Analysis of Temporal and Spatial Variability and Statistical Modelling of Soil Water Reserve in Forest Stands

1988

S. 323-340

4 Abb., 5 Tab., 9 Lit. Ang.

Unselbständiges Werk

200014561

Annales des Sciences Forestieres 1988 ; 45 (4), S. 323-340

Soil water reserve distribution measured with a neutron moisture meter in forest stands presents a great variability, both in space and in time. An important problem in forest stand water consumption studies, at present based upon water balance methods, is the sampling for soil water content measurement. An answer to this question is to study the variability of soil water reserve. The authors propose a statistical analysis model allowing the separation of total variation into a time variation and a space variation. The model for interaction separation is based on Principal Component Analysis. Water stock measured at a point x at time t is decomposed in a general mean to which is added the product of a term linked to time with a term linked to space, and a random term. this expresses the fact that, once the variables are centred, measurements taken at each point of the stand fluctuate proportionally to one another. Random errors are assumed to be independant and to follow a single model. Themodel parameters and their variance-covariances are estimated, assuming th at the random contribution to variance is small compared to global variance, which has actually proved to be right in the present study. In a first stage the model parameters are estimated with their statistical characteristics. A strong heterogeneity generally appears between mensurations points, both for the aver age reserve at each point and for the multiplicative terme linked to space. In a second stage these parameters are analysed in conjunction with specific characteristics of several different stands. The existence of an important spatial gradient has thus been shown in one of the stands, leading to the elaboration ifa map. It has also been possible to quantify the influence of planting density on water reserve spatial distribution. In the case of a recently thinned stand this analysis lead to study the evolution with time of spatial variations. A s trong heterogeneity appears after thinning, followed b a progressive homogenization in conjunction with soil colonization by the roots of remaining trees. Spatial distribution of the studied parameters in relation to the distance from point of measure to tree has however not been modelled here, due to its very highvariability in conjunction with the number of available points of measure in e ach stand.