Moose, Bighorn Sheep, and Mountain Goat Hunting and Harvest Estimates License Year 2005 Research and Technical Services Bureau Wildlife Division Montana Department of Fish Wildlife, and Parks 1400 East 6thAvenue Helena, Montana 59620 Justin Gude, Biometrician Candy Hinz, Survey Coordinator Jenny Dykstra, Statistical Technician Barry Beardslee, Programmer/ Analyst Keith Aune, Research and Technical Service Supervisor Introduction Since 1941, the State of Montana has used post-season surveys of hunting license and permit (LPT) holders to estimate wildlife harvest (Thompson 1947). Results of these surveys are used to develop hunting season regulations, evaluate and develop wildlife management strategies, for wildlife research, for hunt planning by the public, as well as other applications. Over the years, methods used to conduct the surveys have gone through several changes. Some of these changes have resulted from major program reviews (Gooch 1971, Cada 1983, Cada et al. 1986), and some have resulted from the increasing complexity of hunting regulations (ex. Gooch 1971 vs. Martin 2005). Several program changes have also resulted from changes in staff responsible for conducting the harvest surveys. License years 2004 and 2005 mark the beginning of a major change in staffing as well as a reorganization of the harvest survey team and program. For the previous 20+ years, the survey team has consisted of staff from the Research and Technical Services Bureau (RTS) in the Wildlife Division. This staff included the Bureau supervisor, a programmer/ analyst, and a survey coordinator to oversee a staff of approximately 60-80 survey interviewers that collected the data. The team has often had additional database programming support, and in recent years this support has come from non-agency contractors. Beginning in license year 2004, the RTS programmer/ analyst position was vacated and then eliminated. Duties related to data entry and data management, previously a responsibility of this position, have been transferred to the new Application Development Bureau in the Information Services Division. This permits increased focus on these activities and will lead to substantial program development in coming years. Subsequently, RTS created a biometrician position, which will have significant time devoted to the harvest surveys program. This position will focus on procedures for sampling LPT holders and estimating hunting and harvest-related quantities, which will add increased focus on these areas. This will lead to program development related to statistical issues in coming years. Additionally, lead survey interviewer positions were created to add more structure to the team of interviewers and to share some of the supervisory responsibilities held by the survey coordinator. Finally, a working group was formed to oversee and influence the direction of the harvest surveys program. This working group consists of approximately 15 biologists, researchers, and other FWP personnel that make use of the harvest surveys estimates and represent various regions and divisions of the agency. The scope of the group is broad, but includes program directions as well as estimation methods. This reorganization has resulted in several changes that affect estimates of hunting- and harvest- related quantities. As this is the first year these changes were implemented, methods are described with some detail in this report. Despite these changes, tables of estimates generated from the survey data look similar to those received by FWP staff in previous years. Those tables are presented in the results section. 2 Methods Parameters estimated In an effort to define the parameters of the LPT holder population that need to be estimated during the annual hunting and harvest survey, an effort was made to explicitly identify what quantities related to moose, sheep, and goat (special big game) hunting and harvest are necessary for annual use by FWP staff and the public. Due to the straightforward licensing process for special big game (one license per hunter, which can only be used to harvest one animal in one area), all of the quantities listed below are estimated by hunting district and LPT for both residents and non-residents, as well as by residendicy at the regional and statewide levels. Hunting-related parameters Hunter success. For special big game populations, this parameter is often used as an index of population trend in a particular area, either in conjunction with population survey data or in lieu of population survey data when such data are unavailable due to funding limitations. Hunters also use this parameter in planning which licenses to apply for. This parameter is technically defined as the ratio of the number of LPT holders that harvested an animal to the number of LPT holders that actively participated in hunting. Number of hunters. This quantity is often useful for hunting season regulation changes, for example changes in total harvest may be due to fewer or more hunters rather than changes in animal population sizes. It is also required to estimate the economic impact of hunting in different areas, as these calculations rely on the number of hunters and hunter days. This quantity is defined as the total number of LPT holders that spent at least a portion of one day hunting in the area of interest. Hunter days. This quantity is often useful for hunting season regulation changes, for example changes in total harvest may be due to fewer or more hunter days rather than changes in animal population sizes. As with the number of hunters, it is also required to estimate the economic impact of hunting in different areas. This quantity is defined as the total number of days that LPT holders spent at least a portion of the day hunting in a hunting district, in given region, or in all hunting districts in the state. Days per hunter ratio. This quantity is used to monitor changes in the average hunter effort as an index of the density and distribution of animals in an area. The quantity is defined as the ratio of the number of days hunted to the number of hunters for a given area, and thus requires no additional questions beyond those already asked of LPT holders during the survey interview. Harvest-related parameters Total harvest. This quantity is fundamental to wildlife management decisions. It is defined as the total number of animals harvested during the license year. Management success. This quantity is useful for making changes to quotas of licenses and permits in particular areas. It is defined as the total harvest estimate divided by the total number of LPT’s issued. 3 Harvest by age-sex classes. This quantity is also fundamental to wildlife management decisions. Estimates of the total number of adult males, adult females, and young of the year are needed on an annual basis to determine if adjustments to hunting season regulations or quotas need to be made in order to meet management goals. Harvest by time period. This quantity is sometimes useful for creating hunting regulations to ensure desired distribution of harvest across time during the hunting season. In previous years, this quantity was defined as the total harvest in each of seven time periods, including prior to the general season, during each week of the four-week general season, after the general season, and an unknown category for reported kills with missing data for time period of kill. For this report, the categories were lumped into four classes, including prior to the general season, during the general season, after the general season, and an unknown category for reported kills with missing data for time period of kill. This will give us a greater ability to have precise estimates of harvest at the hunting district level for relatively small estimates of harvest in each of these categories. Sampling Residency and LPT were used to define sampling strata. This resulted in formation of 133 sampling strata for moose LPT holders, 59 sampling strata for bighorn sheep LPT holders, and 65 sampling strata for mountain goats LPT holders. Sampling rates for these strata were determined in large part using methods from previous years. Although the accuracy and precision goals for these sampling rates were not articulated, these sampling strata are relatively small, with the largest being 93 LPT holders. This means that all individuals must be sampled in order to maximize the accuracy and precision of estimates. So, we attempted to contact every LPT holder in every stratum (1370 people total). Data collection A team of approximately 65 telephone interviewers that were supervised by the survey coordinator collected the hunting and harvest data. Interviewers were hired and trained during November and December of 2005. The majority of interviewers were hired in Bozeman, and a survey calling center was developed in the Bozeman FWP Regional office. Five lead interviewers were hired in Bozeman to structure the interviewing staff at the call center. Additionally, approximately a third of the interviewers were hired in other areas of the state to call resident LPT holders residing in their local calling areas. Calls were made to the sample of special big game LPT holders during December through March of 2006. Calls were made primarily from cell phones to avoid long-distance charges, though calls in local calling areas were made from landline phones to avoid cell phone minute charges. Sampled LPT holders were called up to three times in attempt to conduct the interview. If no contact was made at this point, uncontacted LPT holders were treated as non¬ respondents to the survey. When a contact was established with an LPT holder, interviewers asked license holders a set of pre-defined questions about special big game hunting and harvest activities during the 2005 license year (Appendix 1). Answers to these questions were recorded on paper datasheets. During and following the calling period, lead interviewers and the survey coordinator edited completed datasheets in an effort to monitor the quality of data and to correct obvious mistakes. 4 After the data were collected from LPT holders, datasheets were bundled and delivered to a data entry contractor. The contractor entered the data into a centralized Oracle database using data entry forms developed by FWP programmer analysts using Oracle Forms. Data quality control and management Quality control Once data were entered to the database, additional quality control was performed using SQL in the Oracle database. This included eliminating duplicate records and enforcing data integrity constraints. In many cases this involved the creation of unknown categories for analysis and estimation, including an unknown age-sex class, unknown hunting districts within regions and statewide, and an unknown time period of harvest. While these categories may not be of direct interest, these data are useful for estimating hunting and harvest parameters at higher levels, for example total harvest including the unknown age-sex class and regional and statewide hunting and harvest statistics in the case of unknown hunting districts. Analysis table creation Separate tables to estimate hunting and harvest statistics were created for moose, sheep, and goat surveys. These tables contained a record for every LPT owned by every license holder that was selected for the survey. Each record includes information on the LPT, which hunter owned the LPT, the residency of that hunter, the total number of that type of LPT issued to hunters of that residency, whether the hunter that owned that LPT responded to the survey or not, as well as the following information for hunters that responded to the survey: whether a license holder hunted, the number of days they hunted, whether an animal was harvested, as well as details about animals that were harvested including the age-sex class and the time period when the animal was harvested. Data analysis Response rate Non-response can bias estimates of hunting effort and harvest parameters. The potential for bias increases with the amount of non-response, and therefore higher response rates protect against non-response bias. In an effort to be clear about the potential for non-response bias in estimates of hunting effort and harvest statistics, response rates were calculated for each stratum that was sampled for each species. We assumed that non-respondents were missing at random (MAR) from the sample (Lohr 1999). For this method to produce unbiased estimates, there must be little or no relationship between the probability that an LPT holder responded to the survey and the hunting effort and harvest statistics being estimated. For example, if non-respondents were less likely to harvest an animal than respondents, then estimates of total harvest will be biased high. 5 Estimation of hunting effort and harvest parameters Estimates were generated for each stratum using estimators for simple random sampling, and for groups of strata using estimators for stratified random sampling (Thompson 2002). Estimated variances and confidence intervals were corrected for large sampling fractions using a finite population correction (fpc) factor (Thompson 2002). To estimate the combined hunter success and days:hunter ratios in a particular area, estimates for various LPT’s were weighted by the estimated number of hunters in that area rather than the total number of LPT holders. This was done in order to avoid down-weighting estimates of the ratio in the cases where not all respondents reported hunting in an area. Therefore, combined residency estimates of both the hunter success and days:hunter ratios are conditioned on the LPT holders that reported hunting in a particular area, and can be interpreted as the expected success or number of days that an LPT holder hunted in a particular area, given that they hunted in the area. Following tradition from the previous decade, confidence intervals were estimated around all hunter effort parameters except for hunter success. The only harvest parameter for which an 80% confidence interval was estimated was total harvest. Confidence intervals around these parameters were calculated using the normal distribution approximation (Thompson 2002). The absolute lower boundary of confidence intervals was set to the number of LPT holders in the sample with the population characteristic of interest, as this is the minimum number of LPT holders known to have the characteristic. Confidence intervals were not adjusted for multiple comparisons, and therefore cannot be interpreted as simultaneous confidence intervals. We expect 20% of the intervals around point estimates not to contain the true population value. In some cases, there was only one response in a stratum that contained more than one individual. In such cases, the stratum contribution to the total variance was taken to be the average of all the strata with a responding sample size of more than one individual. This adjustment results in a conservative estimate of the total variance (Lumley 2004). Estimates of population parameters and associated variances were not made for strata in which there were no respondents; all parameter estimates for these strata were set equal to zero. All data analyses were conducted using the R statistical software program (R Development Core Team 2006). Extensive use was made of the survey analysis package written for R (Lumley 2004). Results Response rates were calculated for each LPT that was issued for moose (Table 1), sheep (Table 2), and goat (Table 3), by residency. Hunter effort and harvest statistics were estimated separately for moose (Table 4), sheep (Table 5), and goats (Table 6). 6 Table 1. Response rates for moose hunting and harvest survey strata, license year 2005. residency LPT number issued number sampled responses response rate N 100-00 1 1 1 1 N 101-00 1 1 1 1 N 102-00 1 1 1 1 N 105-00 2 2 2 1 N 106-00 1 1 1 1 N 110-00 1 1 1 1 N 111-00 1 1 1 1 N 311-20 1 1 1 1 N 323-20 1 1 1 1 N 331-10 1 1 1 1 N 331-20 1 1 1 1 N AUCTION 1 1 0 0 N SUM 13 13 12 0.92 R 100-00 14 14 13 0.93 R 101-00 14 14 13 0.93 R 102-00 11 11 10 0.91 R 103-00 5 5 5 1 R 104-00 5 5 5 1 R 105-00 18 18 17 0.94 R 106-00 14 14 14 1 R 110-00 9 9 9 1 R 111-00 9 9 9 1 R 112-00 5 5 5 1 R 121-00 6 6 6 1 R 122-00 6 6 6 1 R 125-00 2 2 2 1 R 130-00 5 5 5 1 R 140-00 5 5 5 1 R 141-00 6 6 5 0.83 R 150-00 2 2 2 1 R 210-10 4 4 4 1 R 210-20 4 4 2 0.5 R 211-10 3 3 3 1 R 211-20 3 3 3 1 R 212-10 5 5 5 1 R 212-20 7 7 7 1 R 214-00 2 2 2 1 R 214-10 2 2 2 1 R 215-10 2 2 1 0.5 R 215-20 3 3 3 1 R 220-20 3 3 3 1 R 230-20 4 4 4 1 7 Table 1. Response rates for moose hunting and harvest survey strata, license year 2005. residency LPT number issued number sampled responses response rate R 240-00 5 5 5 1 R 250-20 5 5 4 0.8 R 261-20 2 2 2 1 R 270-00 10 10 10 1 R 280-20 2 2 1 0.5 R 285-20 2 2 2 1 R 291-00 1 1 1 1 R 292-20 6 6 6 1 R 293-20 2 2 2 1 R 300-10 7 7 7 1 R 300-20 5 5 5 1 R 301-10 5 5 5 1 R 301-20 8 8 8 1 R 302-10 5 5 5 1 R 302-20 5 5 4 0.8 R 303-00 4 4 4 1 R 303-10 2 2 2 1 R 304-20 4 4 4 1 R 306-20 3 3 2 0.67 R 307-20 1 1 1 1 R 308-20 4 4 4 1 R 309-20 7 7 7 1 R 310-20 1 1 1 1 R 311-20 14 14 14 1 R 312-20 5 5 5 1 R 313-20 5 5 5 1 R 314-20 6 6 5 0.83 R 315-10 1 1 1 1 R 315-20 6 6 5 0.83 R 316-20 3 3 3 1 R 317-20 2 2 2 1 R 318-20 3 3 3 1 R 319-10 1 1 1 1 R 319-20 4 4 4 1 R 320-10 3 3 3 1 R 320-20 3 3 3 1 R 321-10 1 1 1 1 R 321-20 3 3 3 1 R 322-20 5 5 5 1 R 323-00 5 5 5 1 R 323-10 12 12 12 1 R 323-20 9 9 9 1 R 324-00 1 1 0 0