Ammospiza nelsoni subvirgata (Acadian subspecies of Nelson’s Sparrow) breeds in saltmarshes from northern Massachusetts to New Brunswick and eastern Quebec. In the Canadian Maritimes, this subspecies also successfully breeds in diked agricultural lands (i.e., “dikeland”) that were originally created by Acadian settlers in the 1600s. Little is known about the reasons for or consequences of using dikeland for breeding. To fill this knowledge gap, we tracked male and female sparrows, and monitored nest fates in natural saltmarsh and human-made dikeland habitats. We collected fecal samples from adults and nestlings to examine which habitat type they were foraging in, and we also quantified vegetative cover. We hypothesized that flood risk in saltmarshes played an important role in the decision of A. n. subvirgata to nest in dikeland given that the saltmarsh is regularly inundated with tidal water. Based on nest monitoring, we estimated higher overall nest success in dikeland than saltmarshes. Fecal sample analysis showed distinct differences in diet between individuals using dikeland compared to saltmarshes. We also observed differences in vegetation. These results suggest that A. n. subvirgata are able to take advantage of a readily available human-made habitat for breeding. With rising sea levels and increased storm events threatening coastal habitats, it is important to understand if coastal-breeding birds can adapt to changes and what trade-offs exist for individuals who shift to alternative habitats.

Study site

We conducted our study near Beaubassin Research Station (hereafter “Beaubassin”) in Aulac, New Brunswick (45°50'51.08", -64°17'7.27") from 2021 to 2023, and in Dorchester, New Brunswick (45°53'32.01", -64°32'27.2") in 2023 only. Beaubassin is surrounded by dikeland, with natural and restored saltmarshes along the shore of the bay (Figure 1). The site in Dorchester was added to the study after we found a A. n. subvirgata nest there in 2023. The Dorchester site includes two large impoundments (human-made diked wetlands), directly adjacent to a large dike that separates a narrow saltmarsh along the Memramcook River. Beyond the impoundments are dikeland used for cattle pasture and hayfields. We did not conduct any radio-tracking or banding of adults at the Dorchester site, but monitored several nests at this site in 2023. The area we covered in Dorchester had a much narrower saltmarsh, measuring approximately 130 m wide, than the one near Beaubassin, where the saltmarsh extends up to 700 m from the dike. Although the dikeland or agricultural land area was over 7 km², we restricted our nest searching efforts to the narrow area (approximately 10% of the total) on or near the dike because these lands were publicly accessible.

Radio-telemetry home range estimation

A. n. subvirgata are difficult to monitor: they have soft songs, often sing and move within dense vegetation, and nest near the ground. We therefore used a combination of radio telemetry and color-banding to track female and male sparrows in 2021 and 2022. To capture sparrows, we set up multiple mist-nets in either dikeland or saltmarsh. In the saltmarsh, most nets were set inside ditches because A. n. subvirgata, when disturbed, would often fly into the ditches and could be coaxed into the net. In the dikeland, if a ditch was not present, we set multiple nets parallel to the dike or near shrubs or small trees where the net would be camouflaged. We had limited success with playback or passive netting and did not use these techniques beyond a few initial attempts.

Once captured, we tagged individuals with Lotek nano tags (Lotek NTQB2-2-M) registered with the Motus network (Taylor et al. 2017). In 2021, we used the clip-and-glue method (Raim 1978; Diemer et al. 2014), but switched to leg-loop harnesses (Rappole and Tipton 1991) in 2022 due to retention issues with the clip-and-glue method. The clip-and-glue method involved clipping some interscapular feathers, gluing a small piece of cotton fabric to the tag to improve adhesion, then gluing the tag to the clipped area on the individual’s back. The leg-loop harness method involved creating two adjacent loops out of elastic cord then gluing on the tag. Harnesses were sized to fit the individual, therefore, we made the harnesses in the field while holding the individual. All individuals successfully flew away after tagging and we confirmed that individuals continued their normal activities after relocating them using radio telemetry. The weight of the tags with the glue, cotton fabric, and cord was < 0.4 g, which is <5% of handled individuals’ weights (range: 14.5 – 22 g). With the leg-loop harness method, we had one mortality due to entanglement of the tag antenna in grass, a problem that has been reported with this method in the past, particularly for species that spend a lot of time on the ground among grass (Hill and Elphick 2011, Choi et al. 2021).

We also banded each individual with a unique combination of color bands to ease identification during resighting, and one metal band with a unique federally issued number. For each individual, we collected information on weight (g), wing length (mm), fat score (0-5), age (i.e., hatch-year or after hatch-year) and sex, determined by the presence of a cloacal protuberance or brood patch. In 2022, we also measured tarsus length (mm), tail length (mm), and collected a fecal sample from each individual. All birds were handled and banded following animal use protocols of the University of New Brunswick (AUP #s: 21034, 22028, 23025) and under banding permit #10801 E.

We recorded locations of individuals every 2-3 days throughout the breeding season and at different times of day. We used handheld radio telemetry equipment (Lotek SRX800 receiver and 8-element Yagi antenna) ensured the tag was on the individual by resighting the color bands or tag or flushing the individual from the location and confirming that the tag location moved. We also included locations of individuals of individuals that died or whose tags dropped if we could confirm the color band combination (n = 68 locations). Once located, we marked the location using a handheld Garmin GPS and recorded coordinates, time, and date. We did not track individuals on days with rain, heavy fog, or a high risk of lightning

We created kernel density estimates derived from each individual’s utilization distribution using “adehabitatHR,” an R package for home range analysis (Calenge 2019). We only included individuals with ≥ 5 locations (n = 50), which is the minimum number of locations required by “adehabitatHR” to create home ranges. We used R package “sp” to convert the individuals’ locations to a SpatialPointsDataFrame object (Pebesma and Bivand 2005), one of the two data types that can be used with “adehabitatHR”. Then we used “adehabitatHR” to create 50% kernel density estimates to look at home ranges for each individual (with ≥ 5 locations; range 5 – 26 locations; mean = 13 locations). The 50% kernel density estimate is usually considered representative of the core range of an individual, and the area where it spend most of its time. We also created kernel density estimates at the 95% level to represent the whole home range with 5% removed to deal with extreme values. We focussed on the 50% core home ranges for analysis because the 95% whole home ranges included substantial overlap that made it difficult to interpret visually.

Nest searching and monitoring

In 2023, we searched for A. n. subvirgata nests in dikeland and saltmarsh habitats from 16 June - 23 August. Nest searching was both opportunistic (i.e., while completing other tasks in the field) and systematic (i.e., dedicated searching). Nearly all nests were found by first flushing the female off the nest. We used a variety of methods to find nests: walking through the habitat, rope-dragging, searching with the assistance of a thermal camera, and following a female carrying food. We found that a combination of multiple methods was the most useful. For example, in many cases, we would split up and walk around the habitat watching for behavioral signs of females with nests nearby, such as carrying food or flushing late (i.e., taking flight when someone was very close to the bird). When one was encountered, we would leave a marking flag and return to the area with the rope. Then, after flushing the female close to the nest with the rope, we would use the thermal camera to try and pinpoint the exact location of the nest. We used an AGM Global Vision Thermal Monocular ASP-Micro TM160 to assist with nest searching efforts. This follows the recommendation of Galligan et al. (2003) of pairing a handheld thermal camera with traditional nest-searching methods. Once we found a nest, we marked the location with two flags placed 2m on either side of the nest.

We checked nests once every three days, unless weather or logistics prevented a check on the third day, in which case we would check on the second or fourth day. During each nest check, we determined if the nest was still active and gathered information on the number of eggs or nestlings, age of nestlings, and if eggs were warm or cold. We also noted if the female flushed when we approached, which can suggest the nest is still active. If the nest was not active, we tried to determine the cause of failure, for example, if the nest itself was destroyed and eggs were smashed, we assumed it was depredated, or if the nest was wet and the area around the nest was flattened from a recent high tide, then we assumed the nest was flooded. We also monitored tides to determine when flooding was likely, and we monitored weather, particularly rain events, to determine when nest wetness might be due to rain rather than flooding. Two high tide events resulted in failures at many of our nests; on 6 July at approximately 0145h, the height of the tide at a local tidal station (Pecks Point, 45°45'0", -64°28'58.8") reached 13.02 m, and on 4 August at approximately 0130h, the tide reached a height of 13.28 m (Government of Canada 2024).

Nest survival

We concluded whether each nest had failed or fledged based on observations at the nest. If we determined that at least one nestling had fledged, then we assigned a fate of “fledged’ (i.e., a successful nest). For each failed nest, we assigned a nest failure type: flooded, depredated, or unknown. Flooded nests were due to high tides and could be determined by nest wetness during nest checks (especially if there was no precipitation in the past 24 hours) and by using tide tables. We considered a nest to be depredated if there were signs of disturbance at the nest or broken shells. If we could not determine the nest fate, we assigned a nest fate of “unknown”. Some nest failures occurred after an extreme rain event, and some of “unknown” nest failures may have been due to this extreme weather event, which did not coincide with a particularly high tide. Nest failure due to heavy rains and winds has been reported in other tidal marsh nesting sparrows (Greenberg et al. 2006)

We weighed, measured, and collected fecal samples from nestlings once they were 7-8 days old. We used chick aging guides from the Saltmarsh Sparrow Research Initiative (Robinson 2021) and the Saltmarsh Habitat and Avian Research Program (SHARP 2021). The former reference guide is specific to A. caudacata, and the latter uses examples of A. caudacata, A. n. subvirgata, and A. maritimus (Seaside Sparrow). We found that our population of A. n. subvirgata were developmentally more advanced (by one day) than the examples in the two guides . We weighed nestlings (g), and measured wing chord (mm), tarsus (mm), and the length of the outermost primary feather (P9; mm).

We included 35 nests in our survival analyses. After removing a nest from the analysis that we suspected had already failed (and the clutch had not been completed by the female) at the time we found the nest. We created several models looking at nest success with a combination of three variables: exposure days, habitat, and nesting cycle. We measured exposure days from the first day an egg was laid, and therefore exposed to threats or could potentially fail, to the day the young successfully fledge. Habitat was a binary variable of saltmarsh or dikeland. Nesting cycle was also a binary variable; the first-cycle nests (nests that failed or fledged before 5 July), and second-cycle nests (nests that failed or fledged after 5 July). In our study area, the first series of very high tides since the beginning of the nesting season occurred around 5 July, and six saltmarsh nests failed. About 10 days after these high tides, we found 10 new nests, which we suspect were due to females renesting after failures. In three cases, the nests were <5 m from a recently failed nest. We considered any nest in the egg stage after 6 July to be a second-cycle nest. We included nesting cycle in the models because, although the nesting cycle of A. n. subvirgata are not known to synchronized with lunar cycles (Shriver et al. 2007), it appeared that nest timing was at least loosely associated with the tide, and, therefore, we tested for an effect. For A. n. subvirgata, nesting cycle is estimated to be 23 days (4 days for egg-laying, 9 days for incubating, and 10 days to fledging; Shriver et al. 2007).

We used a discrete proportional hazards approach and fit binomial regression models with a complementary log-log link and exposure days as an offset term to estimate daily nest survival probability (de Zwaan and Martin 2018). We compared models using corrected Akaike’s Information Criterion (AICc), which corrects for small sample sizes to avoid overfitting (Hurvich and Tsai 1993), and we considered models to  have different support if ΔAICc was >2. We used the complementary log-log link because it fit the data better than a logistic curve; the complementary log-log link is appropriate for time series data when the timing of an event (e.g., nest failure) cannot be precisely determined, but can be assigned to a time interval (e.g., a calendar day; Heisey et al. 2007). We used parameter estimates from the best-supported model to calculate overall nest success. We then back-calculated daily nest survival by taking the inverse exponential of the overall nest success based on the average nesting cycle for A. n. subvirgata.

Diet

We collected fecal samples from all adults banded in 2022 and all nestlings banded in 2023 to determine if adults were foraging in saltmarsh or dikeland habitats, both for their own dietary needs and to feed nestlings. To determine foraging location, we analyzed δ¹³C. We also looked at d¹⁵N values, which provide information about the trophic position of prey items. In 2022, adult sparrows were placed in a brown paper bag instead of the usual cloth bag. We placed all nestlings from the same nest into a single paper bag. Usually, individuals released a fecal sample into the bag before we removed them to band, tag, and measure. The paper bags with the samples were then labeled and frozen.

At the Stable Isotope in Nature Laboratory (SINLAB) at the University of New Brunswick, we analysed carbon isotopic signatures of fecal samples from adult and nestling A. n. subvirgata collected in 2022 (adults: n = 27) and 2023 (nestlings: n = 16). Samples were dried in an oven at 60°C for 48 – 72 hours, then were ground, prepared in tin capsules (8 x 5mm; Elemental Analysis), and weighed (± 0.001 g). We submitted 49 samples, of which four were blind duplicates, to the SINLAB who used Continuous Flow-Isotope Ratio Mass Spectrometry to obtain d¹³C and d¹⁵N ratios. To ensure quality and comparability of data, samples were measured alongside standard and reference materials. Carbon and nitrogen ratios were calibrated to international standards (Vienna Peedee Belemnite for carbon, atmospheric air for nitrogen), and are reported in delta notation (d) parts per thousand (per mille ‰). The SINLAB calculated analytical error as 0.2‰ for δ¹³C and 0.3‰ for δ¹⁵N based on repeat analyses of certified standard materials: bovine liver standard (developed by SINLAB, δ¹³C = −18.76‰, δ¹⁵N = 7.17‰), muskellunge muscle (developed by SINLAB, δ¹³C = −22.30‰, δ¹⁵N = 14.00‰), Nicotinamide (batch 237264, δ¹³C = −32.53‰, δ¹⁵N = 2.10‰). These standards were calibrated against secondary reference materials: USGS61 (caffeine, certified by USGS, δ¹³C = −35.05‰, δ¹⁵N = -2.87‰), CH7 (polyethylene foil, certified by IAEA, δ¹⁵N = 20.3‰), N2 (ammonium sulfate, certified by IAEA, δ¹³C = −32.15‰).

In habitats like dikeland and saltmarsh, which are isotopically different, d¹³C values from fecal samples can provide information on where the individual was foraging because of dominant photosynthetic pathways by freshwater or terrestrial (C₃) and saltmarsh (C₄) vegetation (Brittain et al. 2012). Lower δ¹³C values would suggest a diet with a terrestrial or freshwater origin indicative of vegetation using the C₃ photosynthetic pathway, while higher δ¹³C values suggest a diet with a saltwater origin from vegetation using the C₄ photosynthetic pathway (Kelly 2000). Nitrogen values provided information on trophic position of prey items, which can be used as a proxy for food quality, as protein content is generally positively correlated with trophic position in arthropods (Wilder et al. 2013).

To compare carbon and nitrogen values from nestling fecal samples between habitats, we used Welch’s Two-sample t-tests with ɑ = 0.05. We did not run analyses on fecal samples from adults because sample sizes for individuals using mainly dikeland (n = 4) or using both habitats (n = 2) were too small, so we instead report the means.

Vegetation surveys

We sampled 300 vegetation quadrats (0.5 m x 0.5 m square; Jones et al. 2021) at 100 locations between 6-15 June 2023, when females were choosing nest site locations for first-cycle nests. In each quadrat, we made the same measurements at five points: each of the four corners, and the center of the quadrat using a meter-long ruler (SHARP 2019). For each point in the quadrat, we measured the average height of vegetation crown (an estimate based on the height of continuous vegetation around the ruler; cm), species and height (cm) of tallest plants, and thatch depth (cm). We defined thatch as the horizontal layers of dead vegetation from previous years. We randomly selected one location for each adult A. n. subvirgata that was tagged in 2021 and 2022 and had at least five locations based on radio tracking or resighting. Therefore, half of the vegetation quadrats were completed at a location, where a unique individual was observed during the 2021 or 2022 breeding season (n = 50). For the other 50 quadrat locations, we created a polygon outlining the study area, then overlayed a grid of points on the polygon with 100 m spacing. We used random sampling with no replacement to select locations and removed any points that were located in water. To assess differences in vegetation structure between saltmarsh and dikeland habitats, we used Welch’s Two-sample t-tests to compare means, and F-tests to compare variances, with ɑ = 0.05.

All analyses were completed in R version 4.3.1 (R Core Team 2023). All outliers were defined as values above Q3 + 1.5*IQR or below Q1 – 1.5*IQR.

Literature cited

Brittain, R. A., A. Schimmelmann, D. F. Parkhurst, and C. B. Craft (2012). Habitat use by coastal birds inferred from stable carbon and nitrogen isotopes. Estuaries and Coasts 35:633–645.

Calenge, C. (2006). The package “adehabitat” for the R software: a tool for the analysis of space and habitat use by animals. Ecological Modelling 197:516-519.

Choi, D. Y., A. C. Fish, C. E. Moorman, C. S. Deperno, and J. M. Schillaci (2021). Breeding-season survival, home-range size, and habitat selection of female Bachman’s Sparrows. Southeastern Naturalist 20:105–116.

de Zwaan, D. R., and K. Martin (2018). Substrate and structure of ground nests have fitness consequences for an alpine songbird. Ibis 160:790–804.

Government of Canada (2024). Pecks Point (00190). Retrieved January 30, 2024, from https://tides.gc.ca/en/stations/00190.

Greenberg, R., C. S. Elphick, J. C. Nordby, C. Gjerdrum, H. Spautz, W, G. Shriver, B. Shmeling, B. Olsen, P. Marra, N. Nur, and M. Winter (2006). Flooding and predation: Trade-offs in the nesting ecology of tidal-marsh sparrows. Studies in Avian Biology 32:96-109.

Heisey, D. M., T. R. Shaffer, and G. C. White (2007). The ABCs of nest survival: theory and application from a biostatistical perspective. Studies in Avian Biology 34:13-33.

Hurvich, C. M., and C. Tsai (1993). A corrected Akaike Information Criterion for vector autoregressive model selection. Journal of Time Series Analysis 14:271–279.

Jones, S. F., C. N. Janousek, M. L. Casazza, J. Y. Takekawa, and K. M. Thorne (2021). Seasonal impoundment alters patterns of tidal wetland plant diversity across spatial scales. Ecosphere 12:1–19.

Kelly, J. F. (2000). Stable isotopes of carbon and in the study of avian and mammalian trophic ecology. Canadian Journal of Zoology 78:1–27.

QGIS.org (2023). QGIS Geographic Information System. QGIS Association. http://www.qgis.org

R Core Team (2023). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/.

Raim, A. (1978). A radio transmitter attachment for small passerine birds. Bird-banding 4:326-332.

Rappole, J. H. and A. R. Tipton. (1991). New harness design for attachment of radio transmitters to small passerines. Journal of Field Ornithology 3:335-337.

SHARP (2019). Nest vegetation measurements – standard operating procedure. Saltmarsh Habitat and Avian Research Program.

SHARP (2021). Chick Aging Protocol. Saltmarsh Habitat & Avian Research Program Standard Operating Procedure. Saltmarsh Habitat and Avian Research Program.

Shriver, W. G., P. D. Vickery, T. P. Hodgman, and J. P. Gibbs (2007). Flood tides affect breeding ecology of two sympatric Sharp-tailed Sparrows. Auk 124:552–560.

Taylor, P. D., T. L. Crewe, S. A. Mackenzie, D. Lepage, Y. Aubry, Z. Crysler, G. Finney, C. M. Francis, C. G. Guglielmo, D. J. Hamilton, R. L. Holberton, P. H. Loring, G. W. Mitchell, R. D. Norris, J. Paquet, R. A. Ronconi, J. R. Smetzer, P. A. Smith, L. J. Welch, and B. K. Woodworth (2017). The Motus Wildlife Tracking System: a collaborative research network to enhance the understanding of wildlife movement. Avian Conservation & Ecology 12:8.

Wilder, S. M., M. Norris, R. W. Lee, D. Raubenheimer, and S. J. Simpson (2013). Arthropod food webs become increasingly lipid-limited at higher trophic levels. Ecology Letters 16:895–902.