Aim of this thesis is analysis of available data focused on 10-years lasting monitoring of growth groups in area of interest in eastern part of Czech Republic, and calculate prediction of random logging in between years 2009 and 2025. Growth groups were also divided into units with similar attributes.
First part of this thesis deals with description of random logging, causing agents and also with short theoretical work about used mathematical and statistical methods suitable for modelling random phenomena in forestry. As a source was used czech and foreign literature.
Next part is focused on basic information about area of interest and available data with detail description of particular attributes. Basic statistics were carried out to discover information about mean, variation, distribution and relationships between variables.
Practical part of thesis is formed by calculation of prediction in random logging by logistic regression, which is basis for modelling of probability of increase of random logging in every year until year 2025. This method is focusing on relationship between independent variable (year) and depend variable indicating value of cumulative random logging in interval 0 - 1. Value 0,1 is 10% and 1 means 100%. Some of this numbers had to be corrected, because their value was more than limit of 100%, or wasn't numerical therefore couldn't be used in calculation. Results of prediction revealed fact, that most of the growth groups will probably reach in year 2025 the 100% limit in random logging. Anothers will reach this limit before 2025 and some have already reached limit in year on which was calculation based.
Work revealed fact that it would be better include in calculation also information about changes of size of areas over the years. This problem could be solved by analysis of surviving, but this method wasn't used because of lack of time and suitable data.
Growth groups were divided into 5 main clusters with similar characteristic from several points of view: location, vegetation type, percentage of damage, and agents causing random logging.
Cluster analysis was used as a suitable statistical method for analysing similarity between groups of big data. On every cluster, basic statistical analysis was performed for description of basic characteristics. Visualisation of clusters in several approaches (spatial, vegetation type, percentage of damage, and agents causing random logging) helped to find out dependence between clusters, thus decide if there is a any system in classification of growth group, or it is a random process. Using of multiple regression for discovering correlation between attributes would increase quality of prediction in random logging.
Results of this thesis could be used for forestry economy and ensuring of steps for reduction increasing of random logging, which means a lost of forests. Characteristic of growth groups could be useful for selection of areas predisposed to NT with respect to spatial characteristic, vegetation and damage.
© Renáta SLEZÁKOVÁ - 2014 -