Contents:
1.1 Counting birds
Box Field methods
Box Detectability
Box Selection of sample plots
1.2 Production of national indices and trends
Box Missing values I
Box Trend interpretation and classification
Bird monitoring in Europe is organized on a national level; most countries have developed their monitoring schemes rather independently. National monitoring schemes are organized mostly by NGOs with various involvements of other institutions and individuals (governmental agencies, universities, research institutes etc).
The national bird monitoring schemes employ an array of different methods that are approved by PECBMS. The national coordinators provide PECBMS with the results of their schemes according to agreed standards and formats. Each national scheme delivers indices of numbers of individual bird species and trends therein as its main output; the index is the time series of the numbers, the trend is the change in these numbers over the years. These national indices and trends are the source of data for PECBMS, which uses them to compute supranational species indices and trends and multispecies indicators.
For details and contacts to national schemes´ coordinators visit Common bird monitoring schemes in Europe
.
1.1 Counting birds
Birds are counted using standardised field methods (for details see Box Field methods). Since complete counts are nearly impossible for large spatial units, birds are counted at sample plots selected across a territory of a country (for details see Box Selection of sample plots). Although field methods, selection of sample plots, and also number of years covered differ among European countries, statistical methods can handle this (see chapter 2.2. Combining national data into supranational outputs). Birds are counted within national generic breeding bird monitoring schemes, where all species registered are counted. However, some species are not covered well by these schemes, such as species with nocturnal activity (e.g. owls) or cryptic life style, some clustered and colonial species or extremely rare species. For such species or groups of species, specific surveys need to be set up, but these are not the focus of PECBMS.
Survey results can be affected by the fact that only part of the birds present at a particular site at the moment of counting is detected by an observer. This ´detection probability´ is variable over space and time and may also differ between observers. This should be addressed in field methods and data analysis. New methods are currently being developed to do so.
The majority of field work (i.e. bird counts) is done by volunteers and managed by coordinators. Since bird watching is a widespread activity across Europe and elsewhere, it is often no problem to recruit a high number of volunteers for bird surveys, and this is relatively easy in comparison to other taxa. Just as professionals, volunteers must be able to identify the birds in the field properly, record field data accurately and in proper format and deliver them timely to the coordinators. Since the volunteers do their work for free, one might fear that their work suffers from this, but such fear is unnecessary. Coordinators use a wide array of methods to check the skills of the volunteers and to guarantee a high standard of the data delivered.
One possible problem connected to working with volunteers is the selection of sampling plots. Volunteers might prefer to count in areas that are rich in birds rather than to be directed to plots which have been selected randomly. To solve this problem several national monitoring schemes select sample plots in a stratified random manner. Another problem is that volunteer fieldworkers can leave a scheme at any moment, causing a turnover in the sites counted and missing values. This occurs in any longterm monitoring scheme and statistical techniques and software are widely available and used to solve this problem too.
While the potential risks linked to the involvement of volunteer fieldworkers have been solved, several advantages remain: the running costs of a scheme are relatively low and largescale schemes are feasible (Greenwood, 2007).
For more details on each national monitoring scheme visit Common bird monitoring schemes in Europe.
Box Field methods
There is no uniformly best field method to count birds. What method is to be selected depends on, among others, the goals of a scheme, the sampling design and the availability of fieldworkers. Three main standard types of methods are available and used by national schemes within PECBMS, sometimes slightly modified for national purposes: territory mapping, line transect and point counts. Point counts along a transect are called point transect counts.
The territory mapping method is probably the most precise, but it is also very time consuming and laborious. It can be used on a limited spatial scale, unless simplified version is used. Nowadays, most national schemes apply either line transect or point counts methods. Each method has its strengths and weaknesses and there is no single rule to choose from them. Standard textbooks (e.g. Bibby et al., 2000; Sutherland et al., 2004; Sutherland, 2006) and also the Best Practice Guide (Voříšek et al., 2008) give detailed overviews of the differences between these two methods.
PECBMS works with national indices rather than with raw data. Therefore, the field method used to produce the national indices is of minor concern as long as this method is standardized through years and provides a reliable, representative picture of a species’ national trend. Learn more in chapter 2, Box Missing values II.
Details on methods and monitoring schemes for each country can be found in Common bird monitoring schemes in Europe.
Box Detectability
For many types of bird survey detectability is an issue because any comparison of the raw ´unadjusted´ counts between sites and through time must assume that the probability of detecting birds is the same. However, some birds present in a study area will always go undetected, regardless of the survey method, how well the survey is carried out, and the competence of the observers. Comparison of unadjusted counts will only be valid if the numbers represent a constant proportion of the actual population present across space and time. Detectability is an important concept in wildlife surveys and has been a matter of much debate (Buckland et al., 2001; Rosenstock et al., 2002; Thompson, 2002) and recent statistical developments.
A solution is to ´adjust´ counts to take account of detectability and a number of different methods have been proposed (Thompson, 2002). The ´doubleobserver´ approach uses counts from primary and secondary observers, who alternate roles, to model detection probabilities and adjust the counts (Nichols et al., 2000). The ´doublesampling´ approach uses the findings from an intensive census at a subsample of sites to correct the unadjusted counts from a larger sample of sites (Bart & Earnst, 2002). The ´removal model´ assesses the detection probabilities of different species during the period of a point count and adjusts the counts accordingly (Farnsworth et al., 2002). ´Distance sampling´ models the decline in the detectability of species with increasing distance from an observer and corrects the counts appropriately (Buckland et al., 2001). The ´binomial mixture´ model uses counts from repeated visits within a period of closed population sizes (Royle & Dorazio, 2008).
Distance sampling is a way of estimating bird densities from line or point count transect data and of assessing the degree to which our ability to detect birds differs in different habitats and at different times (Buckland et al., 2001; Rosenstock et al., 2002). The software to undertake these analyses is freely available at RUWPA website. This method is often recommended because distance sampling in the field, e.g. recording a distance to each bird, or more often recording birds in distance bands (e.g. 025 m, 2050 m, 100 m and over for line transects, 030 m and 30 m and over for point transects) is often practical when alternatives are not. While we flag the issue of bird detectability, most breeding bird surveys do not routinely adjust counts when assessing trends. Distance sampling and other methods are useful to provide improved estimates of population sizes, but so far, there is little evidence that detection probability adds significant bias to bird trends (Johnson, 2008).
Box Selection of sample plots
Monitoring schemes contributing their data to PECBMS are based on sampling, i.e. population indices and other results are inferred from a sample of sites distributed across a country. The selection of sampling plots (sites) determines how representative the results are.
The most common methods to select sample plots in generic breeding bird monitoring schemes are free choice, systematic selection, stratified random selection and random selection. Definitions according to Sutherland et al. (2004).
 Free choice – there are no rules, each fieldworker is allowed to select the plot arbitrarily. This method is prone to bias. Fieldworkers can, for example, prefer to work in areas that are rich in birds. Also, observers can abandon a site that has become less attractive because birds have declined.
 Systematic selection – plots are uniformly distributed on a grid every kilometer or hundred kilometers (or whatever scale is appropriate). Although this method is considered much better than a free choice, it still might pose a problem for representativeness if the location of plots coincides with a systematic pattern in the landscape.
 Random selection – sample plots are selected by the generation of randomly distributed coordinates within the study boundary. Random sampling is the ideal method to select sample plots, although with some practical limitations, e.g. some randomly selected plots can be inaccessible. These limitations can be solved by stratified random selection.
 Stratified random selection – the area of interest is broken down into different subareas, known as strata (singular stratum), according to predefined types of habitat, altitude, land use, bird abundance, accessibility of survey sites, administrative or geopolitical boundaries, observer density, etc. Within each stratum, plots are selected at random.
Free choice was the common method in older schemes, but nowadays most of these schemes have been replaced with schemes with some element of randomization. Stratified random selection is the prevalent method of newly established monitoring schemes in Europe.
In 9 countries a scheme with free choice was in place by 2008. In four of them, the old schemes with free choice have been replaced by new schemes using stratified random or systematic choice of sampling sites; these new schemes are combined with data from the old schemes. Improvements in scheme design are ongoing in two other countries. In the Netherlands poststratification and weighting has been used as the method to reduce potential bias (Van Turnhout et al., 2008). Czech Republic coordinators analyzed the main habitats and their coverage by the monitoring, and discovered that only urban habitats are slightly oversampled; important bias is unlikely. Nevertheless, improvements in the sampling design are planned here too.
There are only three schemes where potential bias needs to be addressed better. They will be focus of further efforts to improve sampling design in the near future.
All in all, thanks to the improvements in plot selection and increased rigour that have been applied, we believe that bias which could affect results at the European level is unlikely.
Information on selection of sample plots in national monitoring schemes can be found in Common bird monitoring schemes in Europe.
For more details on sampling strategy see Best Practice Guide (Voříšek et al., 2008) or standard textbooks on monitoring (e.g. Bibby et al., 2000; Sutherland et al., 2004; Sutherland, 2006).
1.2 Production of national indices and trends
Population yearly indices and trends are the most important outputs of national monitoring schemes. The index gives bird numbers in percentages relative to a base year, when the index value is set at 100%. Usually, but not necessary, the first year of a time series is chosen as the base year. Trend values express the overall population change over a period of years.
National species indices are produced by the coordinators of the monitoring schemes. They assess yearly allsites totals per species and compute the individual national species indices in a prescribed way. The count data usually contain missing values, and to impute these they use the predominant statistical technique, that is, Poisson regression, as implemented in the TRIM software (Trends and Indices for Monitoring data, Pannekoek & Van Strien, 2001). TRIM is a widely used freeware program (available via http://ebcc.birdlife.cz). To facilitate the use of TRIM, the software tool BirdSTATs is available too.
Statistically spoken: the basic TRIM model contains both site effects and year effects and estimates missing values from the data of all surveyed sites:
ln μ_{ij} = α_{i} + γ_{j},
with
α_{i} the effect for site
i and
γ_{j} the effect for year
j on the natural log of expected counts
μ_{ij}. Missing counts for particular sites are estimated (´imputed´) from changes in all other sites, or in sites with the same characteristics if the basic model is extended with covariates. The assumption is that changes observed in surveyed sites also apply to nonsurveyed sites.
The program produces imputed yearly indices and totals for each species. These yearly scheme totals, together with their standard errors and covariances, are collected by the PECBMS coordinator.
For details on TRIM and BirdSTATs, see Box Missing values I.
In case there is more than one monitoring scheme within a country, e.g. an old scheme and a new one (i.e. schemes differ in time span) or different regional schemes (i.e. schemes differ in spatial coverage), the coordinator combines the results per scheme to produce new combined indices per species and per country. A tailormade software tool called Combine has been developed for this purpose, which also takes into account standard errors of indices of the constituent schemes. The procedure used resembles the one to produce supranational indices from national results (see below).
In addition to national indices, trends are computed to indicate whether long term changes in bird populations are strongly increasing, moderately increasing, stable, uncertain, moderately declining or steep declining (learn more in Box Trend interpretation and classification).
Box Missing values I
Ideally, in a breeding bird monitoring scheme all sites are surveyed every year. If so, it is easy to assess the changes in the yearly allsites totals of breeding pairs. These totals are usually represented as indices by setting the first year at value 100. But the reality in largescale monitoring schemes is that many sites are skipped once or several times during the lifetime of a scheme because some fieldworkers enroll years after the start of the scheme, while others drop out after a number of years. Missing counts thus arise and simple comparisons of yearly allsites totals of breeding pairs give misleading inferences on trends, as a simplified example shows.
The number of breeding pairs of a given species in the example has declined in sites 1 and 2, which were sampled each year. Site 3 was only surveyed in the third year. As a consequence, the yearly total as well as the index would be highest in year 3 if they are based on the simple sum across all sites, which is of course an artifact caused by the enlargement of the monitoring scheme in year 3. Taking the mean numbers of the sites is also incorrect. This is because, in this case, site 3 happens to have more breeding pairs of the species.

year 1 
year 2 
year 3 
site 1 
4 
3 
2 
site 2 
4 
3 
2 
site 3 
missing count 
missing count 
8 
yearly allsites total 
8 
6 
12 
yearly indices 
100 
75 
150 
yearly mean 
4 
3 
4 
To solve the problem by simply disregarding site 3 would be a waste of useful information, especially if site 3 continues to be surveyed in the years to come. It is a better solution to estimate (impute) the missing counts with sound statistical methods. Such an imputation makes it possible to compare the years in a fair way, ruling out artifacts and producing more reliable figures.
We use the predominant statistical technique to impute missing values in count data, viz. Poisson regression (loglinear models), as implemented in TRIM software (TRends and Indices for Monitoring data; Pannekoek & Van Strien, 2001). Poisson regression is also available in the generalized linear model modules of many other statistical packages. TRIM is an efficient implementation of Poisson regression to analyze timeseries of count data collected in many sites and to produce indices and associated standard errors. It is a widely used freeware program (available via www.ebcc.info/trim.html.
TRIM implements several loglinear models to impute missing data. The basic model contains both site effects and year effects and estimates missing values from the data of all visited sites. The key assumption is that changes observed in surveyed sites also apply to nonsurveyed sites. The next example shows the result.
TRIM produces the following values for the sites in the example:

year 1 
year 2 
year 3 
site 1 count 
4 
3 
2 
site 2 count 
4 
3 
2 
site 3 count 
estimated: 16 
estimated: 12 
8 
yearly allsites total 
24 
18 
12 
yearly indices 
100 
75 
50 
Changes in site 3 have been based on the changes in site 1 and 2. It is clear that the yearly totals and indices now make sense. The same procedure is applied to impute values for sites that had been surveyed in past years, but are not surveyed any more.
Note that such imputation does not affect the trend estimation, just because missing values are calculated from the changes in sites with observations. Estimating missing values only serves a fair comparison between years. Also note that it is not the aim to get reliable information on changes in site 3, but only to get reliable information on trends based on all available information. Imputed values are less valuable than real observations. The major drawback is that the more missing values occur in the data, the wider the confidence intervals of indices will be. This is because imputed values don´t enlarge the sample size; in the example, the sample size for the first two years still is 2.
The basic model may be elaborated by including covariates, such as habitat or region. Any changes between years for nonsurveyed sites then are derived from changes in surveyed sites with a similar habitat or within the same region, thereby relaxing the assumption mentioned above. The incorporation of covariates may lead to better model fit, better imputations and smaller confidence limits of the resulting indices and trends. The penalty of not using such elaborate TRIM models is having larger standard errors of indices.
The usual approach to statistical inference for loglinear models is maximum likelihood estimation and associated calculations of standard errors and test statistics. These estimation and testing procedures are based on the assumption of independent Poisson distributions for the counts. Such an assumption is likely to be violated when animals are counted because the variance may be larger than expected for a Poisson distribution (overdispersion), for instance when the animals occur in colonies. Furthermore, counts are often not independently distributed because the counts at a particular point in time may depend on the counts at the previous timepoint (serial correlation). TRIM uses procedures for estimation and testing that take into account these two phenomena.
Apart from TRIM, a software tool called BirdSTATs is also available for computation of population indices and trends. BirdSTATs is an open source Microsoft Access database, which is programmed to use and automatically run the program TRIM in batch mode to perform the statistical analysis for series of bird counts in the dataset.
BirdSTATs is capable of importing different kinds of counts data, enables stratification of count sites and selection of subsets of counts data, produces standardised TRIM input and command files and runs TRIM in batch mode for all or a selection of strata, and it collects the output of the batched TRIM runs in a convenient and standardised format to fit the requirements of PECBMS. It can be downloaded on www.ebcc.info/trim.html.
Box Trend interpretation and classification
In addition to yearly indices, it is relevant to assess the trend over the whole study period. This trend is the slope of the regression line through the logarithm of the indices. This slope, the standard trend estimate used in scientific papers, is called the additive trend in TRIM.
In addition, TRIM calculates the multiplicative trend, which is easier to interpret for laymen. This multiplicative trend reflects the changes in terms of average percentage change per year. If this trend is equal to 1, then there is no change. If the trend is e.g. 1.08, then there is an increase of 8% per year. This means: in year 2, the index value will be 1.08, in year 3 1.08 x 1.08 = 1.17 etc. If the trend is e.g. 0.93, then there is a decrease of 7% per year.
Both trend estimates are different descriptions of the same estimates: the additive parameter is the natural logarithm of the multiplicative parameter.
The multiplicative trend estimate (trend value) in TRIM is converted into one of the following categories to facilitate its interpretation further. The category is not only determined by the value of the multiplicative trend itself, but also by its uncertainty, here its 95% confidence interval (= trend estimate +/ 1.96 times the standard error of the trend).
 Strong increase – increase significantly more than 5% per year (5% would mean a doubling in abundance within 15 years). Criterion: lower limit of confidence interval > 1.05.
 Moderate increase – significant increase, but not significantly more than 5% per year. Criterion: 1.00 < lower limit of confidence interval < 1.05.
 Stable – no significant increase or decline, and most probable trends are less than 5% per year. Criterion: confidence interval encloses 1.00 but lower limit > 0.95 and upper limit < 1.05.
 Uncertain – no significant increase or decline, and unlikely trends are less than 5% per year. Criterion: confidence interval encloses 1.00 but lower limit < 0.95 or upper limit > 1.05.
 Moderate decline – significant decline, but not significantly more than 5% per year. Criterion: 0.95 < upper limit of confidence interval < 1.00.
 Steep decline – decline significantly more than 5% per year (5% would mean a halving in abundance within 15 years). Criterion: upper limit of confidence interval < 0.95.
See the TRIM manual (Pannekoek & Van Strien, 2001) for technical details on the computation.