Biomass equations and biomass expansion factors (BEFs) for pine (pinus spp.), spruce (picea spp.) and broadleaved dominated stands in Norway

Abstract The objectives of this study were (1) to develop models for estimation of stand-level tree biomass for spruce (picea spp.)- pine (pinus spp.)- and broadleaved-dominated forest in Norway and, (2) develop biomass expansion factors (BEFs; ratio of stem volume to biomass) which convert stem volume to whole tree biomass for Norwegian forest conditions. A dataset from a 5 year period (2006 – 2010) from the Norwegian National Forest Inventory (NFI) were used to develop the BEFs and models. For construction of BEFs the whole dataset was used, while for the development of models, the data was divided in two sets. One dataset for model development (80%) and a validation dataset (20%). Swedish tree-level biomass equations were used for the construction of the models and BEFs since the existing biomass equations in Norway are based only on data from local conditions in parts of the country. Three tables with BEF-values were constructed. One general table for all areas within “Other wooded land”, “Productive”- and “Non-productive forest”, and two tables for Productive forest in development class III – V. The two tables for productive forest were divided into spruce, pine and broadleaved dominated forest, and showed BEFs varying with site index in combination with age classes or volume classes per hectare. In general, the BEFs decrease as stand age or volume per hectare increases, and the BEFs are lower at high productive sites compared to low productive sites. Since there are rather large differences in BEF ratio between low- and high-productive sites, the inclusion of site index classes in the tables most likely makes the BEFs more applicable in Norway compared to developed BEFs from Finland which frequently have been applied in Norway. Stand-level models for estimation of biomass from the different tree components; stem, bark, living branches, dead branches, foliage, below-ground for bioenergy use, total below-ground and total biomass were developed. Volume per hectare and site index were chosen as independent variables to be included in the models, and since the relationship between volume and biomass was slightly nonlinear, a nonlinear function was used. 4 The selected functional form was: Y = + ×Volume + × Siteindex 1 3 ˆ β β β 2 β Where Yˆ is the predicted biomass while 1 2 3 β,β ,β ,β are the estimated regression parameters. In order to account for the heteroskedasticity the models were fit with a normal probability density function (error distribution) where the variance increased proportionally to the predicted value. The new models have high r2 values ranging from 0.975 to 0.998 for the components; stem biomass-, total above-ground biomass- and total biomass. Living branches, dead branches and foliage components had lower r2-values, which varied from 0.575 to 0.962. A t-test based on the validation-dataset comparing the estimates from the new stand-level models to the total biomass calculated from the Swedish equations showed that the new models predict quite similar total biomass estimates for a wide range of stand characteristics such as stand age, volume per hectare and site index. However, the models for stands dominated by coniferous species estimated significantly lower total biomass compared to the Swedish tree-level equations in low-productive stands (site index class 6), on the west-coast, and frequently in the southeast region at elevation higher than 750 meters above sea. In general, at elevation lower than 250 meters in the southeast region, the new coniferous models predicted higher biomass than the estimates from the Swedish equations. The total biomass and below-ground biomass estimates from the stand level models developed in Finland were in general substantially lower than the estimates from the tree level equations from Sweden, and will most likely result in an underestimation of biomass when applied in Norway