A national-scale, stand-level model to predict total above-ground tree biomass from growing stock volume

Interest in the accurate measurement of forest biomass has increased remarkably in the decades following the agreements made at the United Nations Framework Convention on Climate Change, particularly because of forest biomass' role in carbon storage. Information on forest biomass over large areas is generally obtained from National Forest Inventory (NFI) statistics; however, these data are usually estimates of stem volumes and only rarely estimate tree biomass through direct measurements. Conversion from stem volumes to biomass is commonly achieved through factors known as biomass expansion factors (BEFs). BEFs may also expand the estimation of tree components and, in such cases, are also referred to as conversion and expansion factors (BCEFs). Experience has proven that biomass factors vary with stand characteristics such as composition, age, volume and others. For this reason, models that predict such factors as a function of stand variables have been developed. In Italy, the total above-ground tree biomass (AGB) of forests was estimated by the second Italian NFI through direct measurements and, on the basis of the stock estimates of tree volume and biomass, BCEFs representative of Italian broad-spectrum forest conditions (e.g., composition, density, silviculture treatment and others) for reference year 2005 were published. This paper presents a stand-level model developed to predict AGB density (Mg ha−1) from growing stock volume (GSV) density (m3 ha−1) by forest category. The model was calibrated using NFI plot level data. The fitted model was tested to estimate total AGB from aggregated data, which was represented as the mean GSV per hectare at the plot level, and two broader aggregation scales using the mean GSV per hectare of forest category in each administrative region and at the national level. The accuracy of the estimates generally decreased with the level of data aggregation, but the upscaling exercises only overestimated the national overall AGB value by a maximum of 2.00%, an acceptable difference given the broad datum used as the independent variable. The upscaling exercises proved that the model is also suitable for predicting AGB by forest category, particularly given that the estimated values always fell within the 95% confidence interval of the NFI reference estimates.