Review of eastern grey kangaroo counts and derivation of sustainable density estimates in the Australian Capital Territory
|
|  Yüklə 0.61 Mb.
|
Counting kangaroos
Kangaroos are culled during the March-July shooting season. The counts of eastern grey kangaroos used to determine the required density reductions should be conducted as close as possible to the culls so as to minimise potential population chage (e.g. due to natural recruitment of mortality) between count and cull. Currently, counts of eastern grey kangaroos used to determine the required density reduction in Canberra Nature Park have sometimes been conducted in spring and sometimes in summer. We suggest that the Canberra Nature Park culls be scheduled for say a 4-week period within this shooting season in a way that allows all the counts to be conducted as close as possible to the culling season such that the estimated number of animals to be culled is as accurate as possible. For example, the counts could be conducted in February and March (when nearly all young born in the previous year have emerged from the pouch) with a view to starting culling in May.
A target density of 1 kangaroo/ha has been stipulated for kangaroo management units (KMUs) in Canberra Nature Park. Kangaroo counts are currently conducted once annually. Above we recommended that the counts be conducted as close as possible to when culling occurs. Hence, assuming the counts are accurate and that the cull target is attained, the population would be at the target density immediately after the cull. During the remainder of the year the population could be expected to increase (through births but also potentially immigration) if conditions (e.g. high per capita food availability) are good but could, more rarely, decrease (through deaths and emigration) if conditions are poor. These possibilities are currently addressed by allowing for expected population kangaroo growth in deciding the cull target (see ‘Calculation of the number of kangaroos to cull’ document). It is unclear to us how the expected annual population growth rate is estimated and applied to each culled population. A more transparent way of addressing this issue is to conduct a second annual count six months after the first count, and use the mean of this and the pre-cull count as the annual density. Conducting a second count would obviously have financial implications.
Kangaroo surveys are currently conducted by staff, contractors and volunteers with a wide range of experience, training and likely skill. The accuracy of counts would almost certainly be improved if greater emphasis was placed on the selection and training of people collecting these data. We acknowledge that conducting field work is costly but using poorly trained staff, contractors and volunteers could substantially affect estimates of kangaroo population size, particularly in sweep counts which often require a large number of people to follow detailed instructions and to respond in real-time to changing circumstances during the sweep. Volunteers with a bias could also deliberately bias the sweep counts (e.g. by deliberately undercounting kangaroos). This potental bias applies mostly to the two ‘total count’ methods (direct and sweep counts) as the two ‘sampling’ methods (pellet counts and walked line transects) are conducted entirely by trained staff. We suggest that serious consideration be given to having a core team of staff/contractors that conduct all kangaroo counts (i.e. volunteers are not used). The core group should receive regular training and only people that meet minimum competency standards can undertake surveys. We note that replacing sweep counts with walked line transects would reduce the need for large numbers of people to undertake field work, and hence the cost of field work. Methods to count kangaroosFive methods have been used to count the number of kangaroos: (i) direct (or vantage point counts); (ii) sweep (or drive) counts; (iii) walked line transect counts; (iv) pellet counts; and (v) driven line transect counts. The reason for using different count methods is that each method has advantages and disadvantages that vary with site features such as vegetation height and density, topography, size, personnel availability and skill level, and cost. We address each method in turn. A trial to compare the costs, accuracy and precision of the methods, particularly the four methods that count kangaroos, is recommended.
Direct counts are used in small reserves in which it is thought that all kangaroos present can be counted from a vantage point by one or more persons. This method should be termed ‘vantage point count’ (Ratcliffe 1987, Mayle et al. 1999) rather than ‘direct count’ because the latter applies to any method in which the number of animals is directly rather than indirectly counted (Thompson et al. 1998). Hence, vantage point counts, walked line and driven transects and sweep counts are all ‘direct count methods’ and pellet counts are an ‘indirect count method’. Vantage-point counts are the most preferred method for counting kangaroos within Canberra Nature Park because they are cheap and are thought to provide an accurate and easy-to-compute method for estimating kangaroo abundances in small reserves with vantage points (e.g. ridge or roadside) that enable all kangaroos thought to be present to be seen. Direct counts involving 1−3 people were used in three KMUs in 2013; Crace, Dunlop and Mulanggari. Our visit to Mulanggari KMU (7 April 2014) suggested that direct counts are a cost-effective method for estimating kangaroo abundances there. To our knowledge there are no published studies evaluating this method for eastern grey kangaroos. Direct and sweep counts conducted by ACT are always replicated to achieve a mean count as follows. At least three replicate counts are required if the team of counters has not previously used the method on the site, and the result is accepted only if (i) there is no suspicion of over- or under-counting when the data are perused, and (ii) the counts are consistent, with the highest and lowest count being within 10% of the mean count. Two replicate counts are required if the team has previously used the method at the site and met points (i) and (ii) above. For sites with at least three replicates the operational details are changed (e.g. the sweep direction is changed) until the rules are met. Potential sources of variation include the observer (likely to vary with individual experience and inherent ‘skill’) and the time (some kangaroos may be lying down) and date (the number of kangaroos present may vary with time) of the sampling. We suggest that a minimum of four (but preferably more) direct counts be conducted in sites at which this method is used, and that a mean and 95% confidence intervals be reported for these four samples.
Sweep counts are more commonly referred to as drive counts in the literature (e.g. Coulson & Raines 1985; Southwell 1989), but the term ‘sweep’ reflects the local preference as care is made to avoid forcing the kangaroos ahead of the observers. This method was used in at four KMUs in 2013: Gungaderra, Mt. Painter, The Pinnacle, and Mulligans Flat Woodland Sanctuary. Sweep counts can provide a very accurate estimate of population size when animal movement is restricted to within a small area (Southwell 1989). Coulson & Raines (1985) evaluated the accuracy of sweep counts of eastern grey kangaroos on 266-ha Rotamah Island, Victoria. The abundance estimated in one sweep count (547) was well within the 95% confidence intervals of estimates from walked line transects (Coulson & Raines 1985). The sweep counts conducted in Canberra Nature Park involve one or more lines of people moving through the site and counting kangaroos as they pass back through the line(s), and are used in areas considered too large, too heavily forested, or where the topography is too complex for vantage point counts to be effective. Mixed count methods are used in some reserves. For example, sweep counts are used in the heavily wooded part of Gungaderra reserve and vantage point counts in the more open grassland part of the reserve. Our visits to all four sites (7 and 8 April 2014) at which sweep counts were used in 2013 indicated some of the difficulties associated with this method. The main difficulty with this technique is ensuring that all kangaroos present in the reserve prior to the count are counted, although one factor in selecting the mathod for a reserve is the degree of isolation of the population – by major roads and suburbs. There are two main potential problems with sweep counts in terms of accuracy. First, there is potential for kangaroos to ‘leak’ out of the reserve ahead of the line(s) and hence not be counted. Second, observers in the line may lose position and/or otherwise be spaced such that not all kangaroos moving through the line are counted. Kangaroos that have moved behind the line of people counting them are recorded in one column (‘positives’) seperately from those that move in the opposite direction back through the line (‘negatives’). A key issue with sweep counts is the arbitrary nature of deciding whether the count was successful (and hence used in subsequent analyses) or unsuccessful (and hence discarded). The ACT managers are aware of this issue and attempt to deal with it by applying the ‘rules’ and criteria described above, and on the absence of ‘incidents’ relating to the intactness of the line of counters and ‘leakage’ of kangaroos. Either 2 or 3 sweep counts were conducted at each site during late 2013, and these were often on the same day (i.e. morning and afternoon). These 2 or 3 counts were used to estimate the mean population size and its 95% confidence interval. It does not seem sensible to estimate a 95% confidence interval from only 2 or 3 sweep counts: as for direct counts, we suggest a minimum of 4 (and preferably more) sweep counts be conducted. It would be desirable to replicate the counts on at least different days (rather than morning and afternoon of the same day) so that any daily variation in the number of kangaroos present in the reserve is captured. We note daily variation is unlikely to be an issue in most reserves that are more-or-less isolated by suburbia. The large number of people (e.g. about 50) required for some sweep counts has meant that they are conducted by a mix of ACT Parks staff, contractors and volunteers. The accuracy of sweep counts are likely to be significantly affected by the wide variation in observer experience and ability (including ability to follow instructions), particularly on larger KMUs where there is greater potential for leakage and animals moving through the line to be missed. A potentially significant risk with sweep counts is kangaroos being pushed out of the site onto roads and colliding with vehicles. This has not yet happened and without directly observing sweep counts we cannot assign a likelihood of this occurring. Measures are taken to minimise this risk by (a) not sweeping towards a road and by sweeping inwards toward the center of the reserve, (b) where the sweep is beside a road two people move ahead of the count line along the road edge to encourage the kangaroos to move away from the road and in the path of the sweep line, and (c) an experienced wildlife ranger patrols roads during the count to radio through to the sweep leader if any kangaroos approach or cross the road. However, the risk of vehicles colliding with kangaroos must be explicitly considered in the decision to conduct sweep counts (rather than some other count method) at a site and in the planning of the conduct of the sweep count. The current briefing document for sweep count observer (which is given to each participant) states that ‘If it appears that kangaroos are likely to go onto a road, stop the line and report.’ However, other risk-reduction options are possible, including (but not limited to): temporary road signs, temporary reduced speed limits, and temporary road closures.
Walked-line transect counts have been widely used to estimate the abundance of kangaroos (e.g. Coulson & Raines 1985; Southwell 1994; Clancy et al. 1997; le Mar et al. 2001; Glass 2013). Studies evaluating the accuracy and precision of walked line transect estimates of eastern grey kangaroo abundance in populations of known size indicate that the method is accurate and has reasonable (i.e. ≤ 20%) precision (Southwell 1994, Glass 2013). Importantly, the assumptions underpinning the method, the field procedures for collecting data, and methods for analysing the data, are all well described and widely accepted in the scientific community (Buckland et al. 2001, 2014; Thomas et al. 2009, 2010). Walked-line transects were conducted in the two largest KMUs, Callum Brae-Isaacs-Hume (2033 ha) and Goorooyarroo (1391 ha), and also at Mulligans Flat Woodland Sanctuary (487 ha), Googong (c. 700 ha but not a cull site) and Wanniassa Hills (498 ha). The methods used to collect and analyse the walked line transect count data (described in Fletcher 2013) seemed highly appropriate. We were encouraged to see that KMUs had been stratified by vegetation type (from aerial photographs) and that sampling was conducted systematically (i.e. along parallel equally-spaced transects) and with the required transect length being calculated in advance, with the aim of obtaining estimates of abundance with precision ≤ 15% (although this may not always have been achieved). The two observers conducting the field work were ‘experienced’ (Fletcher 2013). It is good that the most recent version of program DISTANCE (Version 6.0. Release 2, which has some major analytical advantages over previous versions) was used to analyse the walked line transect data. Walked-line transects may be considerably cheaper than sweep counts and have a much lower risk of pushing kangaroos onto adjacent roads and hence colliding with vehicles. In view of the limitations and risks of sweep counts (see above), we suggest that walked-line transects be considered for all kangaroo management units in which sweep counts are conducted.
Pellet counts have been widely used to estimate large herbivore abundances (Marques et al. 2001), including eastern grey kangaroos (Coulson & Raines 1985, Johnson & Jarman 1987, Johnson et al. 1987) and the method can be partitioned into three aspects: (a) pellet accumulation rate estimation, (b) defecation rate estimation, and (c) analysis. When defecation rate is estimated for the population of interest (rather than using an estimate from captive animals) then the method has been shown to be reasonably accurate (Johnson & Jarman 1987). Pellet counts were used to estimate kangaroo abundances in three KMUs in late 2013: Mt Majura, Kama Nature Reserve and Jerrabomberra East (a planned reserve). Kama Nature Reserve and Jerrabomberra East Reserve are research sites at which kangaroos are not subject to culling: density estimation at those sites is for research purposes. Pellet counts rather than the more preferable direct counts (i.e. direct, sweep and walked line transects) were used at Mt Majura because thick vegetation meant that most kangaroos could not be observed, and at Kama Nature Reserve (155 ha) and Jerrabomberra East (97) because of their connectivity to adjacent rual lease habitats mean animals easily move in and out of the reserve and pellet counts would better represent ‘average kangaroo usage’ than direct counts (Fletcher 2013). We believe that walked line transects could be conducted at Kama Nature Reserve and Jerrabomberra East, and that repeat sampling be conducted on different days and perhaps different times of the day (e.g. midnight, dawn, dusk and mid-day). A methodology for conducting walked-line transects in a smaller area (76 ha) with high densities of eastern grey kangaroos was successfully implemented in part of Wilsons Promontory National Park by Glass (2013). Walked-line transects could also be conducted at night with spotlights (with star pickets with reflectors placed out before the count was conducted). The size (3.1 m2) and number (n = 240) of quadrats used to estimate pellet crops at the sites within the Canberra Nature Park seem appropriate in the sense that the estimates of mean pellets had moderate precision (ranging from 13% to 32%). The 21-day accumulation period seems appropriate, any longer and pellets deposited at the start of the interval might have decayed by the end of the interval, and any shorter may decrease the precision of the estimates. It is assumed that observers can differentiate eastern grey kangaroo pellets from those of other macropods. Although swamp wallaby (Wallabia bicolor) is currently present at low density relative to eastern grey kangaroos at Mt Majura, if eastern grey kangaroos were culled there then the relative density of swamp wallaby would increase. The absolute density of swamp wallaby might also increase if culling of eastern grey kangaroos increased food availability for the former and the pellets of juvenile feral pigs and some deer can sometimes be mistaken for kangaroo pellets. Defecation rate can either by estimated empirically or assumed equivalent to other studies (e.g. from captive animals). The latter is likely to be substantially biased because defecation rate will vary with factors including such as season, sex-age class of the population and food availability (Johnson & Jarman 1987, Southwell 1989). Research has indicated that the daily defecation rate of eastern grey kangaroos in Mulligans Flat Woodland Sanctuary and Goorooyarroo KMU varied over short periods (Howland 2008). Empirical estimation of defecation rate, as is currently done by ACT Parks, is therefore preferred. The current approach of using a population of known size (i.e. fenced with open grassland habitat) with apparently similar food availability to the Mt Majura, Kama Nature Reserve and Jerrabomberra East survey areas is appropriate, with the following two caveats. First, there needs to be standardised replication (e.g. n = 4 or 6) of the direct counts used to estimate the number of kangaroos present over the 21-day sampling period (see above). Second, the pellet quadrats should be randomly or systematically located (see above). The third and final aspect of pellet counts is the analysis. The analyses estimating the number of kangaroos present in Mt Majura, Kama Nature Reserve and Jerrabomberra East from pellet counts are appropriate but should be improved by addressing the following issues. First, uncertainty in the number of kangaroos present in the population of ‘known’ size should be propagated into the analysis of defecation rate such that the estimate of daily defecation rate has a standard error and 95% confidence interval that is contributed to by the numerator and denominator. Second, the uncertainty in the defecation rate addressed in the previous point needs to be properly accounted for in the estimated number of kangaroos. Current measures of uncertainty (i.e. standard errors and 95% confidence intervals) surrounding the estimated mean abundances of eastern grey kangaroos are overly conservative (narrow) because uncertainty in the estimated defecation rate (from both variation in the direct count and the rate of pellet accumulation) is not considered. This is a straightforward issue to address. The driven-line transect method has only been used once in the Canberra Nature Park before being replaced with walked line transects. There are two key problems with this method. First, sampling along roads is ‘convenience sampling’ (i.e. not a random or systematic sample of the area because roads tend to be on flat areas and on ridges) and hence could provide a highly biased estimate of kangaroo abundance (Marques et al. 2013). Second, kangaroos would be expected to use roads less than surrounding areas because they learn to avoid cars on roads and also because there would be less forage and/or protective shelter there relative to surrounding areas (the former would be the case for gravel roads); kangaroos may also hear the vehicle approaching and move away before they are observed from the vehicle, violating a key assumption of walked-line transect. Marques et al. (2013) proposed a method by which some of these problems could be overcome, but this used GPS telemetry data from a sub-sample of kangaroos and hence would be logistically and financially difficult to use regularly. We believe that walked- line transect should always be used in preference to driven-line transects. |