Multi-source inventory methods for quantifying carbon stocks and stock changes in european forests

Soil organic carbon (SOC) data from typical large-scale soil inventories in 6 test countries were analysed. A quantitative approach to representativity analysis was developed in a test area. Additional sampling to optimise the representativity of the inventory was conducted. In another test area, a model was applied as an alternative approach to come to SOC baseline values. High-resolution soil C baseline maps were developed in 3 test countries and 1 test area using regression kriging, for the other two countries, regional estimates are based on geo-matching / class matching. Regional uncertainty estimates were conducted. The plot level error sources to the SOC estimates were reviewed in detail. An unbiased, representative regional SOC inventory was conducted as a model case to show the ability of existing inventories to detect SOC changes from repeated sampling.

Terrestrial ecosystems contain about 3 times the atmospheric carbon mass in living biomass and soil organic matter. Their annual gross fluxes exchange about 1/6 of the atmospheric carbon dioxide. The inter-annual variation in the biospheric net exchange fluxes is in the order of magnitude of fossil fuel emissions. This suggests that options of management for increased carbon storage in ecosystems may exist. The Kyoto protocol, with article 3.4, opened an avenue to search how "additional human-induced activities related to changes in greenhouse gas emissions by sources and removals by sinks in the agricultural soils and the land-use change and forestry categories shall be added to, or subtracted from, the assigned amounts for Parties included in Annex I". If the instrument of sink management within forests is to be used there is a need to account for 2 groups of processes that reduce C stocks in forests, namely disturbances and forest harvesting activities. The aim of WP 5 was to analyse the state of knowledge on the effects of disturbance and forest management on carbon stocks in forests and to identify inventory methods for monitoring of these changes that are available or need to be further developed. The main results obtained are publicly available with 6 reports at the project WEB site: WP 5.1 - Review of state of knowledge regarding carbon dynamics after disturbances in European forests WP 5.2 - Case study data on carbon losses covering 20 years after a storm disturbance in a temperate spruce forest WP 5.3 - State of knowledge regarding effects of forest management practices on ecosystem carbon storage and projected effects of changes in management practices. WP 5.4 - Monitoring standard for detecting carbon sources after major disturbances WP 5.5 - Monitoring standard for detecting carbon sinks following changes in forest management practices WP 5.6 - Documentation on the verifiability of carbon sinks through forest management activities

Methodologies for reporting afforestation/reforestation activities to the Kyoto Protocol are largely based around inventory data from land use, forestry and other national databases. The main issues to be overcome in reporting afforestation activities under Article 3.3 of the Protocol are area identification and carbon stock estimates of the five specified pools (i.e. above-and belowground biomass, soil organic carbon, deadwood and litter). Young forest stands pose difficulties in detection as well as carbon stock development as they have not yet reached the merchantable timber volume upon which most national forest inventories are based (NFI). This results in increased uncertainty in the development of C stocks for reporting. A case study approach was applied to identify and test specific data requirements for afforestation/reforestation reporting under Article 3.3 to the Kyoto Protocol (KP) according to the Good Practice Guidance for reporting Land Use Land Use Change and Forestry (GPG LULUCF). Ireland was selected to host the test area due to the high levels of annual afforestation and the absence of a NFI at the time of the project commencement. The planning of the first NFI gained from the project through the inclusion of trees <7cm dbh in the plot measurements allowing plot based DBH and height distributions to be developed for all trees. Additionally, the presence (and data for developing volume estimates) of dead wood and litter are recorded, and the soil classified. Measurements within NFI plots identified in the test area were used as well as results from national research and other default values were used in a comparison C stocks and uncertainties calculated at the three tiers of reporting outlined in the GPG LULUCF. Aerial photography was used in combination with the national forest administration database and NFI plot points to identify and validate forest boundaries. Field research undertaken during the course of the project resulted in the development of above- and belowground biomass functions for the two main tree species. All work undertaken in the field resulted in increased certainty in estimates. The example has particular national interest in Ireland in the development of methodologies for reporting during the commitment period, and also for the wider community as an example as to how to use existing data sources. A detailed report has been submitted to appropriate national government departments responsible for developing national inventories to the UNFCCC and a scientific paper is in preparation for publication in a peer reviewed journal. This case study will aid in the preparation of national C inventories with improved transparency and reduced uncertainty estimates for reporting to the UNFCCC.

Forests inventories provide useful data for the estimation of carbon stocks in forests, but this information does not cover all budgets necessary for a global carbon assessment. Nevertheless, some forest inventories offer detailed information of the aboveground carbon that may provide good tools for some of such estimations. The Ecological Forest Inventory of Catalonia (IEFC), which is an extensive database, offers such potential application. In this part of the project the main variables involved in the calculation of aboveground biomass expansion factors (ABEF), i.e. the ratio between the aboveground biomass and the stem volume, at stand level for the main forest tree species occurring in Catalonia (NE Spain) and the biomass equations are analysed. The IEFC forest inventory data set is based on 10644 plots where dasometric and dendrometric measurements were carried out. The main results that arise from this analysis are: 1.ABEF values significantly differ between species. 2. Wood density is an important variable to explain differences between species. 3 Differences in the total branch biomass also contribute on these species differences. 4. ABEF values are also influenced by mean stem diameter of the species present in the area; the tendency is that species diminishes their ABEF value as their mean stem diameter value increases. 5. Stem wood production of the species inversely relates to their ABEF values, thus fast-growing species show the lower values and 6. Carbon content shows small variability between species, given a tree component (leaves, bark or wood). Values of ABEF varies from 1.28 (Quercus ilex) to 0.44 (Pinus radiata) Mg m-3 and branches biomass expansion factors varies from 0.42 (Q. ilex) to 0.04 (P. radiata) Mg m-3. Wood density varies from 0.90 (Q. ilex) to 0.43 (Populus nigra) Mg m-3. Wood carbon content varies from 47.2 (Quercus suber) to 51.1 (Pinus pinaster) % of dry weight. It is concluded that is very essential to differentiate between species in the carbon accounting by using biomass expansion factors.

For mapping the local distribution of woody biomass, Landsat ETM remote sensing imagery were combined with sample based field measurements from national forest inventories (NFI's). Alternatively to the NFI reference data, airborne LIDAR data acquired in parts of project areas can be used for operational applications in case that no NFI reference data is available. For classification of woody biomass, the k-Nearest Neighbours Method (k-NN-method) was applied. Specific methods for pre-processing of the remote sensing data are required to optimise the classification results. In previous studies as well as for the applications in the test sites performed within the project CarboInvent, the reported estimation errors are high at the pixel level. However, the estimation error decreases when the size of the assessment unit increases. Therefore it is recommended to aggregate the pixel based classification results to larger assessment units, e.g. at the municipality level. Compared to field assessments it is a very cost effective method for large area mapping of the local distribution of woody biomass.

The 'top-down integration' applies existing forest inventory data in their aggregated form (for example, tree species or tree species group per region in a country) for forest area, standing volume and increment. Volume estimates were expanded to total tree biomass carbon estimates per tree species and age class by using biomass expansion factors (BEFs) from WP2 ("Biomass Expansion Factors"). Carbon budgets were calculated for the six European countries (Austria, Finland, Sweden, Spain, Ireland and Germany) using an existing modelling framework, the European Forest Information Scenario Model EFISCEN. The EFISCEN model was run until 2015, assuming that harvest levels would remain constant after 2005. Carbon stock changes were calculated from stock changes over the period of time considered. The BEFs developed for Finland were also used in Sweden, and those developed for Germany were also used in Austria. BEFs for Spain were not age-specific, due to the lack of suitable data. Data were also scarce for Sitka spruce and Lodgepole pine; the most abundant tree species in Ireland, therefore biomass functions developed in America were applied for these two species. Initial (1995) biomass carbon stocks ranged between 25 and 50 Mg C ha-1 in Spain, Ireland, Sweden and Finland, and between 100 and 110 Mg C ha-1 in Germany and Austria. Differences in the mean carbon stocks per hectare can be related to the mean volumes per hectare. High per hectare volumes in Germany and Austria resulted in high biomass carbon stocks compared to the other countries. In all of the six test countries, carbon stocks increased over time. It was assumed that no changes in the forest area occur during the simulation period; therefore, the trend of the carbon stocks is affected only by fallings and by ageing of the forests. The accuracy of the results for the biomass carbon stocks depends greatly on the quality of the utilised inventory data, and on the adequacy and representativeness of the applied biomass functions. The quality of the forest inventory data varies between countries. Adequate biomass functions were available for boreal conditions, as well as spruce, beech and pine in central Europe, but representative data for other species/regions was scarce. Biomass carbon stock changes depend on the difference between increment and fallings. While average increment rates usually do not change drastically over the course of 20 years when large forest areas are considered, felling levels depend on many fluctuating factors such as market prices, wood demand and the occurrence of natural disturbances like storm. Therefore, our modelled tree biomass changes may deviate considerably from reality. In the EFISCEN modelling framework, the soil carbon stocks are assessed with a modelling approach, where tree carbon pools are combined with compartment-specific turnover rates to estimate the litter input to the soil. This litter is given as input to the dynamic soil carbon model YASSO, which simulates litter decomposition. Mean soil C stocks ranged between 60 and 90 Mg C ha-1 in Spain, Ireland, Finland and Sweden, and were around 130 Mg C ha-1 in Germany and Austria. National soil carbon estimates from WP3 ("Soil carbon inventories") were 50-70% lower than our modelled values for Finland, Sweden, Germany and Austria. For Ireland and Spain, the soil carbon estimates from WP3 were higher than our model results. Among the possible explanations for the variation between model results and soil carbon estimates are: -the fact that WP3 only considered carbon down to 20cm, while YASSO is assumed to simulate soil carbon down to 1m, -the equilibrium assumption in Yasso, which is unrealistic in the light of past changes in forest management and environmental conditions, and - the overestimation of decomposition rates on organic soils within YASSO, which leads to low soil carbon estimates in countries where a large share of the forest grows on peat-land (e.g. Ireland). Soil carbon stock increased over time in all test countries, with an annual change of around 0.01 Mg C ha-1 in Finland, Germany, Spain and Sweden, almost 0.3 Mg C ha-1 in Ireland, and 0.8 Mg C ha-1 in Austria. The high increase in Austria was caused by the strong increase in simulated growing stock, which resulted in more litter-fall. However, we probably underestimated fallings in Austria, and therefore the stock change should be regarded with care.

1) Although substantial progress has been made in recent decades to understand the effects of rising levels of atmospheric CO2, temperature rise and changes in other elements of physical climate, nutrient availability and pollutants, there are still substantial limitations to predictability of the effect sizes. Although hypotheses and elements for understanding of these processes were published and are further developed, a set of rules for prediction of the effect sizes that are commonly accepted within science has not yet been achieved. Thus, the individual factors that make carbon uptake and release vary cannot currently be identified and quantified, i.e. factoring out every single component from a composite flux is to high a burden for Kyoto reporting. The alternative interpretation of factoring out is based on the view that the aim is not to understand every single factor but rather to quantify the change imposed by application of a specific management measure, i.e. factoring out the direct human induced change. Then, the sources of variation due to indirect human induced effects and former management impacts need not to be identified and quantified individually. This can be achieved with paired comparison of stands that are treated in the "standard" way but subjected to all indirect effects to stands that are treated with a changed management regime but apart from this subjected to the same environmental conditions. 2) Many climate protection options within the whole forest/wood industry cluster and by use of wood products are accounted for in other articles of the Kyoto protocol. The reduction of emissions that is achieved by substitution of fossil fuel intensive non-woody materials by wood is reducing emissions from industry and households for the whole country. Simulation results show that managed forests turn from an inferior to a superior choice if substitution effects are included in the evaluation. The generality of these results needs to be further assessed. As opposed to increases in stocks within the forest, substitution leads to cumulative effects (every harvest) - there is no sink limitation for this part of the sink. Thus, an evaluation of the whole set of policies and rules concerning wood use within the wider national concept for climate protection is needed when decisions on article 3.4 are to be taken. The same holds true for a wider range of policies if a sustainable use of forests, nature protection objectives, the provisioning of other goods and services than wood production and carbon storage need to be taken into account. 3) A large portfolio of methods to monitor timber volume changes exists that already allows quantifying a high percentage of FM-induced changes. Additional inventories need - in most cases - to be conducted for wood density, soil, dead wood, leaf litter, and ground vegetation carbon stocks. Biomass equations and BEF's (and the conditions of their use) can still be improved, especially in as much as roots are concerned. However, in most cases the methodology exists and needs just to be incorporated in existing inventory schemes. The main problem is that in many respects the baseline cases against which effects need to be evaluated by monitoring are not yet fixed. The baseline cases cannot be defined in one single way, but decisions to be taken depend on a detailed identification of policy objectives with respect to the Kyoto protocol and the role of articles 3.3 and 3.4 therein as well as with respect to other forestry related policy fields (nature protection, multi purpose forestry, sustainable use of natural resources, role of forest economy). Attempts to find answers to these open questions are part of the continuous negotiations of COP and define alternative policy options.

The cost and cost efficiency estimates within the bottom-up approach of the CarboInvent project were applied in all test sites. Basically the sampling error of the different national inventories was related to the costs for the applied in situ samples and to the results of the stratified samples. The aim was to show the gain in precision with respect to assessment costs for each design alternative. The application of EO data as stratification source proved to be an efficient method to reduce the sampling error with relatively low additional costs. The effect could be shown especially for high in situ measurement costs and even with relatively high EO data acquisition costs. Very low in situ costs might be combined with EO stratifications depending on inventory size and intensity.

The statistical analysis of the carbon stock estimates were applied in the four test sites Catalonia in Spain, Salzburg in Austria, Thuringia in Germany and Hyytiala in Finland. Individual single tree data were available through bilateral contracts between the data providers in the countries and the University of Hamburg. Soil information was provided by partner 9 (University Genth) but could not be merged and up scaled together with the tree information due to different sampling locations. Biomass expansion procedures were provided by partner 5 (METLA). In most of the cases the carbon stocks estimated with BEFs are smaller than the carbon stock estimates obtained by biomass functions. Only in some age classes the BEF based estimates exceeded those obtained by functions. The average carbon stock per hectare is about 108 t/ha in the Finnish, 151t/ha in the Spanish, 210t/ha in the German, and 248t/ha in the Austria test site. The applied combined 2 phases EO/ in situ carbon stock inventory concept proved to be superior to inventory concepts that utilize only in-situ (terrestrial) data. The combined approach consistently resulted in smaller sampling errors and thus more reliable carbon stock estimates.

According to the definition in the Marrakesh Accords Revegetation (RV) result as a buffer activity between the other activities of Art. 3.3 (Afforestation/Reforestation) and of Art. 3.4 (Cropland Management and Grazing land Management) of the Kyoto Protocol (KP). Thus the identification of activities qualifying for RV is subordinated to the election of national definitions and assumptions. The main criteria that must be identified at the national level are: the definition of forest, clarification of the concepts of "human-induced" and "at maturity" and the establishment of the hierarchy among the activities of Art. 3.4. On the basis of the elected criteria the relevance of RV may vary significantly inside the country's land and consequently the election of this additional activity at the national level may be cost effective or not. The convenience of election is connected to the comparison between benefits in terms of carbon-sink rising from RV activities and costs connected to the inventorying and monitoring of RV. The developed analysis highlighted that RV may be significant in helping to fulfil KP commitments only under broad criteria. In particular the inclusion of natural development of vegetation not qualifying for forest in abandoned lands would enlarge the area that may qualify for RV reducing costs per unit area. The process of vegetation expansion was proved to be mainly concentrated in the Mediterranean areas first through the analysis of FAO data on forest land and secondly through the processing of Corine Land Cover (CLC) maps in Italy that provided an estimate of land undergoing RV under broad assumption (around 8400ha year-1). The comparison to a more detailed scale study in the South of Italy confirmed that the CLC is not a suitable tool for the monitoring of RV activities, since the undetected land-use change area is not negligible in terms of surface. In addition the definitions of the cover classes and the minimum area used in the CLC are not compatible with the parameters requested for the reporting under the KP. The CLC validity is limited to a preliminary analysis on the presence or absence of RV under broad definitions and to a first comprehension of the distribution of RV at the country level in order to evaluate the significance of the RV processes and the effectiveness of election. Besides the uncertainties connected to the RV significance in terms of land affected, the calculation of benefits connected to RV activities is highly dependent on the knowledge on C dynamics and C-accumulation rates further to land-use changes qualifying for RV. An analysis of studies in the Mediterranean areas was developed. The uncertainties are many especially as concern succession processes in abandoned areas due to the numerous factors involved. The abandonment of agricultural areas usually lead to a carbon-sink but at the same time adverse processes have been documented, leading to soil organic matter depletion (thus, GHG emission). Additional studies should be developed to understand the extension of these processes on the territory in order to correct the estimate of C accumulation after land abandonment. But the most common trend seems to be an increase of C in the biomass and in the soil. The biomass data are also very variable and limited in number so the identification of general trends was not possible. Further analysis would be needed and additional information should be gathered on the biomass and carbon dynamics to fill the gaps existing at the moment.

The role of data from forest soil inventories and monitoring programmes was reviewed. Options to connect soil C inventories with ecosystem information were discussed. The requirements for verifying changes identified using the available soil C inventory data were compiled. The review includes options to link soil inventories with ecosystem research sites, such as ICP Forests Level II, Integrated Monitoring and the CarboEurope flux measurement network. A second report discusses the soil-related results of CarboInvent in the context of the reporting requirements on carbon sequestration in forest soils.

Typical soil organic carbon values were calculated for the typical soils in 6 test countries and 4 test areas. The data were calculated from various national and EU-wide data sources, and provided to other project partners for further evaluations. The data will be made accessible through the CarboInvent-JRC web portal.

Biomass and stem volume equations for tree species in Europe A review of stem volume and biomass equations for tree species growing in Europe was compiled. The mathematical forms of the empirical models, the associated statistical parameters and information about the size of the trees and the country of origin were collated from scientific articles and from technical reports. The total number of the compiled equations for biomass estimation was 607 and for stem volume prediction it was 230. The analysis indicated that most of the biomass equations were developed for aboveground tree components. A relatively small number of equations were developed for southern Europe. Most of the biomass equations were based on a few sampled sites with a very limited number of sampled trees. The volume equations were, in general, based on more representative data covering larger geographical regions. The volume equations were available for major tree species in Europe. The collected information provides a basic tool for estimation of carbon stocks and nutrient balance of forest ecosystems across Europe as well as for validation of theoretical models of biomass allocation.

The objective of this study is based on SOC analysis, evaluate the relation between forest site factors and the soil organic carbon (SOC) stock of the organic layer and the 0-50cm of the mineral soil in Podzols, which is the most predominating soil type. In this study we used data from the Swedish National Soil Survey period 1993 - 2002 from which data were available from 8580 plots on forestland (production exceeding 1m3 yr-1 ha-1), and that were sampled during 1993 - 2001. Of this number almost 4777 were podsols, which was the most predominating soil type and common in almost all parts of Sweden. The results showed that the coefficient of variation in the forest floor was ca 85% and with a mean SOC content in the forest floor of 3.3kg, a minimum of 3500 samples are needed to detect a significant change of 5% on p = < 0.05 level (*). The coefficient of variation in the forest floor plus mineral soil 0 - 50cm was 68%. With the mean SOC content in the forest floor plus mineral soil 0 - 50cm for all sub types of podsols of 8.2kg a minimum of 3500 samples are required to detect a significant change of 5% on p = < 0.05 level (*). The moisture conditions, ie the depth to the groundwater level, were of a profound significance for SOC stocks. For the forest floor the gradient was from 2.0 to 5.3kg C m-2, whereas corresponding figures for total soil to 50cm was 6.8 - 9.7kg. Other important site factors were latitude, temperature, site capacity, stand age and tree species.

For mapping the local distribution of stem volume, Landsat ETM remote sensing imagery were combined with sample based field measurements from national forest inventories (NFI's). Alternatively to the NFI reference data, airborne LIDAR data acquired in parts of project areas can be used for operational applications in case that no NFI reference data is available. For classification of stem volume, the k-Nearest Neighbours Method (k-NN-method) was applied. Specific methods for pre-processing of the remote sensing data are required to optimise the classification results. In previous studies as well as for the applications in the test sites performed within the project CarboInvent, the reported estimation errors are high at the pixel level. However, the estimation error decreases when the size of the assessment unit increases. Therefore it is recommended to aggregate the pixel based classification results to larger assessment units, e.g. at the municipality level. Compared to field assessments it is a very cost effective method for large area mapping of the local distribution of stem volume.

Within the project CarboInvent, the forest type maps were used to reduce the sampling error for the estimation of stem volume, tree biomass and carbon stocks for large areas by stratification. This method is especially relevant when estimates at the national level have to be performed e.g. for Kyoto Protocol reporting, where a high estimation accuracy is required. To derive the required strata, a supervised maximum likelihood classification and an unsupervised k-means classification was performed. As the results have shown, both methods have a high potential for reduction of the sampling error when the classification results are used in combination with field measurements from National forest inventories for stratification. E.g., the sampling error of the tree carbon stock estimates could be reduced to one third by integration of the forest type maps. Compared to the field assessments, the application of the remote sensing methods can be achieved at very low cost. The integration of forest type maps for large area estimation of forest parameters by stratification is therefore recommended.

The objective of work package 6.3 ("uncertainty estimate of the national level biomass and soil carbon stock adn stock change") was to estimate the overall uncertainty of carbon stocks and stock changes in forests on the national level when applying the top-down approach. A step-by-step estimation of uncertainty of the improved carbon stock estimates of D6.1 ("improved regional and national level estimates of the carbon stock and stock change of tree biomass for six European countries") and D6.2 ("improved national level estimates of the carbon stock and stock change of the forest soils for six European countries") was carried out for Finland, Sweden, Ireland and Spain using Monte Carlo simulation. The results for Germany and Austria had to be discarded, because erroneous BEFs had been used in the analysis. Unfortunately it was not possible to repeat the uncertainty assessment with the correct input data due to time constraints. In addition, a soil model comparison was conducted to estimate the effect of the choice of the soil carbon model on the overall carbon assessment. UNCERTAINTY ANALYSIS. A large-scale forest scenario model (EFISCEN) in combination with a dynamic soil model (YASSO) was used to estimate biomass and soil carbon stocks and stock changes. EFISCEN uses aggregated inventory data as input and models the development of forest resources in 5-year time steps. The national forest inventory data used in EFISCEN were gathered around the year 1990 for Finland and Ireland, and around 1995 for Sweden and Spain. Sources of uncertainty that were taken into account in the analysis of biomass were: inventory data, biomass allocation, dry wood density and carbon content. Simulations of biomass uncertainty were done in two steps: 1) the inventory data was converted into total biomass taking into account the error related to data itself and the error of BEFs; and 2) the total biomass was converted into carbon so that the uncertainty of carbon content was taken into account. The uncertainty analysis for the stock change from the initial year (1990/1995) to 2010 was done assuming that the data would be based on two inventories - the uncertainty of EFISCEN scenario (i.e. level of thinning and felling, no change in forest area and tree species) was excluded from the analysis since we did not have information about its reliability. The uncertainty was calculated in Monte Carlo simulation by calculating the difference between the stocks of the initial year and 2010 so that the uncertainty of stocks was taken into account. Sources of uncertainty in the soil carbon assessment that were taken into account in the analyses were uncertainty estimates of turnover rates and uncertainty estimates of parameters in the soil carbon model. The soil carbon stocks were assessed with a modelling approach, where biomass estimates of different compartments are converted to litter using compartment-specific turnover rates. This litter is given as input to the YASSO soil model, in which litter decomposition was simulated. Biomass C stock uncertainty ranged between 2 and 5%. The uncertainty of the biomass C stock change ranged between 11 and 27%, and was dependent on the size of the change. When the biomass C stock change was low, the uncertainty was higher, while a large C stock change resulted in a lower uncertainty. C stocks in the soils were much more uncertain than the biomass C stocks, but the C stock change estimate for the soils were more reliable than the soil C stock assessment. The soil C stock uncertainties were very similar between countries - about 45% - because similar assumptions are made in the soil model. Uncertainties in soil C stock changes were highest in Finland (34%) and ranged between 20 and 23% for Sweden, Ireland and Spain. SOIL MODEL COMPARISON. This chapter analysed the uncertainty related to model structure and possible bias in the selected soil model based on a comparison of four models - 4C, EFIMOD (with the soil module ROMUL), RothC, and YASSO. The models were compared with respect to the simulated soil carbon pools of four sites in southern Finland, and two sites in southern and northern Germany. Soil carbon pools were compared for the carbon in the organic layer, the mineral soil, and the total soil carbon. The choice of the model for the assessment of the carbon stock and stock changes in forest soils may considerably influence the uncertainty of the carbon stock assessment. The overall uncertainty due to the selection of simulation tool for the soil carbon assessment at individual sites may be in the same order of magnitude as the uncertainty due to model parameters. The identified differences between models were site and species dependent, and the small number of six study sites with two different litter input time series was not sufficient to generalise the quantitative importance of the uncertainty compared to other sources of uncertainty.

The database of Allometric Biomass and Carbon (ABC) factors was developed by JRC under WP2 of the Carboinvent project in collaboration with the JRC institutional Action GHG Data (2211, see http://ies.jrc.cec.eu.int/Action_2211-GHG-Data.86.0.html) in order to support assessment of carbon stocks (and stock changes) on the basis of inventories of timber volumes or growing stock as available normally from National Forest Inventories (NFIs). The database is available and maintained under the web based information system of the abovementioned JRC Action: http://ghgdata.jrc.it/carboinvent/cidb_bioctrans_abcf.cfm. It contains several types of factors (e.g., wood density, expansion factors, carbon fractions, water content), which are selected depending on the availability of initial values to be converted and/or expanded ("from what") and on the target values to be achieved ("to what"). The biomass compartments considered are the ones proposed by the IPCC Good Practice Guidance for LULUCF (see http://www.ipcc-nggip.iges.or.jp/public/gpglulucf/gpglulucf.htm). The database is accessible to the public through its download mask, where any user can query the database and download selected datasets, and the upload mask, where registered users can insert data. As up to end of 2005, the database contains some 400 datasets focussed on European forests. Users are encouraged to enhance the value of the database by providing own datasets.

The method applied including the two phases and the different data sources is a combined 2-phase multi source inventory for carbon stock. Multi source inventories have been recently applied by Tomppo (2001), Tuominen et al. (2001) for accuracy and efficiency improvement of forest inventories. In the project consistent and harmonized methodologies will be developed to quantify carbon stocks and stock changes in European forests. These methods will be based on linking the currently available inventory data with biomass expansion factors, soil carbon observations. The individual methods (soils measurements, BEFs, stem wood volume measurements, remote sensing) will be integrated in test sites representing the major climatic regions in Europe.

For mapping the local distribution of tree carbon stocks, Landsat ETM remote sensing imagery were combined with sample based field measurements from national forest inventories (NFI's). Alternatively to the NFI reference data, airborne LIDAR data acquired in parts of project areas can be used for operational applications in case that no NFI reference data is available. For classification of tree carbon stocks, the k-Nearest Neighbours Method (k-NN-method) was applied. Specific methods for pre-processing of the remote sensing data are required to optimise the classification results. In previous studies as well as for the applications in the test sites performed within the project CarboInvent, the reported estimation errors are high at the pixel level. However, the estimation error decreases when the size of the assessment unit increases. Therefore it is recommended to aggregate the pixel based classification results to larger assessment units, e.g. at the municipality level. Compared to field assessments it is a very cost effective method for large area mapping of the local distribution of tree carbon stocks.

According to the Kyoto Protocol, Annex B Parties must report carbon stock changes and non-CO2 greenhouse gas emissions during the commitment period on land areas that have been subject to direct human-induced deforestation activities since 1990. The definition of deforestation is given by the Marrakesh Accords. Deforestation for the purposes of the Kyoto Protocol involves the conversion of forest land to non-forest land. Remote sensing methods were applied to evaluate the applicability of Landsat data for monitoring of deforestation in remote areas for which no field data e.g. from national forest inventories is available. As Landsat TM data from the 1990's and Landsat ETM+ data from the 2000's is available as archive data free of cost for most parts of the world (e.g. GLCF archive), this is of main interest for large area applications, especially for remote locations for which no higher resolution imagery is available (or cannot be purchased because of high data costs). Detectability of the deforestation events depends mainly on the minimum area definition, the spatial resolution of the remote sensing imagery and on the change characteristics. To improve the interpretation accuracy, methods for fusion of the panchromatic Landsat ETM+ data with the multi spectral bands were developed. The results show, that on the one hand, the Landsat data allows mapping of deforestation at a wall-to-wall basis at very low cost for most parts of the world with already available archive data, on the other hand, detectability with this data starts with large deforestation events of 1ha. If quantification of deforestation including smaller deforestation events is required, in addition to the wall to wall mapping, estimation of the probability distribution of the aerial extend of the deforestation events is required. This can be assessed by sampling parts of the whole area with very high resolution EO imagery (e.g. IKONOS satellite data). The developed methods allow very cost effective monitoring of large areas at a wall-to-wall basis (full aerial coverage).

The National Forest Inventories are the most detailed national sources for information on forests. They have been a basic source of numerical information used in preparation of the GHG inventories since beginning of the UNFCCC process. Good Practice Guidance for LULUCF confirms that the use of NFI as a source of information is a "good practice" in the GHG inventory preparation, both for the purposes of UNFCCC and the Kyoto Protocol reporting. In Annex I countries, the forested land and ARD represent areal events, which significantly differ in size (by couple of orders of magnitude). Hence, estimates of area for forest and ARD events obtained from the NFI differ in precision and applicability in the context of Kyoto Protocol. The Monte Carlo type model was developed for testing sensitivity of the NFI geometrical arrangement to the presence of ARD events. The model space mirrors geometrical arrangement of the NFI based on circular sampling plots. The simulation was performed for ARD events with fixed area ranging from 0.1ha to 50.0ha. The results proved that the NFI tends to overestimate areas of the events much smaller than a size of the NFI grid. The relative overestimation increases with decreasing area of ARD event.

Forest inventory data and the biomass functions gathered and developed during the CarboInvent project were integrated in two ways: in a top-down approach (WP 6) using aggregated inventory data, and in a bottom-up approach using inventory data at the plot level (WP 7). We compared biomass allocation and tree carbon stocks of the bottom-up and top-down approaches (deliverable 6.4) for the test regions located in Austria (Salzburg), Finland (Hyytiala) and Germany (Thuringia). In the bottom-up approach, biomass functions were applied directly to the dimensions of the inventoried trees, mostly diameter and height, and then age- and tree species-specific carbon estimates were calculated. The singletree results were aggregated by tree species and age class. In the top-down approach, no singletree dimension was available, only aggregated information from the inventory. Growth and yield tables were used in order to assign diameter and height to the age class information of the forest inventory data. Only for the Finnish test region, biomass allocation could be derived directly from the applied biomass functions. The same biomass functions were used in both approaches for the test regions in Austria and Germany. For the Finnish test region, compartment-specific biomass function were used in the top-down approach, while the bottom-up approach used functions from the same source that expanded to total tree biomass directly, or only aboveground biomass in the case of broadleaves. For each of the three test regions, we compared carbon stocks stratified by tree species and age class for the top-down and bottom up approach. Total tree carbon stocks differed only slightly (1-2%) in the Austrian and German test regions, where the same biomass functions were used in both the bottom-up and the top-down approach. In the Finnish test region, the deviation was 6%. The deviation between the two approaches was highest for the youngest age class in the Austrian and German test regions. In the Finnish test region, the deviation between the two approaches increased with age. A more detailed comparison of the top-down and the bottom-up approaches was made for one of the test regions, Thuringia (Germany). For both approaches, the share of each biomass compartment was calculated as percentage of the total tree biomass for stem, branches, foliage, coarse roots, and fine roots separately for the main tree species. The differences between the approaches were generally larger at the species level. For beech, the bottom-up and top-down approach gave very similar biomass allocations. Due to slightly higher stem shares in the top-down approach for older age classes, tree carbon stocks were 5-8% lower than in the bottom-up approach for stands older than 80 years. The difference in stem shares for older stands was even more pronounced for oak, where tree carbon stocks were estimated 19% smaller (35 Mg C ha-1) in the top-down approach. The deviation in biomass allocation for spruce was highest in the first two age classes and decreased for stands older than 30 years. The top-down approach calculated 17% larger tree carbon stocks for the youngest age class, whereas for the oldest age class the result was 8% smaller than with the bottom-up approach. The differences between biomass allocations followed the same pattern also for pine, but the deviation in tree carbon stocks for the youngest age class was with +43% even more pronounced. Deviations in biomass allocation between the bottom-up and top-down approach were caused by differences in tree dimensions between growth and yield tables and the measured tree diameter and height from the most recent national forest inventory in Germany. The effect of using different growth and yield tables in the top-down approach was assessed for the German test region for the case of spruce, the most abundant tree species in Germany. The growth and yield tables varied in the intensity of thinning and in yield class. The deviation in biomass allocation was highest in the first age class of most species, where the stem share of total biomass ranged between 36% and 56% depending on the used yield table. The deviation decreased with increasing age, and the allocation for older stands matched very well. Since the carbon stock is still very low in young stands, the difference in biomass allocation and carbon stocks in young stands between the two approaches had only a small effect when comparing the carbon stock for the total test regions. It can be concluded that the two approaches gave very similar results at the regional level, whereas species-specific tree carbon stocks for different age classes showed larger variation, particularly for the youngest one or two age classes.

The Austrian Federal Environment Agency (Dr Peter Weiss) has taken the opportunity of the CarboInvent Project to refine the biomass expansion factors for Austrian forests. Based on data archives and data evaluations of Austrian data, scientists from the University of Agriculture in Vienna (BOKU; Prof Eckmullner and Prof Hochbichler), from the Forest Research Center Vienna (BFW; Dr Neumann, Dr Schadauer), and from the University of Innsbruck (Prof Cernusca) together with the Forest Administration of Tyrol (Dr Stohr) a new suite of biomass expansion functions was derived. The developed functions cover spruce, pine, fir, larch, beech, oak, hornbeam and are specific for Austrian (Central European) growth regions. The results are in the pipeline for publication in a special issue of the Austrian Journal of Forest Science and are used for Carbon Reporting Purposes, when data from the Austrian Forest Inventory are used to calculate C stocks in Austrian forest ecosystems. The estimated terrestrial C pool is from now on based on improved data and is therefore better than previous ones.

If the instrument of sink management within forests is to be used there is a need to account for 2 groups of processes that reduce C stocks in forests, namely disturbances and forest harvesting activities. Disturbances induce important changes in biogeochemical cycling and population dynamics. In spite of a growing body of information on these processes the current state of knowledge does not allow to set up a monitoring scheme that would allow for tracing all effects caused by the wide array of potential disturbances. However, with respect to lasting major changes in carbon stocks in forests, it is recommendable to focus on stand replacing disturbance events, that lead to a sudden transfer of carbon from live trees to litter or from live trees and soil to the atmosphere. Under European conditions these are mainly fires and wind-throw events. In order to capture C stock changes following fires it is necessary to monitor weather and climate variables that can be measured at a nearby weather station. Together with additional post-disturbance measurements on flame height as well as litter and O-horizon carbon content, carbon lost from living vegetation and soils during the fire can be modelled. Beyond this point the monitoring requirements for fire and wind-throw effects are the same: amount and quality of wood removed from the site and left on-site after the disturbance. The variables to be measured and ways to assess them are reported in the final reports to WP 5.1 and 5.4. A questionnaire sent to researchers and forest managers revealed that interest in and knowledge about carbon dynamics following disturbances are developed in different intensity across Europe. Up to today no country in Europe seems to have a forest inventory in place that can deliver the full range of data needed to assess post-disturbance carbon stock changes. However, in some countries, additional information sources, as e.g. records of disturbance events including location, area affected and effect size existed. These can be combined with classical inventories in order to extract the information needed for Kyoto monitoring. The release of CO2 from decaying wood is one of the main fluxes that need to be quantified when substantial amount of wood is left on site after disturbances. There is little information available on decay constants and the variation with environmental conditions for European forests and tree species. The literature contains some studies conducted in boreal forests; a small number only is available for the rest of Europe. Therefore, a method to estimate decomposition rate constants from published sources has been proposed and is included in the appendices' to the final report of WP 5.4. A database on dead wood dynamics was acquired from M. Harmon, amended with results on European tree species, and is available on the project's result web page at JRC. With regard to changes in forest management the main challenge was to define how "forest management" can be defined and to distinguish management options that in praxis often have shifting boundaries. Two levels of management that need to be distinguished: Forest management in a wide sense (actions that shape external influences and legal constraints that frame the decisions taken at the individual forest management unit level) and forest management in a narrow sense (treatment schedule which best meets the objectives set for forest stands). A list of forest management measures that can be applied at the stand level can be found in the WP5.5 report. The most prominent effects can be expected on changes in rotation length and fertilisation of forests that stock on nutrient poor soils. If forest management effects are to be assessed for detailed actions (e. g. different cutting regimes), it is of paramount importance to define and delimit such actions in a consistent way to be able to distinguish between different categories of FM activities (see final report of WP 5.5 for details). For the estimation of effects the technical systems used for conducting a specific activity and the area subject to this activity must be known. This information has thus to be included in reporting schemes.

Age-dependent BEFs developed in this project (Lehtonen et al. For. Ecol. Managem. 188: 211-224) were also applied for the estimation of biomass carbon stock and carbon sink of vegetation in Finland. In addition to estimation of carbon sink of vegetation, we also derived carbon input to the soil based on these biomass estimates and assessed carbon balance of forest soil with the help of dynamic soil carbon model Yasso (Lehtonen 2005 (available from http://www.metla.fi/dissertationes/df11.pdf), Liski et al. submitted manuscript).

In this project we achieved remarkable results in the field of evaluating methodologies for the assessment of carbon stocks of forested areas. The test area was established to model real afforestation/reforestation situations, where greenhouse gas emissions and removals are to be monitored complying with relevant provisions of the Kyoto Protocol and the Marrakesh Accords. Carbon stock changes in the following carbon pools were estimated: aboveground biomass, belowground biomass, litter and soil. Forest inventory data was used to estimate aboveground biomass carbon stocks and their changes. The forest inventory data were translated to carbon data using site-specific factors that were developed in the study. The biomass campaign was also used to develop biomass equations to model and analyse situations where this data can be better used to arrive at more accurate estimates. The belowground biomass data, and their changes, were estimated using default factors. However, soil and litter carbon stock changes were estimated using site-specific data, using chrono-sequences of stands within the project test site. In order to develop carbon stock change data for periods of five years, i.e. for periods longer than the duration of the CarboInvent project, a model called CASMOFOR was used. An earlier version of this model was developed to be user-friendlier, and to produce more powerful outputs that could be used in monitoring of afforestation/reforestation projects. In addition to the emission and removal estimates, a complete uncertainty analysis was carried out to demonstrate the methods and the extent this uncertainty estimation could be used in projects. Finally, the results include improved guidelines with respect to the methodologies that could be used in monitoring afforestation/reforestation projects. These guidelines can be used not only in joint implementation projects, but also in clean development mechanism (CDM) projects.

Parties to the Kyoto Protocol (KP)must make important policy decisions about specific activities and definitions within the Land Use, Land-use Change and Forestry (LULUCF) sector. These decisions have to take into account scientific understanding, requirements within the KP, environmental integrity, practical applicability and cost effectiveness. In fulfilling Articles 3.3 and 3.4 of the Kyoto Protocol and subsequent decisions in the Marrakech Accords (MA), each Annex I Party to the Protocol must, by 31 December 2006: 1. Adopt a single definition of the term forest by selection of: a. a minimum tree crown cover threshold between 10 and 30%; b. a minimum land area threshold between 0.05 and 1 hectare; c. a minimum tree height threshold between 2 and 5 metres; and d. a minimum width as recommended by the IPCC Good Practice Guidance for LULUCF. 2. Select any or all of the following human-induced activities under Article 3.4 in the first commitment period: revegetation, forest management, cropland management, and grazing land management. The Marrakech Accords provide a rather broad definition of forest management. Upon election of forest management as an activity under Art. 3.4, Parties to the Protocol are asked to interpret this definition under their national circumstances. These decisions may have an impact for the future due to the possibly long-term reporting obligations for lands once they have entered the reporting system under the Kyoto Protocol. The workshop offered a review of scientific issues and a practical guidance to estimation of: - carbon benefits and their uncertainty ranges resulting from the adopted definition of forest; - carbon benefits and their uncertainty ranges resulting from the adoption of each Art. 3.4 activity; - risk of potential need to report carbon liabilities as a result of the adoption of each Art. 3.4 activity, -monitoring/data collection and reporting costs, -trade-offs and synergies with other objectives, such as environmental or socio-economic considerations and -the range of incentives (if any) that may be required to achieve the desired GHG and other objectives.

Generalized equations for biomass estimation in temperate forest Several biomass equations are available for temperate forests, but few of them have regional coverage within their sampling. In this project Muukkonen (2005) developed generalized biomass equations for temperate region based on the biomass equation database (Zianis et al., 2005). The generalized equations can be used in the conditions where local representative equations are not available.

The carbon assessments on test areas were used to study the reliability of carbon figures. The estimated figures were subject to different sources of errors such as sampling errors, assessment errors, classification errors in remote sensing imagery and model errors. Different sources of errors were studied and their propagation to the total estimation error was quantified by means of an error budget. The error budgets present the contribution of each error source in terms of precision, accuracy and bias and allow ranking the components of carbon budgets according to their contribution to the total error. The affect of individual components on the minimum reliable estimate of carbon storage was realised by the error budget. Generally the application of stratification using auxiliary data sources reduced the sampling error remarkably. The ranking identified a high influence of the uncertainty of soil carbon stock within the entire carbon sock estimates. The tree carbon stock estimates were proved to be highly accurate with the application of biomass functions if single tree data are available, while BEFs proved to be highly applicable for stand wise or aggregated forest information. Within the application of the bottom up approach the BEF turned out to be biased in young age classes. Local adaptation of BEF and biomass functions as well as carbon expansion factors is recommended. Further investigation is needed for the improvement of BEFs, biomass function.

Undirected methods to estimate biomass Guidance on use on BEFs and biomass equations in regional/national estimation of tree biomass was prepared (presentation by Z. Somogyi et al. in the final whole action meeting of COST E21 in Dublin. http://www.efi.fi/coste21/ftp/2004-10-06/Somogyi_etal_Oct_2004.ppt) and review paper on this issue prepared (Somogyi et al., in revision). In this paper, a decision tree for selection of biomass estimation method is provided.

Estimation of tree biomass in boreal forests Availability of information related to allometry of trees was evaluated and database of biomass and volume equations for tree species in Europe was developed by Zianis et al. (2005)(available from http://www.metla.fi/silvafennica). According to the evaluation of the existing biomass equations, we conclude that biomass equations of Norway spruce, Scots pine and birch developed in Sweden by Marklund can be applied in nation-wide carbon in northern Europe. His equations are based on large number of trees (>1000) representing entire country, whereas several other equations are based on very limited number of sampled sites and trees. In regional biomass inventories the biomass equations can be applied if the inventory agency have access to tree-level data. If calculations are based on aggregated data (stem volume of growth estimates according to tree species and regions) biomass can be calculated with the help of representative biomass expansion factors. BEFs with uncertainty estimation were developed in this project for large-scale biomass inventories by Lehtonen et al. (2004, For. Ecol. Managem. 188: 211-224). Foliage biomass estimation by BEFs, biomass equations and pipe model theory were tested by Lehtonen (2005, Ecol. Modelling 180:305-315)). It was found that BEFs are suitable for regional biomass assessments, but not for estimating foliage biomass of single plots. Applicability of the developed BEFs was also tested in large-scale inventories (Jalkanen et al. 2005, Ann. For Sci. 62: 845-851)

Exploitable results

Share this page

Download