Methods

Spatial Sampling Framework

Atlas Squares

The basic sampling unit was a 10 km by 10 km (100 km²) square from the Universal Transverse Mercator (UTM) grid. Saskatchewan had 6,914 such squares across three UTM zones: 12, 13, and 14 (Figure 1). Partial squares (less than 100 km² of land mass and/or freshwater) occur along UTM zone lines and provincial boundaries. During data collection, partial squares of at least 10 km² were treated as full squares (i.e., survey effort was comparable to that of full squares). Partial squares with less than 10 km² of land mass and/or freshwater were surveyed less intensely than other squares.

Maps of individual squares (Figure 2) depicted broad habitat types, roads, trails, contour lines, and the 1-km UTM reference grid (with 100 m intervals indicated). Pre-determined, on-road point count locations were mapped with coordinates displayed in the legend. These maps were available online for field-use by atlassers, and helped participants to navigate the square and to allocate appropriate sampling effort (e.g., visiting all habitat types; conducting on- and off-road point counts).

Figure 1. Atlas squares and UTM zones in Saskatchewan (NAD83)

Figure 2. Sample 10 x 10-km square map.

Administrative Regions

Saskatchewan was divided into 16 administrative regions (Figure 3). Atlassing activities in each region were managed by one or two volunteer Regional Coordinators (RCs) or Atlas Staff. RCs served as the main contact and survey information source for atlassers within a region, and helped ensure that regions were adequately surveyed. How they did so varied, depending on the number of volunteers and the accessibility of squares in their region. RCs spent time assigning squares, training individual participants, coordinating and managing volunteer efforts, and importantly, RCs often conducted most point count surveys themselves as many regions had few volunteers. In addition, RCs participated in yearly planning meetings and reviewed all data submitted by atlassers in their regions. RCs were the backbone of the project and vital to its success.

Figure 3. Administrative regions Figure 3. Administrative regions

Priority Squares

To ensure adequate spatial distribution of survey effort, the Technical Committee opted to designate a portion of the atlas squares as Priority Squares. Volunteers were directed to survey Priority Squares first, before turning their attention to other squares. Naturally, squares around cities, national and provincial parks, and near atlasser homes received attention regardless of priority status.

In the more populated southerly parts (regions 1 through 14) where grid roads and highways facilitate travel to most places, roughly 20% of squares were assigned priority status in an evenly-distributed systematic grid. In northern regions (regions 15 and 16), where both volunteers and access are severely limited, it was necessary to take a different approach to assigning priority squares to establish challenging but achievable coverage goals.

Northern Sampling Strategy

We initially considered designating Priority Squares in northern regions by establishing a ‘systematic grid’ as was done in southern Atlas regions, but at a lower density. While a systematic grid would be consistent with work in southern regions and would be statistically advantageous if achieved, numerous Priority Squares would be impractically difficult and costly to access owing to the lack of roads, navigable rivers, or even suitable lakes for landing floatplanes in most northern squares. Instead, we used a hierarchically structured sampling design known as the Boreal Optimal Sampling Strategy (BOSS; Van Wilgenburg et al. 2020) to select the northern Priority Squares. The BOSS approach considered access related cost-constraints, habitat representation within ecoregions, and spatial balance (evenness of spatial coverage) to provide a pragmatic and cost-effective sampling framework. We allocated the sample size of target Priority Squares amongst ecoregions based on a combination of ecoregion area, historic avian species richness and expected variability, with the aim of placing greater sampling where we anticipate greater sampling variability and potentially more species range boundaries. Since the northernmost regions have little or no history of sampling from which to derive estimates of spatial and temporal variation in bird communities, we employed proxy variables known to correspond with spatial and temporal variation in bird abundance. We used proxy variables with well-documented links between species abundance and distribution reflecting spatial and temporal variation in vegetation, forest age and climatic variables (Van Wilgenburg et al. 2020). In all, 387 Priority Squares were allocated across the two northern regions, and these were further stratified into primary, secondary, and tertiary priorities to establish sampling targets that corresponded to achieving minimal, better, and ideal coverage of remote northern regions.

Establishment of point count stations in northern squares was done in two ways. If the square was accessible using the National Road Network, then up to 30 predefined road-side points were established. Then for all Priority Squares identified using the BOSS approach, we selected a stratified (based on habitat) random sample of systematic grids consisting of nine points spaced 300 m apart. We set a sampling target of two grids of nine stations per Priority Square so that a two-person team could complete the square in one morning. For a detailed description of the sample grid selection procedure using the BOSS approach see Van Wilgenburg et al. (2020).

Distribution and Breeding Evidence

From 2017 to 2021, atlassers systematically surveyed Saskatchewan square by square, looking for evidence of breeding birds. Atlassers followed standardized methodology used for other Canadian atlasses, based on techniques initially devised for the first British atlas (Sharrock 1976), replicated in many places since (Gibbons et al. 2007) and promoted and refined by the North American Ornithological Atlas Council. Atlassers used a checklist approach, termed “general atlassing” detailed in the Instructions for General Atlassing, where participants kept track of the date, time, duration, and location of each checklist, as well as a complete species list of birds encountered and a tally of individuals. For each species detected, the breeding evidence associated with that observation was classified into one of three categories: possible, probable, or confirmed (Table 1). Data gathered from these surveys constituted the bulk of data used to create the distribution maps.

Casual Observations

Atlassers were also encouraged to submit so-called casual or incidental observations, that is, observations of breeding evidence made while not dedicating time to a complete visit to a square. For example, while driving through a square, an atlasser observed an American Robin carrying nesting material. Survey effort was not recorded for casual observations.

Prioritising Squares

Saskatchewan’s access limitations and small population size made obtaining even an adequate level of coverage across all geographic areas and habitats immensely challenging. To ensure surveying was geographically well-spaced, a series of squares were chosen as priority targets in northern and southern Saskatchewan and coverage targets per region were set. In regions with fewer participants, volunteers surveyed priority squares first, before turning their attention to other squares. Extra time and effort on remote expeditions was paid to priority squares.

“Completing” Squares

In order to consider an atlas square complete, atlassers were instructed to spend 12-20 hours of effort surveying each square, outlined in the Instructions for General Atlassing, and a goal of 15 point counts per square was set. Atlassers were strongly encouraged to cover all habitats in the square, vary the times of day at which they surveyed, and extend the 12-20 hours of effort over days, months, and even years, to maximize the number of species detected. Although this type of temporal and seasonal variation in coverage is ideal, it was frequently impractical to achieve, particularly in regions with few surveyors and in more remote, challenging areas to access.

As the atlas progressed, it became clear that the initial square completion goals would be unattainable in most areas of the province given the number and distribution of participants, even when considering solely the priority squares. As a result, during the last two years of the atlas emphasis was shifted to completing 12-15 point counts in each priority square in order to ensure that relative abundance mapping would be possible. This came at a sacrifice to general atlassing hours, since atlassers who could complete point counts often had only one day or less to spend searching for breeding evidence in a given square.

Distribution and Breeding Evidence

From 2017 to 2021, atlassers systematically surveyed Saskatchewan square by square, looking for evidence of breeding birds. Atlassers followed standardized methodology used for other Canadian atlasses, based on techniques initially devised for the first British atlas (Sharrock 1976), replicated in many places since (Gibbons et al. 2007) and promoted and refined by the North American Ornithological Atlas Council. Atlassers used a checklist approach, termed “general atlassing” detailed in the Instructions for General Atlassing, where participants kept track of the date, time, duration, and location of each checklist, as well as a complete species list of birds encountered and a tally of individuals. For each species detected, the breeding evidence associated with that observation was classified into one of three categories: possible, probable, or confirmed (Table 1). Data gathered from these surveys constituted the bulk of data used to create the distribution maps.

Table 1. Breeding evidence categories and codes. Observations recorded under the category “observed” were not considered breeding evidence and were not mapped.

Casual Observations

Atlassers were also encouraged to submit so-called casual or incidental observations, that is, observations of breeding evidence made while not dedicating time to a complete visit to a square. For example, while driving through a square, an atlasser observed an American Robin carrying nesting material. Survey effort was not recorded for casual observations.

Prioritising Squares

Saskatchewan’s access limitations and small population size made obtaining even an adequate level of coverage across all geographic areas and habitats immensely challenging. To ensure surveying was geographically well-spaced, a series of squares were chosen as priority targets in northern and southern Saskatchewan and coverage targets per region were set. In regions with fewer participants, volunteers surveyed priority squares first, before turning their attention to other squares. Extra time and effort on remote expeditions was paid to priority squares.

“Completing” Squares

In order to consider an atlas square complete, atlassers were instructed to spend 12-20 hours of effort surveying each square, outlined in the Instructions for General Atlassing, and a goal of 15 point counts per square was set. Atlassers were strongly encouraged to cover all habitats in the square, vary the times of day at which they surveyed, and extend the 12-20 hours of effort over days, months, and even years, to maximize the number of species detected. Although this type of temporal and seasonal variation in coverage is ideal, it was frequently impractical to achieve, particularly in regions with few surveyors and in more remote, challenging areas to access.

As the atlas progressed, it became clear that the initial square completion goals would be unattainable in most areas of the province given the number and distribution of participants, even when considering solely the priority squares. As a result, during the last two years of the atlas emphasis was shifted to completing 12-15 point counts in each priority square in order to ensure that relative abundance mapping would be possible. This came at a sacrifice to general atlassing hours, since atlassers who could complete point counts often had only one day or less to spend searching for breeding evidence in a given square.

Point Counts, Bioacoustics, and Abundance

Point counts are timed surveys in which the observer records all birds seen and heard at predetermined points, relying on observers having high levels of identification skills who can recognise the songs and calls of local species. Point count data were used in combination with distribution data to plot abundance in relation to habitat and landscape features as well as to map species’ Probability of Observation. Point counts were conducted from May 28 to July 7-10 (June 7 to July 7-10 in region 16), starting no earlier than 30 minutes before sunrise and ending no later than 5 hours after sunrise. The end date for point counts was slightly flexible, and allowed surveyors to extend the survey period by a few days if birds were still singing at full chorus after the July 7 cut-off date in their region. Atlassers also recorded all breeding evidence observed during point counts.

The close collaboration between Environment and Climate Change Canada and the Saskatchewan Breeding Bird Atlas team resulted in the adoption of two in-person point count protocols; one used primarily south of the boreal forest by volunteers, modeled after recent Canadian Atlases in Manitoba, British Columbia, and Ontario, (i.e., Atlas point counts), and another used primarily by staff as part of the Boreal Optimal Sampling Strategy, i.e., BOSS point counts (VanWilgenburg et al. 2020). Technological and statistical advancements allowed for the use of bioacoustic surveys to supplement the in-person point count data, using both autonomous recording units (ARUs) and handheld devices to make recordings emulating both in-person protocols.

Atlas point counts

Atlas point counts were similar to the approach taken in recent atlases (Manitoba, British Columbia, Ontario II) where count duration was 5 minutes, except that the survey was divided into 3- and 2-minute intervals using removal sampling. Species were recorded within an unlimited radius (i.e., as far as the eye could see and the ear could hear). The coverage goal was to complete 15 point counts in each priority square and all other squares surveyed by point counts.

Point count locations were randomly placed along roads at the start of the project and each square map detailed 30 potential point count locations. Atlassers were asked to conduct counts at locations numbered 1 through 15 and, if one or more locations were deemed unusable (e.g., inaccessible or unsafe), to use subsequent points as replacements (16, 17, etc.). Once locations were selected, points were surveyed in any order.

Before data collection began, the habitat type at each predetermined atlas point count location was categorised using geographic information systems (GIS). In squares where on-road points did not adequately sample a representative proportion of habitats in the square, atlassers conducted additional off-road counts in the underrepresented habitat(s) as identified on the atlas square map. The locations of off-road point counts were chosen by individual atlassers, who could ensure that they truly represented the desired habitat type and were accessible. Off-road point count locations were ideally at least 100 m from the road and at least 300 m apart, but road avoidance was not always possible. Atlassers recorded UTM coordinates and habitat for off-road point count locations. Each square map detailed the number of pre-defined and off-road point counts to be surveyed, always totalling 15.

Methodology for atlas point counts is further detailed in the Instructions for Point Counts.

BOSS point counts

The point count effort in Regions 15 and 16 by Birds Canada and ECCC staff, as well as volunteers who undertook remote expeditions, was focused on conducting point counts in the Boreal Optimal Sampling Strategy (BOSS) grids. BOSS point counts were 10 minutes in duration and employed time removal (Farnsworth et al. 2002) and distance sampling methods (Buckland et al. 2001), i.e., observers documented the minute interval of first detection and distance (0-50 m, 50-100 m, and > 100 m) to each bird.

Bioacoustics sampling

Bioacoustic technology functioned as a workforce multiplier and greatly increased capacity for point count data collection, giving more participants and partners the opportunity to contribute to the point count effort. Both handheld recorders and autonomous recording units (ARUs) were used to make recordings which were later transcribed by a skilled observer.

ARUs are weather-proof digital devices that are programmed with custom recording schedules and are designed to be deployed in the environment and retrieved at a later date. During the point count season, ARUs were programmed to make several recordings daily throughout the morning (half hour before sunrise to 5 hours after sunrise, based on location), as well as overnight to capture nocturnal species. Recorders deployed in the fall and winter months were also scheduled to make intermittent recordings from February to early May to capture early vocalizing residents such as owls, woodpeckers, and grouse.

Once the ARU was retrieved, recordings were selected at random for transcription, discarding any that were unsuitable due to interference from wind, rain, or human disturbance (e.g., road noise, trains, industrial machinery, etc.). ARUs were primarily used for the Boreal Optimal Sampling Strategy, where one 10-minute morning recording was selected along with 5 supplemental 3-minute dusk/night recordings. ARUs used by atlas staff and partner organizations included Wildlife Acoustics models SM2, SM2+, SM4, SMmini and Frontier Labs BAR-LT. Recordings made by Prince Albert National Park staff were made using the River Forks E3A system.

Unlike ARUs, handheld recorders are not weather-proof and require operation by a human to start and stop the recording. A small fleet of handheld Zoom H2N recorders and tripods were available for loan, allowing participants who enjoy early morning birding but perhaps lack aural birding skills or have some hearing loss, to contribute to the point count effort along with general atlassing. Observers noted any visual only detections of birds, which were added to the corresponding checklist (but not the point count) after the recordings were transcribed.

Wildlife Accoustics SM Mini Deployed in a Grassland Setting and a SM2 Deployed in Winter in the Boreal Forest. Photos by LeeAnn Latremouille

Rare, Colonial, and At Risk Species

An emphasis was placed on thorough documentation of rare and colonial species, including species assessed as “at risk” (Special Concern, Threatened, or Endangered) by the Committee on the Status of Endangered Wildlife in Canada (COSEWIC), as well as species that are uncommon provincially or regionally. For colonial species and widespread species at risk, like Barn Swallows, extra documentation was only solicited for confirmed breeding records. For rare and colonial species, additional documentation requested included UTM coordinates (either using the square maps, GPS enabled devices, or the online mapping tool), species and habitat descriptions, and the number of adults observed (for rare species) or the number of active nests or pairs (for colonial species). For uncommon species, atlassers were encouraged to thoroughly document how they identified the species (particularly if identification was challenging).

Snapshot of a Large Mixed-Species Nesting Colony on Doré Lake. Found to Contain Nests of American White Pelican, Double-crested Cormorant, Ring-billed, Herring, and California Gulls, and Caspian and Common Terns. Photo by Vicki St Germaine

External and Additional Data Sources

Compatible datasets from other surveys that were conducted during the atlas period, and that focused on recording breeding species, were incorporated into the atlas database by assigning sightings a breeding evidence code and a square. Since most external data sources did not follow atlas protocols to assign breeding evidence, such observations were most often given a possible breeding evidence code. Data that did not include measures of effort were entered as casual observations. External data were subject to the same review process as other atlas data and were omitted if they didn’t meet the standards for breeding evidence or if the location of the observation could not be assigned to a square. The bulk of external data came from the Saskatchewan Conservation Data Centre, the provincial repository for data collected on wild species, which includes industry-commissioned or conducted surveys, as well as the data collected by Nature Saskatchewan through their programming.

Data Review and Preparation

Atlassers submitted data online through Birds Canada’s NatureCounts portal for the Saskatchewan Breeding Bird Atlas. The data management system automatically flagged records requiring additional review and for which additional documentation was required. Flagged records were typically for rare or colonial species but could also be unusually high numbers of individual species detected on single point counts, or species occurring outside their known breeding range. RCs reviewed all data submitted, with special attention paid to flagged records, all of which were further reviewed by the Atlas Coordinator.

References

Buckland, S. T., Anderson, D. R., Burnham, K. P., Laake, J. L., Borchers, D. L., & Thomas, L. 2001. Introduction to distance sampling: estimating abundance of biological populations. Oxford University Press, Oxford, UK.

Farnsworth, G.L., Pollock, K.H., Nichols, J.D., Simons, T.R., Hines, J.E. and Sauer, J.R., 2002. A removal model for estimating detection probabilities from point-count surveys. The Auk, 119(2), pp.414-425.

Gibbons, D.W., Donald, P.F., Bauer, H.G., Fornasari, L. and Dawson, I.K., 2007. Mapping avian distributions: the evolution of bird atlases. Bird Study, 54(3), pp.324-334.

Sharrock, J. T. R. 1976. The atlas of the breeding birds in Britain and Ireland. T. & A. D. Poyser, Hertfordshire, U.K.

Van Wilgenburg, S.L., Mahon, C.L., Campbell, G., McLeod, L., Campbell, M., Evans, D., Easton, W., Francis, C.M., Haché, S., Machtans, C.S. and Mader, C., 2020. A cost efficient spatially balanced hierarchical sampling design for monitoring boreal birds incorporating access costs and habitat stratification. PLoS One, 15(6), p.e0234494.