Modelling height, height growth and site index from national forest inventory data in Norway
MetadataShow full item record
The present study aimed to develop dominant height growth models, site index prediction models, individual tree height growth models, and height-diameter models using Norwegian national forest inventory (NFI) data. Data from other sources such as long-term experiment (LTE), stem analysis and meteorological stations were used as supplementary data. Data from Norway spruce (Picea abies (L.) Karst.), Scots pine (Pinus sylvestris L.) and downy birch (Betula pubescens (Ehrh.)) were used. Since primarily designed for other objectives, NFI data have various weaknesses (measurement errors, small sample plot size, few height sample trees, and short time series) that were challenging modelling in the present study. Despite these challenges, various forest models were developed. The thesis contains four individual papers addressing the individual objectives as pointed out above. Paper I presents dominant height growth models that were developed using the generalized algebraic difference approach. Model parameters were estimated using nested regression techniques. NFI data models showed significant bias for young stands and higher site index classes when compared with LTE data. Therefore, NFI data and LTE data were pooled to develop combined data models. These models showed no significant bias for any ages and site index classes for both NFI and LTE data. The combined data models showed no significant bias when tested on independent stem analysis data and on region-specific model fitting data for Norway spruce and Scots pine. Paper II presents site index prediction models that were developed using the site index as a function of site and climate variables. Significant time trends in observed site indices were included in the site index prediction models. Among various models developed, a model including year of stand origin, temperature sum, understory vegetation type, soil depth, aspect, slope, and latitude described the largest proportion of the total variation in site indices for both Norway spruce and Scots pine. Analyses showed that site index increased after about 1940, which might be due to increased temperature and precipitation, and various other reasons. Paper III presents both spatially explicit and spatially non-explicit individual tree height growth models developed using a potential modifier function that reduces the potential height growth (Paper I) to the expected height growth of individual trees. Parameters in competition indices and parameters in the potential modifier models were estimated simultaneously from the data. Under strong competition, height growth was substantially reduced for both Norway spruce and Scots pine. For Scots pine, height growth was also reduced under very low competition. Paper IV presents height-diameter models, which were developed incorporating stand variables that are independent of thinning as covariates and sample plot-level variations as random effects. For all three species, generalized mixed effects models predicted heights without substantial bias when the random effects were predicted using all measured heights of the focused species (species used to develop species-specific model) per sample plot. The present study successfully developed methods to fit models to the NFI data that were not collected for growth modelling proposes. The models substantially improved the current models, which have been applied in an individual tree based forest simulator-T. Therefore, all models presented in the thesis may be used in future Norwegian forest simulators.