Data from 2,396 SNFI4 plots in Navarra province located in Northern Spain were used in this study. In this area, the three bio-geographical regions, Mediterranean, Atlantic and Alpine, are well represented.

Dead wood components data were recorded in the SNFI 15 m radius subplot. The dead wood components considered were as follows: dead standing and downed trees; dead standing and downed saplings; downed coarse wood pieces; stumps and accumulations. Trees as well as shrub species were recorded and the five decay classes proposed by Hunter (1990) and Guby & Dobbertin (1996) were considered. Downed trees and saplings are recorded when the stump or thickest end is within the plot. Other dead wood components are recorded when more than 50% of the piece is inside the plot ( Alberdi et al. , 2014 ). The methodology to estimate tree volume (either standing or downed) was described in Crecente-Campo et al. , (2015) . The aim of this study was to present the dead wood volume reference values in different forest ecosystems.

Cores of the dominant trees of the three main species per plot are extracted in the SNFI4 plots (25 m radius). With this information, two nonlinear mixed age-diameter models have been developed for the species Populus alba L. and Populus nigra L. ( Eq. [1] ) in order to estimate the tree age depending on its diameter. They incorporate a random effect for each individual tree. Cores were air dried, mounted and sanded and measured with a LINTAB measuring table (Rinntech, Heidelberg, Germany) with an accuracy of 0.01 mm. A total of 1,233 records of a total of 32 trees have been used ( Fig. 2 ) to fit the mixed models with the statistical package nlme (nonlinear mixed-effects models) of the R software program (R Development Core Team, 2014). These trees were distributed in the Mediterranean biogeographical region in plots located in Central-Northern Spain (La Rioja and Madrid regions).

where t _ij is the response variable (age expressed in years), j is the observations for every ring of the tree, i is the tree level, d _ij is the predictive variable (diameter expressed in cm), ß is the fixed estimated parameter, b _i is the mixed estimated parameter and ε _ij is the error term. Ψ ₁ is the error correlation structure and σ ² is the variance.

The significance level considered was 0.05 and the nonlinear mixed models were fitted through maximum likelihook. The Akaike´s information criterion (AIC) was used to compare alternative nonlinear mixed models ( Zhao et al. , 2005 ; Wang et al. , 2007 ) and select the better models for each species (data not shown). The suitability of the models was also evaluated through the significance of the parameters of the models, the analysis of the model residuals and to evaluate the goodness of fit of the model we also used the root mean square error (RMSE) according to Willmott (1982) .

This case study was based on SNFI4 information for the four provinces which comprise the region of Galicia, in the north-west Iberian Peninsula. This region presents a dominant Atlantic climate, acidic soils and a complex topography with altitudes ranging from sea level up to 2,124 m.

To study this indicator, prior to the field work, a list of invasive species likely to be found in forested areas of the monitored province is drawn up for the additional biodiversity monitoring plots ( Table 1 ; Sanz Elorza et al. , 2004 ; Fagúndez & Barradas, 2007 ; Romero Buján, 2007 ). The invasive tree, shrub and herbaceous species previously listed were then recorded in 10 m, 5 m and 1 m radius subplots respectively. In addition, their presence in the 25 m radius SNFI plot was registered.

Data from 900 plots in the SNFI4 corresponding to Balearic Islands, located to the East of Spain, were used in this study ( Fig. 3 ). In the SNFI4, browsing impact data are recorded in the 10 m radius subplot for trees, saplings and shrub species and in the 5 m radius subplot for tree regeneration. For each species, crown cover was estimated as a proxy of browse availability. Average browsing degree, indicating browse utilization, was also recorded by species according to the classification by Fernández-Olalla et al. (2006) , based on Etiènne et al. (1995) and Aldezábal & Garín (2000) . The 6-rank browsing degree classification was as follows: (1) no browsing evidence; (2) slight browsing evidence: only a few twigs browsed; (3) low browsing intensity: plenty of twigs browsed but clearly under 50% of potential browsing biomass; (4) intense, although sustainable, browsing: plenty of twigs browsed, around 50% of potential browsing biomass; (5) high browsing intensity: consumption over 50% of potential browsing biomass and clear shaping of the original plant form; and (6) maximum browsing intensity: no or almost no green parts remain.

To study browsing preferences in the study area, the utilization of each woody species was compared with its availability through the forage ratio index ( Krebs, 1999 ; Fernández-Olalla et al. , 2006):

where w _ij : forage ratio or preference (selection) index for species i in plot j ; o _ij : browsing utilization of species i in plot j; pij : browsing availability of species i in plot j , and n : number of woody species present in plot j .

To estimate cork production at national level, plots from the SNFI2 and SNFI3 were used. Since cork-related variables inventoried in SNFI changed from the SNFI2 to the SNFI3, two different approaches were performed to obtain cork production. In the SNFI2 the variables recorded were debarking height and cork thickness. Debarking height was measured in all the trees in the plot and cork thickness was inventoried on 1 to 6 trees per plot. The cork thickness of the rest of the trees in the plot was assumed to have the same cork thickness which was equal to the arithmetic mean of the cork thickness of the trees measured. With these data the volume of cork for an individual tree ( V _cork ) was estimated through the following equation:

where dh is the debarking height (cm), ct is the cork thickness (cm) and pbh is the stem perimeter at breast height (cm). By multiplying Vcork (cm ³ ) by the average cork density, the weight of cork for every tree was obtained. The cork of each plot was calculated as the sum of the cork weight of all the trees in the plot. The number of plots considered was 2109, with a total of 10,543 cork oaks.

In the case of SNFI3, only the debarking height was measured. Since no direct estimation of the cork volume was possible with these data, a different approach was adopted based on the application of a diameter increment model to data from SNFI2 plots. Hence, to estimate the cork thickness in the trees from the SNFI3, the variables used were the diameter under cork and the cork thickness of all the cork trees from the SNFI2. Using this information, the annual diameter increment of these trees was estimated through the model developed by Sánchez-González et al. (2006) . All the variables included in this model were directly given for SNFI plots except for site index, which is estimated by the dominant height model developed by Sánchez-González et al. (2007) . With this increment data, the theoretical diameter under cork at the SNFI3 was estimated by adding the annual diameter increment in SNFI2 multiplied by the number of years elapsed between SNFI2 and SNFI3 to the recorded diameter at breast height under cork in SNFI2. The difference between the estimated diameter at breast height under cork in SNFI3 and the diameter over cork given for the plots in the SNFI3, is the cork thickness ( ct ) for the trees in the SNFI3 plots. Using the estimated cork thickness data and the dh reported in the SNFI3 plots, the weight of cork per plot was estimated according to Eq. [3] . In order to allow comparisons between SNFIs, the same plots were analyzed in each. In accordance with this criterion 2109 plots (10,543 cork oaks).

In this study, models to estimate the main canopy fuel complex characteristics in maritime pine ( Pinus pinaster Ait.) and radiata pine ( Pinus radiata D. Don) plantations in Galicia (NW Spain) were obtained by combining 554 sample plots from the SNFI4 with information based on ALS over the same plots. These variables, related to the available fuel in the aerial layer, were canopy base height (CBH), canopy fuel load (CFL) and canopy bulk density (CBD). CBH is considered the most important variable in estimating the potential of surface fires to transition to crown fires, and CBD in estimating the potential for active crown fire and crown fire intensity. CFL is used to estimate the amount of canopy material consumed in a crown fire and it is useful not only in fire behavior simulations but also in fire effects simulations ( Keane, 2015 ).

ALS information was provided by the Spanish National Aerial Photography Program (Plan Nacional de Ortofotografía Aérea, PNOA). In total, 39 metrics related to canopy cover and height distribution were extracted from ALS pulses and used as regressors for statistical analyses (see Tables 2 and 3 ). For further details of the procedure used to obtain the ALS metrics, see the steps outlined in González- Ferreiro et al. (2013) .

The average dead wood volume (DWV) for Navarra province was 8.83 m ³ /ha, presenting a high SD due to its high amount of variation (18.11 m3/ha). This volume represents 6.34% of all the wood volume of forests.

Standing and downed trees present in Navarra account for an equal percentage of DWV (35% each), while the percentage of standing saplings is slightly higher than downed saplings (3% and 2% respectively). Furthermore, the large percentage of branches (18%) in this region should also be mentioned. The dead wood was highly decompose, as classes 4 and 5 account for 60% of the total DWV.

The DWV in the three ecoregions of Navarra differed; from 14.05 m ³ /ha (SD=24.10 m ³ /ha) in the Alpine region to 10.55 m ³ /ha (SD=20.18 m ³ /ha) in the Atlantic region and 5.88 m ³ /ha (SD=12.97 m ³ /ha) in the Mediterranean region ( Table 5 ). However, the average percentage of dead wood for the three ecoregions represents between 5.4 and 5.6% of the total wood volume.

Mixed forest types generally presented higher average DWV. This is the case of "Mixed autochthonous broadleaf and coniferous forests in the Alpine biogeographical region" (20.56 m ³ /ha), "Flood plain forest" (19.31 m ³ /ha) and "Mixed forests in the Atlantic biogeographic region" (14.01 m ³ /ha). In contrast, in forests dominated by Mediterranean species, such us "Juniper forests ( Juniperus spp.)" (0.34 m ³ /ha), "Oak forests of Quercus ilex " (1.32 m3/ha) or in managed forests, the figures are lower ( Table 5 ).

For the P. alba the best model ( Table 6 ) among the tested ones had an AIC=2,767.09, being the number of observations 554. For the P. nigra the best model presented ( Table 6 ) had an AIC=3,290.31 and the number of observations 679. The residuals analysis shows a RMSE of 2.66 years for the P. alba model and 2.50 years for the P. nigra model. The simplicity of both models and the fact that they showed a good distribution of residuals ( Fig. 4 ) were considered for their selection.

Even though in terms of composition there were a similar number of invasive herbaceous (7) and invasive tree (5) species recorded, our findings suggests that there is a greater potential for invasion by invasive tree species (mainly Acacia spp.) in forest ecosystems of NW Spain ( Table 1 ). A. melanoxylon and A. dealbata displayed the greatest invasiveness, being present in 3.2% and 2.8% respectively of the total forested area of the study region and exhibiting a high number of trees/ha. Among herbaceous species, Cortaderia selloana, Phyllostachys spp. and Phytolacca americana were the most widespread in NW forest ecosystems with a high number of individuals/ha. This trend according to functional type was patent if we examine the total percentage of forest area invaded (7.4%) by functional type, with 6.9% corresponding to alien tree species and 0.5% to herbaceous species.

As regards the level of invasion per province, Pontevedra, with a high incidence of fires, showed considerably higher rates of invasion than the other regions. As regards forest vulnerability to invasion, the highest rates of invasion in native forests were found in riparian (11.3%) and Quercus robur forests (6.2%) ( Fig. 5 ). Accordingly, riparian forest, together with mixed native forests, presented the highest number of different invasive species of both trees and herbaceous plants. However, the highest degree of invasion was found in altered forests such as non-native mixed forests and Eucalyptus spp. plantations.

Fig. 3 shows the degree of browsing impact for SNFI plots in the Balearic Islands. The greatest browsing damage was found in the northern part of the Tramuntana Mountain range on the island of Mallorca, while moderate browsing was also present on the island of Menorca. Specifically, there were 695 plots in which the browsing degree was greater than 3 for any species, indicating areas that suffer significant difficulties to achieve natural regeneration.

Table 7 shows the preferences of herbivores for tree and shrub species according to forest type in the Balearic Islands.

The cork oak trees that produce cork presented larger dimensions in the SNFI3 than in the SNFI2 ( Table 8 ). However, ct was significantly smaller, hence the volume of cork was much lower in SNFI3 than in SNFI2. The estimation of ct was based on the assumption that the age of the cork (number of years since the last debarking) was the same for all the trees inventoried in the SNFI2 and SNFI3. As a result of this assumption together with the approach followed to estimate cork thickness in the SNFI3, very high negative values for cork thicknesses were obtained.

In a first step, linear and potential models were fitted to estimate the three canopy variables related to crown fire ( i.e , CFL, CBD and CBH), using different combinations of ALS metrics. However, the results obtained for CBD were highly inaccurate for both species. Therefore, we decided to model the mean height () and then estimate CBD indirectly, by dividing the estimations of CFL by the crown length, obtained as the difference between the estimates of mean height and CBH ( van Wagner, 1977 ). The best set of equations for each species is shown in Table 9 .

Linear models provided better goodness-of-fit statistics than potential models for mean height in both species, whereas potential models showed better performance for the other two independent variables (CFL and CBH).

The observed variability explained by the models ranged from 41% to 45% for CFL; from 49% to 68% for CBH and from 62% to 65% for , with better results for radiata pine stands in all cases.

The 99th percentile (h ₉₉ ) was the only predictor in all the models where the dependent variable was, directly or indirectly, related to canopy heights (CBH and ). However, the models of CFL, indirectly related to canopy fuel biomass, also included the ratio of the number of the first laser hits above mean height to the number of first laser hits for each plot (PFRAhmean).

The demands for forest information have been substantially increased and therefore the scope of NFIs has broadened introducing new variables requiring assessments (Tommpo et al. , 2010). NFIs are facing different challenges to provide strategic information for forest related policies.

One of the main challenges for the SNFI is to facilitate reliable forecasts of wood availability, which are highly important in order to define strategies for climate change mitigation or to develop strategies for promoting the use of wood as a renewable energy source ( Barreiro et al. , 2016 ). Although most of the countries in Northern and Central Europe have developed nationwide projection tools based on empirical growth models ( Barreiro et al. , 2016 ), some of the characteristics of Southern European forests hinder the development of nationwide projection systems. As regards Spain, the high variability of the forests, both in terms of composition and structure, makes the development of national scale growth models particularly challenging. Nevertheless, a large number of growth models have been developed in Spain ( Bravo et al. , 2011 ), although they are mainly regional scale models which focus on productive species in monospecific even-aged stands. Hence, these models are not suitable at national scale or for use with SNFI data, which still do not record age. SNFI data are said to have four major weaknesses for forest modelling purposes ( Álvarez-González et al. , 2013 ): excessively long inventory cycles for fast growing species; systematic inventory not allowing comparison between silvicultural alternatives in small areas; the smallest radius plot is too small for ingrowth models; and the absence of stand age.

Sampling optimization is always an important concern for SNFI. It is important to continuously evaluate the use of new technology such as ALS, the satellite constellation Sentinell of the European Union ( EC, 2014 ) or even Unmaned Aerial Vehicles (UAV), although all these technologies cannot replace fieldwork ( McRoberts & Tomppo, 2007 ). It is also crucial adequately communicate the extensive forest information to society through accessible databases, comprehensive reports and publications aimed at the different groups involved (stakeholders, managers, policy makers, etc.).

In the context of different national and international forest policy information requirements, further variables are currently being evaluated for monitoring in the ongoing SNFI cycles such as those related to non-wood products, plot soil characteristics or shrubs and trees in non-forest areas.

Additionally, European NFIs should provide comparable results as European information is obtained by aggregation (Tomppo et al. , 2010). Although many harmonization processes have already been carried out, such as the assessment of forest area ( Vidal et al. , 2008 ) or above ground biomass ( Vidal et al. , 2016 ), further studies are required, especially as regards new recorded variables.