Master's Thesis

On this page you can download the master's thesis I wrote for my Master of Science in Computer Engineering degree. I wrote the thesis while working on the SUSWOOD project at Åbo Akademi University. In the project we designed and implemented a forest simulator, and my thesis discusses the new features I implemented in the simulator.

The thesis contains a 5-page summary in Swedish.

Title: Implementing Computational Statistical Methods in a Scalable Individual-Tree Simulator
Author: Kristoffer Paro
Date: April 2010
Download: PDF (1.2MB)


This thesis describes a set of computational statistical methods that were implemented in the scalable individual-tree forest simulator SPATE-HPC. Firstly, an overview of the simulator is provided. The thesis then describes the underlying theories behind the computational statistical methods and presents the corresponding program implementations. The implemented methods consist of an efficient, parallel method of computing spatially correlated random numbers, a method of transforming tree characteristics to adhere to an arbitrary distribution, a method of estimating tree trunk volumes and wood material yields, and a method of simulating low and selective thinning using Weibull distribution curves.

The numerical accuracy and the performance of the implementation for correlated random numbers were tested with extensive series of simulations. Furthermore, a separate case study on the functionality and synergy of the other implemented methods was conducted. The case study imitated a real-world scenario by simulating two different harvest schemes applied to the same forest area.

The methods were successfully implemented according to specified requirements. The implementation of correlated random numbers produced output of high accuracy in an efficient and scalable way. The case study showed that the other implemented methods behave as intended and deliver usable results.

Keywords: spatial correlation, distribution transformation, timber assortments, thinning methods, parallel programming