DIGITALNA ARHIVA ŠUMARSKOG LISTA
prilagođeno pretraživanje po punom tekstu
ŠUMARSKI LIST 5-6/2010 str. 57 <-- 57 --> PDF |
D. Klobučar: PRIMJENA GEOSTATISTIKE U UREĐIVANJU ŠUMAŠumarski list br. 5–6, CXXXIV (2010), 249-259 the number of trees increases relatively quickly so this structural element shows the poorest spatial correlation (Figure 3c). The omnidirectional experimental semivariogram is approximated with the theoretical (Table 2, Figure 3). Sample distribution maps (Figures 4, 5 and 6) show an estimated value of structural elements on either coordinates (x, y). Structural elements’ assessments through kriging and inverse distance method are tested with cross-validation and a root mean square error was used as an accuracy benchmark (Table 4). The mean square errors of assessment methods are very similar and therefore inconclusive when determining which interpolation method is more acceptable. Thus, a testing of the value differences between the measured data and interpolation methods for the three structural elements (number of trees, basal area and volume) was done by using the analysis of variance of repeated measurements. As visible in Table 5, statistically significant difference between the measurement data and interpolation methods of kriging and inverse distance was not determined. During the assessment of structural elements’ value (Figures 7, 8, 9) it is visible that the kriging assessment is more compatible with range of measurement values for all three structural elements, while inverse distance method measurements have a significantly lower value range (in other words model cells assessment tend to be around the mean value of incoming data). Consequently, this research considers kriging as the acceptable interpolation method when compared to inverse distance method. The making of semivariogram cloud is a useful tool because it allows the observation of each variable (structural element) as a distance function (shown on the x-axis) between measured data (variogram values between pairs are shown on the y-axis) within the analyzed area (view of the forest are with locations where measurements were done) on an interactive interface. In geostatistics the size of area and variable is not a limiting element. Any variable obtained through forest inventory, by tree type or total, can be observed by using a geostatistical analysis. The only condition is that some form of autocorrelation is assumed between them. Since forest inventory is conducted periodically, the geostatistical method which allows the possibility of monitoring forests in space (spatial structure), also allows monitoring forests in time. The changes of variable(s) in space and time (change of structural elements’ values by tree type and total, health of forests, etc.), as well as the forest management itself, can thus be monitored by continuously mapping two or more successive measurements. In addition, the above mentioned approach also enables the control of forest measurements. By doing the forest inventory, a lot of information is gathered on the state of forests. Geomathematical tools (geostatistical and neural) enable the data to be used in a more relevant and rational way in space and time, in order to manage forests in a more optimal way. Key words:forest management, forest inventory, structural elements, geostatistics, kriging, semivariogram. |