High Nature Value Forests in Ireland: A Typology

ABSTRACT

High Nature Value (HNV) forests support biodiversity and the provision of ecosystem services. In recognition of this role, a framework for defining and identifying HNV forest in Ireland has recently been developed and applied to Irish National Forest Inventory (NFI) forest plots. This study aimed to identify different types of HNV forest, using available variables collated by the NFI for the same plots identified as HNV forest. A multiple factor analysis and cluster analyses was performed. A spatial analysis exercise was conducted to quantify the number of HNV forest plots that overlap with designated/protected areas in Ireland. Three primary types and four subtypes of HNV forest were identified. Tree species richness, fern species richness and planting year, woodland subtype, woodland habitat type and nativeness of tree species were the most important variables explaining the difference among the three main forest types identified. Areas with distinct ecological characteristics can fall within the same HNV score range and only a small variation in naturalness is seen among HNV forest types. Results show that less than 50% of NFI HNV forest plots overlapped with designated areas/protected habitats. Results of this work can be an important step for the design of targeted conservation policies.

INTRODUCTION

Across Europe, High Nature Value (HNV) forests support biodiversity and the provision of ecosystem services (Paracchini and Oppermann 2012; Keenleyside et al. 2014; Lomba et al. 2014; EEA 2014; Moran et al. 2021). Despite their vital role in supporting biodiversity, many HNV areas occur outside of protected areas (Matin et al. 2020). In recognition of their value and the historic lack of policy protection, since the early 1990s European and national policies began to focus on the need to protect and maintain HNV systems (Carlier et al. 2023). In 2005, the assessment and monitoring of the extent and quality of HNV farmland and forest at EU Member State level was included in the Common Monitoring and Evaluation Framework (CMEF)—as baseline and impact indicators—of the Rural Development Programme (part of the Common Agriculture Policy (CAP); e.g. IEEP 2007).

The monitoring requirements set out in the EU Performance Monitoring and Evaluation Framework (replacing the CMEF for the CAP 2023–27) includes an indicator which makes specific reference to HNV areas but not a standalone HNV indicator. This might be related to both a lack of scientific stringency behind the concept (Strohbach et al. 2015) and the acknowledged diversity of rural landscapes across the EU—making the use of harmonised indicators difficult (Lomba et al. 2015). In comparison to many other European countries, HNV farmland in Ireland has been relatively well studied and defined (Benedetti 2017; see Boyle et al. 2015; Matin et al. 2016, 2020; Sullivan et al. 2017; Moran et al. 2021). This may be one of the reasons why the Irish Government maintained the inclusion of HNV farming systems in national policy. The CAP Strategic Plan Regulation 2021/2115 highlights that: ‘The CAP should play a role both in reducing negative impacts on the environment and climate, including biodiversity, and in increasing the provision of environmental public goods on all types of farmland and forest land (including high-nature-value areas) and in rural areas as a whole’ (included in the final version of the Key Environmental Targets in Ireland’s CAP Strategic Plan 2021; DAFM 2021).

Since the publication of the EEA (2014) technical report that presented a methodology for the development of a HNV forest indicator, few studies on this theme have been published at either Irish or European level (see a list of studies in Ruas et al. 2023). Furthermore, because the HNV forest concept has only recently been developed for the Irish context (Ruas et al. 2023), it is unknown how these forest areas overlap with habitats listed in the Habitats Directive and other protected areas, and whether there is a difference in ecological characteristics and/or management that can be attributed to occurrence in a protected area.

In the framework for defining and identifying HNV forest in Ireland, the EEA (2014) definition of HNV forest was adopted: ‘forest dominated areas which, in the continuum gradient of naturalness, are located close to natural conditions’—and a Nature Value (NV) index was proposed for the identification of HNV forest. The NV index is based on six indicators of forest naturalness collected by the National Forest Inventory (NFI), namely, 1. nativeness of tree species composition; 2. tree distribution; 3. development stage; 4. age (from planting year); 5. thin status, and 6. evenness of stand age (see Ruas et al. (2023) for a complete description). The NV index ranges from 0 to 11, with HNV forests at the upper end of the naturalness spectrum (i.e. scores >8). Ruas et al. (2023) showed that, broadly, HNV forests in Ireland are largely composed of native tree species, are more than 30 years old (from planting year) and have no obvious signs of thinning management.

The broad category of HNV forest incorporates a wide range of forest types, each providing specific ecosystem services, and requiring specific conservation and management strategies. Categorising types of HNV forest can help ensure that specific HNV forest types are not overlooked, and that conservation and management practices are tailored appropriately. In a comprehensive study on HNV farmland typologies, Sullivan et al. (2017) highlighted that identification of typologies is particularly useful for grouping HNV areas with similar characteristics. This knowledge can then be used to implement practical and tailored ‘actions that can support positive management practices’ (in Sullivan et al. 2017).

Several approaches for the identification of HNV forest typologies can be followed. For example, HNV forest types can be classified following a framework similar to that recommended for HNV farmland types (Andersen et al. 2003)—i.e. separated into Type 1, Type 2 or Type 3, based on the proportion of semi-natural vegetation, farming intensity and rare species presence, respectively. This approach was followed by the Scottish Government (2011), which categorised HNV forests into type A, B and C, mirroring the HNV farmland typology. Type A corresponded to forest areas dominated by semi-natural features and low intensity management (native woodlands with >50% canopy of native species—semi-natural and planted); type B related to forest areas with high diversity of habitats and low intensity management (mostly broadleaved and mixed woods: mainly non-native); and type C related to mixed and non-native forest areas where species of conservation concern are present (see The Scottish Government (2011) for a detailed description). The NV index for identification of HNV forests (proposed by Ruas et al. 2023) is largely based on naturalness variables, thus it can be assumed that the HNV forests identified by the authors will broadly correspond to the type A HNV forest (forest with a higher percentage of native tree species), with only a few corresponding to type B or type C. The Scottish Government HNV forest classification (2011) was based on the IEEP (2007) definition, which relates more directly to biodiversity/conservation value since it was adapted from the HNV farmland concept (EEA 2014). The EEA (2014) definition of HNV forest (applied by Ruas et al. (2023) for Ireland) is based on the naturalness concept, making a direct translation of the HNV farmland types (type 1, 2 and 3) to HNV forest less appropriate. As a result, in this instance, the identification of typologies of HNV via statistical clustering based on multiple variables is deemed more appropriate. Similar approaches were proposed and implemented by Sullivan et al. (2017) and Mądry et al. (2020) for describing HNV farmland types in Ireland and Poland, respectively. Perrin et al. (2008) focused on identifying broad groups of semi-natural woodland in Ireland by vegetation types.

The main objective of this study was to identify types of HNV forest through a statistical approach using NFI variables available at plot scale. We expect to also be able to identify the general attributes, threats and pressures that are specific to certain HNV forest types. Through investigation of the degree of overlap with protected areas per HNV forest type, it will be possible to better understand whether, and to what extent, all HNV forest types are recognised, even if not directly, within existing conservation policies.

METHODOLOGY

SELECTION OF HNV FOREST PLOTS AND NATIONAL FOREST INVENTORY VARIABLES

The definition of forest in this study was in line with that of the (Irish) Forest Service (2018): ‘land under trees with (a) a minimum area of 0.1ha, (b) tree crown cover of more than 20% of the total area, or the potential to achieve this cover at maturity’. The Forest Service provided the complete NFI dataset, which consists of detailed field data collated for 1923 permanent NFI plots. NFI plots are circular areas (25.24m in diameter) that are surveyed on a cyclical basis for a range of variables, with data gathered on forest plot management, soil type, floristic richness etc. (see complete list in: https://www.gov.ie/en/publication/823b8-irelands-national-forest-inventory/). The NFI dataset contained 30 variables which are listed and summarised in Table S1 (in Supplementary Materials (SM)). Ruas et al. (2023) calculated NV score (scores range from 0 to 11) for each NFI plot. This resulted in a total of 178 NFI plots being identified as HNV forest (i.e. those plots with NV scores above 8). Their distribution in Ireland is shown in Fig. S1 (SM).

The data provided by the Forest Service (Table S1) were used to determine whether it would be possible to identify and describe different types of HNV forest in Ireland at a plot scale. Due to a lack of information for some of the 30 variables in the 178 HNV plots, 117 HNV plots were used for statistical analyses (their distribution in Ireland is shown in Fig. S5 (SM)).

STATISTICAL ANALYSES FOR HNV FOREST TYPE IDENTIFICATION

A multiple factor analysis of mixed data (FAMD) and cluster analyses (CA)—was used to identify HNV forest typologies, following the approach used in Sullivan et al. (2017) for HNV farmland. FAMD is a dimensionality reduction method dedicated to analysing a dataset containing both quantitative and qualitative variables (Pagès 2004). FAMD is regarded as a combination of a Principal Components Analysis (PCA) for continuous variables and Multiple Correspondence Analysis (MCA) for categorical variables (Husson et al. 2010). The FAMD was conducted on all 117 NFI plots for all 30 variables, with the function FAMD() from the package FactoM-ineR (R package; Le et al. 2008). Following an initial run of the FAMD and CA, two of the 117 plots were identified as outliers and were subsequently removed, thus a total of 115 plots were considered for the HNV typologies identification.

The final number of dimensions selected was based on the eigenvalue being >1 criterion (Hair et al. 2006). Due to the amount of dimensions with eigenvalue >1, a cut-off point of 60% cumulative variance percentage was used to further decide on the number of dimensions to keep. Thresholds differ across research areas (Henson and Roberts 2006), but a minimum of 60% cumulative variance is commonly accepted in order to guarantee more parsimonious models—i.e. to decide on the number of factors that best reproduce the variables (Hair et al. 2006). The FAMD resulted in a reduction of dimensionality from the initial 30 variables considered, to 14 dimensions, explaining approximately 61% of the variance (eigenvalues and cumulative variance in Table S2 (SM)).

A CA was used to group NFI plots based on their similarity and was carried out using the HCPC() function in R. This function performs an agglomerative hierarchical clustering (using the Ward’s criterion) on results from the factor analysis (Husson et al. 2020) and was applied to the FAMD results (component scores for the 14 most important axes of the FAMD). The final number of clusters was decided based on examination of the dendrogram and factor map outputs (using the ‘factoextra’ R package (Kassambara and Mundt 2020))—in particular, a minimum overlap between clusters in the dendrogram and factor map was the criterion for selecting the final number of clusters. The mean and standard deviation value for each variable were used to describe the types (clusters) of resulting HNV forest.

Since each forest plot has an associated NV score, we summarise the variation of the NV scores per cluster (boxplot) and conducted a Kruskal-Wallis test to determine whether there were significant differences between the forest types identified.

To identify ecologically distinct subtypes within the broad types of HNV forests, we separated the plots belonging to each of the dominant three clusters and reran a separate FAMD and a CA for each cluster, following the same criteria for the final number of dimensions and clusters selection. For the plots belonging to cluster 1, the FAMD resulted in a total of 7 dimensions explaining approximately 60% of the variance (from the initial 30 variables considered; see eigenvalues and cumulative variance in Table S3). The CA was thus performed over the FAMD results for 7 dimensions. For the plots in cluster 2, the FAMD resulted in a total of 11 dimensions explaining 62% of the variance (eigenvalues and cumulative variance in Table S6). The CA was conducted over the FAMD with 11 dimensions. Once again, a minimum overlap between clusters in the dendrograms and factor maps was the criterion for selecting the final number of clusters. Since there were only eleven NFI plots within cluster 3, no additional analyses for subtypes identification were performed. All statistical analyses were conducted in R v 4.0.2 (R Core Team, 2020).

Photographs provided by the Forest Service allowed for a visual representation of each forest type and the distribution of the HNV forest plots per typology is shown in Figure S5 (SM).

INTERSECTION WITH PROTECTED AREAS

The 115 NFI plots that have been identified as HNV forest by Ruas et al. (2023) and used for typologies identification in this work, were spatially intersected (QGIS) with available NPWS datasets that refer to protected areas and habitats of relevance: Annex 17 habitats (91A0_Old_Oak_Woodlands; 91D0_Bog_Woodland; 91E0_Residual_Alluvial_ Forests; 91E0_Residual_Alluvial_Forests); Special Areas of Conservation (SAC_ITM_2024_01); Special Protection Areas (SPA_ITM_2023_05) and Natural Heritage Areas (NHA_ITM_2019_06). All layers were obtained from the NPWS website (https://www.npws.ie/maps-and-data). A HNV forest type, based on the results of the HNV forest type identification, was assigned to each of the 115 NFI plots.

From the results of this spatial analysis, the following were calculated: 1) the percentage of all 178 HNV forest plots that intersect with any protected or designated areas; 2) the percentage of the 115 HNV forest plots used for analyses in this study that intersect with any protected or designated areas; and 3) the percentage of NFI plots, per type and sub-type, in relation to all identified plots per type, that overlap with any protected or designated areas. Even if only a portion of the NFI plot intersected with any of the listed NPWS layers, it was considered a positive overlap.

RESULTS

HNV FOREST TYPES AND SUBTYPES

Three distinct clusters were identified in this study: cluster 1 comprised 30 plots, cluster 2 comprised 74 plots and cluster 3 comprised 11 plots. The factor map (see Fig. 1) highlighted that there was minimal overlap between HNV forest plots for the three clusters. The three most important quantitative variables explaining the difference between clusters were tree species richness, fern species richness and planting year (Table 1); and the three most important qualitative variables explaining the difference among clusters were woodland subtype, woodland habitat type, and nativeness of tree species (Table 2).

Within cluster 1 (n = 30 plots), two distinct subclusters were identified (see factor map in Fig. S3). Subcluster 1.1 included 22 plots and subcluster 1.2 included 8 plots. The most important variables distinguishing these subclusters were bryophyte cover, elevation, forest subtype and mixture (all significant variables are shown in Tables S4 and S5).

Within cluster 2 (74 plots in total) we identified two distinct subclusters (see factor map in Fig. S4). Subcluster 2.1 included 19 plots and subcluster 2.2 included 55 plots. The most important variables distinguishing these subclusters were elevation, tree species richness, soil type, and soil group (all significant variables are shown in Table S7 and S8).

We summarise the characteristics of the three clusters and four subclusters obtained (hereafter simply referred to as types and subtypes of HNV forest), based on the mean values of statistically significant NFI variables. Figure 2 provides a visual representation of each forest type and subtype: i.e. a photograph of a NFI plot representative of each HNV forest subtype and of Type 3 HNV forest.

Type 1 HNV forest - Regenerating bog/wet-woodlands

Relative to the other two types of HNV forest identified, plots in Type 1 are mainly young forest areas (30–50 years) of lower tree height; with lower diversity of tree species (average of 3 species) in comparison to HNV forest Types 2 and 3; with a low to medium diversity of different vegetation layers (ferns, herbs, grasses and shrub species); mainly in peaty, minerotrophic (wetlands with a peat soil of more than 40 cm thick) and alluvial soils. Most of the plots are characterised by being mostly single species (i.e. the dominant species occupies 80% or more of the canopy) and uniform stands (NFI variables: Forest subtype and tree species Mixture – Table S1). In Table S9 a complete statistical description of Type 1 HNV forest is provided.

Subtype 1.1 – Bog woodlands: relative to Subtype 1.2, plots belonging to this HNV forest subtype have higher cover of bryophytes, occur at higher altitudes, predominately in peat soils. Table S10 provides a complete statistical description of the subtype.

Subtype 1.2 – Riparian/wet woodlands: relative to Subtype 1.1, plots belonging to this HNV forest subtype have lower tree species diversity, lower bryophyte cover, taller trees and occur in alluvial soils. Table S11 provides a complete statistical description.

Type 2 HNV forest – Diverse semi-natural forest

Relative to Type 1 and Type 3 HNV forests, plots in HNV forest type 2 have higher diversity in vegetation layers (ferns, herbs, grasses and shrub species), higher diversity of tree species and higher bryophyte cover. These NFI plots are older (on average) than Type 1 HNV forest; they are classified as semi-natural woodland (NFI variable: Woodland habitat) and the most common forests belonging to this HNV forest type were oak-birch-holly woodlands and oak-ash-hazel woodlands; mainly occurring in mineral soils, brown earths, brown podzolic soil and lithosols. Most of the plots are characterised by being individually mixed (i.e. there is more than one species present, with the species mixture occurring in a random manner). See Table S12 for a complete statistical description of HNV forest type 1.

Subtype 2.1 – High altitude diverse forest on shallow mineral soils: relative to Subtype 2.2, these plots are at higher altitudes and on shallower soils (mineral soils (<30 cm)). These plots have relatively lower tree species richness (see Table S13).

Subtype 2.2 – Low altitude diverse forest on deeper mineral soils: by comparison with Subtype 2.1, plots in this subtype had higher tree species diversity, occur at lower altitude and on deeper mineral soils. See Table S14 for a complete statistical description of the cluster.

Type 3 HNV forest - Modified HNV forest

These HNV forest plots are characterised by higher tree species diversity and lower grass cover than Type 1 and 2; are classified as highly modified (NFI variable: Woodland habitat) and a have a higher number of non-native species (but mostly less than 20%) than the other HNV forest types; most were established through reforestation (NFI variable: Establishment type) and are classified as mixed (NFI variable: Woodland subtype). See Table S15 for a complete statistical description of the cluster.

DIFFERENCES IN NATURE VALUE SCORES

Type 2 HNV forest plots (Diverse semi-natural forest) showed the highest median NV scores (median = 10.50), followed by plots belonging to Type 1 (median = 10.25) and finally by plots belonging to Type 3 (median = 10.00) (Fig. S2). However, results of the Kruskal-Wallis test (χ² = 5.34; d.f. = 2; P-value = 0.07) indicate that differences between the medians are not statistically significant.

INTERSECTION WITH PROTECTED AREAS

The spatial analysis indicated that approximately one third of the total 178 HNV forest plots intersected with a protected or designated area (32%). Of the 115 HNV considered for analyses in this study, 41 plots intersected with a protected or designated area (36%). Of the 30 plots belonging to HNV forest Type 1, 33% (10 NFI plots) intersected with a protected or designated area. Of these 10 intersecting plots, 7 are Subtype 1.1 (Bog woodlands) and 3 are Subtype 1.2 (Riparian/wet woodlands).

Of the total 74 plots assigned to HNV forest Type 2 (Diverse semi-natural forest), 40.5% (30 NFI plots) intersected a protected habitat/designated area. Of these 30 intersecting plots, 9 are Subtype 2.1 and 21 are Subtype 2.2.

One out of the 11 plots classified as Type 3 HNV forest (Modified HNV woodland, 9.1%) intersected a protected habitat/designated area.

Table 1.

Most important quantitative variables explaining the difference between clusters (High Nature Value plots from the Irish National Forest Inventory. Eta2 – square correlation coefficient. Significance level: *** P < 0.001; ** P <0.01; * P <0.05.

Fig. 1.

Factor map of the three main clusters from the 115 Irish NFI plots classified as HNV. Cluster 1, red, corresponds to Regenerating bog/wet woodlands; cluster 2, green, corresponds to Diverse semi-natural forest; cluster 3, blue, corresponds to Modified HNV forest.

Table 2.

Most important qualitative variables explaining the difference between clusters (High Nature Value plots from the Irish National Forest Inventory) (χ² results). P-value – significance test (< 0.05). d.f. – degrees of freedom.

Fig. 2.

Photographs of NFI plots representative of HNV forest types and subtypes (provided by the Forest Service). Photograph A: NFI plot of HNV forest Subtype 1.1. — Bog woodland; Photograph B: NFI plot of HNV forest Subtype 1.2 — Riparian/wet woodlands; Photographs C: NFI plot of HNV forest Subtype 2.1. — High altitude forest on shallow mineral soils; Photograph D: NFI plot of HNV forest Subtype 2.2 — Low altitude forest on deeper mineral soils; Photographs E: NFI plot of HNV forest Type 3 — Modified HNV forest.

DISCUSSION

This work provides an ecological characterisation of the NFI HNV forest plots, that goes beyond the six indicators used for calculating the NV index (Ruas et al. 2023). Through a statistical approach, the variability within HNV forest in Ireland was identified at NFI plot level. The approach can easily be adapted for other European member states since most countries already have a systematic gridded permanent plot-level NFI in place (Pucher et al. 2022).

Results from this study indicate that there is no single, archetypical, HNV forest in Ireland. The three main types identified show that forest areas with distinctly different ecological characteristics can fall within the same NV range (i.e. 8–11) (Fig. S2). In addition, most of the NFI plots that were identified as HNV forest do not seem to be under any formal nature conservation protection, and some types and subtypes of HNV forest are less likely to overlap with a designated/protected area than others.

The main characteristics and potential threats associated with the different types/subtypes of HNV forest identified and how current conservation and management policies can be expanded to recognise all HNV forest types are now discussed. The robustness of the analyses conducted and key recommendations for future work are also included.

TYPES AND SUBTYPES OF HIGH NATURE VALUE FOREST IN IRELAND: MAIN CHARACTERISTICS AND THREATS

Using available data from the NFI, it was possible to categorise HNV forests into three main types and four subtypes of ecologically distinct groups. The clusters obtained are primarily separated by differences in plant species richness and cover (e.g. tree species diversity), forest structural characteristics (e.g. height; mixture), forest age, elevation, soil type and indicators associated with human intervention. The results align with existing described phytosociological associations and with a wet-dry environmental gradient (Perrin et al. 2008). It is important to note that Perrin et al. (2008) defined clusters of native woodlands using specific species dominance (Indicator Species Analyses); thus, the types of HNV forest identified in this study do not replace the native woodland groups identified by Perrin et al. (2008), despite some similarities. Furthermore, the typologies here identified refer to HNV forest exclusively, which is a different categorisation and sample from the one used by Perrin et al. (2008). The variables used here are not directly indicative of forest phytosociology and relate more to structural conditions and management (NFI variables), which can provide more direct information on actions and measures for HNV forest conservation and enhancement.

Type 1 HNV forest largely represents regenerating bog woodland and riparian/wet woodlands. The subtypes identified seem to correspond to two previously described forest types that are included in Article 17 of the Habitats Directive (91E0 alluvial woodland; 91D0 bog woodland) (Cross and Collins 2017; NPWS 2019). These forest types are relatively young (average 30 to 50 years old) which might indicate that the land was previously farmed and/or used for peat extraction and that forest has naturally expanded onto these areas (predominantly for bog woodlands; Cross and Lynn 2013). The main threats to conservation of bog woodlands include drainage, invasive species and burning (Miguel Muñoz et al. 2016; NPWS 2019). In general, riparian and wet woodlands are not actively managed and some are inaccessible, making them unattractive for agricultural reclamation and are subject to lower grazing pressure (O’Neill and Barron 2013). However, the loss of alluvial woodland area and its ecological quality due to clear felling and invasive/exotic species has been identified as a threat by O’Neill and Barron (2013) and by the NPWS (2019).

From the subtypologies identified it is apparent that within Type 2 HNV forest there is a less clear separation of subtypes than in Type 1, which can also be seen by a lower contribution (% variance explained) of the first two dimensions of the FADM (see Table S3 and Table S6). Type 2 HNV forest seems to mainly comprise oak-birch-holly woodlands and oak-ash-hazel woodlands, which are potentially protected under the Habitats Directive (91A0 Old Oak woodlands). The forest plots that comprise Type 2 HNV forest areas are older than Type 1 HNV forest and seem to be occurring in both lowlands and uplands. The Type 2 HNV forest here described (mainly oak-birch-holly woodlands and oak-ash-hazel woodlands) is likely to include the old oak woodland habitat (91A0), which seems to still be experiencing some decline, albeit at a very low level (NPWS 2019). Ongoing pressures on these forests relate to expansion of non-native invasive species (mainly Rhododendron ponticum, Prunus laurocerasus L., Fagus sylvatica L.) (Perrin et al. 2008; NPWS 2019). Overgrazing by deer and continued fragmentation of remaining areas have also been identified as threats to old oak woodland habitat by the NPWS (2019).

Type 3 HNV forests are less common (i.e. fewer NFI plots) than Type 1 and Type 2 and includes HNV forest with more obvious signs of human intervention and/or a certain number of non-native species. This cluster comprises plots identified by Fossitt (2000) as mixed broadleaved woodland (WD1), a habitat category that includes forest areas with 75–100% cover of broadleaved trees and 0–25% cover of coniferous trees. As they are identified as HNV by Ruas et al. (2023) the percentage of native species must still be high, but there is evidence of anthropogenic activity (see Table 1 and Table S15). HNV forest Type 3 (Modified HNV forest) face similar pressures to those highlighted for HNV forest Type 2. However, there seems to be some evidence of lower ecological quality, as highlighted by their lower diversity of vegetation layers and higher cover of non-native species relative to Type 1 (see Table S15).

LEVEL OF PROTECTION FOR HIGH NATURE VALUE FOREST IN IRELAND

Despite Type 1 HNV forest seemingly corresponding to forest types that are included in Article 17 of the Habitats Directive (91E0 alluvial woodland; 91D0 bog woodland), most of the NFI plots in this type of HNV forest do not seem to have any level of protection. Similarly, Type 2 HNV forest (which seems to include old oak woodland habitat (91A0)) areas, showed less than 50% overlap with a designated or protected area. These results may relate to the fact that only a portion (500² m plots) of the total forest area is surveyed in the NFI (Forest Service 2018); i.e. it is unknown whether the forest area in which the plots are located have the same characteristics. Nevertheless, because the NFI sampling strategy has been designed to be statistically representative, other reasons may explain the relatively low overlap, including: a) the Habitats Directive only requires core areas of listed habitats under Annex I to be formally designated by Member States, and, by design, does not achieve complete cover of the relevant habitat types; and/or b) the limited capacity to map/survey all forest areas in Ireland results in smaller pockets of forests that should be under Habitats Directive to be overlooked; and/or c) despite similarity to Habitat Directive woodlands, the NFI plots attributed to the HNV forest types and sub-types identified in this work have characteristics (e.g. level of disturbance, total forested area) that do not warrant formal conservation status.

Type 3 HNV forests showed the lowest percentage of overlap with protected/designated areas, possibly due to the more apparent signs of human intervention. Nevertheless, these areas, with appropriate management may reach comparable ecological conditions or nature value to Type 1 and Type 2. The recently published Irish Forestry Programme 2023–27 (DAFM 2023), provides financial incentives to private landowners to maintain and enhance native forest. In particular, the new Payment for Ecosystem Services (PES) premium pilot is targeted for existing forest owners. The payment has a duration of seven years and there are several options to choose from (PES 1 to PES 6). The objective of this intervention ‘is to provide support for Sustainable Forest Management with the potential to deliver ecosystem services and environmental and climate benefits’ (in DAFM 2023). Both PES 1 (Native Woodland Conservation) and PES 4 (environmental enhancement scheme) are likely to be the most relevant for HNV forest management. The PES 1 premium is applicable for all HNV forest types and subtypes, and the PES 4 premium can be applied for if the HNV forest area is ‘adjoining designated areas and specific habitats’ (in DAFM 2023). PES approaches have been gaining traction with policymakers in recent years, culminating in a recent guidance document on developing payment schemes for forest ecosystem services (see EC 2023). However, the pilot programme in the Irish example makes no distinction between the condition of the native forest and does not appear to have a strong results-based approach (RBA). RBA are a form of PES since land managers are paid for delivering a specific environmental result or outcome (e.g. Keenleyside et al. 2014). Recent studies (e.g. Allen et al. 2014; Herzon et al. 2018) on the effectiveness of environmental schemes have highlighted the advantages of RBA relative to traditional action-based payments. In Ireland, a hybrid RBA is currently being implemented in eight priority zones under the ACRES Co-operation Agri Climate Rural Environment Scheme (DAFM 2024) and a specific scorecard is used to score scrub and woodland areas (see scorecard in DAFM, 2022). As with other national and international RBA (see Moran et al. 2021), the payment that the scheme participants receive is thus based on the score obtained. The scorecard includes indicators that relate to nativeness (cover of native tree/scrub species vs non-native tree/scrub species) and human intervention levels and it is expected to provide appropriate incentives to private landowners to maintain and enhance HNV forest. The percentage of native tree species is one of the variables with the highest weighting in the scorecard and also a key naturalness indicator for calculating the HNV index. Furthermore, through other financial incentives available in the ACRES scheme (i.e. supporting actions) the landowners can also avail of funding for the removal of non-native invasive species. Thus, such hybrid-RBA may be better suited for maintaining and improving the naturalness value of HNV forests than those that do not consider RBAs (e.g. the PES premium of the Forestry Programme 2023–27). In particular, RBA may be most beneficial to HNV forest types that show least overlap with designated/protected areas and relatively higher threat levels (e.g. HNV forest Type 3, HNV forest subtype 1.2 and HNV forest subtype 2.2 in the Irish context). Future work could focus on comparing the performance of PES that incorporate RBA and those that do not, and assessing how the scores vary for different HNV forest types.

ROBUSTNESS OF APPROACH AND FUTURE WORK

A statistical approach, similar to the one followed by Sullivan et al. (2017) for HNV farmland, was considered the most appropriate method for identifying HNV forest typologies. This methodological choice was related to the current definition of HNV forest (EEA 2014; and adapted for Ireland by Ruas et al. 2023) being more related to the naturalness concept, hindering a direct translation of HNV farmland types into HNV forest types. In particular, the FAMD made it possible to summarise the similarity/dissimilarity between HNV forest plots by taking into account both quantitative and qualitative variables (Pagès 2004). This type of dimensionality reduction technique is becoming more popular in scientific research (e.g. Tolvanen et al. 2020; Visbal-Cadavid et al. 2020; Fernández-Pascual et al. 2021) and can be easily performed and interpreted in freely available statistical software (e.g. R) (Le et al. 2008; Kassambara and Mundt 2020). However, the results obtained are obviously linked to specific methodological choices when conducting these analyses. For example, the cutoff point defined in this study (60%) for selecting the number of dimensions to use in the CA might be considered by some to be relatively low; the threshold for cumulative variance explained is often above 70% or even 80%, particularly for the natural sciences whereas the 60% threshold is commonly used in the social sciences (Taherdoost et al. 2014). Due to the high number of dimensions with eigenvalues >1 and relatively small number of plots (117 plots) for the number of variables (30 variables), we adopted a more parsimonious approach and used a 60% threshold. This means that there is still 40% of the variance not explained by the FAMD dimensions introduced in the CA. The inclusion of more HNV forest plots (NFI and possibly other areas) would be an important step in improving the strength of the analyses conducted.

In this study, the HNV typologies methodological approach relied on available NFI variables, which are only collated at plot level. Information regarding the whole forest area (within which the plots are located) was not available for the 178 HNV plots. Thus, variables that refer to the diversity of habitats within the total forest area, or even landscape structure related variables (connectivity; shape index variability; fragmentation levels etc.) could not be considered in the analysis. Ultimately, the types of HNV forest identified in this study reflect the variations of ecological conditions at a small scale. When developing this work, some inconsistencies and inaccuracies of forest patch delineation in available maps were noted (e.g. Prime 2; Private Forest 2019 shapefiles), and this hindered a more comprehensive HNV typology analysis. This limitation could be addressed in future studies and other variables could be gathered from available land cover/habitat maps. Digitisation/mapping should therefore follow consistent definitions and procedures for forest patch delineation (see example in Forest Service 2018).

Despite the limitations outlined, the different types and subtypes of HNV forest identified seem to relate to some of the previously described woodland types in Ireland (Fossitt 2000; Perrin et al. 2008; NPWS 2019). Due to the restricted number (and type) of variables available the characteristics of HNV forest types and subtypes will always be somewhat distinct from other existing classifications. The relatively low spatial overlap with protected areas emphasises the relevance of the HNV forest concept in identifying and characterising the nature value of a wider range of forest types outside protected areas.

SUPPLEMENTARY MATERIAL

Supplementary material related to this article is available online here: http://muse.jhu.edu/resolve/251.