Proceedings of the First Symposium on Marsupials in New Zealand
Estimating the Density of possums Trichosurus Vulpecula
* Now Wairarapa Catchment Board.
There are four methods which have been, or are being used by F.R.I, to measure possum populations and/or to assess the effectiveness of possum control operations.
Spot-light counts are applicable in farmland situations but become inefficient as the amount of bush or scrub cover increases. Weather markedly affects the number of animals counted.
Gin grapping, as well as contributing towards control, provides an index of animal density based on the numbers caught on successive nights. Precise estimates are dependent upon the catchability of animals remaining constant, a factor often subject to change because of climatic variations and changes in the structure of the population due to animals being removed.
Interference by possums of non-toxic baits provides an index of animal numbers and is a technique which may be applied in a wide range of habitats. The accuracy of the estimates is influenced by weather changes and changed behaviour as animals learn locations of bait stations.
Faecal pellet counting is a suitable method for sampling large areas, although dense ground vegetation limits its application. Current research is aimed at solving problems associated with variations in defecation rates and pellet decay rates.
It can be seen that there are problems associated with all techniques and that more research is required before any one of them can be unconditionally recommended for any given situation.
In this paper I shall give an evaluation of the four most commonly used methods of assessing common brushtail possum Trichosurus vulpecula populations. Examples have been chosen which illustrate the type of results that can be expected when these methods are applied in their appropriate field situation. With one exception, the assessments were conducted by F.R.I. in the course of their possum research programme.
Spotlight counting is a relatively untried technique. My analysis is based on counts of possums by the M.A.F. rabbit research section, during routine counts of rabbits. In the Waikari district in Canterbury, two 27 km transects were laid out in different areas. Two men on trail bikes with helmet-mounted spotlights, each covered a transect on three or four successive fine nights, starting at dusk and finishing three to four hours later.
During a 10 month period a total of eight samples was obtained from the two areas. The 95% confidence limits of each sample averaged ± 69% of the mean, while the confidence limits of the total samples in each area were ± 27 and ± 34 percent respectively. In other words the counts did not vary much from one sample to the next, but, from night to night there was often significant variation.
Because this technique is relatively appealing to use, is not labour intensive, and because large areas can be covered, it has a high potential use in pastoral areas. While the variation between nightly counts could be reduced by increasing the number of counts and increasing the transect length, more research is required to identify the factors which influence possum activity.
The trap-catch model for estimating possum density was formulated from data collected in 1945 and 1951 from the Pararaki catchment in the Haurangi Forest Park (Batchelor et al. 1967). When the model is used in other areas the estimate must be regarded as an index of density because the probability of capturing animals - an important constant in the formula - may not be the same. The model's principal application is therefore in measuring the effect of control operations, using the change in the density index as a measure of percent kill.
The Pararaki population was measured again in 1975 and in 1976. Each sample consisted of four lines of 20 traps, spaced at 100 m intervals which were run for three consecutive fine nights. Density in 1975 was 11 possums per ha ± 50% and in 1976 12 per ha ± 44%. Smaller errors could be expected if the sample size was increased to six lines of 25 traps - requiring approximately 12 man-days of effort. A typical situation where the trap-catch model could be used would be an easily accessible forested area due to be aerially poisoned. Trap lines would be run before and after poisoning and the mean and 95% page 179 confidence limits of the kill assessed from the percent reduction in the density indices of individual lines. Total effort required would be about 20 man-days.
Non-Toxic Bait Interference
A non-toxic bait interference method is being used in the Haupiri research area to establish the relative levels of pasture utilisation at increasing intervals from the main bush margin.
Monthly samples consist of lines, 100 m apart, which run from 250 m inside the bush to a maximum of 850 m into pasture. Bait stations consist of a small length of alkathene pipe attached to a reinforcing rod, into which is forced a plastic cap to contain the bait. These are placed at 50 m intervals along the lines and filled with a flour and soya bean oil paste. Baiting is carried out for five successive nights regardless of the weather.
Nightly patterns of baits taken are frequently irregular as a result of unfavourable weather, although in general they increase from night to night. During the 18 months of sampling the average number of baits taken has steadily increased even though other sources indicate that a small drop in animal numbers has occurred. My conclusion is that animals have learnt to associate some sign, whether it be the sight or smell of stations, with the presence of a bait which they have come to accept over the months. Because bait stations are removed at the end of a sample and replaced at the beginning of the next, the baits taken increase from night to night as their positions are discovered by animals within their particular home-ranges.
A need to modify Bamford's (1970) sample design is evident from these results. Two options are available: (1) the first to stabilise the estimate by introducing a time based factor accounting for the learning process (Jane this symposium); (2) to side-step the problem by doing a series of one-hit counts. For example, an assessment of a control operation would require before and after samples consisting of five lines of 20 baits set in different localities on three consecutive fine nights. Total effort would equal about 12 man-days. The 95% confidence limits of the estimates would be about ± 25%, depending of course, on the distribution of the population. The technique can be used in a wide range of habitats - from native forests to open pasture, but, in general, is most applicable in areas which have good access.
Faecal Pellet Counts
Pellet counts have been used in many areas to estimate possum densities but, in general, the technique is most suited to bush areas where access is limited. Their most extensive application has been in the Haupiri research area where monthly assessments have been made over the period of a year.
An estimate of possum numbers is obtained by calculating the number of pellets accumulating each day in a given area and dividing this by the average number of pellets voided each day by one animal. Experience in the Haupiri indicates that to obtain a meaningful estimate, a large sample (in excess of 500 1-metre-radius plots) is necessary. Two such samples, normally one to two weeks apart, are required for a single estimate of possum density. In the first, pellet density is measured and pellets are identified in some way, such that when the second sample is run, the density of those which remain can be measured. The number of new pellets recruited into the population is basically the difference between the number of survivors and the total density in the second sample.
Although animal numbers may remain fairly static, pellet densities may fluctuate considerably. For example in the Haupiri, consecutive pellet densities fluctuated, on average by ± 42%. Wind and rain which caused, on the one hand a high rate of pellet disappearance through litter fall and mechanical breakdown, and on the other, lowered activity levels and therefore a reduction in the number of pellets being deposited, probably accounted for most of this variation.
Results from a 'one-hit' sample which do not include an estimate of pellet disappearance rate can only be interpreted in very broad terms. The animal controller who wishes to use pellet counts to measure the effectiveness of an operation is faced with a before and after survey consisting, in each case, of two pellet density estimates and a disappearance rate estimate: something like 30 man-days of effort.
Bamford, J. 1970. Evaluating opossum poisoning operations by interference with non-toxic baits. Proceedings of the N.Z. Ecological Society 17: 118-125.
Batcheler, C.L. , Darwin, J.H. & Pracy, L.T. 1967. Estimation of opossum (Trichosurus vulpecula) populations and results of poison trials from trapping data. N.Z. Journal of Science 10: 97-114.page 181
Jane, G.T. 1981. Application of the Poisson model to the bait interference method of possum Trichosurus vulpecula assessment. In Bell, B.D. (Ed.) Proceedings of the first symposium on marsupials in New Zealand. Zoological Publications from Victoria University of Wellington 74: 185-195.
JANE. Do you think it is practical to use pellet counts in exotic forests when the frequency of pellet counts is about five percent?
D.J. BELL. No. At that level the errors probably make the estimate rather meaningless.
JANE. I have experienced a considerable bias in spotlight counts in estimation of density. Have you found this?
D.J. BELL. No. The results I have seen tend to indicate that you can achieve some sort of stability if you select your nights and maintain them over a period of time. I can't really comment on what bias is involved - it would depend partly on the type of control operation used.
JANE. No it is not so much the type of control operation. Rather it is due to variations in the area that is being observed with the spotlight. Where you have a number of tall stands with understorey your search area or search distance is short; in younger more open stands your search distance can increase ten-fold.
D.J. BELL. I pointed out spotlight counting was a technique with high potential for use in pastoral areas. One usually repeats the same route on a trial-bike each night so the intensity of search does not vary much from one count to another. When you get into forest areas there are many difficulties.
WODZICKI. The decay rate of pellets must vary between different environments, and are you aware of the work by Taylor and Williams* on rabbit pellet density estimates?
D.J. BELL. Actually Taylor and Williams' method is the basis for our pellet count work on possum populations. Pellet disappearance or decay is a very important factor.
WODZICKI. But have you measured decay rate in a range of possum environments?
D.J. BELL. We run a series of plots on a transect through the habitat which we want to sample and we measure the density of pellets on these plots and we also mark pellets on these very plots for decay. So on exactly the same ground that we are measuring pellet density, we are measuring their decay rate. We do this twice to get one estimate of recruitment rate, which is directly attributable to the number of animals.
KEBER. Have you ever tried calibrating these indirect methods of population assessment by say, laying cyanide out over successive nights, collecting dead animals and then calibrating the recorded fall-off in population numbers with changes in pellet density or bait interference?
D.J. BELL. Not yet. We are planning a large kill at the end of the Haupiri study when detailed information on the composition, movements, and density of the possums is available. We will probably try out such experiments there. Trap counts and pellet counts have been compared in the past and the results tend to agree fairly well.
* Taylor, R.H. & Williams, R.M. (1956). The use of pellet counts for estimating the density of populations of the wild rabbit, Oryctolagus cuniculus (L). N.Z. Journal of Science & Technology, Section B 38, 236–256.