Keywords

2.1 Introduction

The object of study in a statistical survey consists of the set of units in which one or more common characteristics are to be studied, and it is indicated by the term population. The population can be divided into study domains, i.e., its different components for which it is required to produce separate estimates (OECD, 2007). The population and its possible divisions must be identified clearly and unambiguously from the start of the survey design. The term inventory domain refers to the main object of the INFC inventory survey, which is represented by the Forest and Other wooded land as defined by the FAO Global Forest Resources Assessment (FRA) report. The divisions of the inventory domain are represented by 21 administrative units into which the Italian territory is divided. These units consist of 15 regions with ordinary statutes, 4 regions with special statutes, and the two autonomous provinces, called regions for brevity. Further divisions for which separate estimates are produced are those relating to the inventory macro-categories and inventory categories and the forest types.

Once the population of interest and its divisions have been identified, the design of a statistical survey requires the definition of a sampling plan or sampling design, i.e., the procedure to select the units to be observed: the structure of the sample, the methods for selecting the population units, their probability of inclusion in the sample itself and the fraction or sampling rate, on which the sample size depends (Fabbris, 1989). The characteristics of the study population, called parameters, are described starting with the data collected in the sample through estimation techniques, i.e., the calculation of appropriate estimators, such as the sample mean and its variance. The set consisting of the sampling design and the estimators is called the sampling strategy (Cicchitelli et al., 1997).

This chapter describes the criteria adopted to define the inventory domain of the INFC and its divisions. The sampling design and the national grid underlying the localisation of observations and measurements on the territory are then described. These correspond with the points or plots of different shapes or sizes named in the complex sampling units, which are listed and described in the concluding part of the chapter. The estimators and the procedures used to calculate the inventory estimates are the subject of Chaps. 5 and 6.

2.2 Inventory Domain and Classification System

Starting with the second Italian forest inventory INFC2005, the areas characterised by a forest cover responding to the definitions of Forest and Other wooded land elaborated on the FAO Global FRA2000 survey (FAO, 1998, 2000) were adopted as inventory domain. These definitions were then further detailed and applied in all subsequent FRAs (FAO, 2010, 2012, 2018; FAO-ITTO, 2003), becoming a reference point for the harmonisation of forest statistics on an international scale. They are based on some objective characteristics, some of which can also be evaluated from aerial photos or other remote images, given sufficient resolution, and others that require observation on the ground for checking or for accurate evaluation. Among the former, the minimum size is determined in terms of extension and width of the areas with tree-shrub cover. Among the latter, the coverage of tree species, distinct from that of shrub species, the height of the mature subjects and the land use are the determinants.

In the INFC, the two classes of Forest and Other wooded land, which together identify the inventory domain of the INFC, are called inventory macro-categories. These are divided into more detailed classes called inventory categories, some of which are distinguished based on the height of the tree species and identified to facilitate the comparison between the estimates produced by the first Italian forest inventory IFNI85 and those of the second one INFC2005. In IFNI85, in fact, the threshold for the height of the trees at maturity is 2 m. Table 2.1 shows the definitions for the inventory macro-categories and inventory categories used in the third Italian forest inventory INFC2015. In the INFC2015 results, the inventory category of Temporarily unstocked areas, originally distinguished in the classification of vegetation by the surveyors, has been merged with that of Tall trees forest. In addition, the list of categories in Table 2.1 includes the residual category of Not accessible or not classified wooded areas, which refers to all situations not otherwise classifiable included by convention in the processing of data, in the Other wooded land macro-category. These are formations characterised by a cover of trees and shrubs greater than 10% verified by photointerpretation, but for which a more detailed classification of the vegetation is lacking, mainly due to their inaccessibility by the surveyors.

Table 2.1 Inventory macro-categories and inventory categories that identify the divisions of the inventory domain in INFC2015 / Macrocategorie e categorie inventariali che individuano le ripartizioni del dominio inventariale INFC2015

The INFC classification system includes two further levels in addition to those described above, the forest types and subtypes, respectively, which identify the inventory domains in greater detail. The formations of the inventory categories Tall trees forest, Short trees forest, Sparse forest and Scrubland are classified according to 17 forest types, Plantations according to 3 other forest types and Shrubs according to 3 further forest types (Table 2.2). The forest types are therefore divided into subtypes, for a total number of 68 for the woods, 7 for the Plantations and 16 for the Shrubs, which are distinguished based on the dominant species or according to ecological criteria. For example, the subtypes Subalpine Norway spruce forests and Mountain Norway spruce forests are distinguished on the basis of an ecological criterion, while those of the forest type of Mediterranean pines (Pinus pinaster forests, Pinus pinea forests and Pinus halepensis forests) differ in species.

Table 2.2 Forest types and subtypes and corresponding codes CORINE Biotopes and EUNIS / Categorie e sottocategorie forestali e corrispondenti codici secondo le classificazioni CORINE Biotopes ed EUNIS

The forest type is assigned according to the dominant species or group of species in terms of crown coverage, the latter assessed in a neighbourhood of the inventory point with an area of approximately 2000 m2 (AdS25, cf. Chap. 4). The prevailing group of species is first determined as conifers, deciduous broadleaved or evergreen broadleaved, and then the forest type based on the prevailing species of the group, again in terms of crown coverage. In INFC2015, in the absence of a classification in the field of the forest type, the inventory point was assigned to the not classified class. The classification system with the related forest types and subtypes (Pignatti, 2003) was defined during the planning of the second Italian forest inventory and also remained unchanged in the third.

2.3 Sampling Design

The design of the Italian NFI includes three sampling phases, with samples extracted according to a stratified sampling (Fattorini et al., 2006; Gasparini & Di Cosmo, 2016; Gasparini & Tabacchi, 2011). The first phase (or phase 1) consists of the preliminary classification of land use and land cover through the photointerpretation of orthophotos (cf. Chap. 3) at over 301,000 points, one for each mesh of the 1 km × 1 km grid in which the national territory has been divided (cf. Sect. 2.4). This sampling scheme is called tessellated sampling (Särndal et al., 1992). The points are positioned randomly within the meshes and therefore their distribution is of a non-aligned systematic type (Gallego, 1995). The purpose of the first phase is to identify the sampling points of the strata of interest for the subsequent survey phase, represented by the classes of forest land use and cover (Forest formations, Sparse forest formations, Temporarily unstocked areas and Plantations; cf. Chap. 3) and from the class of points not classifiable from orthophotos. Furthermore, for INFC2015, additional strata of interest have been identified, as described below.

The second sampling phase (or phase 2) involves a subsample of the first phase points, over 30,000, selected according to a sampling stratified by region and class of land use and land cover. The points of the second phase sample are visited on the ground, to verify the preliminary classification by photointerpretation, confirming it or not, and to assign the inventory category and the forest type. During the second phase surveys, the qualitative characteristics of forests are also evaluated and classified (cf. Chap. 4), in order to produce estimates of the distribution of the wooded area according to the different characteristics (composition by species groups, degree of coverage, silvicultural system, stage of development, management methods, presence of constraints and protected areas, characteristics of forest stations, road conditions, etc.). These will be presented in Chaps. 7, 8, 9, 10 and 11 of this volume.

The third phase strata are identified by the forest type assigned in the second phase together with the land use and cover class, and the region. For each Forest stratum, a subsample of second phase points is extracted to carry out measurements envisaged for the third sampling phase (phase 3) relating to the quantitative characteristics of forests (cf. Chap. 4). The measurements performed on approximately 7000 points produce the estimates of the totals and densities, or values per unit area, of the quantitative variables presented in Chaps. 7, 9 and 12 of this volume. These include growing stock, biomass, annual volume increment, deadwood biomass, etc., which are important for assessing the state of Italian forests and their role as a carbon pool and biodiversity reservoir.

The sampling design described above, outlined for the second Italian NFI, was also applied for INFC2015 with some adaptations. In fact, with the operational planning, the indications from the beginning of the operational structure responsible for the coordination and implementation of the INFC were incorporated. They constituted organizing the surveys in a single campaign, including as many points already detected in the previous survey as possible in order to reduce the time that the survey was on the ground and to facilitate its organization, while maintaining an unchanged quantity, quality, and level of detail of the produced estimates.

The sample for the new ground survey was then constituted by including: (i) the points of the third phase INFC2005 not affected by significant changes in land use and cover, (ii) the points of the second phase sample INFC2005 affected by significant changes and (iii) an additional sample, stratified by region, for strata consisting of new points in wooded areas and plantations and new non-classifiable points. The significant changes were highlighted by comparing two photointerpretations of INFC2005 and INFC2015 and are related to the points transited from strata of inventory interest to the stratum “other land uses” or vice versa. The outcome of the comparison is shown in Fig.2.1 for the set of 301,271 points on the national territory according to the opinion of the photo interpreters, out of a total of 301,328 points identified by the national grid (cf. Sect. 2.4).

Fig. 2.1
figure 1

Comparison between the results of the photointerpretation of land cover and use in the last two Italian national forest inventories: number of photoplots / Confronto tra i risultati della fotointerpretazione dell'uso e copertura del suolo dei due inventari forestali italiani più recenti: numero di fotopunti

The actual change from forest use and cover to other land uses was verified on a subsample of 2338 second phase points. Of these, almost half belonged to other land uses in the second phase of INFC2005, and therefore it was not necessary to carry out a further verification in the field. The remaining 1303 points were instead detected on the ground during the INFC2015 campaign or classified through a photointerpretation that included more recent images, if they were inaccessible to the surveyors. Points in the second phase sample not affected by significant changes in land use and cover represent the greatest part. For them the third phase stratum attributed during the INFC2005 surveys was considered valid and the new surveys concerned 6597 points of the third phase subsample. Finally, the extracted subsample from the new points transited in strata of inventory interest consisted of 874 points. The sampling rate for these strata is lower than that adopted for the other ones, due to the limits mentioned above. However, it was considered suitable for evaluating the actual change in land use and cover and ensured a balanced subsample compared to the overall sample of points selected for the field survey.

Overall, the INFC2015 second phase sample consisted of 30,877 points of which 8774 were classified for the qualitative characteristics during the new survey campaign or, if inaccessible, evaluated remotely or through recent orthophotos, in order to verify the land use and cover and the forest type. Of these, 6993 points were measured to estimate the quantitative characteristics and represent the third phase sample of INFC2015. The composition of the sample selected for the INFC2015 field campaign is represented in Fig. 2.2; the largest group is that of the sample units already recorded both in phase 2 and in phase 3 INFC2005, which represents more than three-quarters of the total.

Fig. 2.2
figure 2

Distribution of the sample selected for the INFC2015 field campaign in relation to the previous field surveys of INFC2005 / Distribuzione del campione selezionato per la campagna di rilievo INFC2015 in relazione ai precedenti rilievi in campo INFC2005

The number of sample units in the three phases and the resulting sampling rates by region and at a national level are shown in Table 2.3 together with the area represented by each sample point or plot. At the national level, the sampling rate of phase 2 and phase 3 is equal to 0.26 and 0.30, respectively, and the area represented by each sample unit is 3.9 km2 for phase 2 units and 13.0 km2 for phase 3. In the second phase, the sampling rate at regional level varies little, from 0.23 for Lombardia to 0.28 for Trentino, and the area represented by each sample unit varies from 3.7 km2 in Toscana to 4.6 km2 in Lombardia. In the third phase, the regional sampling rate varies from 0.23 for Toscana to 0.44 for Valle d'Aosta and Molise, while the area represented by a sample unit varies from 8.7 km2 in Valle d'Aosta to 15.8 km2 in Toscana.

Table 2.3 Number of INFC2015 sample units selected and surveyed and their representativity, by sampling phase and region / Numero di unità campionarie INFC2015 estratte e rilevate e relativa rappresentatività, per fase di campionamento e regione

2.4 National Grid

The grid used to subdivide the national territory into portions of equal area and identify the points of the INFC sample was built during the design of the second Italian inventory INFC2005. It contains quadrangular meshes, 1 km × 1 km, geometrically coupled to meridians and parallels. Such meshes, although square in plane projection only near the meridian of origin, comply everywhere with the requirement of identical area (1 km2) of each mesh (Fig. 2.3).

Fig. 2.3
figure 3

Pieces of an ellipsoidal grid in three different zones of Italy: a North-Western zone, b along the meridian of origin, c South-Eastern zone. The different shape of the meshes can be observed, each having the same area / Porzioni di grigliato ellissoidico in tre zone diverse del territorio nazionale: a zona nord-occidentale, b centrale lungo il meridiano d’origine, c sud-orientale. Si osserva la diversa forma delle maglie, tutte aventi medesima superficie

The WGS84 datum (DMA, 1991) was adopted for the generation of the grid nodes, due to its perfect congruence with the GNSS systems that would have been used later for finding and positioning the points in the field. After establishing a pair of geographical coordinates, \(\rm{\varphi }_0 ,\rm{\lambda }_0\), at the South-West corner of the quadrangle which contains the whole national territory, further nodes of the grid have been generated moving 1000 m eastwards along the same parallel. Once the South-East node of the quadrangle is reached, the procedure starts again from the initial node, this time moving 1000 m North along the meridian and repeating the previous step. The procedure was repeated until the quadrangle containing the national territory was completed, ending in the last North-East node (detailed algorithms available in Floris & Scrinzi, 2011).

Intersecting the grid with the border of the national territory, the actual inventory grid was identified and consisted of 306,831 quadrilateral meshes, with side 1000 m on the ellipsoid surface. By converting the geographical coordinates of the grid nodes into plane coordinates in the UTM-WGS84 reference system and recalculating the surface of each mesh with the Gauss formula for a closed polygon, deviations of less than one square metre were consistently obtained with respect to the nominal value of 1 km2.

To randomly select a sample point for each grid mesh, a pair of angular positive values \(\rm{ \delta \varphi ,\delta \lambda }\) has been added to the coordinates of the South-West node of each mesh in which a portion of the national territory is included (Fig. 2.4). The pair of angular values were randomly generated within the limits of 1000 m from the node itself. The sample points have again been intersected with the official national administrative borders, obtaining the final sample of 301,328 points.

Fig. 2.4
figure 4

A piece of inventory grid in plane projection, with a mesh detail, a sample point (red dot) and \(\rm{\delta \varphi ,\delta \lambda }\) from the SW node / Porzione di grigliato in proiezione piana, con dettaglio di una maglia, il relativo punto di campionamento (punto rosso) e \(\rm{\delta \varphi ,\delta \lambda }\) dal nodo Sud-Ovest

The plane coordinates of the sample points have also been transformed into the reference systems UTM-ED50 and Gauss-Boaga-Roma40 (aka Italy-Monte Mario) (Cima et al., 2003) in such a way as to make the points identifiable on any kind of map or mapping tool used during the inventory surveys. As described in Chap. 3, the reference system Gauss-Boaga-Roma40 was adopted in the WebGIS platform Geoinfo and used for photointerpretation (cf. Chap. 3).

2.5 Reference Units for the Survey

In this text, the term sampling unit indicates the different reference units (i.e., points and plots) used to collect information on the numerous attributes detected by the INFC in different ways and at different phases of the survey. These comply with the expected accuracy goals and are in relation to the available financial and time resources.

In the first phase for the photointerpretation of land use and land cover, the sampling unit is the sample point with an analysis window consisting of nine contiguous squares, of 50 m side, arranged according to a 3 × 3 scheme, for a total area of 22,500 m2 (Gasparini et al., 2014). The central square of the analysis window is centred on the sample point. The analysis window is functional to the assessment of the dimensional thresholds necessary for the correct application of the definition of Forest and Other wooded land adopted (cf. Sect. 2.2 and Chap. 3).

In the ground survey, besides the check of the classification of land use and land cover, many qualitative characteristics are assessed, including those related to a more detailed classification of the vegetation (Gasparini et al., 2016). The administrative and regulatory attributes, height and distance from the closest road are measured with reference to point C. Coverage of tree and shrub crowns is attributed by observing, both on the ground and in orthophotos, the central quadrant (50 × 50 m) of the analysis window used in the photointerpretation, called photoplot 2500 (FP2500). A circular area centred on point C, inscribed in the same central quadrant and with a radius of 25 m, known as the ground reference area (AdS25, area approximately 2000 m2) is the sampling unit for evaluating accessibility, descriptive characteristics of the vegetation, site characteristics, and the health status of the stand. In AdS25, further observations on qualitative characteristics related to management and silvicultural practices are also conducted (Fig. 2.5).

Fig. 2.5
figure 5

Sampling units in INFC2015 and related attributes to be classified and measured, in brackets / Le unità di riferimento per i rilievi INFC2015 con indicazione degli attributi da rilevare per ciascuna di esse, tra parentesi

The third phase of INFC is mainly addressed to collect the quantitative data necessary for estimating dendrometric parameters (cf. Chap. 4). In INFC2015, the third phase measurements were carried out simultaneously with the classification of the qualitative characteristics in all points visited (cf. Sect. 2.3). Measurements are performed on circular sample plots, useful in maximising the area/perimeter ratio and reducing the number of edge elements (de Vries, 1986). Moreover, they can be easily instituted in the field by measuring the distance from the plot centre. Two concentric plots centred on point C, with a radius of 4 m and 13 m respectively (AdS4, area 50.27 m2 and AdS13, area 530.93 m2), are used to select the standing trees of different size to be measured. AdS4 is used for measuring trees with DBH between 4.5 and 9.4 cm, and AdS13 is used for trees with DBH equal or larger than 9.5 cm. Tree heights, incremental cores, deadwood lying on the ground and stumps are measured and collected in AdS13. In addition, two circular sub-areas with a radius of 2 m (AdS2, area 12.57 m2), located 10 m eastwards and westwards from point C are dedicated to tree regeneration and shrub surveys.