The topic of this thesis is to solve spatial heterogenity in forestry models by means of utilizing linear mixed-effects models (LMM) and geographically weighted regression (GWR) to model a height-diameter curve. Both of these methods were previously tested, and they have a high potential to reduce the minimal necessary amount of data needed, and at the same time, increase precision. The data come from VŠLP Křtiny, LÚ Borky, a complex of forests utilized for educational purposes by Mendel’s university in Brno. We choosed beech as the model species. We split the data into training and validation sets for fitting, and consequent prediction assessment. Resulting models were compared with OLS fitted global model. Local OLS models were unreliable, as only a very few measured trees were available for each plot. Results were different for GWR and LMM. GWR models failed at prediction, but had good results on training plots, especially considering the reduction of autocorrelation of model residuals. LMM provided the best results for both training and validation plots
| Identifer | oai:union.ndltd.org:nusl.cz/oai:invenio.nusl.cz:428295 |
| Date | January 2018 |
| Creators | Forró, Martin |
| Source Sets | Czech ETDs |
| Language | Slovak |
| Detected Language | English |
| Type | info:eu-repo/semantics/masterThesis |
| Rights | info:eu-repo/semantics/restrictedAccess |
Page generated in 0.0023 seconds