Keywords

4.1 Introduction

The harmonisation process of NFIs dates back to the late 1990s when the forest resources assessment reporting program FRA 2000 (FAO 2001) and its regional contribution TBFRA 2000 (UNECE/FAO 2000) introduced a homogenous set of definitions to be used for reporting. The harmonisation continued in subsequent global FRAs when definitions were revised based on the experiences made. At European level, the European Forestry Information and Communication System (EFICS) was established by Council Regulation (EEC) No. 1615/89. The EFICS study had the aim to collect comparable information about the forestry sector in the European Community (EFICS 1997), and revealed the need for comparable information at the international level and the existence of knowledge gaps and unsatisfied information needs concerning the forest sector (Päivinen and Köhl 2005). The harmonisation of NFIs was strongly amplified by COST Action E43 (2010) which integrated the previous efforts by EFICS (1997), FAO (2001), and UNECE/FAO (2000), and developed techniques for common reporting by establishing concepts for European NFIs, and elaborating the methods to obtain comparable estimates (Tomppo et al. 2010). The harmonisation process of NFIs continued in the subsequent COST Action Usewood (2014) as well as in many follow-up projects such as the Specific Contracts within two Framework Contracts with the DG Joint Research Centre (JRC) of the European Commission (JRC 2012). In this chapter the principle of bridge building using national and reference definitions is described and the current state of harmonisation at international level is discussed. Recent advances in the harmonisation of NFIs regarding the assessment of wood resources are presented, as well as large-scale harmonisation implementations and case studies on bridge building.

4.2 Basic Concepts for Harmonising European NFIs

Harmonisation in terms of NFIs is the process of making data and estimates comparable across administrative borders. To achieve this goal, common reference definitions must be agreed among the countries involved in the harmonisation process (see Chap. 3). Then, each country would apply these reference definitions and store the metadata on the methods used in order to transform national data into estimates comparable between countries. These conversion methods are characterised as ‘bridging functions’ (Ståhl et al. 2012). Bridging functions can take different forms and include mere recalculations based on existing data as well as advanced statistical functions to convert the estimates based on existing national definitions into estimates corresponding to common reference definitions. In general, each country participating in the harmonisation process needs to develop its own bridging functions, at least as long as it conducts its basic data acquisition differently from the common reference definitions. Of course, the countries can exchange experiences and expertise among themselves to take advantage of existing data and methods developed by other countries (Fig. 4.1).

Fig. 4.1
figure 1

NFIs harmonisation process

Full size image

An important matter in harmonisation is whether or not necessary ancillary data as mentioned above is available to support the estimation according to the proposed and agreed reference definition. From this point of view, it is necessary to make a distinction between three general cases called reductive bridges, expansive bridges, and neutral bridges (Ståhl et al. 2012):

  1. 1.

    In the case of reductive bridging functions, national data are available in ‘surplus’, i.e. the reference definition is narrower than the national definition. For example, a country may have a national definition that includes more areas than the reference and thus the bridging function becomes a matter of identifying which parts should be excluded in order to meet the reference definition. An example applies for the Third Spanish NFI forest national definition with a minimum crown cover of 5 % instead of 10 % for the reference definition (Alberdi et al. 2010). The situation may be more complex when there are differences in more than one criteria involved.

  2. 2.

    The opposite situation is encountered when an expansive bridge needs to be developed. In this case the scope of the reference definition is wider than the national definition. This situation is often more complicated because the required data are generally not fully available at national level. In this case conversion methods may be computed based on auxiliary data. Such information could be obtained from pilot studies within the concerned country or data available from other countries with similar conditions. For example, when estimating growing stock (GS) the Swiss NFI measures only trees with a diameter at breast height (dbh) superior or equal to 12 cm whereas this threshold is 0 cm for the reference definition of GS agreed under COST Action E43 (Lanz et al. 2010). So an estimation of the missing part of Swiss GS is necessary to correspond with the reference definition.

  3. 3.

    With neutral bridges, the scope of the reference definition is the same as the national one, but internal subdivisions between different features may have been applied. In the case of the total biomass of a tree for example, the whole tree is partitioned into above ground biomass (AGB) and below ground biomass (BGB). The splitting between AGB and BGB is different among countries. Some countries specify AGB to include all tree parts above the ground level while others consider the tree parts above stump height. Concerning the subpart of AGB, the reference definition strictly specifies the ground level as dividing line. Thus conversion factors are necessary for estimating and including the stump biomass in AGB and excluding the stump biomass from BGB. The total biomass remains unchanged and therefore this adjustment is considered as a neutral bridge.

The methodological approach of bridge building not only needs a reference definition but also a description of the full process involved. The general way of working includes an initial screening exercise of the state of the art in each country (see Chap. 1). An enquiry is conducted, for example, using a questionnaire in order to explore and highlight the differences between the definitions applied by the NFIs. The enquiry is followed by an analytical decomposition of these national definitions into criteria and a description of the differing national thresholds (Vidal et al. 2008) (see Fig. 4.1). As an illustration the example of an individual tree, the basic entity for GS and biomass estimation, can be taken. First, a reference definition of a tree is needed for the distinction from other woody plants. Second, unambiguous specifications of the elements or parts of a tree are required, because definitions related to GS and biomass estimation are usually formulated on the basis of tree parts to be considered. Thus, a reference definition of each tree element used for defining GS and biomass is also necessary. To establish a clear specification for each element, a multi-step analytical partitioning approach was proposed leading to an unambiguous definition of each tree element (Gschwantner et al. 2009). These tree elements are distinct parts of the tree that are clearly defined. Combined, these elements can describe a whole tree. In this way, a flexible scheme could be achieved that (i) serves as a basis for a broad range of reference definitions (ii) permits further partitioning into additional, smaller elements and thus (iii) facilitates the adaptation to future developments and harmonisation requirements. The accuracy and precision of this harmonised information is an essential point to be considered. It depends on many factors, e.g. the conciseness of the definitions involved, the availability and quality of data for developing and applying bridging functions, the ancillary information used, and the method applied to perform the conversion from a national to a reference definition.

4.3 Harmonisation and International Reporting

International reporting programs such as the FAO’s Forest Resources Assessment (FRA) and FOREST EUROPE are important drivers for the harmonisation of forest-related information. Following the introduction of commonly applied definitions in FRA 2000 and TBFRA 2000 (FAO 2001; UNECE/FAO 2000) the comparability of estimates is still an aim to be strived for. The harmonisation in international reporting and its deficiencies are mainly linked to the differing national situations. The differences are due to different tools for forest assessment, historic backgrounds or climatic conditions as well as forestry policy issues. As a result, agreed definitions at international level must be broad enough to accommodate the circumstances of each country (e.g. Lund 2015). This lack of harmonisation which appears in international reporting can be highlighted on selected key examples. As a consequence, the effectiveness of a reference definition is ultimately tested when applied by individual countries.

4.3.1 The Example of Growing Stock Definition

The FAO (2010) defines growing stock by referring to living trees, specifying a dbh-threshold and the included tree parts. Additional specifications are given in the explanatory notes regarding the thresholds, wind-fallen trees and certain tree parts:

Growing Stock is the volume over bark of all living trees more than X cm in diameter at breast height (or above buttress if these are higher). Includes the stem from ground level or stump height up to a top diameter of Y cm, and may also include branches to a minimum diameter of W cm.

Explanatory notes:

  1. 1.

    Countries must indicate the three thresholds (X, Y, W in cm) and the parts of the tree that are not included in the volume. They must also indicate whether the reported figures refer to volume above ground or above stump. These specifications should be applied consistently through the time series.

  2. 2.

    Includes windfallen living trees

  3. 3.

    Excludes smaller branches, twigs, foliage, flowers, seeds, and roots

This definition is pragmatic, because NFIs used the criteria dbh, top diameter, and branch diameter and their associated thresholds X, Y and W according to national purposes and forest industry for a more or less long time and following country-specific traditions. Furthermore, countries consider GS either above ground level or above stump height. This point is only explicit in the FAO definition under a footnote but has some importance. Proposing a growing stock definition with flexible thresholds was the compromise to obtain a global agreement on what to consider as growing stock for the FRA reporting. Nevertheless, working this way leads to a lack of comparability between national figures. This lack of comparability can be illustrated by opposing two extreme interpretations of the definition which could lead to remarkable differences in estimates: (i) dbh-threshold = 10 cm, included tree parts = bole (wood + bark) from stump height to top diameter of 7 cm and (ii) dbh-threshold = 0 cm, included tree parts = stump, bole (wood + bark) stem top, large branches. In FRA 2010 (FAO 2010) the authors note the following about the growing stock figures presented in Table 2.13 of the FRA report: “It is interesting to note that the growing stock figures presented in FRA 2010, including growing stock per hectare, are in general higher than those contained in the FRA 2005 report. This is because many countries have collected new and better data for FRA 2010, more countries have reported, and more effort has been made to help countries provide the best possible estimates with the weak data they often have available.”

4.3.2 The Example of Forest Definition

The definition of forest agreed by FAO and national correspondents under the FRA process is less ambiguous than it is for GS because the threshold values for the criteria of minimum area, canopy cover and minimum height are specified. Nevertheless, as each country has its own national definition used by the NFI for national purposes, the country had to find a way to provide estimates according to the international FAO definition. Often, countries approached the problem of adjusting national estimates by reclassification from national land cover/land use classes into the FAO categories (Gabler et al. 2012).

For example, in the Third Spanish NFI, the minimum crown cover for defining a forest at national level was 5 %. Therefore, a reductive bridge was applied to estimate only the part with a crown cover more than 10 % for FAO tables. The Spanish NFI uses a stratified sampling plan. The forest strata are defined with the National Forest Map elaborated by photo-interpretation using several ancillary data (such as previous NFIs, National Forest Maps , Natura 2000 information or ownership). Forest area is delimited into homogeneous “tesserae” (homogeneous in the main forest species, crown cover and stand stage) (MMA 2006). As crown cover of all tesserae is known, the ones with a crown cover lower than 10 % can be identified and therefore not considered for international reporting. The area of tesserae with a crown cover between 5 and 10 % is then reported as other wooded land (OWL).

Another example is the Swedish case, where forest was traditionally equal to productive forest and defined by law in the Swedish Forestry act (Swedish Forest Agency 2015). This definition is linked to a productivity criterion that requires an average annual productivity over the rotation period of more than 1 m3/ha/year of stemwood over bark. Forest according to the FAO definition that did not reach the productivity criterion was defined as forest impediments. To report all the forest area to FAO, Sweden was required to build a bridge to convert national definition into the FAO one. The complete difference in the criteria used in the two definitions made the exercise somewhat difficult. The Swedish NFI thus decided since 1998 to assess both definitions of forest in the field, to facilitate reporting at a national level with the national definition and at the international level with the FAO definition. In 2008 the Swedish Forestry Act was revised in order to harmonise with the international definition and since then, Forest is defined according to the FAO definition and productive forest is defined according to the traditional definition as a sub-class of forest. This shows how NFIs can harmonise their estimates as well as their legislation by adopting international definitions but still continue to use the traditional national ones. As a consequence of this harmonisation process some countries have adopted the FAO definition and the situation has greatly improved.

4.4 Recent Harmonisation Advances at European Level

Harmonisation of forest estimates and derived information is a long term process. The duration of this process is due to the broad diversity in general ecological conditions, in the approach to forest policy and in the importance of the forest sector in different countries. A common forest policy in Europe would motivate the implementation of this process, but without this common policy approach, only small incentives can help to support the harmonisation. Nevertheless, important steps towards harmonisation at an EU level could be reached in the recent time in the short to medium term under the auspices of several EU projects.

4.4.1 Demonstration Project in Building Harmonised Information for EU Purposes

Following the 2010 Green Paper on forest protection and information in the EU (COM 2010) and a proposal of the European Parliament, the European Commission (EC) launched a Preparatory Action (PA) in 2012 titled “Future legal base on harmonised EU forest information” with the objective to contribute to the development of a European level framework for comparable and harmonised forest information. For this purpose, the DG Joint Research Centre (JRC) of the EC established two four-year Framework Contracts with a consortium of National Forest Inventory services. Under the Framework Contracts numerous studies on various topics were conducted. The Specific Contract Biomass (SC13) was one of the first implementation steps towards a future Forest Information System for Europe (FISE) (San Miguel Ayanz et al. 2005).

The aim of the SC13 project was to develop and apply a methodology for the harmonised assessment of forest biomass at European level and to provide spatial information on biomass and other forest parameters according to the Infrastructure for Spatial Information in Europe (INSPIRE) grid at the Nomenclature of Units for Territorial Statistics (NUTS) level units. The work included the development of a definition for above ground biomass (AGB) to include the biomass of all living trees higher than 1.3 meters, including stem with bark, above ground part of the stump, living branches, dead branches and foliage. Seedlings and shrubs should be excluded.

4.4.2 COST Action Usewood Achievements

The overall objective of Action Usewood was to improve and harmonise data and information on the potential supply of wood resources at the European level. In consideration of future scenarios of supply and demand of forest products and predicted supply deficits COST Action Usewood integrated three working groups with topic-specific objectives. WG1 aimed at harmonising the assessment and estimation techniques of state and changes in wood resources based on NFIs, WG2 aimed at improving wood resources estimates combining remote sensing and NFI field data , and WG3 aimed at predicting the use of wood resources.

The Action made significant progress to achieve its objectives (i) the harmonisation of European NFIs and establishment of reference definitions on increment, harvest and assortments, forest available for wood supply , other wooded land, other land with tree cover and trees outside forest (ii) the characterisation of tools available in European countries for predicting wood resources combined with an economic analysis study on wood availability in Europe.

Below some thematic fields for harmonisation progresses are presented. Information on the forest area available for wood supply (FAWS) is an essential piece of information for international forest reporting (UNECE/FAO-FRA and SoEF) which is mainly derived from NFI data. In order to meet the information needs concerning climate change mitigation and the increasing demand for wood in Europe, it is crucial to assess the actual quantity of wood available to harvest. Economic restrictions were considered by the NFI experts participating in COST Action Usewood (2014) as one of the main group of restrictions to define and estimate FAWS.

Trees outside forest gain increasing attention due to their environmental and economic roles. The estimation of wood resources outside the forest and OWL is encouraged by international organisations (de Foresta et al. 2013) especially in the context of increasing energy wood demand as a source of renewable energy. Therefore, COST Action Usewood developed an approach to partition the total land area. This approach allows countries to prioritise their efforts towards estimating wood resources outside forest and OWL depending on their importance in the national context (Vidal et al. 2016).

Besides the general availability of wood from in or outside the forest the quality and distribution of wood assortments is a major factor for setting up targets for several environmental policies. Thus it is necessary to distinguish e.g. between fire wood resources or higher value sawn wood timber. There are currently few European NFIs that consider assessment of stem quality or that directly estimate assortments (Bosela et al. 2015). Therefore, harmonisation through bridging functions is needed.

Under the scenarios of increasing demand for wood (e.g. UNECE/FAO 2011; Mantau et al. 2010) the question of sustainable wood supply from European forests has gained considerable attention. For the availability of wood at present and in the future the balance between increment and fellings is considered as decisive, and the fellings should not exceed increment in the long run for shaping a stable growing stock (FOREST EUROPE, UNECE/FAO 2011). In COST Action Usewood (2014) the harmonisation efforts were extended from estimates of state towards the classical change estimates regularly provided by NFIs such as increment, fellings and natural losses. European NFIs apply different sampling methods for sample tree selection as e.g. concentric circular plots and angle-count sampling. Therefore, the components of change were systematically described and defined independently of the sampling method on the basis of individual trees that constitute gains or losses to the growing stock. Harmonised reporting of change estimates builds upon the common definitions of change components and agreement about how they are taken into account for comparable estimates.

4.5 Large-Scale Harmonisation Implementations and Case Studies

Within the ENFIN group the adopted way of working is (i) to agree upon reference definitions (ii) to conduct case studies in order to check if building bridges is feasible and (iii) to make large-scale harmonised estimates. Some examples of different levels of harmonisation implementations are presented.

4.5.1 Biomass Calculation

In the SC13 project (see Sect 4.1) harmonised above-ground estimates were calculated for 17 European countries. The adopted approach followed the common method developed during COST Action E43 (2010) and included the description of national biomass definitions and estimation methods, the agreement on a reference definition and the development of bridging functions to close the gaps between the national and reference definitions. To define an operative harmonised AGB definition differences in the national definitions were mapped at a first step of the project. Following differences we found:

  1. 1.

    Some countries have relatively good data for above ground biomass of trees but no data for estimating below ground biomass

  2. 2.

    The minimum criteria for the size of trees included in the biomass estimates vary from seedlings below 1.3 m height to dbh of 12 cm

  3. 3.

    Some countries exclude stump biomass and other countries include it

  4. 4.

    Dead branches are often excluded from the biomass estimates

  5. 5.

    Some countries have no models for including the biomass of foliage.

In the next step AGB adjustment factors for different components of AGB (by countries able to provide such factors) were collected from countries. For this purpose it was necessary to map in detail which components of biomass countries are able to estimate and which components they may use as explanatory variables to estimate the missing components. It was equally important to identify countries that could derive different correction factors for as many tree species as possible. As a second step countries that could not estimate some components of AGB may use these correction or expansion factors to complete their data.

The following correction factors were required:

  • factors to include trees below a specific diameter (dbh) threshold. Several different threshold values for measuring trees are applied in different countries. Therefore, correction factors were developed to include:

  • the biomass of trees below the common threshold values (2.5, 5.0, 7.5, 12.5 cm dbh)

  • the biomass of branches and foliage with the help of stem biomass,

  • the biomass of dead branches with the help of living branch biomass,

  • the biomass of foliage with the help of living branch biomass.

Table 4.1 shows an example of the harmonisation process from two countries. Most of the correction factors developed by countries were calculated using national data, where available, and were used by other countries with similar environmental conditions; some correction factors were also obtained from literature (Ruiz-Peinado et al. 2011, 2012).

Table 4.1 Examples of filled gaps by countries
Full size table

4.5.2 Assessing Availability for Wood Supply Considering the Slope Restriction: A Case Study

For this study, the slope of the terrain was defined as the percent slope of the NFI sampling plot or of the immediate area surrounding the plot. To test the possibility of using common reference values for terrain slope to assess the availability for wood supply of NFI plots, fixed thresholds were defined. The maximum and minimum thresholds of terrain slope were defined through expert judgement and based on NFI data. The thresholds applied for FAWS vary among countries between 35 and 140 %, respectively. The objective was to investigate the possibility of using a common reference percent slope above which the forest should be considered not available for wood supply for all countries.

To analyse the effect of the thresholds, the area and the growing stock of forests characterised by the different thresholds were assessed for the five countries:

  1. 1.

    Min is by percent slope lower than the minimum—i.e. the area and growing stock present on forest with a terrain slope ≤35 %;

  2. 2.

    Max is by percent slope lower than the maximum thresholds—i.e. the area and growing stock present on forest with a terrain slope ≤140 %;

  3. 3.

    National is by percent slope lower than the national threshold for FAWS, if any.

For Spain, the national slope threshold is 45 %, with the exception of the Cantabrian range where it is acceptable to consider a slope threshold of 75 % for FAWS. For Ireland the national threshold is 66 %, and for Italy, Sweden and Switzerland no national threshold is defined.

In addition, Italian NFI data was analysed to assess the importance of terrain slope for forest available for wood supply. In Italy, NFI field crews classify forest as available or not available for wood supply (FNAWS) in each plot in the field while in the other four countries the classification is made a posteriori. The distribution of the Italian NFI plots into the two availability classes and by percent slope class was determined. The chi-square test was used to test for the difference between FAWS and FNAWS in relation to different thresholds of percent slope.

Table 4.2 provides an estimate of the forest area and the growing stock which could be classified as FAWS according to the two reference values and national thresholds, in the five countries. Looking at the relative forest area and growing stock below the minimum threshold, such a threshold would exclude a major portion (>65 %) of the forest area and growing stock of Italy, Spain and Switzerland. The maximum threshold would cause the exclusion of a relatively large portion of forest area (~10 %) and growing stock only for Italy (~5 %).

Table 4.2 Forest available for wood supply, forest area and growing stock (% of national total) estimated by applying the thresholds proposed for percent slope
Full size table

Figure 4.2 shows the distribution of Italian NFI plots by percent slope classes and by FAWS and FNAWS. Compared with FAWS, FNAWS show a large proportion of plots located in steep or very steep slopes (classes 60–80 % and >80 %) and a smaller proportion in flat areas or moderate slopes (classes 0–20 % and 20–40 %). The distribution of NFI plots by percent slope classes differ significantly between FAWS and FNAWS (p-value < 0.001).

Fig. 4.2
figure 2

Distribution of Italian NFI plots (n = 19,287 FAWS, n = 1647 FNAWS) by percent slope classes; FAWS forests available for wood supply , FNAWS forests not available for wood supply

Full size image

From this case study we can conclude:

  1. (i)

    Slope is a key variable for economic profitability of harvesting, and is commonly available information. It can be calculated using a Geographic Information System or assessed during field measurements. For these reasons it would be a good indicator of availability for wood supply, although it should be considered together with other variables. The use of one common threshold for terrain slope over the whole Europe is not easy. Profitability of wood harvesting is influenced by the terrain morphology in the country, harvesting techniques available locally and market demand. Therefore, a common slope threshold on its own is not enough to encapsulate the variation between countries.

  2. (ii)

    For harmonised reporting of FAWS/FNAWS terrain slope could be selected as reference variable together with country-specific thresholds. They could be assessed using the strong relationship between slopes and harvesting cost as proposed by Fischer et al. (2016) for Switzerland.

4.5.3 Detection of Trees Outside Forests (TOF)

Nowadays, new perspectives for estimating wood resources are emerging through high quality remote sensing data , new techniques regarding processing of large data sets, and regularly repeated image acquisition (Koch et al. 2008; Barrett et al. 2016). However, problems arise regarding the land use criterion, which is a key parameter in forest definitions but hardly assessable when using remotely sensed data (Waser et al. 2015). For example, a temporary unstocked area, e.g. after impacts such as fire, storm, or harvesting will be identified as non-forest area when using remote sensing data and techniques, but will in fact maintain its land use status as a forest.

This case study introduces a highly automated approach of estimating wood resources from TOF based on image-based point clouds from aerial photography, Airborne Laser Scanning (ALS) data, and the CORINE 2006 land cover (CLC2006) in a test site located at the border of Austria, Germany and Switzerland (see Fig. 4.3). The main emphasis was laid on the joint handling of national high resolution datasets based on different aerial cameras, different software packages and different digital terrain models.

Fig. 4.3
figure 3

The test site covers an area of approx. 250 km2, is located at the borders of Lake Constance and is characterised by mixed forests, rural and urban areas with woodland embedded in a pre-alpine topography. Source Swisstop (From Vidal et al. 2016)

Full size image

In this approach, wood resources were classified as shown in Vidal et al. (2016): (1) forest, (2) agricultural or urban land use class with the same criteria of land cover than forest (AUF), (3) OWL, and (4) Agricultural or Urban land use class with the same criteria of land cover than OWL (AUWL).

The workflow incorporates five main processing steps and starts with merging the different data from the three countries sets to one single dataset. Second, a Canopy Height Model (CHM) was calculated by subtracting a Digital Terrain Model (DTM) derived from ALS from the Digital Surface Model (DSM) which is derived from stereo-image based point clouds. More detailed information on the remote sensing input data and used image matching techniques can be found in Ginzler and Hobi (2015), or Straub et al. (2013). Third, wooded areas were extracted using the Normalized Difference Vegetation Index (NDVI) image and the CHM for the minimum tree height criteria from the three different national forest definitions (Tomppo et al. 2010) and the FAO forest definition (FAO 2010; Polley 2011). Fourth, GIS operations were applied on the wooded area using the four criteria: minimum tree height, minimum crown cover , minimum width and minimum area for each of the four forest definitions. The resulting maps included the four categories: Forest, AUF, OWL and AUWL which fulfilled the respective definitions. Fifth, the land use information from the CLC2006 data set was included in order to separate areas under forest-use (according to the CORINE definitions) from AUF, OWL and AUWL. Eleven CLC2006 classes: 111 (continuous urban fabric), 112 (discontinuous urban fabric), 121 (industrial or commercial units), 123 port areas), 124 (airports), 132 (dump sites), 133 (construction sites), 141 (green urban areas), 142 (sport and leisure facilities), 222 (fruit trees and berry plantations), 512 (inland water bodies) were used as non-forest use classes. For more detail see EEA (2015).

The maps were validated using an independent reference data set, which consisted of visual image interpretation of 1100 points in a regular grid (Fig. 4.4 as an example based on the FAO forest definition). Correspondence rates from 70 to 95 % were achieved for the different wood resources classes Forest, AUF, OWL and AUWL and were highest for Forest and AUF.

Fig. 4.4
figure 4

Example of a true color orthoimage with the forest mask based on the FAO definition, estimated AUF, and AUWL and OWL (From Vidal et al. 2016)

Full size image

It seems that the relative coarse spatial resolution (compared to the used aerial images) of the forest and AUF land use information obtained from the CLC2006 data set had no significant effect on these two categories. Lower correspondence rates were obtained for smaller tree patches that belong to AUWL and OWL. Thus, for the estimation of other wood resources classes, e.g. small stands , groups of trees such as Linear Tree Formations (LTF) and the remaining category of this classification Land with Few Woody Plants (LFWP) the CLC2006 data set would be too general and the use of alternative data sets, e.g. topographic maps—as available for many countries—should be taken into account.

To summarise, the present study clearly revealed that besides forest the extraction of information on OWL resources is feasible with a high degree of automation for cross-border regions. No border effects were detected that influence the quality of the present approach—although country-specific input data sets were used. Thus, high resolution remote sensing data is reasonably promising to classify clearly defined wood resources classes such as forest, AUF, AUWL and OWL. It is the first step into the direction of a harmonised reporting and useful for cross-national comparisons.

4.5.4 Stem Quality and Assortment Assessment: A Slovak-Czech Case Study

There are several approaches to achieve harmonised estimates of timber assortments in European forests using European National Forest Inventories (Bosela et al. 2015). One possible way is using a national model developed in a neighbouring country that has similar geomorphologic and climatic conditions and similar forest structure and management practices (tree species, thinning methods applied, etc.).

This case-study aimed at testing the approach of adopting a model for assortment estimation from a neighbouring country. The study used regression models which were developed for Slovakian forests in 1990s (Petráš and Nociar 1991a, b; Mecko et al. 1993, 1994a, b). Data for the development of the models were obtained from 167 research plots which were established over the main growth regions of Slovakia taking into account tree species. In each plot, a minimum of 70–80 trees were assessed.

The tree-level model of assortments was built up using regression modelling of the proportion of quality classes and assortments dependent on tree dendrometric characteristics. From many possible parameters, the dbh, stem quality assessed visually in the field (categories A, B, C, and D) and damage to the tree were found to significantly explain the variability and were thus used as the independent variables. For taking into account tree damages, any kind of stem damage which may allow fungi to infect the wood and is older than 2 years was considered (i.e. visual signs of stem base rot, or visible rot of the knots spreading to the stem, mechanical stem damage or root swelling caused by logging, frost crack, etc.). As mentioned above, stem quality was visually assessed in the field. Only the lower 1/3 of the stem was assessed into one of four categories:

A—High quality stems, almost without knots (only healthy knots under 1 cm in diameter at the base), twisting (spiral growing), and without other technical defects. B—Average quality stems, with small technical defects. In the case of hardwood species all of the healthy or unhealthy knots with diameters under 4 cm are allowed. For spruce and fir healthy or unhealthy knots under 4 cm and for Scots pine less than 6 cm are allowed. C—Low quality stems with large technical defects, with high frequency of branches (densely branched trees), twisting up to 4 % of straight length axis. Healthy knots without limit for the size (diameter) are allowed, unhealthy knots up to a diameter of 6 cm in the case of softwood species, and up to 8 cm for hardwood species. D—Poor quality stems with unhealthy knots over 6 cm for softwood species and over 8 cm for hardwood species, which are also affected by rot. The stems are only utilised as fuelwood.

The mathematical model gives the proportion of assortment classes in each standing tree. The regression models have the following form:

$$ \begin{array}{*{20}l} {Spruce} \hfill & {AP_{spruce} (dbh,q,p) = b_{1} + b_{2} q^{b6} + b_{3} dbh^{b7} p^{b8} + b_{4} dbh\;q + b_{5} p\,q} \hfill \\ \end{array} $$
(4.1)
$$ \begin{array}{*{20}l} {Beech} \hfill & {AP_{beech} (dbh,age,q) = b_{1} + b_{2} q^{b6} + b_{3} dbh^{b7} age^{b8} + b_{4} dbh\,age + b_{5} age\,q} \hfill \\ \end{array} $$
(4.2)

where AP—the percentage of the particular assortment; dbh—tree diameter at breast height; q—quality class (as dummy variable: A = 1, B = 2, C = 3, D = 4); p—damage to stem (as dummy variable: undamaged stem = 1, damaged stem = 2); and age—tree age. To overcome large variability in the proportions of assortments, cumulative proportions were calculated instead (in a stepwise procedure: A1–B1, A1–C1, A1–C3, and A1–D1). Each cumulative proportion was then fitted separately and the final proportion of the respective assortment was obtained by calculating the difference between two respective regression functions.

To test whether the models have a potential to be applied to a neighbouring country the data collected in the forests of Czech Republic during a special study (Adolt and Zapadlo 2010) were used as an independent empirical material. In total, 748 spruce and 773 beech trees were used for the study. Each stem, after cutting, was assorted into the six assortments. For this case study, the assortments were aggregated into three groups: 1 (A1 + B1), 2 (C1, C2, C3, C4 and C5), and 3 (D). Dbh, height, stem quality class and stem damage occurrence were measured and recorded for each tree. Although the stem quality class assessment on standing trees has the same meaning and purpose, there are slight differences in criteria used between Slovak and Czech methodology (see also the methodological guide of Adolt and Zapadlo (2010) used in Czech Republic case). However, we considered them to be the same for this case study as they both have the same range and apply the same principles. These parameters were then used to estimate the proportion of assortments for each tree using the Slovakian regression models. The proportions were then converted to volume. Finally, the measured and estimated assortment volumes were compared.

Total bias was calculated as the sum of absolute differences between measured and modelled volumes in the particular assortment (e.g. veneer category). Then the relative bias was calculated so that the total bias was divided by total volume sum in the respective assortment category.

The total relative bias, when measured and modelled values were compared, was found to depend on assortment classes and species (Table 4.3). In the case of beech, the lowest bias was found for the group including pulpwood and firewood (−6.2 %, 773 stem sections were classified in this category), while the largest one for veneer assortments (app. 164 %, only 31 stem sections classified in the category). The bias for roundwood was −8.8 % (351 stem sections). For spruce, the bias for veneer was even larger and reached 226 % (only 15 stem sections classified in the category), for roundwood −9.9 % (including 528 stem sections) and for firewood and pulpwood 18.5 % (746 stem sections).

Table 4.3 “Goodness-of-fit” of the Slovakian regression model applied to assortment data from Czech Republic
Full size table

The case study tested a potential way (one of several others) to produce harmonised estimates of timber assortments. However, the present status of European NFIs does not allow for developing a general way towards harmonisation of assortment estimations (Bosela et al. 2015). This is because only few NFIs go beyond stem quality assessment and, in addition, because different grading standards are used across Europe. This study showed a potential of adopting a model for assortments estimation from a neighbouring country. However, only few such models exist in Europe, thus this will certainly not be the way to develop bridging functions at European level.

4.6 Evaluation of the Recent Harmonisation and Future Work

The main role of ENFIN (2014) and thus, of COST Action E43 (2010) and COST Action Usewood (2014) was the establishment of a scientific knowledge sharing network on NFIs. This network worked on the harmonisation in terms of definitions, NFI methods, collection of data, and statistical methods. The focus of this work was on harmonisation and not on standardisation as defined by Köhl et al. (2000), as the latter is not a realistic option (McRoberts et al. 2012). Nevertheless, the side effects of these exchanges over the years were to (i) adopt closer approaches through replacing some national definitions by commonly agreed reference definitions, (ii) implement new methodologies of NFIs, thus being more in line with the general approaches and (iii) estimate several variables by using reference definitions and common methodologies.

The advantage of harmonisation, compared to standardisation, is to give NFIs tools that are more flexible, something that is highly relevant when serving all the numerous information needs. Harmonisation also allows parties to maintain their national data collection protocols and consistency of their sometimes very long time-series according to national definitions. This is important since many users of NFI data use the information at a national level and benefit from being able to use information according to ‘well-known’ definitions, which were developed in accordance with national information needs. Many international processes demand information on forest but the definitions of different processes may vary or even change over time as for example the Growing Stock definition between 2000 and 2005 (FAO 2000a, b, 2004). Thus, there will always be a need to harmonise information coming from “support variables” so that the requirements under different agreements are fulfilled. Still, the application of standardised protocols for the acquisition of such support variables often used in international definitions simplifies the establishment of bridging functions. Thus, inventories that concentrate on the provision of high quality support variables directly are well placed TOF report according to many different needs (Ståhl et al. 2011).

The possibilities for standardisation are restricted by the wide variability of ecological and forest conditions and their political impact. For example, the minimal crown cover for the definition of forests is dependent on ecological circumstances relevant for the growth of trees. In hot and dry regions exhaustive roots systems are necessary for sufficient water supply. Possible tree densities are by far lower than in regions with high water supply. Therefore, the minimum crown cover in such dry regions is often defined by low threshold values and these ecological conditions are mirrored in national forest policy. Thus, standardisation over large areas (e.g. all of Europe) is likely to fail.

Increased collaboration among countries offers possibilities for scientific knowledge sharing and development of adequate bridging functions. In Europe, the ENFIN network provides such opportunities (ENFIN 2014). At the international scale UNECE and FAO are likely to continue to be drivers towards harmonised forest statistics through the periodic Forest Resources Assessment exercises conducted (FAO 1998, 2000a, b, 2012 and UNECE 2000).

Even if it is a difficult exercise to harmonise data, the knowledge sharing is an important step towards common approaches.

When expansive bridges are developed based on case studies or existing information from similar regions in other countries, it is important that potential sources of bias should be carefully evaluated. If case studies are conducted these could be directed to the specific areas from which data are needed. Careful selection and application protocols need to be applied so that potential selection bias does not incur bias in the resulting estimates (Gregoire and Valentine 2008).

While this chapter has focused on the development of bridging functions for applications using forest data and information, the bridging function approach should be applicable to other requirements for harmonised statistics.

Bridging functions can be applied at different levels due to the nested sampling strategies and the hierarchical estimation process. It starts from the single tree level, and finally ends up at the highest aggregation level, i.e., the general adaption of the national estimate (Ståhl et al. 2012). In many cases, especially when simple reductive bridges are applied, it is straightforward to apply the bridging function at the level of individual sampling units and then aggregate these units in order to obtain an estimate according to the reference definition. This is not only straightforward when reductive bridges are applied, but also for expansive bridges in cases where auxiliary data is available from all potential units to be included under the reference definition. For categorical data, logistic regression functions can be applied in order to obtain probability estimates. For continuous data ordinary regression techniques (Yu and Ranneby 2007) can be applied to predict values at the level of sampling units. Imputation techniques (Tomppo and Halme 2004) may also be applied to both types of data.

In case auxiliary data at the level of sampling units are not available the bridging function generally would imply to apply a conversion factor or function to the aggregated estimate according to the national definition. Thus, depending on what data are available the bridging functions are applied at different levels. In general bridges applied at a low level (e.g. sampling units) would result in higher accuracy, although this cannot be taken as a rule.

The quality of an estimate derived using a bridging function depends on many different factors, including what bridging scheme was applied. In the case of reductive bridges, estimation of variances in general should be straightforward as standard procedures should be possible to apply. In this case the application of a bridging function is not likely to deteriorate the accuracy of the estimate. However, in some cases different results have been obtained when direct assessments of categorical variables have been made as compared to using support variables for their derivation (Gregoire and Valentine 2008). With neutral bridges, often the only difference would be that new allometric functions are applied upon existing data. Provided that these functions are properly developed, there is no reason why the accuracy of estimates should be lower than in the estimate according to the national definition. But whenever models are applied to derive variables both sampling and model errors ideally should be accounted for in the estimation of variances (Ståhl et al. 2014; Berger et al. 2014).

However, when expansive bridges are applied it is likely that the accuracy of estimates will decrease. This is due to the fact that in general rather coarse model relationships or direct conversion factors need to be applied. In such cases, stratified approaches to the variance estimation can be applied where one stratum is composed of the portion of sampling units where target data are directly available; the other units are put in the second stratum. For the latter, approximate variance estimation methods may be available; e.g., this is the case when imputation techniques are applied (Tomppo and Halme 2004).

To sum up, for future work on harmonisation based on the idea of bridging functions:

  • NFIs should try to include assessments into their protocol, which enable the derivation of high quality bridging possibilities.

  • NFIs should concentrate on the lower aggregation levels when establishing bridges.

  • A sound statistical framework and formulation of the estimation process which includes different types of bridges for the harmonisation should be developed as a basis for quality assessment.