3.4 Relative Abundance

Recommended method: Variable radius point count.

The main assumptions of relative abundance surveys are:

1. Identical or statistically comparable methodologies are used when comparison between areas or monitoring trends in one area over time is an objective of inventory effort.
2. Environmental, biological, and sampling factors are kept as constant as possible to minimize differences in survey bias and precision between surveys.
3. Surveys are independent; one survey does not influence another.

If these assumptions are met then each replicate survey should show (on average) the same relative bias allowing calculation of trends and comparison between areas.

However, each assumption should be scrutinized carefully when investigating the applicability of count-based methods, like point counts. Of particular importance is the assumption of equal bias between surveys. Factors such as variable weather and changes in observers can influence whether this assumption can be met. As an alternative, surveyors may opt to use distance-based point count methods as described in the section 3.5.2. In comparison to traditional point counts, the only additional measure needed for the use of distance methods is the distance of individuals (or clusters of individuals) from the survey center. Distances can be measured in terms of distance categories or groupings (e.g., 40-50 m) when it is difficult to get exact distances. Buckland et al. (1993) believe that in many cases distance methods can be employed in field situations with little extra effort.

When distance methods are used, all animals sighted (regardless of distance from the survey center) are counted and the sightability of animals is estimated. This allows calculation of an actual estimate of population density (and associated variance). In addition, the assumption that the number of birds counted is linearly related to true abundance is less likely to be violated when distance methods are used as opposed to traditional counts (Burnham and Anderson 1984, Buckland et al. 1993). This may potentially contribute to an increase in power and statistical validity with distance methods.

3.4.1 Variable Radius Point Count

Point counts are the most widely used survey method for estimating songbird abundance. They are easy to conduct with an observer recording birds from a single point for a designated time period. Many modifications and alterations of this method have occurred over the years. Most of these were reviewed in the compilation by Ralph and Scott (1981) and by Verner (1985). An important new publication by Ralph et al. (1995) reviewed the various point count strategies and attempted to produce a method which could be used for 1) providing trend data for monitoring population changes; and 2) predicting population changes in response to habitat change. The goal of the method is to be flexible enough to accommodate a variety of study objectives in a variety of habitats.

The chief benefit of using a variable radius point count is the ability to accommodate a wide range of bird species, each which possesses a different singing style and each which may occur in a variety of acoustically-different habitats. The variable radius point count operates by essentially allowing the habitat to determine the size of the area being surveyed. This flexibility eliminates the problem of fixing a provincially-common radius for all point counts regardless of application. The maximum detectable distance to a bird may change between different habitats, but the radius of the survey will also change. As an example, surveys of grassland birds can generally cover a larger area per point because of the absence of a screen of trees, and because bird species may be flushed at greater distances in open habitats. The response of birds to the observer will vary in different habitats resulting in certain gregarious species being counted more frequently than more cautious, "long distance callers". In some cases, fixing a provincially-standard radius may exclude some shy species from surveys altogether. In contrast, a well-designed variable radius point count allows the project biologist to manipulate the sampling design and the maximum point radius to suit the environment and project objectives. As an illustration, very localized sampling of a riparian area might require small radius points and this would be reflected in the sampling design through the spacing of transects and the points along them.

It is important recognize that the objective for this survey is relative abundance and that the distance estimates are only intended to allow the observer to differentiate between the individual birds being counted in order to avoid "double counting". Thus, the accuracy of the distance estimates is not as critical to a successful survey as it might be to derive an estimate of density. However, it is worth mentioning that if distance estimates can be done with a fair reliability, they will enable comparisons with point count data from a variety of studies by making it relatively simple to distinguish between bird observations made inside or outside of any specified radius. Accurate distance measurements may also improve the quality of data analyses possible, as discussed earlier.

Office procedures

• Review the introductory manual No.1, Species Inventory Fundamentals.
• Obtain maps for Project and Study Area(s) (e.g., 1:5,000 air photo maps, 1:20 000 forest cover maps, 1:20 000 TRIM maps, 1:50 000 NTS topographic maps). Any map which is used to record data should be referenced to NAD83.
• Outline the Project Area on a small to large scale map (1:250,000 - 1:20,000).
• Determine Biogeoclimatic zones and subzones, Ecoregion, Ecosection, and Broad Ecosystem Units for the Project area from maps.
• Delineate one to many Study Areas within this Project Area. Study Areas should be representative of the Project Area if conclusions are to be made about the Project Area. For example, this means if a system of stratification is used in the Sampling Design then strata within the Study Areas should represent relevant strata in the larger Project Area.
• Compile a list which includes all potential songbirds for the Study Area.

Sampling design

• Based on the maps and other knowledge of the Study Area (previous reports, local resource specialists), stratify by habitat types (with different bird densities).
• To ensure coverage throughout the Study Area use a systematic sampling design. Establish the first point count randomly and all other point counts following at a set distance. [The Design Components for this survey are Point Count Stations.]
• Barker et al. (1994) conducted a detailed investigation into modeling optimal point count study designs. They found that the optimal study design and sample size were a function of 1) variance of estimated population change; (2) variance of average count; (3) maximum expected total count; or (4) power of a test for differences in average counts. They found that optimal duration of counts was different for trend analysis over time compared with comparisons between areas at one point in time. It is recommend that biologists consult Barker et al. (1994) when formulating point count strategies.
• If the objective is to estimate bird abundances for large areas such as watersheds or management areas then stratification may not be necessary. However if their is prior knowledge of abundance varying with habitat types than stratify accordingly.
• To understand bird habitat relationships, there are two approaches. One is to describe the habitat at the point count and correlate bird abundance with the habitat post hoc. Another approach is to categorize the available habitat types and place the point counts in these habitat types. However, this approach ignores important habitat gradients such as edges and ecotones. Ralph et al. (1995) recommend stratifying the habitats and placing edges and other questionable areas into a separate strata. Location of stations should be constrained so that they are no closer than 100 m from a habitat boundary.
• When possible, randomize or change observers between stations when resampling to minimize recurring bias in any segment of a survey.

Sampling effort

• If the objective is a snapshot of the community at any given time:
• Conduct a minimum of two visits per site as close together as possible. This will accommodate a nested ANOVA design so that variability among observers can be estimated as well as variability among sites. This is important as observer variability is one of the greatest sources of imprecision in bird surveys.
• If the objective is to compare abundances between habitat types:
• It is necessary to sample at each point throughout the breeding season, migration period or winter season. During the breeding season spacing of visits is important to accommodate the different breeding phenologies. If the objective is to document abundance over the breeding season, a minimum of three to four visits well spaced in time (>1 week apart) are required, and six to eight visits are better.
• Surveys should be replicated over time if general habitat inference is an objective. However, it will also be important to stratify by season, or appropriate unit in the eventual analysis to avoid excessive variance caused by temporal effects. The actual number of visits can be determined by power analysis. Most power analysis packages have the ability to perform power analysis for stratified ANOVA designs.
• The minimum number of point counts needed for a habitat type:
• This is driven by objectives and restrained by logistics and personnel. Sufficient point counts are needed to properly assess the number and distribution of birds in an area. A general rule is that a minimum of 30 point counts are needed per surveyed area such as a watershed (Ralph et al. 1993). However, a better approach is to take advantage of a power analysis software package to determine sample sizes needed both for trend analysis and comparison between areas. These are discussed in Species Inventory Fundamentals, Appendix G.
• Replicating counts from an individual site can identify the influence of within-site variability on results using the methods of Link et al. (1994). Within-site variability can be defined as variation due to factors such as differences between observers, and short-term variation in population size at a count station or monitoring site. This is not to be confused with between-site variability, which is due to large-scale differences in the spatial distribution of species, and forms the basis for most experimental designs. In general, Link et al. (1994) found that if the proportion of within site variation is large, and the cost of replicating a site is small compared to setting up a new site, then it is optimal to replicate counts. If the proportion of within site variation is small, and the cost of replicating a site is equal to that of setting up a new site, then it is optimal to not replicate. Not surprisingly, Link et al. (1994) found that counts for birds with lower abundance had the highest percentage of within count variation. It is suggested that biologists consult Link et al. (1994) when designing monitoring studies, especially for birds, which show low abundances.

Personnel

• All crew should be competent in identifying birds by sight and song. Training should be provided to assist with properly estimating the distance to a bird which is seen or heard. Competence in these two areas should be demonstrated by each crew member before field data are collected. Crew should be rotated evenly between stations and areas to minimize observer bias in the data.

Equipment

• Maps
• GPS (use NAD83)
• Compass
• 100 m tape
• Hip chain (Remember to gather all the string between stations; failure to do so can result in the death of many birds)
• Flags
• Data forms or tape recorder
• Clipboard
• Digital watch
• Thermometer
• Binoculars
• Field guide(s) for species confirmation
• Bird song tapes of local/regional bird songs for training

Field procedures

Establishing point count stations in the field

• Effort should be made to place the point count entirely within the identified strata, with a minimum of 100 m from any edge. It is acceptable to use less of a buffer when necessary.
• The distance between any two points will depend on the maximum detection radii. The distances between two adjacent detection radii should be a minimum of 50 m, but 100 m is better to avoid double counting of individuals. For example, a survey using a 50 m maximum detection radii with a 100 m buffer would have point counts separated by 200 m.
• Establish stations in the field using a compass and a hip chain. Flag the route between the points if the points are to be revisited throughout the season. Flag the points, and write the number of the point on the flag.
• If possible, place flags at measured intervals (for habitats where they will be visible) or note the distance to prominent landmarks for aiding distance estimates. Tie a marker at the point count station for map orientation. Orientation is usually upslope. This consideration is important in data analysis since birds are more likely to be heard from upslope than downslope.
• Note that if distances are measured in increments (i.e., 0-2 m, 2-5 m, 5-10 m, etc.) then it may be possible to apply distance methods which are vastly superior to point counts. See Buckland et al. (1993) for a determination of optimal sampling design. Note that with a proper sample design, it is possible to get reliable density estimates with sample sizes below 60 individuals (John Boulanger, pers. comm.).
• The number of stations visited in one day will depend on distance between stations, terrain, total time spent per station, and sample size requirements. A typical goal for point counts should be to complete between eight and ten stations in a morning.

Surveying at a point count station

• Surveys should start at sunrise and may continue for four hours.
• Wait one minute after arriving at the station to allow for return of bird activity before beginning to count detected birds. Use this time to record the required Point Count Station information and weather conditions at the top of the field dataform: Point Count Form. Indicate orientation of the field form (and thus bird observations) by marking a `north' arrow on the form.
• Particular care should be given to not disturbing birds when approaching stations. The movement of birds from the survey area can cause biases especially if distance-based methods (such as program DISTANCE) are used for analysis.
• The recommended time period is five minutes per point count. However, it is acceptable to use a shorter or longer time period provided detections are categorized into time periods. To integrate with the Breeding Bird Survey data set birds detected in the first three minutes at a station must be recorded separately from subsequent time intervals. The standard procedure is to record birds in three time intervals, depending on the total length of the point count: 0-3, 3-5, and 5 plus minutes (refer to Animal Observations Form-Songbird Point Count).
• Use the Point Count Form to plot bird observations. Mapping is an efficient way to record data, especially when counts are longer than three minutes. Shortcuts to keeping track of birds include using standardized behaviour codes for separating birds, recording movements, and showing simultaneous observations. A sample of a schematic map and a system of symbols is given in Figure 3. Note that only one observer counts birds at a single station at one time.
• Estimate the distance from the station center to each bird (at the location it is was first detected). Record bird location as accurately as possible onto the detection circle on the Point Count Form. The detection circle is subdivided into concentric rings of 10 m intervals to aid in marking detection locations.
• Note that the concentric rings will need to be adapted if it to be used with program DISTANCE, as distance methods require tighter rings towards the survey center and an expansion of ring width as the distance from the center increases. See Buckland et al. (1993) for more details on point count transects.
• Transcribe data onto the Animal Observations Form-Songbird Point Count.
• When a nest(s) of interest is located during a survey, refer to the Nest Site Description Form to determine the data attributes to be collected.

Figure 3. Sample schematic map for plotting the location of songbirds during point counts and the corresponding mapping symbols (from Ralph et al. 1993).

Data Entry

• The Design Component for this survey is point count stations. When digitally entering the survey data, choose `Point Count Station' from the `Design Component Type' picklist.
• Transect information for the transects used to layout the stations may also be recorded by choosing `Transects' from the `Design Component Type' picklist. This information can be used on its own as a location reference or it can be used as a way to group associated stations to a transect (optional).

Data Analysis

The quantification of sampling intensity and effort is fundamental to the use of indices and relative abundance measures. This way the assumption of equally bias surveys between areas and over time can be met. In addition, the usefulness of indices depends on the precision of estimates. It is strongly recommend that power analysis procedures be integrated into the study design of all these techniques. As described in manual No.1, Appendix G, programs such as MONITOR, POWER AND PRECISION, and NQUERY are user friendly, and can be easily used in an adaptive fashion to calculate sample sizes needed for the ultimate analysis questions.

If studies are designed appropriately the following general analysis methods can be used (Table 5).

Table 5. RIC objectives and analysis methods for relative abundance data

Objective	Analysis method1	Programs2
Trends in abundance over time	Sample methods Regression techniques Power analysis	DISTANCE, Generic statistical packages MONITOR
Comparison in abundance between areas	ANOVA, method Power analysis	DISTANCE Generic statistical packages Power analysis software
Determine whether habitat modifications have altered population size	T-test method Power analysis	Generic statistical packages Power analysis software

¹See manual No.1, Section 5, for more details on analysis techniques.

²See manual No.1, Appendix G, for more detail on software packages.

Difficulties with count data: One inherent problem with count data is that they are rarely normally distributed, which makes the application of parametric statistical methods risky, especially if sample sizes are low. Before data are used in parametric tests, the assumption of normality should be tested. Transformations may make frequencies nearly normal in some cases. For a detailed discussion of analysis of count data, see manual No. 1, Species Inventory Fundamentals, Section 5. White and Bennets (1996) introduce an alternative method for point count analysis and use songbird counts as an example of this analysis technique.

Trend analysis: The basic method for determination of trends is linear regression and associated techniques. There are a variety of refinements to linear regression that can be used with data depending on sampling assumptions and other characteristics of the data. Manual No.1, Section 5, provides a detailed discussion of these techniques.

Comparison between areas: Parametric tests and other methods can possibly be used to compare areas if surveys are conducted concurrently. If surveys are conducted non-concurrently (such as in different years), then the results might be biased by population fluctuations. See manual No.1, Section 5, for a thorough discussion of analysis of count data.

Habitat based inference: Logistic regression or similar methods can be used to test for habitat associations, but this approach requires that habitat units be the primary sample unit as opposed to population units.