At-Sea Survey Methods

In 2020 and 2021, USGS-WERC completed nine at-sea boat-based surveys per year for marbled murrelets between Half Moon Bay and Santa Cruz, California (fig. 1). We performed surveys during the previously established survey window (June 1–August 24; Henry, 2017) and allocated surveys to two periods within this window: three surveys during June 1–July 9 and six surveys during July 10–August 24. We used data from all surveys to estimate marbled murrelet abundance, and we used the six surveys during the second survey period to estimate juvenile ratio (following Henry, 2017). Surveys were almost exclusively performed when viewing conditions were excellent to good (table 1).

In 2001–03, a pool of approximately 40 unique survey routes were designed by Peery and others (2007) as continuous, approximately 100-km long zig-zag transect lines to sample nearshore (200–1,350 meters [m] from coast) and offshore (1,350–2,500 m from coast) strata, with approximately four times greater effort within the nearshore stratum owing to greater known marbled murrelet densities nearshore (Henry, 2017, and references therein). Routes originally were drawn starting at a random distance (200–2,500 m) from shore, and an equal number of routes were drawn using starting points at the north and south ends of the survey area. Survey routes that are drawn from the south typically result in a greater amount of habitat surveyed within south-facing, leeward bays that often have greater relative abundances of marbled murrelets than more exposed stretches of the coast; annual survey replicates have typically included an equal number of routes drawn from north and south (Henry, 2017).

In 2020 and 2021, we randomly selected (without replacement) nine unique survey routes each year (five drawn from the south and four from the north) from the pool of available routes used by Henry (2017) and Peery and others (2006) during 2001–16 surveys. We performed all surveys by following the selected route from north (Pillar Point Harbor, Half Moon Bay) to south (Soquel Point, Monterey Bay) using a Global Positioning System (GPS). When the survey route intersected land or crossed hazardous areas (for example, high surf areas nearshore), we maintained survey effort while safely navigating as close as possible to the transect line. We performed surveys from a small boat, using line-transect methods (Becker and others, 1997; Peery and others, 2006; Henry, 2017). Two observers, standing on either side of a 6-m open skiff (Boston Whaler Guardian 20; research vessel Lucy M) traveling 22–28-km per hour, recorded the observation time, angle off the transect line, distance from the vessel, and group size of all marbled murrelets detected. The skiff was operated by a third crew member whose sole responsibility was piloting the vessel. Skiff size and travel speed were consistent with those used in surveys since 2007. From 1999 to 2003, surveys were performed in a 4-m open inflatable skiff at approximately 18-km per hour.

Observers counted marbled murrelets as a group when individuals were within 2 m of each other or if they showed behavior indicative of group status (for example, co-diving or vocalizing with one another; Strong and others, 1995). Observers recorded the age-class of each marbled murrelet based on three plumage classifications: (1) “after-hatch-year” (AHY), (2) “hatch-year” (HY), or (3) “unknown.” Behavior was recorded as “resting” on the water or “flying,” with flight direction noted. Distance and angle were estimated at the time of first detection, regardless of behavior. Before each survey, observers calibrated distance estimation using a laser rangefinder on buoys and other targets in the harbor. To facilitate estimations of sighting angles, we placed marks along the bow of the boat in 10-degree increments. The vessel occasionally paused or deviated from the transect line to assess marbled murrelet age-class; no additional observations were counted when deviating off the transect line. Observers also recorded observations of all non-murrelet marine birds and mammals encountered along transects, consistent with surveys performed since 2007. Marine mammal observations were recorded using line-transect methods as described for murrelets, and non-murrelet bird observations were counted in a 75-m aside strip transect (no distances or angles recorded). Non-murrelet observations were archived but are not presented or analyzed in this report.

Observers recorded all observations and observation times using digital voice recorders, including survey start and end times, ocean conditions (Beaufort Sea state), viewing conditions (table 1), and periods when effort was paused for any reason (for example, vessel deviated from the transect line to identify marbled murrelet age-class). Observers reviewed their own recordings and transcribed and tabulated their sighting data into a single spreadsheet that was examined for quality assurance and quality control and then merged into a combined spreadsheet. We acquired a continuous 1-second GPS track during each survey using a handheld GPS unit; this track was used to georeference observations based on matching date/time using custom scripting in R (R Core Team, 2016).

A comprehensive database of all marbled murrelet survey data (1999–present) is available in a U.S. Geological Survey data release (Felis and others, 2022) to allow scientists and managers to evaluate established survey methods and assess trends in abundance estimation and juvenile ratios.

Abundance Estimation Methods

We calculated perpendicular distance for each detection (sine of the sighting angle multiplied by observation distance, where the sighting angle is relative to the direction the boat is traveling). Consistent with all previous years (1999–present), we included sightings of flying birds in our analysis, despite the potential that flying birds might have a different probability of detection and including these could affect abundance estimates. Historical protocol (1999–2003) for this monitoring program indicated that flying birds should only be counted if they cross the beam of the vessel and the distance and angle to flying birds was only to be estimated when flying birds crossed the beam (90-degree angle and a distance estimate; Henry and Tyler, 2017). Re-examination of historical data (Felis and others, 2022) showed that this protocol was followed in earlier years (1999–2003) but that flying birds were given a distance/angle estimate for when they were first detected in later years (2007–16). We maintained the methodology of 2007–16 for estimating distance and angle of flying birds at the time of first sighting and counted them whether or not they crossed the beam of the vessel during 2017–present.

We created a spatial representation of strata in ArcGIS based on the same coastline shapefile used in 2007–present (California Department of Fish and Wildlife, 2004) and calculated linear effort for each survey within the nearshore and offshore strata consistent with previous years by using the hypothetical survey route delineated by the zigzag segment nodes (table 2). We assigned marbled murrelet observations to each stratum in ArcGIS based on spatial overlap. The calculation of stratum-specific linear effort and the assignment of observations to strata from 1999 to 2003 was done with an older coastline representation (Peery and others, 2006). We maintained the use of the modern stratum delineation to facilitate comparison of our abundance estimates to recent data (2007–present) but we advise caution when comparing recent estimates with older abundance estimates from 1999 to 2003.

Following Henry (2017), we used the program DISTANCE v7.1 (Thomas and others, 2010) to model our detection function and estimate marbled murrelet abundance using conventional distance sampling (see Buckland and others, 2015, for detection function modeling, model selection, and line transect abundance estimation methods). For each year, we pooled observations from all surveys to create an annual detection function and applied this function to each survey to calculate stratum- and survey-specific density estimates based on the linear effort sampled during each survey. Consistent with Henry (2017), we grouped perpendicular detection distances into 20-m bins, used a 120-m truncation distance, and evaluated the half-normal function, with or without cosine expansion (expansion orders automatically varied by DISTANCE); of these iterations, DISTANCE automatically selected the detection function model with the smallest Akaike information criterion (AIC; Burnham and Anderson, 2004) value (that is, most parsimonious fit). We used the mean of observed cluster size to estimate cluster (group) sizes at the sample (transect) level. We calculated survey-specific abundances by multiplying the stratum-specific density estimate by the total area of each stratum in the study area (104.65 square kilometers [km²] for each) and then summed the two stratum-specific abundance values for a total area abundance estimate. Variances of abundance estimates incorporate variances from the overall detection model, encounter rates across surveys, and cluster (group) size (Buckland and others, 2015).

We report annual abundances and 95-percent confidence interval (CI) estimates for 2017–21 (current program of study), and previously reported abundance estimates from 1999 to 2016 (as reported by Henry, 2017). Estimates from 1999 to 2016 were calculated similarly, with a few exceptions. Peery and others (2006) included covariates (observer, view conditions) in annual detection function model selection from 1999 to 2003 but this practice was not continued from 2007 to 2016 (Henkel and Peery, 2008; Peery and others, 2008; Peery and Henry, 2010; Henry and others, 2012; Henry and Tyler, 2017). Additionally, detection functions were modeled annually except for 2008, when at-sea murrelet observations and number of surveys were too few to model a year-specific detection function. As a result, Peery and others (2008) modeled a detection function using pooled observation data from 2007 to 2008 and applied it to 2008 data to estimate abundance. Peery and others (2006) used detection distance binning schemes other than 20-m when analyzing data from 1999 to 2003. Finally, the spatial boundaries of strata are different for 1999–2003 and 2007–21 analyses because different coastline representations were used, resulting in subtle changes in stratum-specific linear effort calculations and observation assignments.

Juvenile Ratio Methods

We estimated the juvenile ratio (the ratio of HY to AHY individuals) for marbled murrelet surveys completed during the fledging period. The previously established fledging period ranged from July 10, when an estimated 34 percent of HY birds are thought to have fledged, to August 24, about the time when HY and AHY murrelets become indistinguishable at sea because AHY birds begin pre-basic molt (Long and others, 2001; Peery and others, 2007). Thus, we included only surveys between July 10 and August 24 to estimate annual juvenile ratios (following Peery and others, 2007 and Henry, 2017). We followed techniques outlined by Long and others (2001) to identify HY birds. We also reviewed resources provided by the Alaska murrelet group (K. Nesvacil, Alaska Department of Fish and Game, written commun., 2017) and photographs of HY birds taken during past surveys before starting new surveys. We included only those birds confidently identified to age class to estimate the juvenile ratio. Raw counts are used for the ratio, as opposed to estimates of density or abundance from detection function modeling, with the assumption that HY and AHY murrelets have equivalent detection functions (Peery and others, 2007).

We adjusted HY and AHY counts to account for birds expected to have been inland during the time of the survey. A certain percentage of AHY birds are still incubating young during the fledging period and, therefore, are not on the water during at-sea surveys, potentially creating a positively biased juvenile ratio. The proportion, $p_{A_i}$, of AHY birds incubating in survey $i$, is reported to be less than 6 percent between July 10 and July 17 and less than 1 percent after July 17 (Peery and others, 2004, 2007) and estimated by linear regression model (Peery and others, 2007) shown in eq. 1:

Therefore, to correct for the number of AHY birds counted at sea between July 10 and July 17, we calculated, as the date-corrected number of AHY individuals, $A_i$, as

For surveys after July 17, we assumed no birds were incubating, and the observed number of AHY birds was not date-corrected.

In addition to adjusting for incubating adults (to avoid positive bias in the estimated ratio), the juvenile ratio calculation can be negatively biased by not accounting for HY birds that have not yet fledged by the time of the survey. Based on 47 observed fledging events in California, Peery and others (2007) estimated the daily percent of juveniles expected to have fledged during the study timeframe. Therefore, to adjust the number of HY birds observed during a given at-sea survey, we calculated the proportion, $p_{H_i}$, of fledged HY birds in survey $i$, estimated by linear regression model (Peery and others, 2007) as

To correct for the number of HY birds not counted at sea because they had not yet fledged, we calculated the date-corrected number of HY individuals, $H_i$, as

For each year, we used $A_i$ and $H_i$ to estimate the juvenile ratio, $\widehat{R}$, as

For each year, we estimated the variance of the date-corrected juvenile ratio, $v\widehat{a}r\left(\widehat{R}\right)$, following van Kempen and van Vliet (2000) and Peery and others (2007) as

We calculated annual juvenile ratios for 1999–2021 using only standardized zig-zag transects (using data compiled in Felis and others, 2022) and report those results herein. We calculated date-corrected juvenile ratios and associated variance using R (R Core Team, 2016).

Abundance Estimation Results

We detected 174 marbled murrelet groups consisting of 303 murrelets on all surveys combined in 2020 (range was 9–55 birds per survey; table 2). We detected 129 marbled murrelet groups consisting of 228 murrelets on all surveys combined in 2021 (range was 7–77 birds per survey; table 2). In 2020, murrelet sightings were concentrated from around Point Año Nuevo to Pescadero Creek; sightings were sporadic between Half Moon Bay and Pescadero Creek and rare from just south of Point Año Nuevo to Santa Cruz (fig. 1). Similar to 2020, murrelet sightings in 2021 were concentrated from Point Año Nuevo to Pescadero Creek and rare from south of Point Año Nuevo to Santa Cruz; however, birds were seen more frequently between Half Moon Bay and Pescadero Creek, and an aggregation of murrelets was seen near Soquel Point on a single survey (July 28, 2021; fig. 1).

Detections approached zero at 120-m horizontal sighting distance for both years (fig. 2); therefore, consistent with Henry (2017), we excluded from analysis observations that were greater than 120 m from the transect line. We included flying birds (4 and 13 percent of all detections in 2020 and 2021, respectively; following Henry, 2017). For all surveys combined in 2020, the half-normal detection model with a cosine adjustment (order 2) was the best-fitting model, and the observed number of sightings was not significantly different from the number predicted using this detection model (chi-squared [χ²]=2.00, degrees of freedom [df]=3, probability value [P]=0.57; fig. 2). The abundance estimated for the entire study area in 2020, using all surveys, was 470 birds (95-percent CI, 313–707 birds; percent coefficient of variation [CV], 19.0; table 3). For all surveys combined in 2021, the half-normal detection model with no adjustments was the best-fitting model, and the observed number of sightings was not significantly different from the number predicted using this detection model (chi-squared [χ²]=1.93, degrees of freedom [df]=4, probability value [P]=0.75; fig. 2). The abundance estimated for the entire study area in 2021, using all surveys, was 402 birds (95-percent CI, 219–737 birds; percent coefficient of variation [CV], 27.8; table 3). Estimated abundances for all study years (2001–21) are shown in table 3 and figure 3.

Juvenile Ratio Results

We detected three HY marbled murrelets in 2020: two on August 13, 2020, and one on August 15, 2020 (table 2; fig. 1). In 2020, the corrected juvenile ratio ($\widehat{R}$) was 0.018±0.011 standard error (SE). We detected four HY marbled murrelets in 2021: one on July 28, 2021, two on August 9, 2020, and one on August 23, 2020 (table 2; fig. 1). In 2021, the corrected juvenile ratio ($\widehat{R}$) was 0.041±0.024 SE. Estimated corrected juvenile ratios using standardized zig-zag surveys (1999–2021; calculated in this study) are shown in table 3 and figure 4.