Monitoring woody landscape features in Dutch rural landscapes

Original Articles Monitoring woody landscape features in Dutch rural landscapes Nicolas Grondard a,* , Lenny van Bussel b,1 , Lars Hein a,2 a Earth Systems and Global Change Group, Wageningen University & Research, Wageningen, the Netherlands b PBL-Netherlands Environmental Assessment Agency, PO Box 30314, The Hague 2500 GH, the Netherlands ARTICLE INFO Keywords: Agro-ecosystems Woody Landscape Elements Hedgerows Ecosystem Accounting Common Agricultural Policy LiDAR ABSTRACT Woody landscape features (WLF) declined sharply in the Netherlands in the twentieth century, with negative consequences on biodiversity and ecosystem services. EU and national policies aim to reverse this trend. However, recent assessments of WLF in the Netherlands are lacking. Ecosystem Accounts (EA) could support the monitoring of WLF but data on WLF in the Dutch EA is incomplete. Therefore this paper has the dual objective to assess WLF in the Netherlands, and to analyse how WLF can be monitored with EA, complemented with Laser Imaging Detection and Ranging (LiDAR) data. We used three Dutch rural regions, with different landscape characteristics and WLF densities, as case studies. We assessed the current extent and recent trends of WLF with LiDAR data, and compared the LiDAR based estimates with existing data in the Dutch EA. LiDAR based results indicated a density of WLF about 15% higher than the Dutch EA. However, even the WLF density as measured with LiDAR does not reach the EU and national policy targets for agricultural land in two of the three study areas. Moreover, both LiDAR data and the Dutch EA indicated that the area of WLF decreased over 2011–2021 in all study areas. LiDAR data can be used to enhance the Dutch EA for the purpose of monitoring WLF in rural landscapes, allowing a more complete representation of small features in agricultural fields and surrounding areas, such as farmyards, canals and road verges. Further efforts are needed to monitor and maintain Dutch WLFs. 1. Introduction Landscape features are small natural or semi-natural areas in rural landscapes that are not used for agricultural production. They include natural or semi-natural vegetation (e.g. hedges, trees), surface water (e. g. ponds, ditches) and anthropogenic structures (e.g. terrace walls). Landscape features are an essential component of the identity of European rural landscapes (Arnaiz-Schmitz et al., 2018). They offer breeding sites, shelters and food to farmland species (Carlier and Moran, 2019) and provide a wide range of ecosystem services (ESs) (Drexler et al., 2021; England et al., 2020; Montgomery et al., 2020; Weninger et al., 2021). However, landscape features have drastically declined in the EU in the twentieth century due to land consolidation, the substitution of hedges with barbed wire and urbanization (Koomen et al., 2007; Robinson and Sutherland, 2002), with negative consequences on farmland biodiversity and ESs supply (Denac and Kmecl, 2021). To counter this, the EU biodiversity strategy for 2030 aims to maintain existing and plant new landscape features, in order to achieve a minimum of 10 % of agricultural land under high diversity landscape features by 2030 (European Commission, 2020). Consequently, the enhancement of landscape features is one of the main environmental objectives of the European Nature Restoration Law (European Union, 2024) and of the Common Agricultural Policy (CAP) legislation for 2023–2027. The share of agricultural land covered with landscape features is also an indicator of the CAP performance (European Commission, 2018). However, mapping of landscape features remains a challenge that prevents monitoring whether these EU policy targets are achieved. EUwide datasets are available but have limitations (Czucz and Baruth, 2022): field samples detect detailed landscape feature types of small size but do not have a wall to wall coverage, while the Copernicus land monitoring service maps do not detect small elements, such as woody features with a width less than three meters (Kleeschulte et al., 2023). Few EU countries have datasets specifically aimed to monitor landscape features, and available national topography maps do not provide all inputs required for delineating and monitoring landscape features. A recent review on the EU INSPIRE geoportal found relevant datasets in * Corresponding author. E-mail addresses: [email protected] (N. Grondard), [email protected] (L. van Bussel), [email protected] (L. Hein). 1 0000-0003-3801-9754. 2 0000-0002-4651-2875. Contents lists available at ScienceDirect Ecological Indicators journal homepage: www.elsevier.com/locate/ecolind https://doi.org/10.1016/j.ecolind.2025.113853 Received 23 May 2025; Received in revised form 27 June 2025; Accepted 3 July 2025 Ecological Indicators 178 (2025) 113853 Available online 15 July 2025 1470-160X/© 2025 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ). two third of EU countries but mostly on wet landscape features, whereas only five countries had specific datasets for woody landscape features (WLF) (Czúcz et al., 2022). Also in the Netherlands farming intensification and the accompanying land consolidation resulted in the loss of landscape features (van Apeldoorn et al., 2013). It was estimated that more than half of landscape features have disappeared from rural landscapes during the twentieth century (Koomen et al., 2007). WLF, such as forest patches, hedgerows, tree rows and isolated trees declined the most (Koomen et al., 2007). To address this decline, a Dutch national plan (Hagendoorn et al., 2021) for landscape features was conceived by a coalition of stakeholders from the public and private sector. This plan aims to increase the share of total area of rural landscapes covered by landscape features to 10 % by 2050, with WLF contributing to half of this target (5 % of the total rural landscapes area) (Hagendoorn et al., 2021). Whereas the EU policy target focusses on agricultural land (i.e. arable land, permanent grassland and permanent crops), the Dutch national plan target covers the whole rural area, including agricultural land but also other land uses (e.g. road verges, farmyards, settlements) within rural areas. In 2016, the total coverage of WLF in the Netherlands was estimated to be about 117,000 ha, corresponding to 3.5 % of the rural area (van Doorn et al., 2016). Despite the importance of WLF in agri-environmental policy and their decline in the past, there are gaps in the monitoring of WLF in the Netherlands. No recent evaluation of changes in WLF coverage has been undertaken since the works of Koomen et al. (2007), which covered the period 1990–2003, and van Doorn et al. (2016) whom did not assess temporal changes. It is therefore not known whether the loss of WLF is continuing and how much effort is needed to meet the policy targets, neither in the agricultural areas nor in other land uses within rural areas. The most complete dataset recording WLF over the whole country is the National Topographic (Top10NL) data (Kadaster, 2024; van Doorn et al., 2016), updated every year, with records of forest patches, hedgerows and tree lines, as well as isolated trees. However, comparing this data with exhaustive records for a sample of small landscapes, van Doorn et al. (2016) found that forest patches were covered but that one third of hedgerows/tree lines and 80 % of isolated trees were missing in the Top10NL data. Other datasets provide exhaustive records, but for only a sample of landscapes (LandschappenNL, 2024), or only record features declared by landowners to receive subsidies (RVO, 2023). Another possibility to enhance the monitoring of WLF in the Netherlands is the System of Environmental Economic Accounting Ecosystem Accounting (SEEA EA) framework. The SEEA EA is an integrated statistical framework for organizing biophysical data on ecosystem assets and services (Hein et al., 2020a). Ecosystem accounts (EA) track changes over time of ecosystems extent (extent accounts), condition (condition accounts) and services flows (ES accounts) in a spatially explicit manner and with a full coverage of the national territory (United Nations et al., 2024). The Netherlands is one of the most advanced countries in SEEA EA implementation (Hein et al., 2020b; Statistics Netherlands & WUR, 2022). In particular, ecosystem extent accounts have been compiled over the period 2013–2022 with a yearly update frequency and comprise 52 ecosystem type classes, including forest and hedgerows. Moreover, ecosystem condition accounts include the density of woody linear features in ecosystem types as condition variable. Monitoring WLF with the SEEA EA framework could have several potential advantages: (1) accounts are continuously updated by Statistics Netherlands, the Dutch statistical office, in consistent time series; (2) WLF are integrated in ES assessments, which allows to estimate ESs delivered by WLF and the influence thereof of WLF changes; (3) time series of ecosystem extent accounts combined with land use data can provide insights on the drivers of WLF changes. However, at the moment, forest patches and linear features (hedgerows and tree lines) in the Dutch ecosystem extent accounts map are derived mainly from the Top10NL data and therefore incomplete (van Doorn et al., 2016). This data gap could be filled using airborne Laser Imaging Detection and Ranging (LiDAR) data. While first LiDAR acquisiton campaigns took six to seven years to cover the Netherlands, the frequency of updates has recently increased to three years, which makes it possible to use this data for monitoring purpose. Several studies have demonstrated the use of airborne LiDAR data for mapping WLF in EU countries (Broughton et al., 2021; Estrada et al., 2017; Lucas et al., 2019; Luscombe et al., 2023), but those studies did not assess the use of LiDAR data time series for monitoring purposes. Therefore, this paper has the dual objective to assess the current state and recent trends of WLF in the Netherlands, and to analyse how WLF can best be monitored with EA complemented with LiDAR data. The study is carried out in three regions of the Netherlands with different landscape characteristics and densities of WLF: Rivierenland, Zeeland and Limburg. We address four specific research questions: (1) What is the current extent and recent trends of WLF based on LiDAR data?, (2) What is the current extent and recent trends of WLF based on EA data?, (3) What are the differences between LiDAR-based and EA data?, and (4) How do the current state and recent trends of WLF differ between agricultural land and other land uses? We estimate the current state and recent changes of WLF in the three regions, using, first, LiDAR data (https://www.ahn.nl), and second, EA data. Next, we analyse the differences between these two data sources to assess how LiDAR data could complement EA. Furthermore, we integrate LiDAR based WLF maps with the ecosystem extent accounts and agricultural land use data (RVO, 2024). This integration allows to assess how the current state and recent changes of WLF differ between agricultural land and other types of land use in rural areas. Based on these analyses, we provide recommendations for the monitoring of the Dutch landscape plan targets and the use of the Dutch EA completed with LiDAR data to monitor WLF. 2. Methods 2.1. Case study areas We carried out our study in three Dutch case study areas (Fig. 1): the province of Zeeland in the South West, the Rivierenland region in the center of the country, and the province of Limburg in the South East. The three case study areas cover four main types of landscape (Raap et al., 2022): (i) The South Western marine clay landscape, which covers most of Zeeland. It is an open landscape with a characteristic polder structure. (ii) The southern cover sand landscape is a slightly undulating landscape crossed by rivers. It covers a large part of Limburg and a minor part of Rivierenland. Characteristic WLF are forest patches. (iii) The river landscape is an open landscape with a characteristic structure of river floodplains, river banks, dikes and swamp soils. It covers most of the Rivierenland region along the Rhine river, and part of Limburg, along the Maas river. On river banks, plots are small and often used as orchards in Rivierenland. Willow coppices are characteristic WLF in this landscape. (iv) The Limburg south hilly landscape is a loess area with relatively large height differences. In valleys, plots are small and often surrounded by rows of trees or hedges. On slopes, characteristics features named graften, hedges with steep edges, have been planted to counteract erosion. Our study was focused on rural zones, thus excluded urban areas, nature areas (Natura 2000 and National Nature Network areas) and large infrastructures (e.g. motorways) (see 2.4.1). These rural zones consist of agricultural areas, i.e. parcels dedicated to agricultural production, but also house gardens, farmyards, small roads and canals and their verges, and fragmented (semi-) natural areas such as WLFs and N. Grondard et al. Ecological Indicators 178 (2025) 113853 2 natural grasslands. The national plan targets for WLF cover rural zones, but EU policies targets for WLFs only cover agricultural areas. Therefore, we further distinguished the state and trends of WLFs in agricultural areas from the state and trends of WLFs in other land uses within rural zones (see 2.6). 2.2. Overview of the methodological approach Fig. 2 presents the overall methodological approach and main datasets used (see 2.3) to address our research questions. We developed an automated LiDAR data processing workflow (see 2.4) to assess the Fig. 1. Location of case study areas in the Netherlands. Fig. 2. Overview of the methodological approach. N. Grondard et al. Ecological Indicators 178 (2025) 113853 3 current extent and recent trends of WLF (RQ1). The current extent and recent trends of WLF based on EA data (RQ2) were quantified through processing of existing datasets (see 2.5). We compared the results of these two analyses (see 2.5) to assess differences between LiDAR-based results and EA data results (RQ3). Finally, we used LiDAR-based results to assess differences in WLF extent and recent trends between agricultural land and other land uses (RQ4, see 2.6). 2.3. Data sources Table 1 provides an overview of each dataset, with full name and acronym, description of the main characteristics and how it was used in this study. 2.3.1. Ecosystem accounts and Top10NL data The Dutch ecosystem extent accounts provide a time series of yearly (period 2013 – 2021) ecosystem extent maps (NL ET maps), as vector maps or raster maps of 2.5 m resolution, with 52 ecosystem type (ET) classes (Statistics Netherlands & WUR, 2022). For this study, we simplified the 52 ET classes into 8 aggregated ET categories (Table 2). Two of these categories correspond to WLF: forest (aggregation of forest ET classes), when located in rural landscapes, and hedgerows. Both forest and hedgerows ET classes in the NL ET maps were obtained from forest polygons of the Top10NL database (Kadaster, 2024). The Top10NL is an object-based topographic database, yearly updated through photointerpretation and digitalization of stereoscopic aerial photography (Beeldmateriaal Nederland, 2024). Top10NL forest polygons with a width smaller than 10 m. and a length higher than 100 m. were classified in the NL ET maps under the ET class “Hedges and tree lines”, distinct from other forest ET classes. Besides forest polygons, the Top10NL database records narrow linear features covered by trees as polylines, which are classified as hedgerows or tree lines in the Top10NL database (see Table 3 for definition and recording criteria). These polylines are not included in the NL ET maps, but are included in a hedgerows density indicator in the ecosystem condition accounts. The Top10NL database also records isolated trees (as point data, see Table 3 for definition and recording criteria), but these are not included in any of the EA. In our analysis, we combined data on WLF from the NL ET maps with Top10NL polylines and points, and we compared this NL ET maps – Top10 NL data with LiDAR-based WLF maps. 2.3.2. LiDAR data We used the Digital Terrain Model (DTM) and Digital Surface Model (DSM) at 50 cm resolution generated from airborne LiDAR data (Actueel Hoogtebestand Nederland −AHN −data, ahn.nl), acquired during the AHN2, AHN3 and AHN4 campaigns between 2011 and 2021 during winter (December to March). The point density ranged from 6 to 10 points per m 2 for AHN2 and AHN3, and from 10 to 14 points per m 2 for AHN4. For each 50 cm pixel, the DSM estimates average elevation (taking into account buildings, infrastructures and vegetation) while the DTM estimates ground level elevation. Because it takes several years to acquire airborne LiDAR data over the Netherlands (ahn.nl/kwaliteitsbes chrijving), our studied regions are covered by different acquisition years (Table 4). 2.3.3. Agricultural land use data We used the 2024 version of the Dutch Land Parcel Information System (LPIS) data (RVO, 2024), which records all agricultural land parcels, with a distinction between parcels used for agricultural production and parcels declared as landscape features (classified as woody, water and other) by farmers. The LPIS data was used to (1) complete the map used to mask forest polygons located outside rural zones (see 2.4.1.) and (2) for the assessment to detect WLF current state and trends in agricultural areas within rural areas (see 2.6). Table 1 Overview of all datasets used in the study. Acronym Full name Description Use in the study NL ET map Netherlands SEEA EA ecosystem extent maps 2013–2021 Vector maps or raster maps of 2.5 m resolution, with 52 ecosystem type (ET) classes, wall to wall national coverage of the Netherlands ▪ “Forest” ET located in rural areas: Comparison of WLF state and trend between the NL ET map +Top10NL data and LiDAR based WLF maps ▪ ET “Business park”, “Mining, land fill, etc…”, “Sport park” and “Residential recreation”: Input data to the rural areas mask ▪ Assessment of LiDAR-based WLF state and trends in non-agricultural areas within rural areas Top10NL Netherlands national topographic database Object-based topographic database (scale 1:5000 to 1:25000) ▪ Linear objects hedgerows and tree lines, tree points: Comparison of WLF state and trend between the NL ET map +Top10NL data and LiDAR based WLF maps ▪ Powerlines and windmills objects: masking of powerlines and windmills in the LiDAR derived CHM BAG Basisregistratie Adressen en Gebouwen Object-based database of buildings Building polygons: masking of buildings in the LiDAR-derived CHM LG map Landelijk Gebied kaart Basis map for the delineation of the rural areas mask LPIS Agrarisch Areaal Nederland Vector map of agricultural parcels boundaries ▪ Input data to the rural areas mask ▪ Assessment of LiDAR-based WLF state and trends in agricultural areas within rural areas LiDAR Airborne LiDAR data DTM and DSM at 50 cm resolution generated from airborne LiDAR data (Actueel Hoogtebestand Nederland −AHN −data, ahn.nl) Production of LiDAR based WLF maps Aerial photography Aerial photographic images, covering the whole area of Netherlands every year (Beeldmateriaal Nederland, 2024) RGB ortho photographic images from 2006 (50 cm resolution), 2011 (25 cm resolution), 2021 & 2023 (7.5 cm resolution), color-infrared images from 2013 (25 cm resolution) Validation of LiDAR-based WLF types maps and change maps N. Grondard et al. Ecological Indicators 178 (2025) 113853 4 2.4. Current state and recent changes of WLF with LiDAR data 2.4.1. Current state of WLF with LiDAR data We assessed the current state of WLF based on LiDAR data. The LiDAR AHN data is available for download as 5x6.5 km tiles in public repositories. For each tile covering our regions of interest, the DSM and DTM files were downloaded and processed according to the processing chain described in Fig. A1. First, we generated a canopy height model (CHM) by subtracting DTM pixel heights from DSM pixel heights. Values exceeding the maximum height of trees in the Netherlands (set at >50 m) and negative values were reclassified as NA. Then, the CHM was smoothed with the mean of a 3x3 pixels (1.5 x 1.5 m) moving window to fill empty pixels (function focal of the R terra package, v.1.8.21, Hijmans et al. (2024)). Next, we used the BAG dataset (Kadaster, 2023) to mask buildings and the Top10NL dataset (Kadaster, 2024) to mask powerlines and windmills in the smoothed CHM. Smoothing was done before masking to avoid that pixels overlapping with the edge of masked polygons are filled with a value during the smoothing step. Next, individual tree tops detection was performed using a local maximum filter with variable windows size (locate_trees algorithm of the R lidR package, v.4.1.1, Roussel et al. (2020)). Then, the final CHM and detected individual tree tops were combined to segment individual trees using the method from Silva et al. (2016) implemented in the R lidR package. This step generated a raster map (0.5 m resolution) with delineated tree crowns. This tree crowns map was resampled to 1 m resolution to reduce processing time in the next steps. After resampling, adjacent tree crowns were aggregated in patches using the patches function in the R terra package (Hijmans et al., 2024), to produce a tree crowns patches map. Next, we classified this tree crowns patches map in a WLF map. We used a minimum bounding rectangle to estimate the length of each patch and deducted the patch width by dividing the patch area by its length. We removed very small patches (area below 12.57 m 2 , corresponding to the area of a tree with a 4 m diameter crown) and classified the remaining patches under three types of WLF: (1) isolated trees/groups of trees, (2) forest patches and (3) hedgerows. The classification was based on the size and length −width ratio of patches, following the decision tree presented in Fig. 3. To focus our assessment on rural zones, we masked tree crowns located in urban and nature areas, using the Landelijk Gebied (LG) map produced by Roelofsen (2022), which delineates rural zones, with some modifications. First, we excluded from the LG map the areas classified as “Business park”, “Mining, land fill, etc…”, “Sport park” and “Residential recreation” in the 2021 ET map, so that tree crowns located in these land use classes are excluded and confusions with non vegetation vertical structures are avoided (e.g. cranes in mining, land fill areas). Second, we updated the agricultural areas of the LG map using the most recent LPIS data (see 2.2.3). A sample of the final mask is shown in Fig. 4. Note that the resulting rural zones still include built-up and infrastructure areas, mainly corresponding to house gardens, farm yards, roads and canal verges. The resulting rural zones also include fragments of nature areas, including forest patches and hedgerows, located next to fields, in farmyards and on roads and canals verges. Urban and nature areas masking was carried out after classifying WLFs to avoid classification artefacts on (non-masked) edges of masked forest patches. These non-masked edges have a linear shape and would have been classified as hedgerows instead of forest patches if the WLF classification had taken place after masking. Using this LiDAR-based WLF map produced in each region, we calculated the area of each WLF category (forest patches, isolated/ groups of trees, hedgerows) for the year 2021 in Limburg and Rivierenland, and 2020 in Zeeland. 2.4.2. Quality control of LiDAR-based WLF maps The quality of WLF maps was checked visually, using the CHM and aerial photography for comparison. Moreover, we compared the most recent WLF maps (2021 in Limburg and Rivierenland, 2020 in Zeeland) to a reference sample of WLFs visually identified on aerial photography in each region. To collect the reference sample, we randomly selected 30 forest polygons, 30 tree lines, 30 hedgerows and 30 trees from the Top10NL data in each region. We identified these reference WLFs on aerial photography and assessed by manual delineation how well the reference WLFs were covered by our LiDAR-based WLF maps. We recorded the percentage of the reference WLF covered by the LiDARbased WLF map, and under which WLF categories (forest patches, isolated/groups of trees, hedgerows) the reference WLF were classified in the LiDAR-based WLF map. We used as references the area covered by Table 2 Aggregation of NL ET maps ET classes into ET categories. ET category ET classes Agriculture Cropland, regular; Cropland, extensive; Pasture, permanent; Pasture, temporal; Pasture, extensive; Greenhouse horticulture; Nursery container fields; Fallow land; Arable field margins; Other, agricultural grassland; Perennials, regular; Perennials, extensive. Built-up/ infrastructure Built-up (urban); Built-up (rural); Business park; Mining, land fills, etc.; Infrastructural; Marine, other; Sport park; Residential recreation; other; Other terrain Public green space Landscape garden; Public park (large); Public park (small); Public green space, Semi-public green space. Forest (Semi-)natural forest; Plantation forest; Other forest; Swamp forest. Hedgerows Hedges and tree lines Open nature Tall herbs; Heathland; Drift sand; Semi-natural grassland; Biodiverse cropland; River floodplain; Other, nature grassland Water & wetlands Bogs; Fens; Streams and rivers; Lakes; Brackish; Coastal & marine Coastal dunes; Salt marshes; Beach; Intertidal and mud flats; Shoals; Estuarium; North sea; Wadden sea; Table 3 Definition and recording criteria of Top10NL linear and point WLF. Hedgerows (linear features) 1) rows of trees, either alone or in combination with shrubs, in which the distance between trees or the undergrowth obstructs the view up to at least human height, with a maximum planting width between trunks of 3 m; or 2) a row of shrubs planted next to each other (privet, hawthorn, conifers, hornbeam, etc.) that are trimmed at set times. Top10NL hedgerows have a minimum height of 1 m and minimum length of 100 m. Hedgerows within built-up areas and on farmyards are not recorded (unless a farmyard hedgerow continues beyond the yard). Hedgerows located on earthen walls (houtwal in Dutch) are recorded as forest polygons and not linear features. Tree lines (linear features) A number of trees (minimum 3) that stand in a row, where the distance between the trees is such that the line of trees does not form a visual obstruction up to human height. Top10NL tree lines have a minimum length of 100 m. A line of trees on or around a yard may be indicated if it continues into the adjacent or surrounding terrain. Isolated trees (point features) A woody plant with a single, sturdy, woody, and persistent trunk that branches at a certain height above the ground. Trees within built-up areas, on or along a street, in a garden, or on a farmyard are not included. A small number of trees standing close together are counted as one tree. Table 4 AHN data acquisition dates and number of tiles per region. Region AHN data – acquisition years (source: https://www. ahn.nl/kwaliteitsbeschrijving) Number of tiles Zeeland 2014 (AHN3), 2020 (AHN4) 113 Limburg 2012 (AHN2), 2018 (AHN3), 2021 (AHN4) 112 Rivierenland 2011 (AHN2), 2015 (AHN3), 2021 (AHN4) 76 N. Grondard et al. Ecological Indicators 178 (2025) 113853 5 tree canopies on the aerial photography for patchy features (forest polygons and trees) and the length covered by tree canopies on the aerial photography for line features (tree lines and hedgerows). In parallel, we recorded the width of the reference trees lines and hedgerows (width of tree canopies on the aerial photography), and the area of the reference trees, which we used to estimate the total area of Top10NL features (see 2.5). 2.4.3. Recent changes of WLF based on LiDAR data 2.4.3.1. Net changes. The WLF mapping workflow (see 2.4.1) was repeated for each region and acquisition year to calculated changes in WLF area for the years 2012–2018-2021 in Limburg, 2011–2015-2021 in Rivierenland, and 2014–2020 in Zeeland. 2.4.3.2. Gross changes. WLF maps of the first and last LiDAR acquisition years were subtracted to produce WLF change maps (pixels classified as loss, gain and no change) for the periods 2012–2021 in Limburg, 2011–2021 in Rivierenland and 2014–2020 in Zeeland. Using these change maps, we calculated gains and losses of WLF area during these periods. All WLF types (forest patch, hedgerows, trees) were aggregated in the WLF maps used to produce these change maps. Thus, no distinction between WLF types was made at the subsequent steps of validation and analysis of the types of changes (see below). 2.4.3.3. Validation. We validated the change maps using aerial photography as reference data. We aggregated WLF change pixels in gained and lost patches (patches function in the R package terra) and calculated the area of these change patches. We categorized change patches per size in order to get samples of change patches potentially representing different types of change (e.g. trimming vs. clear-cutting): less than 12.57 m 2 (about the size of a tree crown), between 12.57 m 2 and 100 m 2 , between 100 m 2 and 200 m 2 , between 200 m 2 and 1000 m 2 , between 1000 m 2 and 1 ha, more than 1 ha. We randomly selected 467 change patches spanning the range of change patches size categories (see Table 5) and visually checked the nature of changes using aerial photographic images, covering the whole area of Netherlands every year (Beeldmateriaal Nederland, 2024). We classified change patches as false or true changes by visually comparing change patches to aerial photography before the first date (2006, 2011 and 2013) and after the last date (2021 and 2023) of the change map period. Change patches were labelled as true if the change patch actually corresponded to a gain or loss of tree vegetation confirmed by aerial photography comparison, and as false on the contrary. We used the results of the validation to adjust the estimated net change area taking into account classification errors in the change maps. We calculated pˆ i , the proportion of correctly classified change patches in the validation sample, for each category i of change patch size. We excluded gain/loss patches from size categories with a pˆ i , less than 0.8. This led to excluding all changes with a size less than 100 m 2 when calculating the adjusted net change area. For the remaining change size categories, we calculated the total areas of gain/loss by summing areas of all gain/loss patches from the WLF change maps in the size category. Then we multiplied these map-based sums by the percentage of correct change detection in the relevant size category. Thereby, we obtained adjusted gain and loss areas (Eq.1), with a 95 % confidence interval estimated with the normal approximation method (Eq. (2). Adjusted Gain/Loss Areai=Map Gain/Loss Areai. pi ∓ΔAdjusted Gain/Loss Areai(1) Fig. 3. Decision tree for the classification of tree crown patches in types of WLF. Fig. 4. Sample of mask (red shaded) used to select rural areas and exclude urban and nature areas, in Limburg, around the city of Weert. All red shaded areas are considered rural areas and are used to assess WLFs. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) Table 5 Number of change patches compared to reference data in each region per size category (e.g. in Limburg, we compared to reference data 10 gain patches and 10 loss patches of a size less than 12.57 m 2 ). Category of change patches size Limburg Rivierenland Zeeland Total Gain Loss Gain Loss Gain Loss Gain Loss Less than 12.57 m 2 10 10 10 10 10 10 30 30 12.57 to 100 m 2 19 18 20 18 19 19 58 55 100 to 200 m 2 10 10 10 10 10 10 30 3. 200 to 1000 m 2 20 20 20 20 20 20 60 60 1000 m 2 to 1 ha 10 10 10 10 10 10 30 30 More than 1 ha 10 10 10 10 4 10 24 30 Total 79 78 80 78 73 79 232 235 N. Grondard et al. Ecological Indicators 178 (2025) 113853 6 ΔAdjusted Gain/Loss Areai=Map Gain/Loss Areai.1.9599.  pi(1− pi) ni √ (2) Where, Map Gain/Loss Area i is the sum of gain/loss patches area in gain/loss map for patches of size category i, pˆ i is the proportion of correct changes in gain/loss patches sample in size category i, n i is the number of gain/loss patches compared to reference data in size category i. Finally, we summed the adjusted gains and losses areas to obtain the adjusted net change area (Eq. (3), with a 95 % confidence interval taking into account the propagation of uncertainties of adjusted gain and loss areas (Eq. (4). Adjusted Net Change Area =∑ 4 i=1 (Adjusted Gain Areai −Adjusted Loss Areai) ∓ΔAdjusted Net Change Area (3) Where, i is the change size category, Adjusted Gain/Loss Area i is the output of Eq. (1) for the change size category i, Δ Adjusted Gain/Loss Area i is the output of Eq. (2) for the change size category i. 2.4.3.4. Types of changes. In addition to the validation of changes, we used aerial photography to identify the types of changes and obtain insights into the nature and cause of WLF changes. We labelled each change patch as new or clear cut features, part of a coppice regrowth or cutting, canopy cover loss (gaps, trimming) or growth. We also noted whether the observed changes corresponded to a change in land use, and the location of the changed WLF (e.g. in farmyard or on field border). 2.5. Current state and recent changes of WLF with NL ET maps and Top10NL data To compare results obtained from LiDAR data to the data available in the EA, we assessed WLF recorded in the NL ET maps and Top10NL features. We calculated the area of forest and hedgerows ET categories in the NL ET maps. We calculated the length of Top10NL line features (hedgerows and tree lines) and the number of Top10NL points (trees). To convert linear features length and number of trees to area, we estimated the average width of Top10NL linear features and average area of Top10NL trees on a sample of 60 linear features and 30 trees in each region (see 2.4.2). To be consistent with the LiDAR data, we used the same years (2021 in Limburg and Rivierenland, 2020 in Zeeland) and same masks (see 2.4.1 and Fig. 4) to exclude buildings, areas under powerlines, windmills and non-rural areas from the NL ET maps and Top10NL features. We repeated this assessment for each year of the 2013–2021 time series. Finally, to understand differences of WLF assessment between the LiDAR based maps and the NL ET maps, we calculated the area of LiDAR WLF located within each ET category of the NL ET maps of 2021 (in Limburg and Rivierenland) and 2020 (in Zeeland). 2.6. State and trends of WLF in agricultural and other land use To determine the state of WLF in agricultural land within rural zones, we calculated the area of LiDAR WLF located within the LPIS parcels at the last LiDAR acquisition years in each region. To assess trends of WLF in agricultural land within rural zones, we calculated the area of WLF changes (gains and losses of WLF) within the LPIS parcels, using the LiDAR based WLF change maps excluding small size changes (see 2.4.3). We distinguished between LPIS parcels dedicated to production, parcels declared as WLF, and parcels declared as other types of landscape features. To determine the state of WLF in non-agricultural land within rural zones, we used the NL ET maps. After excluding the agricultural land, we calculated the area of LiDAR based WLF located within each ET category at the last LiDAR acquisition years in each region. To assess trends of WLF on non-agricultural land within rural zones, we calculated the area of WLF gains and losses located in rural zones (excluding agricultural land), using the LiDAR based WLF change maps (excluding small size changes, see 2.4.3). 3. Results 3.1. Quality of WLF maps derived from LiDAR data Visual control showed a good agreement between our WLF maps Fig. 5. Extract of LiDAR-based WLF map in the Rivierenland region. Top: Aerial photography, 2021; Bottom: LiDAR-based WLF 2021 map features overlaid on aerial photography (CRS: EPSG:28992 −Amersfoort / RD New). ΔAdjusted Net Change Area = ∑ 4 i=1 ΔAdjusted Gain Area2 i+ΔAdjusted Loss Area2 i) √ √ √ √(4) N. Grondard et al. Ecological Indicators 178 (2025) 113853 7 derived from LiDAR data and GoogleEarth images, with most of WLF in the landscape being captured in our maps (Fig. 5). A few not masked vertical infrastructure elements, such as train lines power poles (Fig. 6), were misclassified as WLF. Comparison of LiDAR-based WLF maps with the reference WLF on aerial photography showed that the LiDAR based maps detected 85.8 % of the forest polygons, 95.8 % of the tree lines and 92.7 % of trees, but only 49.8 % of hedgerows (see Table 6 and Table A1 in appendix for detailed results per region). The low detection of hedgerows was due to the failure of the LiDAR based map to detect low hedgerows with a narrow (less than 2 m) width (see Fig. 7). There was a good agreement between the Top10NL WLF classification and the LiDAR-based WLF types classification (see Table7) in the case of forest polygons: 96.1 % of the reference forest polygons were classified as forest patches in the LiDAR-based maps. That was not the case for the other types of WLF. Top10NL tree lines and trees were more classified in the LiDAR-based maps as forest patches, respectively 71 % and 59 %, than as trees, respectively 20.2 % and 37.3 %. Top10NL hedgerows were mostly classified as hedgerows in the LiDAR-based maps (56.8 %), but also as forest patches (31.1 %) and trees (12.1 %). 3.2. Current state and recent trends of WLF in rural landscapes based on LiDAR data The three regions differed in their density of WLF, with Zealand having a relatively low density (2.4 % of the total rural area in 2020), Limburg a relatively high density (8.2 % of the total rural area in 2021), and Rivierenland being in between (4.8 % of the total rural area in 2021), according to LiDAR based estimates (Table 8). According to WLF LiDAR maps, the WLF area increased by 0.2 % per year between 2012 and 2021 in Limburg, was stable between 2011 and Fig. 6. Incorrect classification of trainlines power poles as hedgerows in LiDAR-based WLF 2021 map, overlaid on aerial photography 2021 (CRS: EPSG:28992 −Amersfoort / RD New). Table 6 Comparison of WLF maps (2020 in Zeeland, 2021 in Limburg and Rivierenland) with reference aerial photography. Top10NL feature Average size WLF cover* detected by LiDAR-based WLF maps (%) Qty Unit (group of) Trees Hedgerow Forest Patch Any WLF type Forest polygons 2229 m 2 1.7 1.6 82.5 85.8 Tree lines 7.3 m 19.3 8.4 68.0 95.8 Hedgerows 4.6 m 6.0 28.3 15.5 49.8 Trees 85 m 2 34.6 3.4 54.7 92.7 * % presented in this table correspond to the WLF cover detection % areaweighted average (forest polygons and trees) or length-weighted average (tree lines and hedgerows) of reference features. Fig. 7. Several Top10NL hedgerows, corresponding to low and narrow features, are not detected by the LiDAR-based WLF map. Left: Aerial photography 2021; Middle: Top10NL linear features overlaid on aerial photography; Right: LiDAR-based WLF 2021 map and Top10NL linear features overlaid on aerial photography (CRS: EPSG:28992 −Amersfoort / RD New). Table 7 Agreement (%) between the WLF types classification of the Top10NL data and the LiDAR-based WLF maps. Top10NL feature classification WLF cover classified by LiDAR-based WLF maps (%) (group of) Trees Hedgerow Forest Patch Forest polygons 2.0 1.8 96.1 Tree lines 20.2 8.8 71.0 Hedgerows 12.1 56.78 31.1 Trees 37.3 3.7 59.0 N. Grondard et al. Ecological Indicators 178 (2025) 113853 8 2012 in Rivierenland, and increased by 0.8 % per year between 2014 and 2020 in Zeeland. However, the comparison of change patches with aerial photography revealed that most changes of small size (<100 m 2 ) were false changes. When these small size patches are excluded and areas of change are adjusted for false detections (see 2.4.3 and Tables A2, A3 and A4 in Appendix), the WLF area actually declined in all regions (Fig. 8). In Limburg, the WLF area decreased by 0.35 ±0.12 % per year (349 ±120 ha); in Rivierenland, the WLF area decreased by 0.95 ±0.08 % per year (440 ±40 ha); and in Zeeland, the WLF area decreased by 1.2 ±0.15 % per year (248 ±31 ha). In the sample of change patches checked with aerial photography, 66 % of the new WLF area corresponded to new features and 34 % to growth of existing features (canopy growth or regrowth of a coppice). A minor part (about 32 %) of new WLF were due to a land use change, which consisted mostly of agricultural land converted to a forest patch. 74 % of the lost WLF area corresponded to clear cut features and 26 % to loss of canopy cover in existing features (gaps in canopy, trimming or coppice cut). Most clear cut WLF (about 90 %) were due to a land use change, which consisted mostly of the conversion of forest patches or hedgerows to agricultural land (55 %), but also to semi-natural grassland (25 %) and other land uses (e.g. road infrastructure). Overall, 87 % of net changes of WLF area resulted from new/clear cut WLF, and 13 % from growth/canopy losses in existing features. 3.3. Current state and recent trends of WLF in rural landscapes based on NL ET map and Top10NL data The WLF area estimated from the NL ET maps and Top10NL data was lower than the WLF area estimated from the LiDAR data, by 14 %, 5 % and 29 % in Limburg, Rivierenland and Zeeland respectively (Table 9). On average, NL ET map – Top10NL data captured 85.4 % of the LiDAR based WLF area: 51.8 % was captured by the NL ET maps and 33.6 % by the Top10NL linear and point features. Similarly to the LiDAR based trends, the area of forest and hedgerows ET in the NL ET maps decreased in all regions, by 0.8 % per year in Zeeland (2014–2020), 0.4 % per year in Rivierenland (2013–2021) and 0.7 % per year in Limburg (2013–2021) (Fig. 9). The area of Top10NL features, i.e. hedgerows, tree lines and trees, showed a large increase in all regions, but this was likely due to an uncomplete registration of these features before 2015. 3.4. Differences between LiDAR-based and NL ET maps – Top10NL data LiDAR-based WLF maps did not cover 30 to 40 % (depending on the Table 8 LiDAR based WLF area (in ha) and WLF cover (%) in rural areas, agricultural land, and rural areas outside agricultural land in the three regions (last LiDAR acquisition date). Rural area −in agricultural land Rural area −outside agricultural land Rural area −total Area WLF (in ha) Total area (in ha) % WLF cover Area WLF (in ha) Total area (in ha) % WLF cover Area WLF (in ha) Total area (in ha) % WLF cover Limburg (for 2021) 6,247 112,272 5.6 % 4,919 24,691 19.9 % 11,166 136,963 8.2 % Rivierenland (for 2021) 2,283 84,213 2.7 % 2,561 17,161 14.9 % 4,844 101,374 4.8 % Zeeland (for 2020) 1,546 128,797 1.2 % 2,001 16,933 11.8 % 3,547 145,730 2.4 % All three regions 10,076 325,282 3.1 % 9,481 58,786 16.1 % 19,557 384,067 5.1 % Fig. 8. Adjusted total gain, loss and net change WLF areas in hectare in Limburg (2012–2021), Rivierenland (2011–2021) and Zeeland (2014–2020), according to Lidar data adjusted for false change detections. Error bars show 95% confidence intervals. N. Grondard et al. Ecological Indicators 178 (2025) 113853 9 Table A1 (continued) Region Top10NL feature Average size % of WLF cover* detected by LiDAR-based WLF maps as … Qty Unit (group of) trees (in %) hedgerow (in %) forest patch (in %) any WLF type (in %) Trees 81 m 2 25.0 0.0 64.8 89.8 Zeeland Forest polygons 2657 m 2 1.3 2.3 80 83.6 Tree lines 4.6 m 32.1 18.4 45.9 96.4 Hedgerows 4.0 m 8.5 47.0 15.0 70.5 Trees 79 m 2 19.5 2.3 72.1 93.9 Table A2 Detailed results of LiDAR WLF net change area adjustment in Limburg. Change size category Change n i Map change area (in ha) pˆ i (CI 95% ) Adjusted change area (CI 95% ) in ha less than 12.57 m2 gain 10 888 0.1 (0, 0.29) 89 (0, 254) less than 12.57 m2 loss 10 −546 0.1 (0, 0.29) −55 (0, −156) between 12.57 & 100 m2 gain 19 1063 0.68 (0.48, 0.89) 727 (505, 949) between 12.57 & 100 m2 loss 18 −782 0.61 (0.39, 0.84) −478 (−302, −654) between 100 & 200 m2 gain 10 178 0.8 (0.55, 1) 142 (98, 178) between 100 & 200 m2 loss 10 −269 0.8 (0.55, 1) −215 (−148, −269) between 200 & 1000 m2 gain 20 236 0.95 (0.85, 1) 224 (201, 236) between 200 & 1000 m2 loss 20 −426 0.9 (0.77, 1) −383 (−327, −426) between 1000 m2 & 1 ha gain 10 212 1 (1, 1) 212 (212, 212) between 1000 m2 & 1 ha loss 10 –333 0.9 (0.71, 1) −300 (−238, –333) more than 1 ha gain 10 39 0.9 (0.71, 1) 35 (28, 39) more than 1 ha loss 10 −81 0.8 (0.55, 1) −65 (−45, −81) Total net change area 214 a −350 (−470, −230) a Net change area obtained without removing patches in change size categories with correct change detection less than 80 % and without adjustment for false detection. Table A3 Detailed results of LiDAR WLF net change area adjustment in Rivierenland. Change size category Change n i Map change area (in ha) pˆ i (CI 95% ) Adjusted change area (CI 95% ) in ha less than 12.57 m2 gain 10 397 0.1 (0, 0.29) 40 (0, 113) less than 12.57 m2 loss 10 −258 0.2 (0, 0.45) −52 (0, −116) between 12.57 & 100 m2 gain 20 692 0.6 (0.39, 0.81) 415 (267, 564) between 12.57 & 100 m2 loss 18 −468 0.5 (0.27, 0.73) −234 (−126, −342) between 100 & 200 m2 gain 10 145 0.8 (0.55, 1) 116 (80, 145) between 100 & 200 m2 loss 10 −165 1 (1, 1) −165 (−165, −165) between 200 & 1000 m2 gain 20 189 0.95 (0.85, 1) 180 (162, 189) between 200 & 1000 m2 loss 20 −288 1 (1, 1) −288 (−288, −288) between 1000 m2 & 1 ha gain 10 138 1 (1, 1) 138 (138, 138) between 1000 m2 & 1 ha loss 10 −330 1 (1, 1) −330 (−330, −330) more than 1 ha gain 10 41 1 (1, 1) 41 (41, 41) more than 1 ha loss 10 −132 1 (1, 1) −132 (−132, −132) Total net change area −2 a −440 (−480, −400) a Net change area obtained without removing patches in change size categories with correct change detection less than 80 % and without adjustment for false detection. Table A4 Detailed results of LiDAR WLF net change area adjustment in Zeeland. Change size category Change n i Map change area (in ha) pˆ i (CI 95% ) Adjusted change area (CI 95% ) in ha less than 12.57 m2 gain 10 378 0.1 (0, 0.29) 38 (0, 108) less than 12.57 m2 loss 10 −187 0.1 (0, 0.29) −19 (0, −53) between 12.57 & 100 m2 gain 19 442 0.58 (0.36, 0.8) 256 (158, 354) between 12.57 & 100 m2 loss 19 −234 0.47 (0.25, 0.7) −111 (−58, −164) between 100 & 200 m2 gain 10 67 0.8 (0.55, 1) 54 (37, 67) between 100 & 200 m2 loss 10 −77 1 (1, 1) −77 (−77, −77) between 200 & 1000 m2 gain 20 77 0.85 (0.69, 1) 66 (54, 77) between 200 & 1000 m2 loss 20 −150 0.9 (0.77, 1) −135 (−115, −150) between 1000 m2 & 1 ha gain 10 51 0.8 (0.55, 1) 41 (28, 51) between 1000 m2 & 1 ha loss 10 −156 1 (1, 1) −156 (−156, −156) more than 1 ha gain 4 5 1 (1, 1) 5 (5, 5) more than 1 ha loss 10 −46 1 (1, 1) −46 (−46, −46) Total net change area 206 a −248 (−279, −217) a Net change area obtained without removing patches in change size categories with correct change detection less than 80 % and without adjustment for false detection. N. Grondard et al. Ecological Indicators 178 (2025) 113853 16 Data availability Woody landscape features maps derived from LiDAR data are accessible on the following data repository: https://doi.org/ 10.5281/zenodo.14841121. References Albrecht, M., Kleijn, D., Williams, N.M., Tschumi, M., Blaauw, B.R., Bommarco, R., Campbell, A.J., Dainese, M., Drummond, F.A., Entling, M.H., Ganser, D., Arjen de Groot, G., Goulson, D., Grab, H., Hamilton, H., Herzog, F., Isaacs, R., Jacot, K., Jeanneret, P., Sutter, L., 2020. The effectiveness of flower strips and hedgerows on pest control, pollination services and crop yield: a quantitative synthesis. Ecol. Lett. 23 (10), 1488–1498. https://doi.org/10.1111/ele.13576. Arnaiz-Schmitz, C., Herrero-J´ auregui, C., Schmitz, M.F., 2018. Losing a heritage hedgerow landscape. Biocultural diversity conservation in a changing socialecological Mediterranean system [Article]. Sci. Total Environ. 637–638, 374–384. https://doi.org/10.1016/j.scitotenv.2018.04.413. Beeldmateriaal Nederland. (2024). Beeldmateriaal Nederland. https://www.beeldmate riaal.nl/. Broughton, R.K., Chetcuti, J., Burgess, M.D., Gerard, F.F., Pywell, R.F., 2021. A regionalscale study of associations between farmland birds and linear woody networks of hedgerows and trees. Agr Ecosyst Environ 310. https://doi.org/10.1016/j. agee.2021.107300. Carlier, J., Moran, J., 2019. Hedgerow typology and condition analysis to inform greenway design in rural landscapes [Article]. J. Environ. Manage. 247, 790–803. https://doi.org/10.1016/j.jenvman.2019.06.116. Czúcz, B., Baruth, B., Angileri, V., Prieto, L.A., Terres, J.-M., 2022. Landscape features in the EU Member States. Czucz, B., Baruth, B., TERRES, J.-M., Gallego, P.F., Hagyo, A., Angileri, V., Nocita, M., Perez-Soba, A.M., Koeble, R., Paracchini, M.-L., 2022. Classification and quantification of landscape features in agricultural land across the EU. Denac, K., Kmecl, P., 2021. Land consolidation negatively affects farmland bird diversity and conservation value. J. Nat. Conserv. 59, 125934. https://doi.org/10.1016/j. jnc.2020.125934. Drexler, S., Gensior, A., Don, A., 2021. Carbon sequestration in hedgerow biomass and soil in the temperate climate zone. Reg. Environ. Chang. 21 (3). https://doi.org/ 10.1007/s10113-021-01798-8. England, J.R., O’Grady, A.P., Fleming, A., Marais, Z., Mendham, D., 2020. Trees on farms to support natural capital: an evidence-based review for grazed dairy systems. Sci. Total Environ. 704, 135345. https://doi.org/10.1016/j.scitotenv.2019.135345. Estrada, J., S´ anchez, H., Hernanz, L., Checa, M., Roman, D., 2017. Enabling the use of sentinel-2 and LiDAR data for common agriculture policy funds assignment. ISPRS Int. J. Geo Inf. 6 (8). https://doi.org/10.3390/ijgi6080255. European Commission, 2018. Regulation of the European Parliament and the Council establishing rules on support for strategic plans to be drawn up by Member States under the Common agricultural policy (CAP Strategic Plans) and financed by the European Agricultural Guarantee Fund (EAGF) and by the European Agricultural Fund for Rural Development (EAFRD) and repealing Regulation (EU) No 1305/2013 of the European Parliament and of the Council and Regulation (EU) No 1307/2013 of the European Parliament and of the Council. Retrieved from https://eur-lex. europa.eu/legal-content/EN/TXT/?uri=COM:2018:392:FIN. European Commission, 2020. Communication from the European Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions, EU Biodiversity Strategy for 2030 Bringing nature back into our lives, COM/2020/380 final. https://eur-lex.europa.eu/legal-content/ EN/TXT/?uri=celex:52020DC0380. European Union, 2024. Regulation (EU) 2024/1991 of the European Parliament and of the Council of 24 June 2024 on nature restoration and amending Regulation (EU) 2022/869. Retrieved from https://eur-lex.europa.eu/eli/reg/2024/1991/oj. Garratt, M.P.D., Senapathi, D., Coston, D.J., Mortimer, S.R., Potts, S.G., 2017. The benefits of hedgerows for pollinators and natural enemies depends on hedge quality and landscape context [Article]. Agr. Ecosyst. Environ. 247, 363–370. https://doi. org/10.1016/j.agee.2017.06.048. Graham, L., Gaulton, R., Gerard, F., Staley, J.T., 2018. The influence of hedgerow structural condition on wildlife habitat provision in farmed landscapes. Biol. Conserv. 220, 122–131. https://doi.org/10.1016/j.biocon.2018.02.017. Hagendoorn, L., Souwerbren, P., van Herwaarden, G.-J., van Korven, T., Geertsema, W., Sengers, G., 2021. Aanvalsplan versterking landschappelijke identiteit via landschapselementen. https://edepot.wur.nl/544668. Hein, L., Bagstad, K.J., Obst, C., Edens, B., Schenau, S., Castillo, G., Soulard, F., Brown, C., Driver, A., Bordt, M., Steurer, A., Harris, R., Caparros, A., 2020a. Progress in natural capital accounting for ecosystems. Science 367 (6477), 514–515. https:// doi.org/10.1126/science.aaz8901. Hein, L., Remme, R.P., Schenau, S., Bogaart, P.W., Lof, M.E., Horlings, E., 2020b. Ecosystem accounting in the Netherlands. Ecosyst. Serv. 44. https://doi.org/ 10.1016/j.ecoser.2020.101118. Hijmans, R.J., Bivand, R., Cordano, E., Dyba, K., Pebesma, E., Sumner, M.D., 2024. terra: Spatial Data Analysis. https://cran.r-project.org/package=terra. Kadaster, 2020. Basisregistratie Topografie: Catalogus en Productspecificaties 1.2.0.3). https://kadaster.github.io/imbrt/. Kadaster, 2023. Dataset: Basisregistratie Adressen en Gebouwen (BAG). https://www. pdok.nl/introductie/-/article/basisregistratie-adressen-en-gebouwen-ba-1. Kadaster, 2024. Dataset: Basisregistratie Topografie (BRT) TOPNL. https://www.pdok. nl/introductie/-/article/basisregistratie-topografie-brt-topnl. Kakoulaki, G., Martinez, F.A.M., Florio, P., 2021. Non-commercial light detection and ranging (lidar) data in europe. Kleeschulte, S., Ruf, K., Marin, A., Schr¨ oder, C., Maucha, G., Kosztra, B., Hazeu, G., Elbersen, B., Schwaiger, E., Weiss, M., 2023. Quantification of landscape features in agricultural areas using copernicus products: an overview of recent developments. ETC-DI Report. Koomen, A., Maas, G., Weijschede, T., 2007. Veranderingen in lijnvormige cultuurhistorische landschapselementen: resultaten van een steekproef over de periode 1900-2003. (WOt-rapport ; No. 34). https://edepot.wur.nl/43122. Kremen, C., M’Gonigle, L.K., Ponisio, L.C., 2018. Pollinator community assembly tracks changes in floral resources as restored hedgerows mature in agricultural landscapes. Front. Ecol. Evol. 6. https://doi.org/10.3389/fevo.2018.00170. LandschappenNL, 2024. Meetnet Agrarisch Cultuurlandschap. http://www.landschaps observatorium.nl/Meetnet-Agrarisch-Cultuurlandschap/. Lucas, C., Bouten, W., Koma, Z., Kissling, W., Seijmonsbergen, A., 2019. Identification of linear vegetation elements in a rural landscape using LiDAR point clouds. Remote Sens. (Basel) 11 (3). https://doi.org/10.3390/rs11030292. Luscombe, D.J., Gatis, N., Anderson, K., Carless, D., Brazier, R.E., 2023. Rapid, repeatable landscape-scale mapping of tree, hedgerow, and woodland habitats (THaW), using airborne LiDAR and spaceborne SAR data. Ecol. Evol. 13 (5), e10103. https://doi.org/10.1002/ece3.10103. Ministerie van Landbouw, Natuur en Voedselkwaliteit, Interprovinvial Overleg, & Unie van Waterschapen, 2022. The Netherlands National Strategic Plan CAP 2023-2027. Retrieved from https://www.toekomstglb.nl/documenten/publicaties/2022/11/0 2/definitief-nationaal-strategisch-plan. Montgomery, I., Caruso, T., Reid, N., 2020. Hedgerows as ecosystems: service delivery, management, and restoration [review]. Annu. Rev. Ecol. Evol. Syst. 51, 81–102. https://doi.org/10.1146/annurev-ecolsys-012120-100346. Olofsson, P., Foody, G.M., Herold, M., Stehman, S.V., Woodcock, C.E., Wulder, M.A., 2014. Good practices for estimating area and assessing accuracy of land change. Remote Sens. Environ. 148, 42–57. https://doi.org/10.1016/j.rse.2014.02.015. Raap, E., Brinkkemper, O., Baas, H., 2022. Panorama Landschap en de structuurdragers. Karakterisering Van Het Nederlandse Landschap in 78 Regio’s. Rijksdienst voor het, C.E., Landschappen, N.L., Het, K., 2024. Monitor Landschap - Opgaand groen. https://monitor-landschap-2-monitorlandschap.hub.arcgis. com/pages/opgaand-groen-v2. Robinson, R.A., Sutherland, W.J., 2002. Post-war changes in arable farming and biodiversity in Great Britain. J. Appl. Ecol. 39 (1), 157–176. https://doi.org/ 10.1046/j.1365-2664.2002.00695.x. Roelofsen, H, 2022. Ecosysteemdiensten & landschapselementen: De rol van landschapselementen in de levering van ecosysteemdiensten in het landelijk gebied. Doi: 10.18174/567379. Roussel, J.-R., Auty, D., Coops, N.C., Tompalski, P., Goodbody, T.R., Meador, A.S., Bourdon, J.-F., De Boissieu, F., Achim, A., 2020. lidR: an R package for analysis of Airborne Laser Scanning (ALS) data. Remote Sens. Environ. 251, 112061. https:// doi.org/10.1016/j.rse.2020.112061. RVO, N., 2023. Basisregistratie Gewaspercelen (BRP). https://www.pdok.nl/atom-do wnloadservices/-/article/basisregistratie-gewaspercelen-brp. RVO, N., 2024. Referentie percelen (Dutch Land Parcel Information System). https ://www.pdok.nl/atom-downloadservices/-/article/referentiepercelen. Silva, C.A., Hudak, A.T., Vierling, L.A., Loudermilk, E.L., O’Brien, J.J., Hiers, J.K., Jack, S.B., Gonzalez-Benecke, C., Lee, H., Falkowski, M.J., Khosravipour, A., 2016. Imputation of individual longleaf pine (Pinus palustris Mill.) tree attributes from field and LiDAR data. Can. J. Remote. Sens. 42 (5), 554–573. https://doi.org/ 10.1080/07038992.2016.1196582. Statistics Netherlands, 2017. Bestand bodemgebruik. https://www.cbs.nl/nl-nl/dossier /nederland-regionaal/geografische-data/bestand-bodemgebruik. Statistics Netherlands, & WUR., 2022. Natural Capital Accounting in the Netherlands - Technical report. https://www.cbs.nl/en-gb/society/nature-and-environment/natur al-capital/technical-report. United Nations, European Union, Food and Agriculture Organization of the United Nations, International Monetary Fund, Organisation for Economic Co-operation and Development, & The International Bank for Reconstruction and Development, 2024. System of EnvironmentalEconomic Accounting – Ecosystem Accounting (Statistical Papers. Series F No. 124, Issue. https://seea.un.org/sites/seea.un.org/files/do cuments/EA/seea_ea_f124_web_12dec24.pdf. van Apeldoorn, D.F., Kempen, B., Sonneveld, M.P.W., Kok, K., 2013. Co-evolution of landscape patterns and agricultural intensification: an example of dairy farming in a traditional Dutch landscape. Agr. Ecosyst. Environ. 172, 16–23. https://doi.org/ 10.1016/j.agee.2013.04.002. van Doorn, A., Nieuwenhuizen, W., Meijer, M., Snepvangers, J., Herwaarden, G.J., 2016. Samen naar een registratie van groene en blauwe landschapselementen : een haalbaarheidsstudie. Wageningen Environ. Res. https://doi.org/10.18174/392483. Weninger, T., Scheper, S., Lack´ oov´ a, L., Kitzler, B., Gartner, K., King, N.W., Cornelis, W., Strauss, P., Michel, K., 2021. Ecosystem services of tree windbreaks in rural landscapes—a systematic review. Environ. Res. Lett. 16 (10). https://doi.org/ 10.1088/1748-9326/ac1d0d. N. Grondard et al. Ecological Indicators 178 (2025) 113853 17