Browsing by Subject "inventory of rare phenomena"

There is an ever growing interest in coarse woody debris (CWD). This is because of its role in maintaining biodiversity and storing atmospheric carbon. The aim of this study was to create an ALS-data utilizing model for mapping CWD and estimating its volume. The effect of grid cell size change to the model's performance was also considered. The study area is located in Sonkajärvi in eastern Finland and it consisted mostly of young commercially managed forests. The study utilized low-frequency ALS-data and precise strip-wise field inventory of CWD. The data was divided into two parts: one fourth of the data was used for modeling and the remaining three fourths for validating the models that were constructed. Both parametric and non-parametric modelling practices were used for modelling the area's CWD. Logistic regression was used to predict the probability of encountering CWD in grid cells of different sizes (0.04, 0.20, 0.32, 0.52 and 1.00 ha). The explanatory variables were chosen among 80 ALS-based variables and their conversions in three stages. Firstly, the variables were plotted against CWD volumes. Secondly, the best variables plotted in the first stage were examined in single-variable variable models. Thirdly, variables to the final multivariable model were chosen using 95 % level of significance. The 0.20 ha model was parametrized to other grid cell sizes. In addition to parametric model constructed with logistic regression, 0.04 ha and 1.0 ha grid cells were also classified with CART-modelling (Classification and Regression Trees). With CARTmodelling, non-linear dependecies were sought between ALS-variables and CWD. CART-models were constructed for both CWD existence and volume. When the existence of CWD in the study grid cells was considered, CART-modelling resulted in better classification than logistic regression. With logistic model the goodness of classification was improved as the grid cell size grew from 0.04 ha (kappa 0.19) to 0.32 ha (kappa 0.38). On 0.52 ha cell size, kappa value of the classification started to diminish (kappa 0.32) and was futhermore diminished to 1.0 ha cell size (kappa 0.26). The CART classification improved as the cell size grew larger. The results of CART-modelling were better than those of the logistic model in both 0.04 ha (kappa 0.24) and 1.0 ha (kappa 0.52) cell sizes. The relative RMSE of the cellwise CWD volume predicted with CART-models diminished as the cell size was enlarged. On 0.04 ha grid cell size the RMSE of the total CWD volume of the study area was 197.1 % and it diminished to 120.3 % as the grid cell size was enlarged to 1.0 ha. On the grounds of the results of this study it can be stated that the link between CWD and ALS-variables is weak but becomes slightly stronger when cell size increases. However, when cell size increases, small-scale variation of CWD becomes more difficult to spot. In this study, the existence of CWD could be estimated somewhat accurately, but the mapping of small-scale patterns was not successful with the methods that were used. Accurate locating of small-scale CWD variation requires further research, particularly on the use of high density ALS-data in CWD inventories.