THE DEVELOPMENT OF SAMPLING TECHNIQUES FOR FOREST INSECT DEFOLIATORS, WITH PARTICULAR REFERENCE TO THE SPRUCE BUDWORM

The problems that arise in the development of sampling techniques are treated in sequence, the solutions illustrated being based on population studies on the spruce budworm in northwestern New Brunswick. Insect populations may be expressed in different ways, depending upon the objects of the sampling, and it is essential that these objects be carefully defined. In the spruce budworm studies the preparation of life tables is the primary objective, and population is expressed in terms of a basic unit (branch surface) and an absolute unit (the acre). The correct timing of sampling requires a knowledge of the insect's life history, and of the stability of the population in place and time. When insect signs (pupal cases, empty egg masses) are sampled, the retention factor gives rise to certain non-sampling errors. The mechanics of collecting foliage from tall trees is solved with the aid of aluminum pole pruners, extension ladders, tree trestles, and platforms. The mechanics of counting the insects by ocular examination of the foliage samples also gives rise to non-sampling errors, which can be minimized by adequate supervision and by check examination.The universe for which each life table is prepared is a homogeneous forest stand. It is shown that a collection unit smaller than a whole branch, or its longitudinal one-half, is unlikely to be suitable for the measurement of absolute population. Intertree variance is the major source of population variance for the budworm, and for most other insects that have been studied intensively. Significant variance is also associated with crown levels, and the pattern of vertical distribution of the budworm is not predictable. The criterion of representativeness can be satisfied, however, by drawing samples from four crown levels in such a way that the intensity of sampling is equal in each level. The design found suitable for the budworm consists of proportionate sampling within the crown, stratified sampling by crown stories within the stand (with sub-strata, when necessary, according to host species or flowering condition), and cluster sampling within strata. Samples may be drawn from the same trees during successive budworm generations, and show correlations which appear to arise from the preferences of ovipositing adults. The effect of the correlations on sampling design and analysis is discussed.The relation between mean and variance indicates that population data are represented satisfactorily by the negative binomial distribution; variance may be stabilized and additivity provided through the use of logarithmic transformation. Methods of calculating optimum sample size are illustrated for both the original and the transformed data. The cost function for the sampling design used in the budworm work is presented, and its use in the comparison of designs and the detection of limiting factors is demonstrated. The estimation of population per acre is achieved by means of regressions showing the relationships between foliage quantity, crown volume, and diameter of the trunk. Useful incidental data, including a measure of population intensity, and biological data for life tables, may be obtained during sampling. Also, the design can readily be modified to suit purposes of extensive insect survey or the biological assay of a single stimulus.It is concluded that neither sampling nor non-sampling errors are likely to be prohibitive in forest insect population work. The relative magnitude of sampling errors varies inversely with the population mean, however, and the preparation of complete life tables may not be practical at endemic population levels. In any one generation of the insect it is generally necessary to accept error limits that seem large in relation to those of laboratory experimentation. But using the same sampling plots over a period of years, replication is achieved in time as well as in place, so the formation of erroneous conclusions on epidemiological phenomena seems improbable.