This study presents data from aerial surveys conducted in Yellowstone National Park spanning from 1970 to July 1997, aimed at monitoring the bison population. Surveys were initially conducted four times per year from 1970 to 1990, with a consistent pilot and observer, weather permitting. In 1991, the frequency increased to 9-10 surveys per year. Ideal flying conditions required clear-to-partly-cloudy skies and minimal wind, typically in the early morning to avoid stronger winds later in the day. The flight path commonly began in the northern range and proceeded sequentially southward, though weather conditions such as wind or ground fog sometimes necessitated route adjustments. In instances where surveys could not be completed in a single day, the remaining areas were surveyed the following day, and data were combined. During the early years, data from an elk researcher were sometimes used for the northern range when timing allowed, minimizing overlap in efforts. The objective of the surveys was to locate and count all visible bison, excluding those wandering outside population-use areas. A sharp-eyed pilot made the initial count of bison and calves, while the observer cross-checked, noting group sizes, animal locations, and any significant observations, such as mortality or unusual habitat factors. Data were recorded manually and later transcribed to field notebooks. Location descriptions were converted into UTM coordinates using the 1961 U.S. Geological Survey Topographic Map of Yellowstone National Park with a 10 km UTM grid overlay.

The study area encompassed most of Yellowstone National Park, USA. Aerial survey efforts concentrated on the locales used by the mixed groups (cows with young, usually some mature bulls). Historically, the winter locales used by bison occupied three areas designated the northern range (Lamar Valley), Pelican Valley, and Mary Mountain. The latter encompassed both the centrally located Hayden Valley and the Firehole to the west as one wintering unit because of the movements in both directions across Mary Mountain.

Beginning in the 1980’s as bison use patterns began to change, the west side was considered as part of the Mary Mountain geographic unit. Because adult bulls may wander widely, extra efforts were not made to locate these scattered individuals. The data span 1970 through July 1997. There were four aerial surveys per year (Piper Supercub, same pilot and observer with few exceptions), weather, and desired timing permitting from 1970 through 1990. Thereafter surveys increased to 9-10 per year, circumstances permitting, for a total of 151. Two surveys were not complete; these were made to verify Pelican area calf numbers (flight 3, June 3, 1970) and the Mary Mountain winter-spring distribution change (flight 11, May 1, 1972). Flying weather required reasonably wind-free, clear-to-partly cloudy conditions. Occasionally wind and/ or ground fog over open valleys dictated a survey could not be completed in one day, in which case the rest of the areas usually were flown the following day. In the data, these pairs of flights were merged and considered a single flight. Further, flights usually were broken only between the major geographic wintering areas of the northern and central regions. There are indications (to be discussed below) that there is little population transfer between these areas in general. Also, during the earlier years the bison numbers observed on flights over the northern range by an elk researcher (D. B. Houston) occasionally were used when the timing was suitable, rather than duplicate the effort. Surveys were intended to locate all visible bison within the YNP, with the exception of wandering bulls that were beyond population-use areas in unknown and unpredictable sites. An exceptionally sharp-eyed pilot (D. Stradley) made the overall and calf counts.

The observer (MM) cross-checked and recorded those numbers, locations, groups size and cohesiveness, general conditions, and miscellaneous data such as mortality, travel routes, and so forth. All initial data entries were recorded manually and transferred to field notebooks as soon as possible, usually the same day. Over the course of 27 years of data collection, 11 field journals were cataloged with detailed information concerning seven key variables: location, group size, demographics, dispersion, group shape, snow cover, and habitat use.

Processing the data occurred in three major steps. After agreeing on a spreadsheet format, data was hand-translated by the original observer (MM) from journal to spreadsheet outline. Location descriptions were translated into UTM coordinates by the original observer (U.S. Geological Survey Topographic Map of Yellowstone National Park, Wyo.-Mont.-Idaho. 1961ed. with 10 km UTM grid overlay). Secondly, the data was entered into a computer spreadsheet. Upon completion, on December 15, 1998, over 22,000 lines containing 25 variables had been entered into the data set. Finally, the electronic data set was checked and validated as thoroughly as possible. Because of the massive size of the data set, several methods of checking were implemented. Data transformations looked for spatial continuity in observation sequence. Furthermore, each flight was summarized for the number of observed bison. All anomalies suggested by the transformations and summaries were back-checked to the original observer’s log. Five major passes were made to reconcile our computerized data representation with all aspects of the aerial surveys recorded in the field notebooks. This work was completed on September 2000.

The precision of the location description translation process was checked for a random sample of a hundred and fifteen observations, stratified over the five principle localities. For each observation, a maximum deviation from the assigned UTM coordinates consistent with the observer’s description was determined in both east/west and north/south directions. From these a maximum possible error was calculated for each sampled observation. The mean maximum possible error was found to be only 0.324 km (s.d. 0.156) with no significant difference among localities. Errors of this magnitude are inconsequential for all analysis performed in this study. Fig. 32 plots all bison observations made during the study. The points in this figure have radii about 50% larger than the mean maximal error. The second pass through the notebooks also gives us a check of our transcription accuracy. Of the 230 transcribed UTMs, 229 agreed exactly with the spreadsheets. The final UTM had a small discrepancy in trailing digits amounting to an error of about a quarter of a kilometer. Errors of this magnitude would not be caught by our continuity checks.