Table 6).
https://doi.org/10.1371/journal.pone.0306267.g014
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 23 / 35
between burned and unburned sites. Among species, California buckwheat increased rapidly,
and by 13 years postfire, cover of buckwheat exceeded that in unburned habitat. California
sagebrush cover increased initially, but changed little between 9 and 13 years postfire, at which
time it was still only 60 percent of that in unburned habitat. Laurel sumac and black sage, char-
acterized as vigorous resprouters after fire [48], also increased rapidly. Laurel sumac cover
exceeded that in unburned habitat within 5 years postfire, and was over twice as high by 9
years postfire. Black sage increased more slowly, but by 13 years postfire, was comparable in
Table 7. Logistic regression models of California Gnatcatcher occupancy in 2016 as a function of habitat covariates in postfire habitat.
Model ΔAICc AICc Weight logLik Number Parameters Deviance
1ARCA+MALA+SAGE+SUNFL 0.00 0.29 1.00 9 719.40
2ARCA+MALA+ERFA+SUNFL+GRASS 0.34 0.25 0.85 10 717.64
3ARCA+MALA+ERFA+SUNFL+OTHHRB 0.57 0.22 0.75 10 717.87
4ARCA+MALA+SAGE+SUNFL+GRASS+OTHHRB 1.62 0.13 0.45 11 716.81
5ARCA+MALA+ERFA+SUNFL 2.82 0.07 0.24 9 722.23
Detection probability (p) constant across models. Models are ranked from best to worst based on Akaike’s Information Criterion for small samples (AIC
c
), the
difference between the model’s AIC
c
and the highest-ranked model’s AIC
c
(ΔAIC
c
), and AIC
c
weights. AIC
c
is based on -2 x log
e
likelihood (logLik) and the number of
parameters in the model. Minimum AIC
c
= 737.83. Only models with ΔAIC
c
4 presented. Bold means 95% confidence interval of the beta estimate for covariate does
not include 0. ARCA = California sagebrush, MALA = laurel sumac, SAGE = sage (= black sage in this dataset), SUNFL = sunflowers, ERFA = California buckwheat,
GRASS = non-native grass, OTHHRB = other herbaceous.
https://doi.org/10.1371/journal.pone.0306267.t007
Table 8. Comparison of covariates in best supported models (ΔAIC
c
or ΔQAIC
c
2) predicting California Gnatcatcher occupancy in 2016 in the regional, San
Diego (SD) subregional, Orange County (OC) subregional, and postfire datasets. Effect of covariate: “+” = positive, “-”= negative, “+/-”= a threshold effect where the
initial effect is positive, “-/+” = a threshold effect where the initial effect is negative. Asterisk means 95% confidence interval of the beta estimate for covariate does not
include 0 in at least one model. Parentheses mean that ΔAIC
c
or ΔQAIC
c
>2; see Tables 4and 5.
Covariate Regional SD Subregion OC Subregion Postfire
California sagebrush +*++*+*
California buckwheat +*++*+*
Laurel sumac (-) -*
Sunflowers + +*+*
Star thistle
White sage
Dead
Sage +*
Total herbaceous -*
Other herbaceous -*
Grass -*
Total shrub/tree +/-*
Bare +
Shrub height (+)
Slope
Elevation +-*
Time since last fire +/-*+/-*
Effect of covariate: “+” = positive, “-”= negative, “+/-”= a threshold effect where the initial effect is positive, “-/+” = a threshold effect where the initial effect is negative.
Asterisk means 95% confidence interval of the beta estimate for covariate does not include 0 in at least one model. Parentheses mean that ΔAIC
c
or ΔQAIC
c
>2; see
Tables 4and 5.
https://doi.org/10.1371/journal.pone.0306267.t008
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 24 / 35
Fig 15. California Gnatcatcher occupancy in relation to cover of California sagebrush,laurel sumac, sage,
sunflowers, and California buckwheat at unburned points in 2016. Gray lines represent 95% confidence interval.
California sagebrush, laurel sumac, sage, and sunflowers from model 1; California buckwheat from model 2; Table 7).
https://doi.org/10.1371/journal.pone.0306267.g015
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 25 / 35
Fig 16. California Gnatcatcher occupancy in relation to cover of grass and other herbaceous at unburned points
in 2016. Gray lines represent 95% confidence interval. Grass from model 2; other herbaceous from model 3; Table 7).
https://doi.org/10.1371/journal.pone.0306267.g016
Fig 17. California Gnatcatcher occupancy in relation to cover of California sagebrushin four postfire categories
in 2016. Gray lines represent 95% confidence interval. From model 1; Table 7).
https://doi.org/10.1371/journal.pone.0306267.g017
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 26 / 35
cover to unburned habitat. These patterns of rapid postfire increase in non-native annual
grasses and slow increase in shrub cover are consistent with findings of previous studies and
create the conditions that can lead to vegetation type conversion [3,4,11,12,49].
Gnatcatcher occupancy varied across the region and was related to fire history. Occupancy
was highest in unburned habitat, reflecting the effect of fire at different spatial and temporal
scales. At the subregional scale, occupancy was higher in Orange County than in San Diego
County, mirroring differences in vegetation composition associated with different fire histo-
ries that affect habitat suitability for gnatcatchers. Not only have fires in San Diego County
burned more acreage of suitable habitat than in Orange County (see above), they have been
more recent, with the median time since last fire at San Diego County points equaling 2007
compared to that of 1993 at Orange County plots. LASTFIRE emerged as the primary predic-
tor of occupancy in the San Diego subregion, reflecting the effects of large and recent fires on
vegetation attributes that predicted occupancy in the Orange County and regional datasets,
such as woody cover of shrubs favored by gnatcatchers.
At the local scale, occupancy in 2015 was related to time since fire as predicted, with the
lowest occupancy in recently burned sites and the highest in habitat burned longest ago
(2003–2006). This pattern was observed again in 2016, with occupancy in unburned habitat
higher than that in any other postfire category. However, unexpectedly, occupancy in the two
intermediate fire categories (2003–06, 2007–10) differed little from one another in 2016. While
occupancy in the two most recently burned categories approximately doubled between 2015
and 2016, occupancy in the 2003–06 category was stable across the two years. This could be
evidence of a non-linear relationship between occupancy and time since burn, as suggested by
our analysis of LASTFIRE at the regional and subregional scales, or other factors limiting
occupancy of otherwise suitable habitat. Our habitat assessment revealed little difference in
vegetation structure and composition between the 2003–06 and 2007–10 postfire sites, suggest-
ing that something other than habitat condition is limiting occupancy, such as recolonization
potential. Winchell and Doherty [19] proposed that recolonization of burned areas proceeds
from the perimeter inward, and is dependent on proximity to a source of colonizers. Burned
Fig 18. Occupancy (±SE) of California Gnatcatchers at regional and subregional points in 2016. Occupancy
estimated as model-averaged occupancy across best supported models (ΔAIC
c
or ΔQAIC
c
2) for each dataset; see
Tables 4–6.
https://doi.org/10.1371/journal.pone.0306267.g018
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 27 / 35
areas close to occupied habitat are likely to be recolonized faster than those more distant from
a source of colonizers [50]. Differences between our two postfire categories in this regard war-
rant further investigation.
Previous studies have shown that gnatcatchers can recolonize burned sites within 2 to 5
years near the coast [50,51], and 5–10+ years inland, depending on fire intensity [52] and con-
ditions affecting vegetation recovery like precipitation, temperature, and aspect [53]. We
observed gnatcatchers occupying habitat near the coast that had burned two years previously.
While it is possible that recolonization can begin quickly as patches of suitable habitat emerge,
occupancy will be limited by available suitable habitat. We found that it took at least 9 years
postfire for woody vegetation to attain the structure and cover favored by gnatcatchers, and
our modelled relationships between occupancy and LASTFIRE at the regional and subregional
scales suggest it may take decades for burned coastal sage scrub to achieve occupancy levels
typical of unburned habitat. However, we caution that this conclusion is based on analysis of a
variable (LASTFIRE) obtained from a spatial database of fire perimeters (CalFire) that may be
incomplete, and thus this conclusion should be considered preliminary. Several investigators
have used CalFire data to examine the effects of wildfire on vegetation in southern California
[4,6,54], acknowledging the shortcomings of the database. In particular, small fires (<25 ha)
may be missed, and perimeters may not be accurate. These errors grow the further back in
time the records go, and many studies use 1919 as the cutoff for reliable data [55–59] or use a
more recent subset of the data [60]. We used records dating back to 1878 in order to assign a
value for LASTFIRE to every point in our regional and subregional datasets and thereby meet
our samples size requirements for detecting differences in occupancy. In our application,
missed fires, especially before 1919, would cause us to over-estimate the length of time since a
particular point burned, and thus over-estimate the modelled time required to achieve occu-
pancy comparable to that in unburned habitat. Although our regional and Orange County sub-
regional datasets may be subject to this potential bias, we consider our analysis of LASTFIRE
for San Diego County to be robust as only 1 value for LASTFIRE in that dataset preceded 1919,
and only 8 preceded 1950. We have high confidence in our postfire dataset where we analyzed
recovery of points that burned since 2003, and compared them to an “unburned” category of
points that last burned before 2002 where the potential bias created by missed fires would not
affect our results. We suggest that continued monitoring of the postfire points will yield valu-
able insights into the recovery process during the first few years and decades following fire.
Fig 19. Occupancy (±SE) of California Gnatcatchers as a function of time since burn in 2015 and 2016.Unburned
points were not surveyed in 2015. Occupancy in 2015 estimated from model with constant occupancy and detection
probabilities. Occupancy in 2016 estimated from model-averaged occupancy across best supported models (ΔAIC
c
2) for postfire dataset; see Table 7.
https://doi.org/10.1371/journal.pone.0306267.g019
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 28 / 35
Gnatcatchers occupied habitat non-randomly, and modelling occupancy using habitat
covariates clarified bird-habitat relationships. Cover of California sagebrush was the strongest
predictor of gnatcatcher occupancy, and appeared in every dataset. Previous studies have doc-
umented the association of gnatcatchers with sagebrush-dominated sage scrub [7,61], and
recent modelling of gnatcatcher occupancy in unburned habitat in San Diego County [20]
identified California sagebrush as a positive predictor. In our study, gnatcatcher occupancy at
regional and subregional scales increased with increasing cover of California sagebrush up to
about 40–60 percent, similar to the relationship documented by Winchell and Doherty [20]
for occupied habitat.
California buckwheat was another strong predictor of gnatcatcher occupancy in all of our
datasets. Typically co-occurring with California sagebrush in stands of coastal sage scrub, par-
ticularly inland stands [62], California buckwheat has not previously been identified as a pre-
dictor of occupancy, although gnatcatcher association with buckwheat co-dominated habitats
has been reported [7,63–65]. We found that gnatcatcher occupancy increased with cover of
California buckwheat up to about 50–60 percent at regional and Orange County subregional
points. Buckwheat had less influence on occupancy within the San Diego subregion as a
whole, but was a very strong predictor of occupancy in postfire habitat, where cover of buck-
wheat sometimes reached 75 percent. This may explain why California buckwheat did not
emerge as a predictor of occupancy in a previous investigation [20] since only unburned sites
were included in that analysis.
Sunflowers were a positive predictor of occupancy in the Orange County subregion and
postfire sites, with some support in the San Diego subregion. Weaver [65] found high gnat-
catcher abundance and density in coastal sites dominated by bush sunflower (Encelia califor-
nica) and California sagebrush. The inland counterpart, brittlebush (E.farinosa) co-occurs
with California sagebrush and California buckwheat, and is a rapid colonizer of disturbed sites
[62]. Gnatcatcher occupancy in the Orange County subregion increased with sunflower cover
up to about 20 percent, the highest cover we recorded for this region. At postfire sites, sun-
flower cover sometimes reached 60 percent, the upper limit on its representation in coastal
sage scrub plant communities [62].
Laurel sumac was found to be a significant negative predictor of gnatcatcher occupancy in
the postfire dataset, and received some support for an influence in the San Diego subregion.
Previously hypothesized to be a positive predictor of occupancy at the landscape scale because
its thermal tolerance coincides with that of the gnatcatcher [66], laurel sumac was found
instead to be a negative predictor in San Diego County [20]. Laurel sumac resprouts readily
after fire, and at our plots co-dominated the recovering woody vegetation along with Califor-
nia sagebrush and California buckwheat. However, relative cover of laurel sumac in postfire
sites, at 30 percent of combined cover of sumac, sagebrush and buckwheat, exceeded that in
unburned habitat, where sumac makes up less than 20 percent of the combined cover. While
providing shrub cover initially, rapid spread of laurel sumac may inhibit the recovery of species
more favored by gnatcatchers, and reduce habitat suitability long-term. Our results showed
declining occupancy as cover of laurel sumac increased to about 50 percent, above which the
probability of occupancy was minimal.
Black sage-dominated habitat in coastal areas has been reported as supporting low abun-
dance of gnatcatchers, although use of such habitats at inland sites may be more extensive [61,
65]. Winchell and Doherty [20] tested for a negative effect of black sage on gnatcatcher occu-
pancy, hypothesizing that the closed canopy of this plant community might limit foraging hab-
itat, but found no relationship. SAGE did not emerge as a predictor in our study at regional or
subregional scales; however, it was a strong positive predictor of gnatcatcher occupancy in
postfire sites. Like laurel sumac, black sage resprouts quickly after fire, providing a shrub
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 29 / 35
component to the recovering vegetation. However, unlike laurel sumac, at our plots sage cover
remained comparable to that in unburned habitat even 13 years postfire, and did not appear to
be excluding other shrubs favored by gnatcatchers.
Although we found relationships between gnatcatcher occupancy and cover of individual
plant species, coastal sage scrub typically includes a mix of shrub and tree species, each con-
tributing to overall habitat suitability. Regionally, occupancy was predicted by total shrub/tree
cover in a non-linear relationship. Occupancy increased with total shrub/tree cover up to
about 40 percent, but declined with cover above that. Winchell and Doherty [20] proposed 40
percent cover of California sagebrush as a target for habitat restoration, which in combination
with other species would exceed the threshold suggested by our results. Because the two studies
differed in geographic location and extent, timeframe, whether burned habitats were sampled
(this study) or not [20], and modelling approach, it is not possible to determine whether our
apparently different findings are the result of methodological or biological differences. Estab-
lishing targets for habitat restoration is an important application of the results of studies
describing gnatcatcher-habitat associations, and a topic for further refinement.
Herbaceous vegetation, which dominated all our study plots, negatively influenced gnat-
catcher occupancy regionwide. The primary component of herbaceous cover was non-native
grass, made more explicit in analysis of the postfire dataset. Postfire invasion of grass may
inhibit seedling establishment of shrubs preferred by gnatcatchers [67], changing the structure
and composition of coastal sage scrub and in extreme cases leading to type conversion to grass-
lands [3,4,11,68]. At our postfire plots, grass and other herbaceous vegetation not only estab-
lished quickly after fires, but increased with time, making it unlikely that it will subside
naturally.
We found little support for effects of physical variables on occupancy outside of a negative
effect of elevation in the Orange County subregion. There, occupancy declined with increasing
elevation over a range of 8–450 m, consistent with findings by Atwood and Bolsinger [69],
who found that all of 68 gnatcatcher occurrences in coastal Orange County between 1960 and
1990 were below 500 m in elevation. Winchell and Doherty [13] found a decline in occupancy
over a range of 0–1000 m in San Diego County; in contrast, our results produced limited sup-
port for a slight positive effect over a narrower range of elevation in the San Diego subregion.
Coastal sage scrub vegetation composition varies along gradients of distance to coast, eleva-
tion, and slope [48,62,70], and it is likely that in our analyses the effects of vegetation covari-
ates outweighed what might be considered indirect effects of these physical variables on
gnatcatcher occupancy.
Comparable data for evaluating trends in gnatcatcher occupancy in our regional study area
are not available. The most recent information on occupancy was reported by Winchell and
Doherty [13], who estimated occupancy for unburned habitat in San Diego County in 2009 at
approximately 41–44 percent for high and very high quality habitat, respectively. Habitat in
these quality classes best corresponds to our unburned category, for which we estimated occu-
pancy in 2016 at 43 percent. Future surveys using our standardized geographic scope and pro-
tocol will facilitate a more rigorous evaluation of population trends.
Collectively, our rangewide results reveal a widespread and long-term impact of wildfire on
California Gnatcatcher habitat, particularly in San Diego County. We documented habitat
associations that allow a determination of specific ways that fire affects habitat suitability.
These data provide a baseline from which future monitoring can evaluate changes in habitat
condition over time, and improve our understanding of the factors influencing gnatcatcher
occupancy. This in turn will guide development of management practices, particularly pre-
and postfire management, to ensure the long-term persistence of California Gnatcatchers in
southern California.
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 30 / 35
Acknowledgments
We are grateful for the large crew of dedicated field biologists too numerous to list who per-
formed gnatcatcher surveys and spent long grueling hours sampling vegetation. Emily Perkins
created the maps presented in this report. Susan Wynn, Will Miller, and Hans Sin were instru-
mental in devising the project and facilitating funding. The manuscript was improved with
helpful comments by David Wiens, Philip Gould, Clark Winchell, and an anonymous
reviewer. We thank the following partners and cooperators who made this project possible: 4-
S Ranch Homeowners Association, Angeles National Forest, Audubon California Starr Ranch,
California Department of Fish and Wildlife, California Department of Parks and Recreation,
California Department of Transportation, Center for Natural Lands Management, City of
Anaheim, City of Carlsbad, City of Chino Hills, City of Chula Vista, City of Escondido, City of
Irvine, City of Laguna Niguel, City of Oceanside, City of Poway, City of San Diego—Parks
Department, City of San Diego—Public Utilities Department, City of Thousand Oaks, City of
Vista, City of Whittier, Conejo Open Space Conservation Agency, County of Los Angeles,
County of Orange, County of San Diego, Endangered Habitats Conservancy, Fallbrook Land
Conservancy, Fallbrook Public Utility District, Fond Land Preservation Foundation, Helix
Water District, H.G. Fenton Company, Irvine Ranch Water District, Lakeview at Highlands
Ranch Homeowners Association, Manzanita Partners LLC, Matt Witman, Mountains Recrea-
tion & Conservation Authority, Native Habitat Preservation Authority, Natural Communities
Coalition, Olivenhain Municipal Water District, Orange County Transportation Authority,
Otay Ranch Preserve, Otay Water District, City of Palos Verdes Preserve, Ramona Municipal
Water District, Rancho Mission Viejo, Rancho Santa Fe Lakes Homeowners Association, Ran-
cho Simi Recreation and Park District, Riverside County Habitat Conservation Agency, River-
side County Parks, San Bernardino Water Conservation District, San Dieguito River Park,
Sanitation Districts of Los Angeles County, Santa Monica Mountains Conservancy, Sweet-
water Authority, The Environmental Trust, The Nature Conservancy, U.C. San Diego, U.S.
Bureau of Land Management, U.S. Department of Defense, U.S. National Park Service, U.S.
Forest Service, U.S. Fish and Wildlife Service, U.S. Geological Survey, Vallecitos Water Dis-
trict, Western Riverside County Regional Conservation Authority, and Western Riverside
County Multi-Species Reserve.
Any use of trade, firm, or product names is for descriptive purposes only and does not
imply endorsement by the U.S. Government.
Author Contributions
Conceptualization: Barbara E. Kus, Kristine L. Preston.
Data curation: Barbara E. Kus, Alexandra Houston.
Formal analysis: Barbara E. Kus, Kristine L. Preston, Alexandra Houston.
Funding acquisition: Barbara E. Kus.
Investigation: Barbara E. Kus, Alexandra Houston.
Methodology: Barbara E. Kus, Kristine L. Preston, Alexandra Houston.
Project administration: Barbara E. Kus.
Supervision: Barbara E. Kus.
Writing – original draft: Barbara E. Kus.
Writing – review & editing: Kristine L. Preston, Alexandra Houston.
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 31 / 35
References
1. Kelly LT, Giljohann KM, Duane A, Aquilue
´N, Archibald S, Batllori E, et al. Fire and biodiversity in the
Anthropocene. Science. 2020; 370(6519), eabb0355. https://doi.org/10.1126/science.abb0355 PMID:
33214246
2. Myers N, Mittermeier RA, Mittermeier CG, da Fonseca GAB, Kent J. Biodiversity hotspots for conserva-
tion priorities. Nature. 2000; 403:853–858. https://doi.org/10.1038/35002501 PMID: 10706275
3. Minnich RA, Dezzani RJ. Historical decline of coastal sage scrub in the Riverside-Perris Plain, Califor-
nia. West Birds. 1998; 29:366–391.
4. Talluto MV, Suding KN. Historical change in coastal sage scrub in southern California USA in relation to
fire frequency and air pollution. Landsc Ecol. 2008; 23:803–815.
5. Dewees SL, D’Antonio CM, Molinari N. Determining potential drivers of vegetation change in a Mediter-
ranean environment. Ecosphere 2022; 13:e4313. https://doi.org/10.1002/ecs2.4313
6. Syphard AD, Brennan TJ, Rustigian-Romsos H, Keeley JE. Fire-driven vegetation type conversion in
southern California. Ecol Appl. 2022 Sep; 32(6):e2626. https://doi.org/10.1002/eap.2626 PMID:
35397185
7. Atwood JL, Bontrager DR. California Gnatcatcher (Polioptila californica). In: Poole AF, Gill FB, editors.
Birds of the world. Ithaca: Cornell Lab of Ornithology; 2020. https://doi.org/10.2173/bow.calgna.01
8. Keeley JE, Fotheringham CJ. Impact of past, present, and future fire regimes on North American
Mediterranean shrublands. In: Veblen TT, Baker WL, Montenegro G, Swetnam TW, editors. Fire and
climatic change in temperate ecosystems of the western Americas. New York: Springer; 2003. pp.
218–261.
9. Keeley JE. Fire as a threat to biodiversity in fire-type shrublands. In: Kus BE, Beyers JL, editors. Plan-
ning for Biodiversity: Bringing research and management together. Gen. Tech. Rep. PSW-GTR-195.
Albany: Pacific Southwest Research Station, Forest Service, U.S. Department of Agriculture; 2005. pp.
97–106. https://doi.org/10.2737/PSW-GTR-195
10. Syphard AD, Radeloff VC, Keeley JE, Hawbaker TJ, Clayton MK, Stewart SI, et al. Human influence on
California fire regimes. Ecol Appl. 2007; 17:1388–1402. https://doi.org/10.1890/06-1128.1 PMID:
17708216
11. Fleming GM, Diffendorfer JE, Zedler PH. The relative importance of disturbance and exotic-plant abun-
dance in California coastal sage scrub. Ecol Appl. 2009; 19:2210–2227. https://doi.org/10.1890/07-
1959.1 PMID: 20014589
12. Keeley JE, Franklin J, D’Antonio C. Fire and invasive plants on California landscapes. In: McKenzie
D, Miller C, Falk DA, editors. The landscape ecology of fire. New York: Springer; 2011. p. 193–
221.
13. Winchell CS, Doherty PF. Effects of habitat quality and wildfire on occupancy dynamics of Coastal Cali-
fornia Gnatcatcher (Polioptila californica californica). Condor. 2014; 116(4)538–545.
14. D’Antonio CM, Vitousek PM. Biological invasions by exotic grasses, the grass/fire cycle, and global
change. Annu Rev Ecol Syst. 1992; 23:63–87.
15. City of San Diego. Final Multiple Species Conservation Program: MSCP Plan. San Diego: City of San
Diego; 1998. FinalMSCPProgramPlan.pdf (sandiegocounty.gov).
16. AMEC Earth and Environmental (AMEC), Conservation Biology Institute (CBI), Onaka Planning and
Economics, and the Rick Alexander Company. Final MHCP Plan, V. 1; 2003. Prepared for the Multiple
Habitat Conservation Program.
17. RJ Meade Consulting, Inc. Central and coastal subregion natural community conservation plan/habitat
conservation plan; Parts I & II: NCCPO/HCP; Final (Administrative Record Copy). Prepared for County
of Orange Environmental Management Agency and United States Fish and Wildlife Service/California
Department of Fish and Game; 1996.
18. California Department of Fish and Game. Findings of fact under the California Environmental Quality
Act and the Natural Community Conservation Planning Act and Natural Community Conservation Plan
permit for the Western Riverside Multiple Species Habitat Conservation Plan. Sacramento: California
Department of Fish and Game; 2004.
19. Winchell CS, Doherty PF. Using California Gnatcatcher to test underlying models in habitat conserva-
tion plans. J Wildl Manage. 2008; 72(6):1322–1327.
20. Winchell CS, Doherty PF. Restoring habitat for coastal California Gnatcatchers (Polioptila californica
californica). Condor. 2018; 120(3):581–595.
21. Preston KL, Perkins EE, Kus BE. Coastal California gnatcatcher habitat suitability model for southern
California (2015). U.S. Geological Survey data release; 2020. Available from: https://doi.org/10.5066/
P9SJRU51
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 32 / 35
22. Technology Associates International Corporation (TAIC). California Gnatcatcher habitat evaluation
model for USFWS. Digital data. U.S. Fish and Wildlife Service, Carlsbad Fish and Wildlife Office, Carls-
bad (CA); 2002.
23. Preston KL, Rotenberry JT, Redak RA, Allen MF. Habitat shifts of endangered species under altered cli-
mate conditions: importance of biotic interactions. Glob Chang Biol. 2008; 14(11):2501–2515.
24. Knick ST, Rotenberry JT. Limitations to mapping habitat use areas in changing landscapes using the
Mahalanobis distance statistic. J Agric Biol Environ Stat. 1998; 3(3):311–322.
25. Rotenberry JT, Knick ST, Dunn JE. A minimalist approach to mapping species’ habitat: Pearson’s
planes of closest fit. In: Scott JM, Heglund PJ, Morrison ML, Haufler JB, Raphael MG, Wall WA, et al.,
editors. Predicting species occurrences: Issues of accuracy and scale. Washington D.C.: Island Press;
2002. p. 281–289.
26. Rotenberry JT, Preston K., Knick ST. GIS-based niche modeling for mapping species’ habitat. Ecology.
2006; 87(6):1458–1464. https://doi.org/10.1890/0012-9658(2006)87[1458:gnmfms]2.0.co;2 PMID:
16869421
27. Vandergast AG, Kus BE, Preston KL, Barr KR. Distinguishing recent dispersal from historical genetic
connectivity in the coastal California gnatcatcher. Sci Rep. 2019; 9:1355, https://doi.org/10.1038/
s41598-018-37712-2 PMID: 30718575
28. Knick ST, Hanser SE, Preston KL. Modeling ecological minimum requirements for distribution of greater
sage-grouse leks: implications for population connectivity across their western range. Ecol Evol. 2013;
3(6):1539–1551.
29. SanGIS Regional Data Warehouse. Transportation: ROADS_FREEWAYS. San Diego Geographic
Information Source—JPA. https://www.sangis.org/.
30. San Diego Association of Governments (SANDAG) and SanGIS Regional Data Warehouse. Land Use:
LANDUSE_2014. San Diego Geographic Information Source—JPA. https://www.sangis.org/.
31. San Diego Association of Governments (SANDAG) and SanGIS Regional Data Warehouse. Ecology:
Conserved_Lands. San Diego Geographic Information Source—JPA. https://www.sangis.org/.
32. GreenInfo Network. California Protected Areas Database (CPAD). https://www.calands.org/get-data/.
33. CalFire (Department of Forestry and Fire Protection). Fire and Resource Assessment Program
(FRAP): Fire perimeters “fire15_1”. https://www.fire.ca.gov/what-we-do/fire-resource-assessment-
program/fire-perimeters.
34. White GC, Burnham KP. Program MARK: Survival estimation from populations of marked animals. Bird
Study. 1999; 46(suppl 7):120–139.
35. Miller WB, Winchell CS. A comparison of point-count and area-search surveys for monitoring site occu-
pancy of the coastal California Gnatcatcher (Polioptila californica californica). Condor. 2016; 118
(2):329–337.
36. Preston K, Perkins E, Brown C, McCutcheon S, Bernabe A, Luciani E, et al. State of the Regional Pre-
serve System in Western San Diego County. USGS Cooperator report prepared for San Diego Associa-
tion of Governments Environmental Mitigation Program; 2022. Available from: state-of-preserve-report-
2022-12-08.pdf (sandag.org).
37. Theobold DM, Stevens DL, White D, Urquhart NS, Olsen AR, Norman JB. Using GIS to generate spa-
tially balanced random survey designs for natural resource applications. Environ Manage. 2007; 40
(1):134–146. https://doi.org/10.1007/s00267-005-0199-x PMID: 17546523
38. Stevens DL Jr, Olsen AR. Spatially balanced sampling of natural resources. J Am Stat Assoc. 2004; 99
(465):262–278.
39. MacKenzie DI, Nichols JD, Royle JA, Pollock KH, Bailey LA, Hines JE. Occupancy estimation and
modeling: inferring patterns and dynamics of species occurrence. San Diego: Elsevier; 2006. 324
p.
40. Preston KL, Grishaver MA, Mock PJ. California Gnatcatcher vocalization behavior. West Birds. 1998;
29(4):258–268.
41. Burnham KP, Anderson DR. Model selection and multimodel inference: A practical information-theoretic
approach. 2nd ed. New York: Springer; 2002.
42. MacKenzie DI, Bailey LL. Assessing the fit of site-occupancy models. J Agric Biol Environ Stat. 2004;
9:300–318. Available from https://doi.org/10.1198/108571104X3361
43. Fiske I, Chandler R. Unmarked: An R package for fitting hierarchical models of wildlife occurrence and
abundance. J Stat Softw. 2011; 43(10):1–23. Available from https://doi.org/10.18637/jss.v043.i10
44. Arnold TW. Uninformative parameters and model selection using Akaike’s Information Criterion. J Wildl
Manage. 2010; 74(6):1175–1178.
PLOS ONE
Rangewide occupancy and postfire recovery of California Gnatcatchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0306267 July 5, 2024 33 / 35
45. Kus BE, Houston A, Rangewide occupancy and post-fire recovery of California gnatcatchers in south-
ern California (ver 2.0, March 2023). U.S. Geological Survey data release; 2021. Available from: https://
doi.org/10.5066/F7PC30JX
46. O’Leary JF, Westman WE. Regional disturbance effects on herb succession patterns in coastal sage
scrub. J Biogeogr. 1988; 15(5):775–786.
47. Allen EB, Sirulnik AG, Egerton-Warburton L, Kee SN, Bytnerowicz A, Padgett PE, et al. Air pollution
and vegetation change in southern California coastal sage scrub: a comparison with chaparral and
coniferous forest. In: Kus BE, Beyers JL, editors. Planning for Biodiversity: Bringing research and man-
agement together. Gen. Tech. Rep. PSW-GTR-195. Albany: Pacific Southwest Research Station, For-
est Service, U.S. Department of Agriculture; 2005. p. 79–95.
48. Kirkpatrick JB, Hutchinson CF. The environmental relationships of Californian coastal sage scrub and
some of its component communities and species. J Biogeogr. 1980; 7:23–38.
49. Keeley JE, Baer-Keeley M, and Fotheringham CJ. Alien plant dynamics following fire in Mediterranean-
climate California shrublands. Ecol Appl. 2005; 15:2109–2125.
50. Bontrager DR, Erickson RA, Hamilton RA. Impacts of the October 1993 Laguna Canyon Fire on Califor-
nia gnatcatchers and cactus wrens. In: Keeley JE, Scott T, editors. Brushfires in California wildlands:
Ecology and resource management. Fairfield (WA): International Association of Wildland Fire; 1995.
p. 69–76.
51. Wirtz WO II, Mayer AL, Raney MM, Beyers JL. Effects of fire on the ecology of the California Gnat-
catcher, Polioptila californica, in California sage scrub communities. In: Greenlee JM, editor. Proceed-
ings—First conference on fire effects on rare and endangered species and habitats. Fairfield (WA):
International Association of Wildland Fire; 1997. p. 81–89.
52. O’Leary JF. Californian coastal sage scrub: general characteristics and considerations for biological
conservation. In: Schoenherr AA, editor. Endangered plant communities of southern California. Clare-
mont (CA): Southern California Botanists Special Publication no. 3; 1990. p. 24–41.
53. Beyers JL, Wirtz WO II. Vegetative characteristics of coastal sage scrub sites used by California gnat-
catchers: implications for management in a fire-prone ecosystem. In: Greenlee JM, editor. Proceedings
—First conference on fire effects on rare and endangered species and habitats. Fairfield (WA): Interna-
tional Association of Wildland Fire; 1997. p. 81–89.
54. Mann ML, Batllori E, Moritz MA, Waller EK, Berck P, Flint AL, et al. Incorporating anthropogenic influ-
ences into fire probability models: Effects of human activity and climate change on fire activity in Califor-
nia. PLoS One. 2016 Apr 28; 11(4):e0153589. https://doi.org/10.1371/journal.pone.0153589 PMID:
27124597
55. Franklin J, Coulter CL, Rey SJ. Change over 70 years in a southern California chaparral community
related to fire history. JVeg Sci. 2004; 15(5):701–10.
56. Keeley JE, Zedler PH. Large, high-intensity fire events in southern California shrublands: debunking the
fine-grain age patch model. Ecol Appl. 2009; 19:69–94. https://doi.org/10.1890/08-0281.1 PMID:
19323174
57. Keeley JE, Syphard AD. Climate change and future fire regimes: examples from California. Geosci-
ences. 2016; 6(3):37.
58. Syphard AD, Keeley JE. Historical reconstructions of California wildfires vary by data source. Int J Wild-
land Fire. 2016; 25(12):1221–7.
59. Keeley JE, Syphard AD. Large California wildfires: 2020 fires in historical context. Fire Ecol. 2021; 17:1.
Available from https://doi.org/http%3A//dx.doi.org/10.1186/s42408-021-00110-7
60. Conlisk E, Syphard AD, Franklin J, Regan HM. Predicting the impact of fire on a vulnerable multi-spe-
cies community using a dynamic vegetation model. Ecol Modell. 2015; 301:27–39.
61. Atwood JL, Bontrager DR, Gorospe AL. Use of refugia by California Gnatcatchers displaced by habitat
loss. West Birds. 1998; 29(4):406–412.
62. Kirkpatrick JB, Hutchinson CF. The community composition of Californian coastal sage scrub. Vegeta-
tio 1977; 35:21–33.
63. Braden GT, McKernan RL, Powell SM. Association of within-territory vegetation characteristics and fit-
ness components of California Gnatcatchers. Auk. 1997; 114(4):601–609.
64. Grishaver MA, Mock PJ, Preston KL. Breeding behavior of the California Gnatcatcher in southwestern
San Diego County, California. West Birds. 1998; 29:299–322.
65. Weaver KL. Coastal sage scrub variations of San Diego County and their influence on the distribution of
the California Gnatcatcher. West Birds. 1998; 29:392–405.
66. Mock PJ. Energetic constraints