Assessing the attributes of scattered trees outside the forest by a multi-phase sampling strategy

Abstract

A sampling strategy to be used with multi - phase forest inventories is proposed for assessing scattered trees outside the forest on large territories. The fi rst phase is carried out by means of a systematic search over the area to be inventoried. The area is partitioned into regular polygons of the same size and points are randomly located, one per polygon. Subsequently, in the second phase, the land cover class of the fi rst-phase points is determined by very high - resolution remotely sensed imagery and a sample of points are selected from each land cover stratum. Then, the number of trees outside the forest lying within plots at the sampled points is recorded on the imagery. Finally, in the third phase, a subsample is selected from the second-phase samples of each stratum and the biophysical attributes of trees within plots are measured in the fi eld. Approximately unbiased estimators of abundance and of totals and averages of biophysical attributes are achieved in the second and third phase , respectively, together with the estimators of the corresponding variances. A simulation study is performed in order to assess the accuracy of the strategy under random and aggregated distributions of trees. The sampling errors achieved in the second phase using sampling fractions of ~ 0.3 per cent of trees vary from 6 to 13 per cent , whereas the errors achieved in the third phase using sampling fractions of ~ 0.15 per cent vary from 15 to 31 per cent . The results obtained from three case studies carried out in Italy confi rm the accuracy levels achieved in the simulation.