Combining field and control team assessments to obtain error estimates for surveys of crown condition

Three field teams undertook a survey of crown transparency in Switzerland in 1993. A fourth team (control) also visited a subset of the trees that had already been visited by the field teams and reported estimates for the same. The problem of estimating the proportion (P_c) of trees with total crown transparency at least equal to a (fixed) level c is addressed. The method assumes that the control team estimate is an unbiased estimate of P_c and therefore if the expected difference between a field team estimate and the control team estimate is non‐zero then the field team estimate is biased. This assumption is most likely to be met in those countries or areas where the control team consists of the instructors. When this assumption cannot be met, as long as the bias of the control team remains the same over the years, the estimates of P_c from different years can still be used to study, for example, whether there is any temporal trend in P_c . A weighted average of the bias‐corrected field team estimates is used to determine the confidence limits. The method is applied to the Swiss national forest damage inventory data from the year 1993. The same approach can also be taken for estimating the proportion of cases where a random variable X, qualitative (such as the social position of a tree) or quantitative (such as the diameter of a tree at breast height), assumes values in a particular set.