PROJECT METHODS

The BCN Survey has three protocols that are recommended for use by our monitors.  Point Counts are recommended during the breeding season. Recommendations ask that monitors conduct a minimum of two visits during June, but they are also encouraged to conduct additional counts in late April and early May, when resident breeders have begun nesting, and in early July, when many young are being fed.  Monitors are asked to start as close as possible to sunrise (between 5:15 a.m. to 5:20 a.m.), and complete counts by no later than about 9:00 a.m.  Point counts are usually set up at points located at least 150 meter apart.  Point counts are five minutes in length and all birds detected within a 75 meter radius of the point are counted.  Birds detected outside the radius are noted separately.  It is also requested that monitors start at different points on different days in order to maximize the number of birds detected during the hours when birds are most active.  Points may be placed at intervals designed to get the best views or coverage of the habitat being surveyed or may be randomly placed throughout the site.  Points are almost exclusively placed in preserved areas that are usually owned by one of the county forest preserve districts, the state, or local municipalities.  The Forest Preserve District of DuPage Co. point count data is also included in BCN eBird and this analysis.  Their data uses similar, but still slightly different protocols.  Most notably, their data is collected in two five minute intervals, and for the sake of consistency we used the only the first interval for our analysis.

After monitoring has been completed, data is entered on the BCN eBird website.

For more details regarding our survey methods, please visit the BCN Survey protocol page.

DATA PREPARATION

Data collected from the BCN eBird website was imported into MS Access where records are filtered, trimmed, and converted into a usable format for analysis in a statistical software package called TRIM (Trends & Indices for Monitoring Data), developed by Statistics Netherlands.  MS Access was also used to substantially decrease manual labor / work time and automate many tasks including the export of records (for each species) for later import into TRIM, creation of a TRIM "command file" to fully automate the analyses performed in the TRIM software, converting TRIM results into MS Excel format (where improved graphs were created), and to automate the creation of web pages for each species (including automatically inserted titles, names, data, linked images, and linked graphs).

DATA ANALYSIS

The number of birds of each species detected at each point was calculated for each year. Records of adults and juveniles are not separated out by the BCN Survey.  Species not recorded on a point count during a visit are assigned a count value of zero in order to distinguish between species absences during survey years and missing data during non-survey years. Data is limited to observations between June 1st and July 15th as a means of attempting to filter out some of the records pertaining to late migrants (common through end of May), post-breeding dispersal (July-Aug.), and other factors. For the 2007 analysis, the first 15 days of July were included in addition to records from June. Based on general consensus from our scientific advisors, this addition will provide better data for breeding species that arrive late in the season and still avoid the period when post-breeding dispersal becomes an issue. Because point counts are visited multiple times during the breeding season, summary statistics must be developed for each species during each year of the analysis. For the 2007 analysis, mean abundance was used. Although maximum abundance, the method used in the 2004 analysis, is fairly widely accepted, mean abundance is used more frequently among researchers. A number of extraneous variables are more associated with maximum abundances. Data could be inflated by the occasional presence of migrants, nomadic individuals, or birds that occasionally wander outside of their established territories. The number of visits and observer effort also have more significant effects on the results when using maximum abundance. Additional valuable information can be found in the recent publication: Betts, M., Simon, N.P.P., Nocera, J.J. 2005. Point count summary statistics differentially predict reproductive activity in bird-habitat relationship studies. Journal of Ornithology 146: 151-159.

Annual population indices for each species are calculated using a program called TRIM (Trends & Indices for Monitoring Data), developed by Statistics Netherlands. TRIM is a comprehensive statistical software package designed specifically for species population trends. In addition to being used by Statistics Netherlands, TRIM has now been adopted by the European Bird survey Council and is being used throughout the European Union on a region-wide scale. This program uses a log-linear regression model with Poisson error terms for the analysis of time series of counts that contain missing observations (a linear trend approach was used in these analyses).  "If observations are missing, TRIM estimates the missing values on the basis of changes observed on plots that were monitored. This means that when a new year is added the index figures on previous years may change." (Statistics Netherlands 2005)  "The model used in TRIM requires at least two points in the time series to estimate parameters and hence, [points] counted in only one year are excluded. Moreover, if the data are too sparse, i.e. contain too many missing values, the model parameters cannot be estimated." (British Trust for Ornithology 2004). Because the actual numbers of populations are often unknown, time series are converted to index numbers with the base year set as the year the survey began, or in this case the first year that the survey had an adequate amount of data for analysis.  The base year is set at 1.  "A time-point index is a total for a time-point divided by the total for the first time-point.  Indices are thus the increase (decrease) factors with respect to the first time-point." (Pannekoek & van Strien 1996)  The resulting index values on the graph show how percentages have changed with respect to the base year and allow for better comparison of changes for various species.   The TRIM program also attempts to take into account the effects of overdispersion and serial correlation. "The usual (maximum likelihood) approach to estimation and testing procedures for count data are based on the assumption of independent Poisson distributions (or a multinomial distribution) for the counts. Such an assumption is likely to be violated for counts of animals because the variance is often larger than expected for a Poisson distribution (overdispersion), especially when they occur in colonies. Furthermore, the counts are often not independently distributed because the counts in a particular year will also depend on the counts in the year before (serial correlation). Therefore, TRIM uses statistical procedures for estimation and testing that take these two phenomena into account." (Pannekoek & van Strien 1996). Outputted errors will be higher for species that do not fit standard assumptions due to either of those reasons.

Although various models were tested, the Linear Trend model was used in TRIM to analyze each species trend because trends across the entire monitoring period were desired and data was usually too sparse to run the Time Effects model.  The effects of overdispersion and serial correlation were both taken into account in these analyses.  In 2007, categorical covariates were added to the analysis.  When habitat covariates are included in the analysis, missing values are more accurately estimated based on the other observed values within the same habitat type.

Two indices are calculated in the program and displayed on the graphs: Imputed and Model-based.

IMPUTED indices: "Observed counts are used when available. Missing counts are predicted. (more realistic course in time)."

• Imputed indices use actual species totals when available.  For point / year combinations without species totals (years when a particular point was not monitored), estimated totals based on other existing data are used.
• Only missing values are estimated.

MODEL-based indices: "All counts are based on model predictions. (often more stable; summary; std. errors available in TRIM.)"

• Model indices always take the existing data and display estimated totals for all point / year combinations (including years when a particular point was monitored).
• All values are estimated

Three tests are performed to determine the Goodness of Fit of each model.  The Pearson’s chi-squared statistic ("Chi-square") and the likelihood ratio statistic ("Likelihood Ratio") tests pass (good fit) if the p-value is greater than 0.05.  The tests fail (poor fit) if the p-value is less than 0.05.  The Akaike’s Information Criterion (AIC) is somewhat different.  "According to this approach, models with smaller values of AIC, or equivalently LR-2df, provide better fits than models with larger values."  (see Pannekoek & van Strien 1996, Sec. 2.5 for further details). The Chi-square p values have been included with the graphs. If the model fails, the "model" line and estimated trends are significantly different from the data and attempts were made to find a better fitting model. In some cases, species that fail the Goodness of Fit test can still be explained by significantly different trends in different habitat types. In other cases, the time-effects model can be run and used to improve the results. This model can describe trends for species with very erratic year to year changes or sudden changes due to high-impact events such as the spread of West Nile Virus into the Chicagoland area.