Handbook for Atlasing
North American Breeding BirdsEdited by Charles R. Smith, Published September 1990
Atlasing Handbook contents page
Survey Methods and Mapping Grids
Chandler S. Robbins and Paul H. Geissler
Maryland Atlas Project and Patuxent Wildlife Research Center
Sampling Design
One of the major problems facing most atlas organizers is a shortage of qualified field observers. This requires making difficult decisions as to the beat compromise between two extremes: dividing the area into a small number of large blocks, many of which will be poorly covered; or selecting a pattern of small blocks for intensive coverage, while leaving moat of the area uncovered. In general, a plan for good coverage of a relatively small percentage of the total area is preferable to poor coverage of large blocks.
Atlas organizers should carefully consider their objectives, and then design a sampling plan that will meet those objectives. If the main objective is to generate species range maps, regular coverage by systematic sampling is effective. If an objective is to estimate frequency of occurrence (percentage of blocks) for each species or to detect future changes, stratified random sampling should be used to permit greater sampling density of rare or disappearing habitats.
Several options for selecting those blocks to be covered have been used successfully. Here are some examples:
Total Coverage. Standard procedure in western Europe, the Maritime Provinces, and several northeastern states. Examples: Maritime Provinces, Maine, Massachusetts, Rhode Island, Connecticut, New York, Pennsylvania, Maryland, Delaware.
Two-level Sampling. Coverage of all blocks in populated areas and of portions of much larger blocks in roadless areas. Examples: Australia and Ontario.
Systematic Sampling. This is widely used in the United States for convenience: the Priority Block is in the same position on every quadrangle; hence, every field worker knows which blocks are to be covered. Theoretically, the position should be determined at random, then applied throughout (example: Colorado). In practice, moat atlas committees have selected the priority position for practical reasons: some chose one of the two center positions (center west or center east), because block boundaries and access are easier to determine when 3 of the 4 adjacent blocks are on the same map; some atlas committees select the lower right block, because the adjacent margin shows the scale of miles and the name of the quadrangle. Examples: Virginia, West Virginia, Illinois.
Random Sampling. Random selection of one Priority Block within each sampling unit (7.5-min. quadrangle or township). Other, non-random, blocks may be designated for coverage for special biological or political reasons, but those are not used in statistical comparisons. Random sampling is a good statistical procedure, but has not been used in atlas studies for two reasons: (1) Because the sample is not evenly distributed, the maps would show holes in coverage, and some of these gaps would fall at limits of species distribution; and (2) Field workers who travel to other parts of the state or province would not know which blocks other than their own were designated for coverage.
Modified Random Sampling. Similar to the above, but with the stipulation that if a selected block shares a aide with a block already selected, it is rejected and a replacement block is picked. Examples: Vermont and New Hampshire. This procedure prevents clusters of Priority Blocks surrounded by large unsampled areas, and so largely solves the first problem under Random Sampling. From the statistical point of view, however, it introduces a new problem in that all blocks do not have equal opportunity of being selected; selection of each block automatically eliminates the possibility of four adjacent blocks being sampled. For moat practical purposes this makes no difference, but if atlas results were subsequently to be used in litigation they might be challenged on statistical grounds.
Stratified Random Sampling. This permits sampling by ecological regions, and the designation of different sampling densities among ecological (or political) regions. For example, one might wish to sample half the blocks containing elevations above 4,000 feet, or all blocks containing tidal wetlands, but only a random one-sixth sample of blocks with less than a 300 foot difference in elevation. This would concentrate field work where it would be moat effective. Another example: South Dakota is using the township as their basic unit, and their atlas blocks are one-quarter of a township, or 3 x 3 miles. Townships are grouped into 62 strata (called "superblocks") of 33-39 townships each, and these superblocks are grouped into 17 ecological regions. Two "Standard Blocks. have been selected at random in each superblock. These 124 random blocks, representing about 1 percent of the state, will be used for statistical analysis. In addition, "Special Blocks" have been designated to include unique habitats, and suburban blocks where habitat change is likely to occur; also, "Managed Blocks. have been established on various State and Federally managed lands.
Sampling Recommendations
1. When there are too few observers to visit all blocks, sampling is strongly recommended (rather than using a grid of larger blocks). This is primarily because with a coarse grid it may not be possible to detect future changes in bird distribution. The larger the grid, the greater the chance that some remnants of a habitat (and all of its typical breeding bird species) may still remain after a period of years.
2. Systematic sampling is satisfactory in most instances. A systematic sample may be supplemented with a stratified random sample of blocks containing rare or restricted habitats. However, any special blocks selected to sample specific sites must be excluded from statistical summaries.
3. In jurisdictions that are ecologically diverse or contain extensive inaccessible areas, stratified random sampling should be considered. A classical stratified random sampling design (scheaffer, et al. 1979, Chapter 25 of Steel and Torrie 1980, or Chapter 21 of Snedecor and Cochran 1980) for detecting changes in bird ranges would start with an inventory of the major habitat types in the state or province and an assessment of their abundance, their relative importance to birds, and their vulnerability to development. Next, the state or province would be stratified, assigning each block to one and only one stratum. Decide how many blocks to select from each stratum, baaed on the importance of the stratum. Then select a random sample from each stratum, including 100 percent samples of strata containing rare or critical habitat. The classical design would not provide an ideal arrangement of spots for mapping, and it might be too complex for some applications. An alternate approach would be to use a systematic sample to provide a base for mapping, and supplement this with a random sample of critical ecological areas, stratified by habitat. The two surveys would be analyzed separately, but any blocks from the systematic survey that fell into the critical habitat strata could also be used for the latter because they were randomly selected.
Mapping Grids
The two items for consideration are: (1) the reference base, and (2) the grid size. Ideally, the same reference base should be used world-wide. There are three major problems, however, that have prevented the global standard, the Universal Transverse Mercator (UTM) grid, from being widely adopted either in North America or elsewhere in the world: (1) The UTM grid has major interruptions with wedge-shaped blocks of various adzes every 6 degrees of longitude, (2) the UTM grid does not appear on most detailed U.S. maps (such as standard topographic maps and county maps), and (3) our major topographic maps have latitude/longitude boundaries, so even when the UTM grid is indicated in the margins, as it is in the newest maps, the observers or the atlas organizers must piece together several maps to make up a given atlas block instead of using a single readily available map. Therefore, in moat countries the atlas organizers have used their national grid or one based on latitude/longitude rather than struggling with UTM.
Grid adze for moat atlases is now well standardized as either a multiple of 5 km, a close English equivalent, or a subdivision of a degree of latitude and longitude (L/L). For comparative purposes, a summary of atlas grids in use is given in Table 1, arranged from largest to smallest.
Table 1. Summary of grids that have been used in Atlas work. Grid is expressed as length of one aide unless otherwise specified. Grid base is Universal Transverse Mercator (UTM), latitude and longitude (L/L), a national grid, or a quarter township (3 x 3 ml).
Grid Size (and Base) Where Used 100 km Northern Ontario, Northwest Territories 1° latilong (L/L) Montana, Wyoming, Utah, Australia (out-back) 1/2° (L/L) Madagascar, Tanzania, Uganda, Turkey 50 km (UTM) Continental atlas of Europe 50 km (UTM) West Germany (later re-atlased with 25 km) 20 x 32 km (nat'l) Portugal 20 x 27 km (nat'l) France 18.5 x 28 km (nat'l) Spain 7.5 min. (L/L) Maine, Texas 10 km (national) Britain and Ireland, Switzerland, Finland, Czechoslovakia, Italy 10 km (UTM) Southern Ontario, Maritime Provinces, Poland 10 min. (L/L) Tasmania, Australia (inhabited regions) 10,000 yd New Zealand 3 x 5 min. (L/L) Lower Saxony (Germany) 5 km New York, Denmark, Netherlands, Sweden About 5 km (L/L) Massachusetts, Connecticut, Rhode Island,New Jersey, Pennsylvania, Maryland, Delaware, Florida, Missouri Random or systematic Vermont, Now Hampshire, Ohio, Virginia, West sample, 5 km L/L Virginia, Illinois, Indiana, Tennessee, Nebraska, Colorado 3 x 3 mi (=4.8km) Michigan, Iowa, South Dakota, North Carolina 2.5 km l/4 blk (L/L) DC, 1/6 of Maryland plus 6 Maryland counties 2 km (nat'l) several British counties 1 km Berlin, Canton of Geneva (Switzerland) Grid Base Recommendations
It does not seem practical to insist that any one grid base must be used in North American (or New World) atlases. Therefore, the NORAC Working Group on Survey Methods and Mapping Grids recommends that atlas organizers give careful consideration to the options that are available regarding detailed maps of their state/province/department, especially considering grids used in adjoining jurisdictions, and maps that are most likely to be available 20 or 30 years later when the next atlas survey may be conducted. Computer programs are available for converting atlas records from UTM to lat/long grids and vice versa, so the choice of a grid base is largely a matter of convenience, not a matter of being either right or wrong.
We strongly recommend the 1:24,000 7.5-minute topographic maps of the U. S. Geological Survey (USGS), because they can be obtained and understood by anyone in the world and will be in archives for centurion in the future. Orthophoto maps (USGS), which are becoming available at the same scale, are recommended because they show the key habitat features. For Canada, excellent topographic maps are available in several scales from the Department of Energy, Mines, and Resources, Ottawa. For Mexico we recommend the 1x2-degree 1:250,000 Carta uso del suelo y vegetation, available from Instituto Nacional de Estadistica Geographica e Informatica, Mexico City; these Mexican maps have a 10 km metric grid.
Grid Size
There is a tendency among atlas organizers to think of grid size as being a function of only the size of the state/province/country and the number of prospective observers. More realistically, grid adze should depend on the purpose(a) of the atlas.
If the main purpose is to show approximate distribution of breeding species in an area where distribution is poorly known, a large poorly covered grid will produce valuable information, especially if actual localities are recorded for rarities or birds at their elevational or geographical limits.
If, on the other hand, a major purpose is to show habitat relationships or to establish baseline data for documenting future changes, it is important to select a grid small enough to accomplish these purposes. A 10-km grid is marginal, 5-km is better, and intensive random or systematic sampling by means of a smaller grid such as 2.5 km quarter-blocks, 2 km 'tetrads' or 1 km squares will give much more satisfactory results. The smaller grids are especially recommended for areas that are likely to be altered significantly prior to the next atlas project.
Grid Size Recommendations
- Canadian provinces—10 km
- U. S. States—5 km
- Latin America—10 km
- Lesser Antilles—5 km
- County or local—2.5 km
Options to consider:
- Stratified random, or systematic sampling when not all blocks can be covered.
- Lower density of sampling in large areas of continuous uniform habitat.
- Designation of certain priority blocks (e.g., 1 out of 6 or 1 out of 100) for intensive coverage with a finer grid, to improve capability of detecting future changes.
- Designation of certain counties for intensive coverage with a finer grid (depending on need to detect changes, and on available personnel).
Alternatives:
Optimum grid size may be dictated by availability of suitable maps or by road patterns (e.g., moat states use 1/6 of a 7.5-mint topographic map published by the U.S. Geological Survey because these are readily available for the entire country and include green overprint for forest, etc., as well as topography); the area of 1/6 of 7.5 min. decreases from south to north, but is very close to 5 km in the middle latitudes of the country. Many states have a road system based on township lines, with roads at 1-mile intervals; some of these states (Michigan, Iowa, South Dakota) are using a quarter-township, which is a 3-mile (4.8-km) grid, for convenience in the field.
Some large tropical and semitropical countries have opted to use a 1/2-degree block of latitude and longitude.
Other considerations:
Check with neighboring jurisdictions to determine what grid they are using, and why. One of the advantages in using the same grid as your neighbors is that information can be shared from blocks along the border.
Survey Methods
Observers:
Trained observers are fundamental to the success of an Atlas project. They are always a limiting resource, so special efforts must be made to solicit and train observers. Some atlas organizers have placed too much emphasis on coverage by a small number of experts, and have given too little attention to the continued training of interested volunteers. Interested beginners can contribute a great deal to an atlas project, and the special efforts required to encourage their training and participation have produced good results. Atlas organizers should recognize that the project itself presents an excellent opportunity for training and developing interested beginners into good field workers, who will go on to become leaders themselves in other survey and conservation work.
Most atlas organizers educate their observers by means of atlas handbooks, training sessions, and newsletters. These concentrate on identification, safe dates, courtesy toward landowners, and finding difficult species. More emphasis needs to be given to improving observer efficiency in the field. It is helpful to hold annual regional training sessions with participants.
Observers can be stimulated through incentive awards such as certificates, plaques, arm patches, pine, or banners, or just through recognition in newspaper or television coverage.
Training Sessions:
Any steps that can be taken to improve the efficiency of observers will help them bring their home block up to satisfactory coverage in a shorter time and stimulate them to tackle additional blocks. We suggest the following topics for training sessions:
1. Purposes of the project.
2. Responsibilities of the participants; goal for each block.
3. A quick course in map reading, so observers will be able to locate unique habitats such as marshes, or ponds that may not be visible from a road. Where possible, use the new Orthophoto (Topographic) 7.5 minute series (Geological Survey), USDI. There is a quantity discount, so if you are sending a small order, consider combining your order with that from another state.
4. Discussion of importance of habitat in limiting the distribution of each species; examples of habitat requirements.
5. Resources, such as bird song records and tapes, Stokes' (1979, 1983, 1989) guides to Bird Behavior, guide books for bird nests (Harrison, 1975, 1979), Ehrlich, Dobkin, and Wheye's (1988) guide to bird natural history, and local field lists showing dates of occurrence and nesting, and habitat requirements. If your a/ate/province does not have such a local field fiat, use one from a neighboring area.
6. Atlas codes, with examples of their use, and other information from Atlas Handbook.
7. How to fill out the field card and summary sheet; explanation of forms cosigned for computer entry; what not to do.
8. Identification and habitat hints. Species with similar appearance, songs, or habitat Use colored slides if available (or consider use of study skins if a small group).
9. Discussion of bird behavior by groups of birds. Misleading code possibilities. Pair formation (year round, during migration); singing during migration; feeding far from the neat (gulls, herons, vultures); mobility of fledglings (gulls, ducks). Importance of safe dates.
10. Instructions about Rare Bird Documentation forma and Neat Cards.
11. Cautions about neat disturbance. Predation rates are high, especially for nests on and near the ground. Avoid beating a trail directly to a neat; walk by at a safe distance. Preferably use a mirror on the end of a pole.
12. Getting permissions (don't trespass); report findings to landowners.
Field Methods:
Always encourage feedback and sharing of experiences. Nothing takes the place of field experience. Even seasoned observers can benefit from field tripe on which an experienced atlaser interprets behavior and tells what codes to use. Time is of the essence! Although many observers take their birding very casually (and some will not change their habits), they should be encouraged to work efficiently, if not vigorously. Their help can always be used in another block in the future, after they obtain satisfactory coverage of their first assignment. The following suggestions are based on experiences of active atlasers in the Middle Atlantic States.
Preliminaries:
1. Become familiar with your block before the atlas season begins.
2. Study topographic maps; find streams, ponds, marshes, large woodlots, access trails.
3. Contact important land owners. Each county Assessors Office has tax maps showing ownership of all lands in the county. A single permission may provide access to large tracts of interesting habitat.
4. When contacting land owners for permission, ask if they have seen quail, grouse, turkey, or pheasant nests, woodcock or hummingbirds, whether pigeons, phoebes, swallows, or Barn Owls neat in their barn, and whether they hear other owls at night. You may wish to verify their answers if you have any doubts. Knowledgeable farmers can save an enormous amount of time finding and confirming difficult species.
Principles:
1. Concentrate on early morning coverage when the birds are moat active.
2. Sample all habitats; topographic maps are essential for locating off-road habitats.
3. Space tripe throughout the season. If one waits until all species are within safe dates, the activity peak for woodcock, owls, and many other birds will have passed. Confirmation of late nesting species such as waxwings and goldfinches may be missed unless there is some late-season effort.
4. Some experienced observers prefer to begin their field work by laying out a "Miniroute" of 15 (or more) 3-minute roadside stops along secondary roads that have little traffic; they set the stops about one-half mile apart at strategic locations such as habitat boundaries and stream crossings so as to maximize the number of apecioa that may be encountered. (Less experienced birders cannot be expected to contribute to a highly structured study of comparative abundance based on Miniroutes, but they can use the technique effectively for Atlasing if they ignore the 3-minute time restriction). Keep a separate tally of birds at each of these numbered stops. Go back a week later, observing from these same stops; this will automatically upgrade many species from Possible to Probable status (as soon as the spring migration is over and the various species are within the summer "safe dates"). Timed stops can also provide valuable information on relative abundance. Return to these same stops later in the summer (not restricting subsequent visits to 3 minutes) to upgrade additional species that were either missed or "unsafe" on one of the previous visits. It helps to highlight ahead of time those species that you hope to upgrade at each stop. After completion of Miniroute coverage, visit other portions of the block to search for other habitats and for additional species.
Extra Effort:
1. By the second year (or the first year if one gets organized early enough), concentrate on early spring birds such as owls and woodcock that may be much harder to find later in the season. Nocturnal and crepuscular species require special effort. Early mornings average better than evenings in areas where traffic or frog noise is a problem. A moonlit night is generally more productive than a dark or rainy night. Use tapes if necessary to stimulate response; playing tapes once or twice at a given location will cause a minimum of disturbance to the birds.
2. Although roadside coverage may give the moat rapid initial results, off-road coverage is essential for access to other habitats. Railroad rights-of-way, power lines, pipe lines, and sewer lines should be considered as means of easy access to off-road habitats. Be sure to obtain permission, especially when lands are posted.
3. Work as efficiently as possible during the first years of the project. Every atlas project has a near crisis in its final year because observers have neglected to keep on schedule, or have moved out of the area. By the final year, new residential development and shopping centers have wiped out nesting areas where some species would have been easy to find and confirm in earlier years.
Assessing Adequacy of Coverage
Setting Total Species Goals:
It is taken for granted that observers must be given goals toward which to work. Goals are hard to define, however, in a large diverse state or province with varied landscape. It might be easy to find 80 species in some blocks, but impossible to find as many as 50 in others.
Adequacy of coverage has been defined in various atlas projects as finding 75% to 90% of the species suspected of being present in a block and confirming 50% or more of these species. The 75% of species believed present has been generally accepted in most North American projects. This 75% rule has been criticized by Kibbe ( 1986 ) as being inadequate for future comparisons, and ho correctly shows this to be true when only a single block is compared; but when hundreds of blocks are compared the problem disappears.
The coverage goals represent a compromise with respect to the allocation of limited resources. For example, from experimental work in England, Sharrock ( 1973 ) found that in 400 hours an experienced observer could obtain atlas data for either 63% of the species in 80 10-km blocks, 75% of the species in 40 blocks, 87% of the species in 25 blocks, 92% of the species in 4 blocks, or 94% of the species in 2 blocks.
The problem with setting goals too low, such as 63% of the expected species, is that only the more familiar, most vocal, and most widespread species may be recorded, while rare, local, and habitat-specific species will be missed. On the other hand, setting goals too high may discourage some observers, and will seriously limit the number of blocks that can be visited. Whatever the specified goal, it is important to point out that regional differences exist and that blocks in urban areas, blocks with extensive agriculture, and blocks with little diversity of habitat will have small species totals. It should also be specified that blocks with extensive wetlands or a rich variety of habitats may have well in excess of the regional goal.
Most North American atlas organizers agree that 75% of the expected species is a legitimate goal, but that if the number of expected species is later found to have been underestimated the goal needs to be raised accordingly.
Estimating the Number of Expected Species:
In some states preliminary field testing is done in the year prior to launching the atlas. This gives an excellent opportunity to test adequacy of estimates of expected species in various environments. When the luxury of actual field experience is not available, data from nearby states or provinces may prove helpful.
Experienced field ornithologists can usually make fairly reliable estimates with the use of topographic maps. After the first full year of field work, estimates can be revised as necessary. Raynor (1982) described 0a method of estimating the number of species in each Long Island block based on his personal knowledge of the birds and the habitats present and on the fiats submitted by the observers.
A method that proved very helpful in Maryland after the first year of field work wee providing county coordinators and observers with a computer-generated fiat of expected species that were missing from each block. This fiat was obtained by comparing species lists from the eight surrounding blocks with the fiat for each target block, and printing the names of those species not listed for the target block, together with the number of surrounding blocks in which they had been detected. Such a fiat, when added to the fiat for the target block, and adjusted for species whose habitat is completely lacking from the target block, gives a good estimate of the potential for the target block.
Confirmation Percentage:
In western Europe, where habitats are more open and birds easier to confirm, it wee not unusual to have 70% or more of the atlas records confirmations. In Switzerland (Schifferli, et al., 1980), for example, the average number of species per block was 86, and 76% of the records were confirmations. In North America, many eastern states and provinces began with a goal of 50% confirmations. Some states later dropped this goal in favor of better species counts; their emphasis became directed toward more Probable or Confirmed records rather than just confirmations. Confirmation of rare species remains a high priority. Everyone agrees that confirmations are important and that they should continue to be emphasized on the atlas maps. A confirmation rate of 50% is still a desirable goal, but should not be a prerequisite if it will jeopardize adequate coverage of all habitats.
Judging Adequacy of Reports from Individual Blocks:
This is beat done by comparison with results from neighboring blocks that have a similar mix of habitats. Some observers may have hearing problems and miss birds with high pitched calls, some may have neglected wetland habitats or failed to include nocturnal coverage. More often, songs of some species may not have been recognized. Failure to detect a species is generally a greater problem than misidentification.
Hours Afield as a Measure of Adequacy:
It is dangerous to rely too heavily on total hours spent in an atlas block because so much depends on the capability of the observer, the time of day, the distribution of effort through the season, the accessibility of habitats, and even weather conditions. The Ontario goal of 16 hours per square was later changed to 75% of the expected species.
In Spain, de Juana (1980) spent 12 hours per 11-km block, finding 70% of the species present, but confirming very few. In Switzerland (Schifferli, et al., 1980), it took about 9 days to find 80-90% of the species in a 10-km block and confirm 80% of them.
For experienced observers, 16 hours proved adequate in New York (Carroll 1986); they hired 2-person teams to blockbust by spending one day in each block, camping there at night.
Numerical Treatment
In the early years of atlasing, much discussion wee devoted to various means of obtaining meaningful estimates of abundance. Trained professionals are often willing to participate in an activity involving standardized counts, but the average atlaser is not. The options are:
1. To make no attempt at numerical estimates;
2. To compute grid frequencies, the proportion of blocks in which a species wee found, preferably done by physiographic regions as in Laughlin and Kibbe (1985);
3. To urge only those participants who feel comfortable about making estimates to do so (through either standardized surveys of relative abundance, counts of birds actually observed while atlasing, or estimates of the number of pairs nesting within an entire atlas block); see Adams (1986);
4. To rely on the regional coordinators to make estimates for their regions;
5. To have a special committee make estimates of abundance for all species on the basis of whatever information is available to them (primarily published sources);
6. To use existing Breeding Bird Survey data to show the mean number of birds detected per 50 3-minute roadside stops. This information is available for the U.S.A. from the Office of Migratory Bird Management at the Patuxent Wildlife Research Center, Laurel, Md. 20708, and for Canada from the National Wildlife Research Center, Canadian Wildlife Service, Environment Canada, Ottawa, Ontario K1A OH3);
7. To supplement BBS data with additional routes in order to increase the data base, especially for rare or local species, or to provide data from areas not adequately sampled by the BBS; and
8. To establish a special network of standardized counts, such as Miniroutes, to be run by hand-picked experts.
The greatest value of numerical estimates is to serve as a basis for future comparisons. Consider carefully whether future changes in bird population will be detectable by the method selected. A population index that is repeatable will probably be preferable to estimates of actual abundance, unless the latter are made in a highly standardized fashion.
Recommendations for numerical Treatment
1. At a very minimum, State/Province means from BBS should be used as indices of abundance for any U.S. or Canadian Atlas (except in the Arctic where not available). For common species you might also consider separate means for the physiographic regions within the state or province. In requesting these (from Breeding Bird Survey, Patuxent Wildlife Research Center, Laurel, MD 20708), you must specify the years you want included in the means. If the State/Province is well saturated with BBS routes, it would be preferable to use means based only on the Atlas years; otherwise, you might select a longer span of years.
2. If (willing) observers or coordinators are asked to provide estimates of total number of pairs of each species in their blocks, we recommend the estimates be in powers of 10 (1, 2-10, 11-100, etc.) as used successfully in the French and Ontario Atlases and proposed for the atlas of all of Europe. A simple check to determine whether such estimates are in the right ballpark is to consider that if the BBS means were based on 100% efficiency of all observers in detecting all birds within 1/4 mile of each of the 50 stops on every route in every year, the mean would represent the number of birds per 25.42 km2. In other words, the area theoretically sampled by a 50-stop BBS route is essentially the size of a 5-km Atlas block (25 km2). It is safe to say that BBS efficiency never reaches 100%. For conspicuous open country birds it might reach 50%; for moat species, and especially those in habitats poorly sampled by the BBS, it would be under 10%. Possibly estimates for some species in certain habitats may be available within the next decade.
3. Another option is to have observers time all their early morning counts. If a efficiently large number of timed counts are available from experienced atlasers, the number of birds per hour can be calibrated, species by species, and adjusted to a standard observation time (e.g., 6-hour day) as was done successfully in the British Winter Atlas (Lack 1986).
4. Systematic Miniroute (Bystrak 1980) coverage by experienced participants such as BBS observers is recommended as an ideal way to obtain a large sample of data on relative abundance. Either actual counts or frequency (percentage of stops) can be used as an index (Robbins and Dowell 1986).
Monitoring Bird Populations with Atlas Data
An important potential for breeding bird atlases is to monitor long-term changes in populations of rare species. Robbins et al. (1989) noted that standardized roadside surveys such as the Breeding Bird Survey (BBS) poorly sample nocturnal and rare species and those found in habitats such as mountain tops and marshlands that tend to be avoided by roads. They considered the use of breeding bird atlases to supplement the BBS. Unless atlases are designed for monitoring bird populations, it is not valid to use them for this purpose. Techniques to monitor special groups such as raptors have been developed (e.g., Fuller and Mosher 1987).
To monitor rare species and those found in habitats poorly sampled from roadsides, atlases must have a random or systematic sample of these species' habitats. It is likely that many of these habitats are rare and thus poorly sampled by random or systematic samples of blocks selected from the whole state or province.
Arbitrarily selected "Special Blocks" that may be added to include these habitats should not be used for monitoring, because one cannot know whether population changes in these hand-picked blocks are typical for the state or province. Instead, it is necessary to list all the blocks that contain each special habitat of interest and then draw a random sample from each habitat (stratum), or visit a 100 percent sample of the rarest habitats. For the rarest species it would be highly desirable to document numerical abundance and to record time and methodology used in counting these species so the effort could be duplicated in the future.
It is also possible to estimate the number of blocks in which a species was present but not found (Geissler and Fuller 1986, White et al. 1982). This method requires observers to make a separate list of the species found on each trip to their blocks, by habitat (excluding tripe made explicitly to look for particular species). Statistical methods can then be used to estimate the probability of finding each species that is present.
In addition, for monitoring even the more common species it is necessary to standardize procedures so that effort can be matched, or at least compared, in the future. The number of blocks in which a species is recorded cannot in itself be used to monitor change, unless the results can be adjusted for differences in effort. Not having an accurate record of hours of coverage, Robbins et al. (1989) used the ratio of total volume of atlas quarter-block records in Howard County, Maryland, for 1973-75 and 1983-87 to compensate for differences in coverage during the two periods. They were able to show that dramatic changes in bird distribution had taken place in a single decade.
For species common enough to be monitored along roadsides, a network of timed Miniroute stops that can be repeated in the future either during a future atlas project or as a special monitoring program would be ideal. Another method would be intensive ("saturation") coverage of a random sample of blocks, with a special effort made to search all habitats where the species of interest might occur.
Literature Cited
Adams, R. 1986. Relative abundance and the Michigan Breeding Bird Atlas. Pp. 41-43 in Sutcliffe, S.M., R.E. Bonney, Jr., and J.D. Lowe. Proc. Second Northeastern Breeding Bird Atlas Conf. Lab. of Ornithology, Ithaca, N.Y. 192 pp.
Bystrak, D. 1980. Application of Miniroutes to bird population studies. Maryland Birdlife 36:131-138.
Carroll, J.R. 1986. Adequate coverage: New York Breeding Bird Atlas. Pp. 16-17 in Sutcliffe, S.M., R.E. Bonney, Jr., and J.D. Lowe, Proc. Second Northeastern Breeding Bird Atlas Conf., Lab. of Ornithology, Ithaca, NY.
Ehrlich, P.R., D.S. Dobkin, and D. Wheye. 1988. The birder's handbook. Simon & Schuster, New York. 785 pp.
Fuller, M.R., and J.A. Mosher. 1987. Raptor survey techniques. Pp. 39- 65 in B.A. Geron Pendleton, B.A. Millsap, K.W. Cline, and D.M. Bird, eds. Raptor Management Techniques Manual. Nat'l Wildlife Fed., Washington, D. C. 420 pp.
Geissler, P.H., and M.R. Fuller. 1986. Estimation of the proportion of an area occupied by an animal species. Proc. Sect. On Survey Research Methods of the American Statistical Assn.: 533538.
Juana de, E. 1980. Atlas Ornitologico de la Rioja. Instituto de Estudios Riojanos, Logrono, Espana. 658 pp.
Harrison, H.H. 1975. A Field Guide to Birds' Nests. Houghton Mifflin, Boston. 257 pp.
Harrison, H.H. 1979. A Field Guide to Western Birds' Nests. Houghton Mifflin, Boston. 279 pp.
Kibbe, D.P. 1986. Determining adequacy of coverage in statewide breeding bird atlas projects: when enough is enough. Pp. 18-22 in Sutcliffe, S.M., R.E. Bonney, Jr., and J.D. Lowe, Proc. Northeastern Breeding Bird Atlas Conf., Lab. of Ornithology, Ithaca, NY.
Lack, P. 1986. The atlas of wintering birds in Britain and Ireland. T & A.D. Poyser, Calton, England. 445 pp.
Laughlin, S.B., and D.P. Kibbe. 1985. The atlas of breeding birds of Vermont. Vermont Inst. of Nat. Science, Woodstock. 456 pp.
Raynor, G. 1982. Evaluation of breeding bird atlas survey data. Pp. 66-69 in Laughlin, S.B., ed. Proc. Northeastern Breeding Bird Atlas Conf., Vermont Inst. Nat. Sci., Woodstock, VT. 122 pp.
Robbins, C.S., and B.A. Dowell. 1986. Use of Miniroutes and Breeding Bird Survey data to estimate abundance. Pp. 28-40 in Sutcliffe, S.M., R.E. Bonney, Jr., and J.D. Lowe. Proc. Second Northeastern Breeding Bird Atlas Conf. Lab. of Ornithology, Ithaca, N.Y. 192 pp.
Robbins, C.S., S. Droege, and J.R. Sauer. 1989. Monitoring bird populations with Breeding Bird Survey and atlas data. Ann. Zool. Fennici 26:297-304.
Scheaffer, R.L., W. Mendenhall, and L. Ott. 1979. Elementary Sampling Theory. North Scituate, Mass. Duxbury Press.
Schifferli, A., A.P. Geroudet, and R. Winkler. 1980. Atlas des Oiseaux Nicheurs de Suisse. Station ornithologique suisse de Sempach, Sempach. 462 pp.
Sharrock, J.T.R. 1973. Rate of species-registration in atlas work. Bird Study 20:88-90.
Snedecor, G.W., and W.G. Cochran. 1980. Statistical methods. Iowa State Univ. Press, Ames. 507 pp.
Steel, R.G.D., and J.H. Torrie. 1980. Principles and procedures of statistics. McGraw-Hill, New York. 633 pp.
Stokes, D.W. 1979. A guide to the behavior of common birds. Little, Brown, Boston. 336 pp.
Stokes, D.W., and L.Q. Stokes. 1983. A guide to bird behavior, vol. 2. Little, Brown, Boston. 334 pp.
Stokes, D.W., and L.Q. Stokes. 1989. A guide to bird behavior, vol. 3. Little, Brown, Boston. 396 pp.
White, G.C., D.R. Anderson, R.P. Burnham, and D.L. Otis. 1982. Capture-recapture and removal methods for sampling closed populations. Los Alamos Nat'l Laboratory, Los Alamos, N. Mex. 235 pp.
Atlasing Handbook contents page
The NORAC website is hosted and maintained by Bird Studies Canada