When the Department of the Interior [DOI] created the National Biological Service [NBS] in 1993, the Bird Banding Laboratory [BBL] was placed into the new NBS organization. At this time the NBS directorate commissioned a review of the BBL. This panel’s report (referred to as the Buckley Report throughout this document) made recommendations in five areas: (1) permitting procedures and practices; (2) operating issues; (3) data management; (4) BBL organization and staffing; and (5) implementation.

The Recapture/resighting Task Force [Task Force] was created to address some of the concerns of data management. The charge of the Task Force is to develop recommendations for which and how many recapture and resighting data are to be collected by the BBL, and how they are to be managed. Specifically, the Task Force should (1) make theoretical considerations such as potential uses of data using contemporary analytical models, and effects of bander efforts in collecting and reporting recapture data; (2) consider the wide variety of banding situations that produce recapture data (e.g., preseason duck banding, standardized multi-site banding such as MAPS, project-oriented research banding, migration banding, winter feeder banding, irregular backyard banding) in different volumes, frequencies, and value; (3) consider how to link recapture and resighting data to banding data; (4) evaluate the current encounter record format and data fields relative to contemporary uses of the data and probable changes to the management of encounter data; and (5) evaluate and make recommendations for archiving encounter records.

The Task Force consisted of the following members: Bill Kendall (chair), Peter Blancher, Fred Fiehrer, Judy Hildenbrand, Jim Kelley, Kathy Klimkiewicz, Wendy Manear, Jim Nichols, and Jeff Spendelow .

Based on the Buckley Report’s recommendations, our charge from the Implementation Team was not to consider whether the BBL should or should not store recapture/resighting data, but which and how many data of this type should be stored, and how they should be managed. Nevertheless in the process of evaluating different types of data the Task Force could not avoid re-evaluating the question of whether any of them should be stored. In evaluating the merits of storing recapture/resighting information for each of the types of banding studies discussed below we considered several questions: (1) What is the cost of storing such data? (2) What is the potential scientific value of such data? (3) Should the BBL give special consideration to storing re-encounter data where the banding program is part of a monitoring program of the Department of the Interior (DOI), or a DOI agency or Canadian Wildlife Service (CWS) requests it? (4) Is there a need to store such data in situations where the principal investigator also plans to maintain the full database? (5) Should resighting data be handled differently from recapture data? (6) How many re-encounters per bird per unit time should be accepted?

Our conclusion was to agree in general with the Buckley Panel that the BBL should store some selected recapture/resighting information. After discussing the six questions listed above we will conclude with more specific recommendations.

We anticipate that the acceptance of selected recapture/resighting data would greatly increase the volume of re-encounter data submitted to the BBL. We cannot estimate this increase with reasonable accuracy at this time. However, depending on the smallest re-encounter time interval allowed (month, week, day), we anticipate that the volume of data submitted could easily increase tenfold for certain species. There are three major facets of storing and providing recapture/resighting data: data storage, data entry/submission, and data processing. These should be considered in light of this potential increase.

The expense of storing data on disc does not appear to be a major concern. Prices for disc space seem to be on a downward trend and have been for some time. If available disc space becomes a problem, data can be stored off-line on CD-ROM or tape, to be retrieved when needed.

This represents a major expense when data are entered by hand by the BBL, which has been the standard operating procedure. However, it is a minor expense for the BBL if it is done by the bander or others who collect the re-encounter data and the data are submitted electronically to the BBL. The major expense here for the BBL would be to develop standard data-entry and editing programs to be used in the field. Given that the data are submitted to the BBL electronically, there are two additional concerns.

To make the process most efficient for the BBL, banders would accumulate their re-encounter data for a project, season, year, or whatever time interval is best suited to their study and submit the entire data set annually. The bander should be certain that all relevant banding data have been submitted prior to submission of the re-encounter data set. As long as the data for initial bandings have been submitted and the re-encounter data set has been adequately edited by the bander, the re-encounter data can be submitted at any time of year.

Various methods are available for electronic submission, such as disk and remote transmission via ftp or a website. It is crucial that the format in which the data are sent be identical for each person contributing data. This again points to the need for the BBL to develop and supply data- entry and editing programs to those in the field, and the data format and editing program needed for submission for those banders who already have data entry programs.

Editing/correcting data incurs a major cost that is directly proportional to the quantity of data submitted, and most of this should be done by the bander prior to submission. Therefore it is also important that the BBL develop and provide software to the bander that performs the same edit checks that would be done in the BBL, including comparing re-encounter data with that investigator’s banding data. A code could then be attached to a record that would indicate that it had been checked and edited. An investigator should not send any of his/her data to the BBL with unresolved errors. In some cases an investigator could encounter a bird from another study, and therefore the band number would not be in that investigator’s banding file. These data would be submitted also, and edited by the BBL. We anticipate that only a small proportion of the data would fall into this category, although the data in this category could provide important information on movements.

A concern of accepting recapture/resighting data is the potential for a prohibitive increase in the cost to the BBL of processing data. These costs consist largely of staff time devoted to resolving perceived errors or inconsistencies in the data. By requiring all data to be checked and edited before submission, we anticipate that the cost to the BBL per datum submitted would decrease substantially, hopefully enough to offset the increased volume of data submitted. Nevertheless this remains an unknown.

In summary, we anticipate that unless recapture/resighting data are pre-edited by the investigator and submitted in electronic form, the cost of incorporating these data into the BBL database will be prohibitive. Given electronic submission of edited data, the major costs to the BBL would be the initial cost of developing the data entry and editing software for those in the field, in addition to the traditional costs of editing and storage, which hopefully will increase very little.

Recapture/resighting Data vs. Recovery Data: Theoretical Considerations

The reasons for banding birds are varied. They might include interest in movement information, or to identify breeding, wintering, or migration stopover areas. They could also include a desire to estimate the extent of philopatry to these areas, or movement probabilities between them. In addition, there might be a desire to estimate demographic parameters, such as abundance, survival, or recruitment. These parameters might be related to predictor variables that may answer questions related to management, as well as to population or evolutionary ecology.

Band recovery and recapture/resighting statistical models have similar structure and share many assumptions, but differ in the nature of sampling. In recovery models the animal is re-encountered only once, usually dead, after the initial release. In recapture/resighting studies the animal is released after each encounter, barring death due to the capture process. Given this difference in sampling methodology, it is logical that there is much more potential for information from recapture/resighting data than recovery data. There are similarities in the types of biological information that can be obtained using these two types of sampling. Both allow the estimation of survival, and in some cases movement probabilities. However, recovery models (where recoveries come from an extensive geographic area, such as hunting recoveries from throughout North America) allow the estimation of true survival, whereas recapture/resighting models generally only provide estimates of apparent or local survival, the complement of which is death or permanent emigration from the study area. In band recovery models it is assumed that wherever an animal dies, it has some chance of being recovered (i.e., found and reported). However, in many recapture/resighting situations, the area covered by sampling effort is limited, and therefore birds that move outside of that area permanently have no chance of being detected. The survival estimate from recapture/resighting models is therefore biased negatively by a factor equal to the probability of permanent emigration. If this probability is constant from year to year or fluctuates completely randomly, one could still obtain an unbiased estimate of the trend in survival. Conversely, temporary emigration, if it is a completely random process, does not bias the estimator for survival from recapture/resighting models.

The recovery process is not only more extensive spatially, but generally also temporally. Many recapture/resighting studies last only for a few years, after which the encounter probability is zero unless the bird enters a different active study area. Conversely, recoveries of dead birds are usually not based on a formal sampling process, but on either the hunting process or incidental re-encounters.

When the population of interest is well-defined and contained by the sampling effort, and each member has at least a similar chance of being recaptured/resighted in a given time period, abundance, recruitment, and other parameters can be estimated that generally cannot be estimated using recoveries alone. However, recent work has shown that recovery data can be used in conjunction with recapture/resighting data to separately estimate survival and philopatry, and in some circumstances (i.e., when there are multiple sampling occasions within a short period of time) breeding propensity.

The implications of these considerations about use of recapture/resighting and recovery data for the BBL are: (1) for recapture/resighting data to be useful, the user frequently needs to know much more about the circumstances under which the data were collected than he/she would need to know to analyze recovery data; and (2) when these circumstances permit, recapture/resighting data can increase the scientific information extracted from a marked bird considerably. Because of this need for more information about the study design by which the data were collected, we anticipate that the user will look first in the published literature, then to the collector of the data, and therefore the BBL will receive an increase in requests for the identity and address of banders.

Here we present various banding situations in which capture-recapture data can be obtained and offer opinions about the utility of the BBL processing and storing records of such data.

The U.S. Fish and Wildlife Service coordinates the preseason waterfowl banding program through the Office of Migratory Bird Management. Information derived from this program is critical to the promulgation of waterfowl hunting regulations each year. Band recoveries of waterfowl banded preseason and shot by hunters are used to estimate both survival and harvest rates. Implementation of rigorous collection of recapture data at traditional banding sites has the potential of enhancing the quality of information derived from the banding program. Because this banding effort is part of an ongoing program and is conducted in the same locations on the breeding grounds each year, there is potential for considerable information to be extracted from recapture information. Initially, by combining recapture with recovery data, the probability of permanent emigration from the banding area can be estimated. For species where philopatry is virtually complete the addition of recapture data increases the precision of the estimate of survival.

Standardized Multi-site Banding Projects

The Monitoring Avian Productivity and Survival (MAPS) program is an example of this type of program designed to produce recapture/resighting data. Under the MAPS program, banding stations with multiple mist nets are run yearly in a standardized manner across the United States and parts of Canada. Sampling is timed to coincide as closely as possible with the breeding season. Recapture data from this program are ideal candidates for BBL storage, as the anticipated use of the data involves aggregation over multiple study sites within larger areas (e.g., a state, a physiographic region, the continent).

Some goose collar resighting studies also fall in this category. Geese are captured and collared on breeding or wintering grounds. Observers are then assigned to areas throughout the population’s known wintering range. The observer is supposed to cover his/her entire assigned area periodically (e.g., bi-weekly), reading as many neck collars as possible, thus dividing the wintering period into well-defined encounter periods.

There are many research programs in North America that involve special banding and recapture sampling of specific bird populations and communities at a local level. This kind of project is distinguished from the standardized multi-site banding described above, in that objectives are developed by the individual project leader and do not necessarily depend on integration or aggregation of data with results of similar banding/recapture efforts in other locations (i.e., the projects are self-contained). Because each project leader will naturally tailor sampling to meet specific study objectives, it is not nearly so obvious as with standardized multi-site banding that such data should be stored at the BBL. A large fraction of the value of such studies will be extracted by the project leader. Any analysis involving aggregation of such project-oriented research data from different projects will require careful attention to the details of sampling in the different studies, and such details can only be obtained from the individual project leaders. Unless special arrangements have been made with the BBL, individual project leaders will likely retain their own data sets, producing an automatic redundancy if the data are to be stored in the BBL as well. These considerations lead to the conclusion that if logistic considerations restrict the amount or types of data that the BBL can store, data from local, project-oriented research could be excluded from storage in the BBL. Alternatively, some minimal amount of data could be stored (e.g., perhaps 1 recapture/resighting per locality per annual sampling season might be stored centrally). In the latter case, project leaders would view the BBL as a secondary storage facility, and records stored at BBL would not contain all of the information of the primary data set. The reason underlying such use of BBL as a secondary data storage facility would involve the possibility of someone other than the project leader aggregating data from various studies in some large-scale analysis.

The migration banding with which we are familiar involves application of relatively large numbers of bands, with very few recaptures or recoveries in subsequent years. These studies typically are not designed to yield data that can be used to estimate survival probabilities or other demographic parameters. Instead, the data are often used to provide information on the timing of migration, and, in some cases requiring very restrictive assumptions, to index abundance. In these two cases recaptures are used simply to identify individuals that have already been encountered and counted, and therefore maintenance of recapture data by the Bird Banding Laboratory is not required.

In some special sampling situations, capture-recapture/resighting at migration stopover sites can provide data with which (1) stopover times, and even (2) numbers of birds passing through the area, can be estimated. These estimates permit identification of important stopover sites and their characteristics. Such analyses require data at frequent sampling periods (sampling every day or every few days throughout the migration period). The potential for meta-analyses of this sort across many banding stations, and hence the potential value of central storage at BBL, is increased only if all stations are using a standardized methodology (and greater still if other standard ancillary data are collected and centrally stored). Without standard methods, the data are not conducive to standardized storage in that records will be tailored to a specific study design. Without standardization these studies can be regarded as local project-oriented studies.

It is difficult to understand how recapture data from irregular efforts could be very useful in bird population studies. One option would be to not maintain such data in the BBL, instead allowing each bander to keep his/her own recapture data. The other option would be to develop a minimal data record (e.g., the first encounter of each calendar year) for storage of such data in hopes that they might be used in meta-analyses using data aggregated over large areas.

Recapture and resighting data are similar in that the banded bird can be encountered multiple times during a study. It is often easier to resight a bird than to recapture it, and therefore the volume of resighting data tends to be greater. Resighting alone does not allow for unbanded birds to be banded and released.

Identifying a marked bird from a distance by resighting it provides both an advantage and disadvantage. Sample size increases over recaptures because the bird does not actually have to be captured. Conversely, there is more concern for the quality of data resulting from resighting encounters than from recapture encounters, because it is more difficult to read the code inscribed in a collar, tarsus band, nasal saddle, or patagial tag from a distance than to read the band number for a bird in the hand. In some cases there is not a single coded marker but multiple markers on each leg, where the individual is identified by the combination of codes or colors. In some cases the code on a single marker consists of both vertical and horizontal characters. In addition to the difficulty of reading a code from a great distance from a bird that might not remain still, marker colors fade with time and in many cases they fall off the bird at a much higher rate than metal leg bands (in areas of salt water the converse is sometimes true).

Given that a code or color combination is read correctly, there is the possibility that the identifier is not unique. In some cases two different investigators use the same colors/codes for different birds of the same species. The individual in this case should be identifiable from its geographic location. In at least one case an investigator has re-used the same code on a new bird if the code had not been observed in ten years.

For the purposes of storing data, it is crucial that the observed marker be correctly linked back to the leg band number. For investigators who submit their own resightings to the BBL, this link should already have been accomplished before submission. For observations submitted directly to the BBL, the marker should be linked to band number by contacting the bander, or by maintaining a database of the investigator’s codes. We anticipate that this will not be a simple process, due to the variety of types of marks and the potential for error at each step of identification. Given that the leg band number is identified, we see no need to include an extra field on the main re-encounter record for the auxiliary marker code that was sighted, especially given the difficulty of standardizing this in the face of multiple marking methods and coding schemes. However, this information could be kept in another part of the relational database. In order that users can judge the quality of these observations, we recommend that resighting data stored with the BBL include a field that classifies the sighting with respect to the circumstances (e. g., observation by professional observer as part of sampling scheme, incidental observation by professional biologist, incidental observation by average citizen , etc.), and another field that identifies which marker or combination of markers (e.g., BBL leg band only, special band, neck collar, combination of partial reading of BBL and special band, etc.) was used for identification. This could be accomplished through the current ‘How Obtained’, ‘Who Reported’, and ‘Why Reported’ codes.

As with recapture data, design considerations are very important to the interpretation of resighting data. For example, it is important to know whether or not certain specified sampling areas are covered periodically by observers, so information about the temporal and geographic characteristics of regular sampling is important ancillary information.

Every encounter that is stored requires storage space, editing, and error verification. Some birds are encountered daily, or even multiple times in a given day. Should every re-encounter for a given band in a given location be stored? In the case of designed studies, this decision should be based on study objectives and the corresponding sampling design. For estimation of annual survival rate using classical capture-recapture models, one need only know whether or not a band was re-encountered at any time during that year’s sampling period. In some cases investigators want to partition annual survival into smaller intervals (e.g., every 3 months). Multiple sampling periods per year are required in such instances, and the required data are whether or not a marked individual is seen during each sampling period. We anticipate that some bird studies are, or will be, conducted using Pollock’s robust design, where each primary period of interest consists of multiple "secondary" samples, taken over a short period of time (sometimes daily). In this case detailed capture information (individual presence or absence) for each of these samples would be important. However, the user would need to know even more detail about the sampling scheme to make proper use of this detailed information.

It is clear that recapture/resighting data will not be useful for analysis without extensive knowledge of the sampling design (to a much greater extent than with recovery data) and other circumstances under which the data were collected. Therefore any investigator willing to submit these data to the BBL must also be willing to provide information on how they were collected. The user can be pointed to this information in any of three ways. (1) If the study has been published, the methods section of that publication should provide this type of information (the BBL could even request that if a publication resulted from the study that an electronic file containing the manuscript text be sent to the BBL) . (2) The BBL could maintain a database of study method descriptions that is linked to a project code in the re-encounter database. (3) The name and address of the investigator can be provided to the user. In any case, another code will be necessary on the re-encounter record to identify the study from which the data came. A banding permit number will not be sufficient, given that multiple banders could be included on one permit, and each bander could be involved with multiple projects. Therefore a unique project number would be necessary.

The aforementioned database of studies and investigators could also include those studies where the investigator does not want to submit his/her data to a central database but is willing to consider making data available to interested parties. In this way the BBL would expand its role as a clearinghouse for data on marked birds. This would also apply in the other cases where the BBL does not store data from studies due to lack of standardization, etc.

We were asked to consider whether all of the fields contained in the re-encounter record are still pertinent to contemporary uses of the data. Three fields came into question: HSS, batch number, and cycle number.

HSS, or hunting seasons survived, is used in the analysis of game bird data as a way to determine how long the bird lived without having to compute it directly. It is also used as a tool for a quick search of the data base for direct recoveries (i.e., those birds shot in the hunting season immediately following their banding). This field does not actually exist in the active BBL data base, and therefore takes up no storage. Instead it is computed and added to the file of re-encounters generated from a data request and sent to the user. Therefore the question is not one of storage, but whether the BBL should continue to supply this field to the user. We have no recommendation for the BBL on this issue, other than to point out that the user could compute HSS from the rest of the information on the re-encounter record.

Batch number and cycle number are used by BBL staff to locate the original paper record of a re-encounter. Currently the batch number is the Julian date on which the re-encounter was reported. For re-encounters reported by telephone, this code provides a check on the date of encounter reported by the caller. In addition, we anticipate a need to continue to access original paper re-encounter records, to check inconsistencies in data. Therefore we believe batch number and cycle number need to be retained in the BBL active database for internal purposes. However, we see no reason that they would need to be included in the re-encounter file provided to external users.

It is our understanding that there are a large number of paper records of re-encounters that are waiting to be archived. These records provide a means to resolve discrepancies that arise in re-encounter data (e.g., a band number is reported for a bird already reported to be dead). Therefore, in the interest of maintaining unbiased data it is important to maintain these records for a reasonable period of time. This period could be species-specific, depending on the maximum life span of the bird. Given that the technology exists to scan these records onto compact disks, we recommend that this be done.

Finally, we recommend that, in addition to the fields already stored, the following information be included in the database and resulting archives: name and address of those that report re-encounters, the species reported, and any auxiliary marker information (e.g., code on neck collar or color marker combination/description). The resolution of data discrepancies can be greatly aided by having access to this information.

The Task Force concurs with the Implementation Team that the BBL should store recapture/resighting data in some circumstances. Submission of these data to the BBL should be done on a voluntary basis, as is the case with recoveries. We summarize our recommendations with respect to each of the specific charges we were given below.

(1) Theoretical considerations - recapture/resighting data potentially provides much more information about demographic parameters than recoveries, due to the potential for more than one re-encounter with each marked bird. State-of-the art statistical methodology exists to exploit these multiple re-encounters. Nevertheless these methods require more information about the circumstances of sampling than do band recovery models, in order to properly interpret the resulting estimates .

(2) Banding scenarios - Our criteria here were not based on the usefulness of data to the investigator running each study, but to other users who would access the data from the BBL (e.g., for meta-analysis). Generalizing based on the amount of standardization and potential amount of re-encounter data, we rank the banding scenarios in descending order of expected usefulness: preseason waterfowl banding, standardized multi-site banding (e.g., MAPS or selected goose neck-collar studies), local project-oriented research banding, migration banding, and irregular banding. However, we could think of several scenarios where this ranking would not hold. The conclusion is that each project or study must be evaluated on its own merits and a description of its methods must be made available to the user. Therefore we cannot make sweeping recommendations that the BBL accept all preseason waterfowl banding but no irregular banding. Nevertheless all re-encounters would have to be classified as belonging to some ‘project’.

(3) How to link recapture and resighting data to banding records - the current re-encounter record includes a ‘How obtained’ field that can be used to identify the re-encounter as a recapture or resighting. It can thus be linked to the banding record using the band number. For resightings the observed mark (e.g., neck collar) must be linked to the band number using information supplied by the investigator (i.e., either direct communication with him/her or a database that includes investigators, studies within investigator, codes within study). This type of system is certainly feasible using a relational database.

(4) Evaluate the current encounter record with respect to recaptures/resightings - the fields in the current re-encounter record are appropriate, and could be further exploited to determine the type of re-encounter (through the ‘How Obtained’ field) and the nature of the report (e.g., re-encountered as part of a sampling scheme, incidental observation by professional biologist, incidental observation by general public), through the ‘Who Reported’ and ‘Why Reported’ fields. The key additional field would be a ‘project number’ that links the observation to a specific investigator and study. Another part of the database would contain the name of the investigator and/or a detailed description of the study (i.e., sampling scheme).

(5) Archiving encounter records - re-encounter records, including the name and address of the person that reported the bird, the species, and any auxiliary marker information, should be archived. The archive is useful for resolving data discrepancies, and should be maintained for as long as one would expect discrepancies to arise (i.e., at least as long as the maximum life span of the bird).

Given that the BBL should store at least some recapture/resighting data, there are two possibilities. The first is to screen various studies to determine whether the data are ‘worthy’ of central storage. This worth could be based on study design, past work of the investigator, whether the data are maintained by another organization, the likelihood that the data will be lost to retirement or death of the investigator, etc. This type of valuation would require a considerable amount of staff time by a biologist and supporting committee. The alternative is to accept data from all studies and provide sufficient information to the user to determine the value of the data (i.e., access to the details of the study protocol). With either approach, given that the data for a study are acceptable to the BBL, a decision would have to be made about how many observations per bird per geographic area per unit time to accept, and whether to accept only re-encounters reported by participants in the study or to include incidental observations by others.

We suggest that the BBL begin the process slowly, focusing on one game bird project (e.g., preseason waterfowl banding or goose neck collar study) and one standardized banding scheme for non-game birds (e.g., the MAPS program), and accept all data from these projects (limiting it to one re-encounter per day). In this way the system can be tested in all respects (utility of the data entry program, the need for error checking when the data reach the BBL, etc.). If the problems that arise are such that BBL staff would be overwhelmed, then restrictions could be imposed by restricting data submission to a coarser temporal scale, limiting the type of project that would be accepted, or both.

We suggest that the BBL concurrently build a database of recapture/resighting projects, the principal investigators, and project descriptions, to be provided to interested users regardless of whether the data from that project are stored by the BBL. This is another important and simpler way in which the BBL can act as a clearinghouse of information for those interested in meta-analysis of these data.