Introduction

Growing demand for wood products, combined with efforts to conserve natural forests, has supported a 65% increase in the global extent of planted forests since 1990 (FAO 2015). As concerns Europe, in recent decades forest plantations for wood production have been supported by structural funds of European Commission (EC), to enhance the valorization of farmland: the main objectives are linked to the production of wood assortments for the timber industry, but also for biomass energy production, in compliance with the policies for climate change mitigation and adaptation. These considerations have supported a significant expansion of forest plantations (EUROSTAT 2019), especially by fast-growing tree species, like willows, poplars, eucalypts, Douglas fir, etc.

As distinctively concerns fast-growing forest plantations within agricultural farms, characterized by relatively short rotations and an intrinsic high dynamism in relation to the temporal variability of farmland destinations for cultivation, there is the need for a frequent updating of accurate and spatially detailed statistical data about tree species composition, stand structure and wood supply attributes (Wulder et al. 2008; Groot et al. 2015). Such needs may exceed the scope of conventional forest inventories, making room for ad hoc information sources (Alam et al. 2014).

Poplar is among the most used species for forest farming worldwide because of its great potential for woody production due to its rapid juvenile growth, high photosynthetic capacity, high growth performance and high wood biomass production in a single growing season (Pointailler et al. 1997). The genus Populus is also cultivated for quality wood production, as it can be used as sawn timber, veneer and plywood (Cardias Williams and Thomas 2006). Poplar includes about 50 species and hundreds of cultivars, which can be cultivated either as high stands under rotations usually between 10 and 30 years or as short rotation coppice (2–5 years) (Johansson and Karačić 2011; Barontini et al. 2014). Some clones of hybrid poplars are characterized by fast-growing rates, which are partly induced by a set of physiological intrinsic characteristics (e.g., Casella and Ceulemans 2002; Dillen et al. 2010). According to several authors, poplar biomass growth ranges from 2 to 25 Mg ha−1 year −1 (Spinelli et al. 2008; Guidi et al. 2009; Bergante et al. 2015; Stolarski et al. 2015, 2017; Krzyżaniak et al. 2019). The development of new poplar clones and the improvement of in cultivation techniques have made possible to sustain high biomass yield (Bergante et al. 2010). Poplar cultivation presents also environmental benefits such as the prevention of soil erosion and protection of soil water (Pandey and Srivastava 2010) and the habitat for many species (Walle et al. 2007; Archaux et al. 2010).

Owing to distinct features (fast growth and short rotation) and the dependency on the wood price market, hybrid poplar plantations have experienced large inter-annual fluctuation in the planted area. This calls for a frequent update of information about the geographical distribution and the potential wood supply from these plantations, since the economic value of poplar production largely depends on projections about local wood availability in the near future.

From a methodological perspective, the large-scale statistically sound estimation of fast-growing plantations requires the use of an effective probabilistic sampling scheme. Owing to the large extension of the study area, the scheme should involve more than one sampling phase. In this scenario, Marcelli et al. (2020) suggest the use of a two-phase scheme in which the first phase exploits the first phase of some conventional NFI, as suggested by Corona and Marchetti (2007). In their first phases, NFIs select a large sample of points onto the study area in accordance with a suitable probabilistic scheme. Then, these points can be checked on the basis of aerial information and those plantations spotted by at least one point are selected. Subsequently, in the second phase, a stratified sample of the plantations selected in the first phase is achieved and these plantations are visited on the ground. In this way, plantation surveys can benefit of the relatively intensive sampling usually performed in the first phases of NFIs. Moreover, taking fixed the first-phase sample of plantations selected by the NFI scheme, the second-phase sampling can allow to update information on a short time.

Regarding the first phase of sampling, spatially balanced schemes are usually adopted in the first phase of NFIs to ensure a uniform sampling coverage over the survey area (e.g., Tomppo et al. 2010). For the first phase of plantation surveys, Marcelli et al. (2020) suggest exploiting those NFIs adopting tessellation stratified sampling (TSS), a scheme that is becoming increasingly popular: actually, both the Italian and the USDA Forest Services are currently adopting TSS in the first phase of their NFIs (Fattorini et al. 2006; Tomppo et al. 2010). TSS is based on covering the survey area by means of regular polygons of equal size, each of them containing at least a portion of the area and then randomly selecting one point in each polygon (Fig. 1).

Fig. 1
figure 1

Example of plantations sampled by TSS in the first phase. Green areas represent tree plantations, and red points represent sample points. Plantations selected in the first phase are those spotted by at least one point

Full size image

The statistical properties of TSS and its superiority with respect to other schemes usually adopted for allocating first-phase points on a large-scale study area have been investigated by Barabesi and Franceschi (2011) and Barabesi et al. (2012). Moreover, the statistical appealing of second-phase strategies subsequent to TSS has been theoretically proven by Fattorini et al. (2017). In this sense, the sampling strategy proposed by Marcelli et al. (2020) is equipped by a substantial theoretical background. In the same paper, performance of the proposal has been empirically checked by means of a simulation study carried out on a set of artificial populations of poplar plantations, providing satisfactory results in terms of precision.

In this paper, we applied the proposal by Marcelli et al. (2020) for assessing the state of hybrid poplar plantations in Northern Italy. In the first phase, we exploited the TSS scheme adopted in the IUTI (from the Italian acronym of ‘Inventario dell’Uso delle Terre d’Italia’) permanent national land-use pure-panel survey (Corona et al. 2012). In the second phase, plantations selected in the first phase were partitioned by the five administrative regions of Northern Italy and a sample of them was selected by means of stratified sampling with proportional allocation.

Materials and methods

The survey here reported refers to the assessment of hybrid poplar plantations in Padan plain (Northern Italy) where poplar cultivation is mainly concentrated, targeted to plywood productions, with plantation density ranging from 200 to 300 trees ha−1 and rotations around 10–12 years. It provides high-quality wood material and represents a major source of timber for the national industry.

Survey area

The survey area is constituted by five administrative regions across the Padan plain: Piedmont, Lombardy, Emilia-Romagna, Veneto and Friuli-Venezia Giulia (Fig. 2) and extends approximately 650 km (400 miles) in an east–west direction running from the Western Alps to the Adriatic Sea, spreading over about 100,000 square kilometers.

Fig. 2
figure 2

Survey area: the Padan plain in Northern Italy with boundaries of the five administrative regions evidenced in red

Full size image

First-phase sampling

In the first phase, we exploited the TSS scheme adopted in the IUTI permanent national land-use pure-panel survey performed by the Italian Ministry of Environment and Protection of Land and Sea (for details, see Corona et al. 2012). The scheme was based on a grid of semi-kilometer quadrats covering the whole Italian territory and as many sampling points randomly selected within each quadrat. Sampling points lying within the five administrative regions of the survey area were photo-interpreted by high-resolution airborne imagery (digital orthoimages with geometric resolution ≤ 50 cm) supported by Google Earth images and tools (i.e., Google Street View). This phase envisaged checking about 390,000 sampling points. The hybrid poplar plantations spotted by at least one sampling point (Table 1) were selected and mapped. Based on the definitions by FAO/FRA (2000) and IUTI (Corona et al. 2012), we only considered those hybrid poplar plantations having a minimum area of 5000 m2 with at least one side longer than 20 meters and composed at least by 3 rows. Canopy cover and the corresponding age class (determined by the level of canopy cover, according to Lapietra et al. 1994) were assessed on the orthoimages.

Table 1 First- and second-phase sample sizes by administrative regions
Full size table

Second-phase sampling

In the second phase, plantations selected in the first phase were partitioned into strata coinciding with the five administrative regions. Then, within each of the strata a subset equal to 3% of the plantations selected in the first phase were drawn by means of simple random sampling without replacement (Table 1).

The plantations selected in the second phase were visited on the ground, recording information such as year of plantation, clone type and tree spacing. Year of the plantations was assessed by Google Earth images of different years and checked also by tree cores (No errors have been detected by the field survey carried out in the second phase with respect to the age classes of the selected plantations as assigned in the first phase according to their level of canopy cover.) Each tree in the selected plantations was enumerated, and diameter at breast height and tree height was recorded. We then reckoned the total wood volume of second-phase plantations as the sum of volumes of single trees that, in turn, were determined by means of allometric equations developed by Chiarabaglio and Coaloa (2002). The mean annual increment in wood volume was calculated for each plantation dividing the total wood volume by the age of the plantation. Finally, a digital terrain model at 20 m resolution was used to create slope, exposure and altitude classes for further elaborations.

Estimation

Any theoretical detail about estimation formulas is given in Marcelli et al. (2020). Regarding the two-phase estimation, it is worth noting that Marcelli et al. (2020) adopt a two-phase version of a one-phase estimator of population totals originally proposed in Baffetta et al. (2011) for surveying woodlot and tree rows at large scale. The estimator avoids the cumbersome quantification of the sizes of the portions of the selected plantations lying in adjacent quadrats and is suitable when the sizes of units to be sampled are small compared with the size of quadrats, in such a way that units are likely to lie within a unique quadrat. Apart from the pedagogical representation of Fig. 1, fast-growing tree plantations within agricultural farms in the survey area are smaller than 25 ha (on average, plantation size is 6 ha) and rarely straddle quadrat boundaries.

In the first phase of sampling, we estimated the total area \(|P|\) covered by plantations by means of

$$\left| {\hat{P}} \right| = \left| Q \right|r\left( P \right)$$

where \(\left| Q \right|\) is the size of the study area, \(R\) is the number of first-phase points, \(R\left( P \right)\) is the number of first-phase points falling within plantations, and \(r\left( P \right) = R\left( P \right)/R\) is the fraction of first-phase points falling within plantations. We estimated the sampling variance by means of

$$\hat{V}^{2} \left( {|\hat{P}|} \right) = \frac{{|Q|^{2} }}{R - 1}r\left( P \right)\left[ {1 - R\left( P \right)} \right]$$

We distinguished poplar plantations in accordance with some criteria (e.g., clone type) and achieved separate estimates of the plantation areas for the subpopulations (usually referred to as domains). We estimated the plantation surface \(|P_{k} |\) in the \(k\)th domain by means of

$$\left| {\hat{P}_{k} } \right| = \left| Q \right|r\left( {P_{k} } \right)$$

where \(R\left( {P_{k} } \right)\) is the number of sample points falling within plantations of the kth domain and \(r\left( {P_{k} } \right) = R\left( {P_{k} } \right)/R\) is the corresponding fraction. We estimated the sampling variance by means of

$$\hat{V}^{2} \left( {\left| {\hat{P}_{k} } \right|} \right) = \frac{{|Q|^{2} }}{R - 1}r(P_{k} )\left[ {1 - r\left( {P_{k} } \right)} \right]$$

In the second phase of sampling, we estimated the total \(T_{Y}\) of any interest attribute \(Y\) by means of

$$\hat{T}_{\left( 2 \right)y} = \frac{\left| Q \right|}{R}\mathop \sum \limits_{l = 1}^{L} \frac{{n_{l} }}{{m_{l} }}\mathop \sum \limits_{{j \in w_{l} }} \frac{{y_{j} }}{{\left| {a_{j} } \right|}}$$

where \(L\) is the number of strata partitioning the plantations selected in the first phase, \(n_{l}\) is the size of the stratum \(l\), \(W_{l}\) is the sample of plantations selected from stratum \(l\) in the second phase, \(m_{l}\) is the size of \(W_{l}\), \(\left| {a_{j} } \right|\) is the size of the plantation \(j\), and \(y_{j}\) is the value of the attribute for the same plantation recorded on the ground in the second phase. We estimated the sampling variance by means of

$$\begin{aligned} &\hat{V}^{2} \left( {\hat{T}_{\left( 2 \right)y} } \right) = \frac{1}{{R\left( {R - 1} \right)}}\left\{ {|Q|^{2} \mathop \sum \limits_{l = 1}^{L} \frac{{n_{l} }}{{m_{l} }}\mathop \sum \limits_{{j \in W_{l} }} \frac{{y_{j}^{2} }}{{|a_{j} |^{2} }} - R\hat{T}_{\left( 2 \right)y}^{2} } \right\} \\ & \quad + \;\frac{{|Q|^{2} }}{{R^{2} }}\mathop \sum \limits_{l = 1}^{L} n_{l} (n_{l} - m_{l} )\frac{{S_{l}^{2} }}{{m_{l} }} \\ \end{aligned}$$

where \(t_{j} = y_{j} /\left| {a_{j} } \right|\),

$$S_{l}^{2} = \frac{1}{{m_{l} - 1}}\mathop \sum \limits_{{j \in W_{l} }} \left( {t_{j} - \bar{t}_{l} } \right)^{2}$$

and

$$\bar{t}_{l} = \frac{1}{{m_{l} }}\mathop \sum \limits_{{j \in W_{l} }} t_{j}$$

Regarding the estimate of the plantation surface \(|P_{k} |\) in the \(k\)th domain, if the domain cannot be determined by photointerpretation but it can only be recorded in the field, we cannot use the first-phase estimator \(\left| {\hat{P}_{k} } \right|\). In these cases, we adopted the two-phase estimator

$$\left| {\hat{P}_{\left( 2 \right)k} } \right| = \frac{\left| Q \right|}{R}\mathop \sum \limits_{l = 1}^{L} \frac{{n_{l} }}{{m_{l} }}m_{lk}$$

where \(m_{lk}\) is the number of the plantations belonging to the domain k that were selected in the second phase from stratum l. We estimated the sampling variance by means of

$$\begin{aligned} \hat{V}^{2} \left( {\left| {\hat{P}_{\left( 2 \right)k} } \right|} \right) = \frac{{|Q|^{2} }}{{R^{2} }}\mathop \sum \limits_{l = 1}^{L} n_{l} \left( {n_{l} - m_{l} } \right) \frac{{\bar{y}_{lk} \left( {1 - \bar{y}_{lk} } \right)}}{{m_{l} - 1}} + \frac{{\left| {\hat{P}_{\left( 2 \right)k} } \right|\left( {\left| Q \right| - \left| {\hat{P}_{\left( 2 \right)k} } \right|} \right)}}{R - 1} \\ \end{aligned}$$

where \(\bar{y}_{lk} = m_{lk} /m_{l}\).

Finally, we estimated the relative standard error of any estimator by dividing the square roof of the variance estimate by the value of resulting estimate.

Results

The first-phase sampling procedure allowed to estimate the area of hybrid poplar plantations in the considered five administrative regions across the Padan plain equal to 43,400 ha, with a relative standard error equal to 2.4% (Table 2).

Table 2 Estimated area (expressed in hectares) of hybrid poplar plantations by administrative regions (percent standard error, SE %, is reported in brackets)
Full size table

Except for Piedmont, in which hybrid poplar cultivation is spread even in hilly areas (mainly between 300 and 600 m asl), most of the hybrid poplar plantations are concentrated in flat areas (below 300 m asl). About 98% of the hybrid poplar plantations is estimated to grow at lower slopes (< 5%), with more variability observed in Piedmont. Most hybrid poplar plantations grow in territorial zones under nature conservation or environmental restrictions (about 45% of the plantation area is estimated to fall within such zones.)

Table 3 illustrates the estimated area of hybrid poplar plantations according to the canopy cover and age: considering the conventional rotation of 10–12 years, over half of the poplar plantations proves to be near the technical maturity (i.e., ≥ 7 years).

Table 3 Estimated area of hybrid poplar plantations according to the canopy cover and age classes by administrative regions
Full size table

Tables 4, 5 and 6 summarize the results of the second-phase sampling, where hybrid poplar attributes have been measured in the field. Overall, tree spacing is predominantly equal or greater than 36 m2 (over 80%; Table 4). The clone ‘I-214’ (Populus x euramericana (Dode) Guiner) is the most used in cultivation: it contributes to about 77% of the total poplar planted area (Table 5), and it is mainly represented in older plantations (54% of the ‘I-214’ clone plantation area and 43% of the total plantation area).

Table 4 Estimated area of hybrid poplar plantations according to tree spacing by administrative regions
Full size table
Table 5 Estimated area of hybrid poplar plantations according to the clone type (‘I-214’; other clones) by administrative regions
Full size table
Table 6 Main wood production of hybrid poplar plantations by administrative regions
Full size table

Piedmont and Lombardy regions show the largest wood production, while Veneto shows the smallest. As expected, the productivity of hybrid poplar plantations is distinctively high: considering the plantations near the technical maturity, the annual increment in wood volume is about 700,000 m3 year−1.

The achieved results evidence the importance of poplar cultivation in Italy as a significant component of the national timber industry and stress its critical role in sustainable development strategies under a green economy perspective. The classification of poplar trees in age classes shows an uneven distribution of the areas invested annually compared to the typical rotation of 10–12 years; in particular, there is a concentration of hybrid poplar plantations close to commercial maturity, but, at the same time, the estimates of number of plantations of the first age class (1 year old) most likely indicate the beginning of a new cycle of poplar plantation expansion. From this evidence, it is possible to hypothesize that, starting from 2020, there will be a progressive reduction in the poplar wood available, due to the decrease in new plantations established in the last years, and then, there will be a new increase after 2023: if the current levels of annually planted area are confirmed in the next years, starting from 2027, at least 4000 hectares per year will be available for cutting, with about 1 million cubic meters of wood for industrial use.

Final remarks

Forest tree monitoring and assessment are rapidly evolving as new information needs arise and new techniques and tools become available. However, the exploitation of the latter, as well as their implementation within operative management processes, should be evidence-based (Corona 2018). Under this perspective, we have applied the two-phase sampling strategy developed by Marcelli et al. (2020) in order to provide a large-scale assessment of hybrid poplar plantations in Northern Italy, where most of the national forest tree crops within agricultural farms are concentrated.

From a practical point of view, a large survey area as Northern Italy (over 46,000 square kilometers) makes a complete investigation of the forest tree crops prohibitive. On the other hand, we completed the first-phase work on airborne imagery in about 60 days/man and the second-phase work in the field in about 15 days/man: such figures, compared with the estimated uncertainty below 3% for the estimation of the area covered by hybrid poplar plantations and below 12% for the estimation of their wood volume, represent a reasonable compromise between available resources and survey efficiency. As assessed during the survey, both the interpretation of remotely sensed imagery and fieldwork prove to be easier than in conventional forest inventory. Forest tree crops recognition within farmland by the high-resolution imagery is straightforward, while the accessibility of the plantation is simple, and field measurements easy. In this sense, the adopted sampling strategy is proven fully feasible to be implemented and replicated, with reduced costs when exploiting freely available remotely sensed data like Google Earth images.

On the whole, the sampling strategy can allow to update information on a short time; it has the advantage of being rapid and simple to be carried out, being particularly suited for fast-growing plantations, which rapidly alternate on agricultural lands, making conventional 5–10 years periodic inventory information temporally obsolete.

The produced information on the extent and characteristics of forest tree crops is of primary importance for developing long-term and large-scale strategies for supporting fast-growing forest tree cropping as well as the timber industry. The necessity of producing reliable estimates among countries allows obtaining harmonized and comparable information (Corona 2016; Fardusi et al. 2017). This can support the development of the sector in view of the forthcoming rural development policy: for instance, considering the importance of the aspects related to the environmental sustainability of hybrid poplar plantations in this application, it may be useful to favor there the supply possibilities on the nursery market of clone varieties with greater environmental sustainability compared to the clone ‘I-214,’ much appreciated by the plywood industry but susceptible to numerous biotic adversities and therefore subject to a relatively intense chemical phytoiatric protection.