On the Use of Statistical Tests with Non-Normally Distributed Data in Landscape Change Detection
Research assessing changes of the main attributes describing forested landscapes often supply contradictory findings when viewed within an unchanging socioeconomic framework (i.e., unchanging forest management policies and no human population increase). Some studies suggest forest decline, others no change, and others expansion. We used ground-measured forest inventory data from Romania to examine the evolution of the forested landscape over a 10-year period (1993‐2003), during which no substantial transformations occurred in the human population or in forest management policies. Within a repeated-measures framework, a range of parametric and nonparametric statistical tests were used to evaluate forest change. Our results revealed that significant structural changes occurred at the landscape level. The changes were detected by several multivariate and nonparametric techniques. Multivariate techniques revealed that “age class,” “phyto-climatic zone,” and “ownership type” attributes had a significant influence in delineating between the 1993 and 2003 forested landscape. The univariate nonparametric tests indicated that there are factors not considered in the planning process (such as ownership or inflation) that can lead to a significant decrease of the main quantitative forest attributes (such as canopy closure or height), a decrease undetected by the corresponding parametrical statistics. Our research revealed that even when the distribution did not exhibit significant skewness, the nonparametric techniques could be more efficient than the corresponding parametric techniques by detecting significant changes of the attributes describing the forested landscape.
No Supplementary Data
No Article Media