No abstract

No abstract

No abstract

No abstract

No abstract

Site index prediction models for Norway spruce (Picea abies (L.) Karst.) and Scots pine (Pinus sylvestris L.) were developed using Norwegian National Forest Inventory data. A number of multiple linear regression models with different combinations of site and climate variables were developed in order to facilitate their application to a range of situations where the accessibility of various explanatory data differs. The best models used year of stand origin, temperature sum, vegetation type groups, soil depth, aspect, slope and latitude to predict site index. These models explained a large part of the total variation (R 2 adj: ¼ 0:86 and 0.72 for spruce and pine, respectively) and had little residual variation (RMSE 0 2.04 and 1.95 m for spruce and pine, respectively). Alternative models using only year of stand origin, temperature sum and vegetation type groups, or soil depth in addition, had slightly lower but still useful predictive power. All the developed models exhibited a strong non-linear effect of the year of stand origin on site indices, which varied when temperature sum was included. The increase in site indices along with increasing year of stand origin was significantly faster after about 1940 for both species. Similar time trends were observed for mean temperature and precipitation sums for the periods of stand growth, but only exhibited a faster increase after about 1960. Even though increased temperature and precipitation after 1990 seem to contribute to increased site indices, increased nitrogen availability and atmospheric CO 2 levels may also be important factors.

We developed nonlinear mixed effects height-diameter models for three major tree species: Norway spruce (Picea abies [L.] Karst.); Scots pine (Pinus sylvestris L.); and downy birch (Betula pubescens [Ehrh.]) in Norway. We used data from four Norwegian national forest inventory (NFI) cycles (7thÀ10th NFI cycle) as model fitting data and data from the 6th NFI cycle as validation data. Among several bi-parametric functions tested as base functions in a preliminary analysis, the N€ aslund function showed the smallest residual variations, and therefore it was extended by incorporating stand variables as covariates that act as modifiers of the original parameters of the N€ aslund function. Sample plot-level random effects were also included in order to account for inter-plot variations within the populations. Unlike a basic mixed effects model, the extended mixed model described larger parts of variations in the height-diameter relationships and predicted heights without significant bias for validation data from the sample plots, where all measured heights of the focused species (species used for species-specific model) were used to predict random effects. For species independent models, when measured heights of other than focused species were used to predict random effects, a significant height prediction bias occurred. This bias could be reduced for certain diameter ranges by applying an extended ordinary least square model. We recommend using extended mixed effects models to estimate the missing heights on NFI sample plots and other sample plots, where measured tree heights of the focused species are available for prediction of random effects. When measured heights are not available, the extended ordinary least square model can be used.

No abstract