L'echantillonnage dans l'etude de la mineralomasse forestiere : l'exemple des taillis ardennais

anglais. The intensification of forestry (including whole tree harvesting) raises the question of its consequences on site fertility. We evaluate the removal of mineral elements related to the degree of intensification by applying tariffs established by stratified sampling. Here we summarise the limitations to the application of these tariffs due to the variability of other factors : species, site, tree and its components. - Three coppices in the Primary Ardennes were sampled by : 1. Species : Oak (Quercus sessiliflora , birch (Betula verrucosa), and mountain ash (Sorbus aucuparia).2. Site : « sol brun acide » with greater (80 cm) or lesser (40 cm) depths of loess over shales. 3. Tree component : leaves, branches, and three parts of the trunk cut at 7 cm, 4 cm and 1 cm diameters. - The dry weight of the material at 65 °C and major elements, N, P, K, Ca, Mg were determined. We used the following tariff model : Log (biomass) = a + b Log (circumference) and applied the correction for bias as used by Baskerville (1972). - The analyses of variance with one or more factors were measured to examine the effects of the factors. We used analysis of co-variance to compare the tariffs.Results : The qualitative examination of the variation in the chemical composition of the different components, showed that each is partly a function of site, but is primarly due to species (table 2). The interaction between species and site is low. We can classify the effects of, and interactions between, factors as a function of their F test value in the following order of decreasing magnitude : Species » tree component »> site > species X site ≥ species X component > station X component. The quantitative results (table 3) confirm that species is the most discriminating factor. There are also significant differences between the tariffs. These must be specific to species, site and tree component. The cost of these studies is high, so it was not possible to regroup the samples. Therefore, is it possible to increase the efficiency of the sampling strategy ? As no general law exists for the relationship between the parameter measured (circumference at 1.30 m or C130) and the mineral content of tree tissue, we calculated the minimum number of individuals, N, necessary by using the formula :(formule document ci-joint) at 5 p. 100 where L is the confidence limit at 10 p. 100 of the mean. The value of ơ is the value of the variance, and we admit that the best estimate is the variance estimated from our samples. Table 4 shows the results calculated for the different tree components. The variability is greater for mountain ash than for birch or oak.In our study, the choice ot 60 trees per species is theoretically insufficient in 17 p. 100 of the cases. At each site, for each species, we propose to sample 30 branches to estimate the mean at 10 p. 100, and 15 trees for the mean at 15 p. 100.

français. La quantité de matière sèche et d’éléments minéraux exportée au cours de l’exploitation forestière est estimée pour 3 taillis d’espèces mélangées dans les Ardennes françaises. Des tarifs, régression linéaire, sont construits à partir d’échantillons. Nous montrons dans cet article la nécessité d’établir des tarifs de biomasse ou de mininéralomasse en fonction de chacun des facteurs étudiés. L’analyse de variance permet de classer, en fonction de la valeur du test F, les effets des facteurs ou de leur interaction soit : Espèce » Compartiment »> Station > Espèce-Station ≥ Espèce-Compartiment > Station-Compartiment. La nécessité d’utiliser les tarifs spécifiques à chaque espèce, à chaque station et pour chacun des compartiments pose le problème de l’optimisation de l’échantillonnage. Nous admettons que la variance générale estimée à partir de nos échantillons est le meilleur estimateur de la variance générale. Nous calculons ainsi le nombre minimum d’individus (arbres) à la probabilité de 5 p. 100 pour satisfaire à une erreur de 10 p. 100 sur la moyenne. Le nombre de 20 arbres échantillonnés par station et par espèce, se révèle insuffisant dans 17 p. 100 des cas.