Using LiDAR data to assess the influence of
fine-scale habitat structure on the abundance of the
bearded reedling (Panurus biarmicus)
Ashleigh Campbell
Name / Title
Specialisation
Institute
Involvement
Dr. W. Daniel Kissling Macroecology, ecological informatics,
global biodiversity change
Biogeography and Macroecology (BIOMAC) lab, Institute for Biodiversity and Ecosystem Dynamics (IBED), University of Amsterdam
Daily supervisor and Examiner
Zsófia Koma (PhD student)
LiDAR, GIS, remote sensing
Biogeography and Macroecology (BIOMAC) lab, Institute for Biodiversity and Ecosystem Dynamics (IBED), University of Amsterdam
Support with Species Distribution Modelling,
bird data and LiDAR data
Ir. Henk Sierdsema Senior researcher and coordinator Species and
Habitats
Sovon Dutch Centre for Field Ornithology
Provision of bird and covariate datasets and
modelling support Dr. Yifang Shi Scientific developer for
ecological applications of LiDAR Remote Sensing (LifeWatch-ERIC Virtual Laboratory Innovation Center) Biogeography and Macroecology (BIOMAC) lab, Institute for Biodiversity and Ecosystem Dynamics (IBED), University of Amsterdam
Support with LiDAR aspects of the project
Dr. A.C. Seijmonsbergen Geomorphology, geodiversity, remote sensing, natural hazards, LiDAR Biogeography and Macroecology (BIOMAC) lab, Institute for Biodiversity and Ecosystem Dynamics (IBED), University of Amsterdam
2
Summary
Wetlands provide an important habitat for a wide range of birds, mammals and invertebrates as well as some of the most threatened species. Reedbeds in particular are in decline globally, with many species relying upon the reed for all aspects of their life cycle. The fine-scale structure of ecosystems is of great importance when considering the distribution of species and level of biodiversity in an area. Collecting this data with field methods is time consuming and impossible to carry out in a spatially contiguous way and over large spatial extents. This has meant that current species mapping methods, such as species distribution modelling (SDM), lack data on fine-scale habitat structure. Advances in remote sensing technology, particularly LiDAR (Light Detection And Ranging), allow for vegetation structure to be measured directly over broad extents and at a high resolution. This study will focus on the bearded reedling (Panurus biarmicus), a reedbed specialist of conservation concern found in wetlands across the Netherlands. Unlike many wetland birds, the species is not migratory, and relies solely upon one species of reed (Phragmites australis) for all aspects of their ecology. This makes them particularly vulnerable to any changes in their habitat composition. In this project, I will use LiDAR metrics to test whether and how fine-scale habitat structure influences the relative abundance of the bearded reedling, beyond other variables such as land use, soil type and climate data. This will be carried out using habitat-specific LiDAR metrics for the bearded reedling, bearded reedling observation data and other relevant covariates (climate, soil type, land use) to create SDMs.
3
Theoretical Framework
Effective conservation management requires a thorough understanding of species distributions, and which factors determine this distribution. The fine-scale structure of ecosystems has proven to be of great importance when considering the distribution of species and biodiversity in an area (Hewitt et al., 2005). At broad spatial scales, factors such as climate are thought to be the main drivers of diversity and distribution of terrestrial species (Pearson and Dawson, 2003). However, this is not representative of how species interact directly with their local habitat, and how the 3D habitat structure shapes species distribution on a local scale. Habitat structure is directly influential in the creation and availability of niches for different species, thus influencing both species diversity and composition within a given area (Davies and Asner, 2014). Many studies have shown the relevance of both vertical and horizontal variability of vegetation in understanding how different species use their habitat, and how this drives both the distribution and diversity of species (Bakx et al., 2019; Davies and Asner, 2014). Diversity in vegetation height was first presented as an important factor correlated with an increase in avian diversity in 1961 (MacArthur and MacArthur, 1961). Since then, multiple studies have highlighted the importance of structural vegetation heterogeneity for birds in particular (Müller et al., 2010; Robinson and Holmes, 1984).
Species Distribution Models (SDMs), sometimes referred to as ecological niche models or habitat models, are a significant factor involved in conservation planning (Austin, 2002). These models are designed either as predictive tools or to better understand the habitat requirements of different species (Vaughan and Ormerod, 2005). They allow for quantitative modelling and mapping of a species’ distribution or abundance using species specific environmental and occurrence data (Elith and Leathwick, 2009). These standardised methodologies have been developed and improved over the years and are now central to many areas of research including biogeography, ecology and conservation biology (Araújo and Guisan, 2006). Climate and land use datasets are the most used variables for SDMs, thought to be the main drivers of species distribution (Pearson and Dawson, 2003). Although usage of these types of data is beneficial in many cases, especially over large spatial extents, it does not take into account the role of habitat structure in species distribution (Koma et al., 2021). As discussed previously, habitat structure is known to directly influence species distribution and biodiversity (Hewitt et al., 2005), therefore the inclusion of data of this kind in SDMs would be invaluable. Measuring habitat structure over large spatial scales using field methods is, however, time consuming and impossible to fulfil in a spatially contiguous way (Bakx et al., 2019). This has meant that, until recently, it was not feasible to include fine scale habitat structure when modelling species distribution. Active remote sensing methods, such as (Light Detection And Ranging) LiDAR, can provide a solution, allowing for vegetation to be measured directly, over broad extents, at high resolution and in a spatially contiguous way (Davies and Asner, 2014). Use of LiDAR for ecological applications and habitat modelling is a relatively new concept. This is due to recent advances in LiDAR technology allowing for detailed 3D modelling of animal habitats (Davies and Asner, 2014). LiDAR data collection involves aerial laser altimetry (usually with a drone or other aircraft), where laser pulses are emitted towards the ground, and the time taken for the pulse to return is recorded (Lohani and Ghosh, 2017). These light pulses can penetrate through vegetation layers, providing multiple pulse returns, mapping the 3D structure of a habitat in the form of a point cloud (Davies and Asner, 2014). As the resolution of LiDAR data is so high (< 1m2 resolution), it can be used to provide a detailed picture of the habitat structure of a species (Vierling et al., 2008). LiDAR data allows for the 3D structure of vegetation to be analysed and classified into vegetation metrics. These metrics can be defined by vegetation part (e.g. total vegetation, single trees, canopy, understory) and structural type (e.g. cover, height, horizontal variability, vertical variability) (Bakx et al., 2019). By categorising the vegetation in this way, you create a conceptual framework within which
4 ecological LiDAR metrics can be compared. The metrics must also be calculated and extracted from the LiDAR point cloud in order to be used for modelling applications. The canopy height could, for example, be calculated as the average height of the first returned laser pulses - these will be the tallest plants (Ackers et al., 2015; Roll et al., 2015). This information can further our understanding of how species use their habitat at the local scale, and how habitat structure is a driving factor in species abundance and distributions.
Most of the LiDAR-based studies for ecological applications have so far been focussed around birds in forested habitats (Davies and Asner, 2014). This is due to LiDAR being well suited for use within forest habitats, capturing the 3D structure of the canopy layers in detail (Maltamo et al., 2014). Subsequently there is a lack of representation of certain habitat types in ecological applications of LiDAR, such as savannas and wetlands, within the relevant literature (Bakx et al., 2019). LiDAR has, however, been used for mapping within wetlands and reedbed habitats, and has shown to improve the accuracy of vegetation mapping in these environments (Koma et al., 2020; Onojeghuo and Blackburn, 2011). One recent avian study has also looked into the use of LiDAR in wetlands for identifying niche separation and overlap of closely related bird species (Koma et al., 2021). Reedbeds are in decline globally and are often subject to intensive anthropogenic management (Malzer, 2017). They provide important habitats for some of the most threatened animal species, many of which - such as the bearded reedling-, rely on the reed for all aspects of their life cycle (Onojeghuo and Blackburn, 2011). The lack of ecological research in reedbeds using LiDAR data, and the relevance for conservation, makes this a highly interesting focal habitat for this study.
The bearded reedling (Panurus biarmicus) is a wetland bird species and a reedbed specialist. They utilise one species of reed (Phragmites australis) for both feeding and breeding, and thus rely on large expanses of reeds and are not found in reedbeds containing trees or shrubs. These specific habitat requirements leave them particularly vulnerable to any changes in habitat composition (Beemster et al., 2010). Unlike many other wetland bird species within the Netherlands, bearded reedlings are not migratory and stay within the reedbeds year-round. To account for the changes in season, bearded reedlings alter their gut morphology, allowing them to be purely insectivorous during the summer and eat Phragmites australis (P. australis) seeds throughout the winter months (Malzer, 2017). As the bearded reedling relies on P. australis for all aspects of its ecology, reed structure is likely to be a highly influential factor in the relative abundance of the species. Literature shows preferential selection of old, dense areas of reed for nesting, with areas of younger reed commonly selected for foraging (Malzer, 2017; Trnka and Prokop, 2006). Nest location is of greatest priority and individuals are known to take foraging flights of up to 500m from their nesting site (Malzer, 2017). Reed density has also shown to positively affect the abundance of the species during the breeding season (Malzer and Hansell, 2017; Trnka and Prokop, 2006). Bearded reedlings can be highly sensitive to local climatic conditions, with warmer springs allowing for earlier breeding and colder winters resulting in high mortality at a local scale (Malzer, 2017; Malzer and Hansell, 2017).
The study will be based on bearded reedling populations in the Netherlands due to the availability of high-quality bird monitoring data (Sovon) and countrywide LiDAR surveys (AHN) for this research. The outcome of this research could be beneficial in advising conservation management on how the habitat structure within reedbeds is affecting the distribution of the bearded reedling, a species of conservation interest in the Netherlands. If successful, the framework in this study could also be applied to other reed dwelling species, increasing our knowledge of how individual species utilise the fine-scale structure of their habitats and how this drives their distribution at the local scale. Additionally, there is also the possibility for future up-scaling of the project to apply the methodologies proposed in this study to bearded reedlings in other countries and compare the results.
5
Research aim
This study will use LiDAR metrics to test whether and how fine-scale habitat structure influences the distribution of the bearded reedling, beyond other data sources such as land cover, soil type and topographic features. As the usage of LiDAR data for ecological applications is a relatively new field with a lack of studies on wetlands, and reedbeds in particular, I aim to expand the current knowledge base in this field. I will use the standardised methodology used by Sovon (Dutch Centre for Field Ornithology) and assess if and to what extent LiDAR data adds to this process. If the results of the project meet our expected outcomes, they could be used to advise conservation management on wetland areas within the Netherlands. As the bearded reedling is a species of conservation concern in the Netherlands, these results could be beneficial for the future monitoring and protection of the species. The methodology in this study could also be used to assess the relative importance of fine-scale habitat for the bearded reedling outside of the Netherlands, as well as having potential applications for other reedbed species.
Research questions
- Does the usage of LiDAR metrics improve the accuracy of species distribution models for the bearded reedling?
- Is the 3D habitat structure of the bearded reedling of greater importance for the relative abundance of the species than factors such as soil type, climate and landcover?
- Which LiDAR metrics are most important for understanding the habitat requirements of the bearded reedling?
Expected results
Based on the research questions above, I expect the outcome of this study to show that the fine scale habitat structure of the bearded reedling has a greater influence on the relative abundance of the species than other, more commonly used, predictor variables.
I also expect that the usage of LiDAR metrics as predictor variables within a species distribution model for the bearded reedling will increase the accuracy of the model output.
Finally, I expect that the LiDAR metrics most important for use in species distribution modelling of the bearded reedling will be the canopy height and horizontal variability metrics (e.g. patchiness of vegetation). This is due to the habitat preferences of this species being extensive reedbeds, relying upon a single reed species (P. australis), with no shrubs or trees.
6
Methods
Workflow
Figure 1 shows the proposed workflow. This is a simplified version of the stages that will be required in the methodology of this study. Each aspect of the workflow diagram and the processes and methodologies involved in each stage are explained and justified later in the relevant methodology section.
Bearded reedling data
The observation data for the bearded reedling is provided by Sovon (Dutch Centre for Field Ornithology) in several different datasets. The first is the bird Atlas field observation data, consisting of 1km data, 5 minute point counts and 10 minute point counts. For the collection of the bird Atlas data, the Netherlands is split into 5km2 plots and within these plots eight 1km2 plots are selected. For surveys, visits of one hour are carried out to each of the eight selected 1km2 plots, point counts occur in the centre of the 1km2 plot for 10-minute and 5-minute periods. Each 1km2 is surveyed several times throughout both the winter and breeding season. The other dataset provided by Sovon is the territory mapping data for the breeding monitoring scheme, collected within specific survey plots across the Netherlands. Unlike the ATLAS data, this data does not give complete coverage of the country but provides bird observation data with a 10m accuracy. From these datasets, I will only use the observation data that aligns with the time period in which the AHN3 LiDAR data was collected, between 2014 - 2019.
The bearded reedling observation data will be prepared and aggregated for use in the SDM, and pseudo-absences will be generated to provide adequate data input for the model. This is because SDMs that use both
Figure 1. The proposed methodological workflow of this study. Blue boxes represent raw input data to be
7 presence and absence data perform better than those using presence data only (Barbet‐Massin et al., 2012). Aggregation of the observation data will allow us to look at whether and how fine-scale habitat structure is affecting the relative abundance of the species throughout the Netherlands. These data pre-processing stages will be carried out in R. It is yet to be decided to what resolution the observation data will be aggregated. This will depend upon the resolution of covariates to be included in the model. Land use data is available from the Land Use Database of the Netherlands (LGN), bioclimatic variables are available from WorldClim and soil data is available from de Bodemkaart van Nederland (the soil map of the Netherlands).
LiDAR data
AHN (Actueel Hoogtebestand Nederland) is an open source digital elevation file for the whole of the Netherlands, produced using laser altimetry. I will use AHN3, the most recent AHN data that is available, this dataset was collected between 2014 – 2019 (‘AHN: The making of | AHN’, 2021). AHN3 was collected during the ‘leaf-off’ season of these years and has a high resolution of between 6 - 10 points per square meter (‘Kwaliteitsbeschrijving | AHN’, 2021). This data will contain the LiDAR point cloud file, stored in LAZ format, where a classification has been applied to each of the points (‘AHN: The making of | AHN’, 2021). Manipulation and visualisation of the LiDAR data will likely be carried out within R in the package ‘lidR’ and in ArcGIS Pro. The ‘lidR’ package allows LAS files to be read into R so that metrics can then be computed (Roussel et al., 2021).
To derive the LiDAR-based vegetation metrics, a thorough review of literature on the ecology of the bearded reedling must first be carried out to identify relevant habitat preferences. Horizontal and vertical vegetation metrics can then be selected and calculated based upon the structural habitat preferences of the bearded reedling. The LiDAR metrics will then be derived through rasterization, which is currently the most commonly used method (Bakx et al., 2019). For the bearded reedling it is likely that canopy height and reed density will be some of the most interesting metrics to extract from the data due to its preference for older areas of reed for nesting which tend to be tall and denser in nature (Malzer, 2017; Trnka and Prokop, 2006). To assess this, I will calculate the horizontal variability of the vegetation and the canopy height. Horizontal variability could be calculated as the standard deviation of canopy cover in a given area (Smart et al., 2012), and canopy height could be calculated as the average height of first returns (canopy layer) (Ackers et al., 2015).
Modelling and analysis
Species distribution modelling (SDM) will be carried out in R, using either the R package ‘sdm’ or the package ‘SDMaps’. The package ‘sdm’ provides an environment within which the whole modelling process can be performed using multiple modelling techniques offered by different packages (Naimi and Araújo, 2016). The ‘SDMaps’ R package was created by Sovon specifically for the analysis of species abundance and distribution data, making it an ideal choice for use in this study. I will use three SDM algorithms suitable for use with presence absence data: Random Forest (RF), Boosted Regression Trees (BRT) and Generalized Linear Models (GLM). GLMs are a standard regression modelling approach used in SDMs (Elith and Graham, 2009). Both RF and BRT are tree-based, machine-learning classification and regression modelling methods that make very few assumptions. Using both algorithms allows for a comparison of two ensemble tree models. RF in particular has been used for similar studies (de Vries et al., 2021). It is also possible that I will decide to use an ensemble modelling approach. This approach is often used by Sovon for their analysis, allowing the use of multiple types
8 of model that, combined, can often give the best output (Araújo and New, 2007). The models will be run with and without the LiDAR data so that I can compare the accuracy of the original model output with the LiDAR adapted model.
To assess and compare the accuracy of the model outputs, accuracy methods such as AUC (Area Under the ROC Curve) and TSS (True Skill Statistics) will be used (Shabani et al., 2018). I will use a standardised protocol for reporting the SDM results such as the ODMAP (Overview, Data, Model, Assessment and Prediction) protocol to ensure transparency and reproducibility of the modelling process (Zurell et al., 2020).
Time schedule
Table 1 below shows the predicted time span for each of the stages of the project. There is room for changes in the timespan of each objective dependent on progress, but this will be monitored throughout the project duration and changed if necessary.
The first stage of this project will involve a literature review on the habitat requirements of the bearded reedling in order to define the appropriate LiDAR metrics for the study. Throughout March and April, I will continue reviewing literature alongside starting the preparation of data and the extraction and calculation of LiDAR metrics. Both the data preparation and LiDAR metric extraction/calculation are likely to take up a large proportion of the designated project time, so I have allowed four months for both objectives. This should ensure I have enough time in case of any delays or issues with the data. I have also allowed three months for the analysis and modelling stage as this will involve learning new skills and I could run into issues in the process.
Objectives February March April May June July August
Literature review LiDAR metric definition LiDAR metric extraction
and calculation Data preparation Modelling and analysis
Introduction and methods
Results Discussion and
Conclusion
Table 1. Gantt chart showing the proposed time schedule of the study with the predicted period of time necessary for completion of each individual stage.
9 I will start writing the introduction and methodology for the project in March so that only minor changes and additions are necessary by July when these sections should be completed. The results section I aim to start writing in April if the modelling and analysis stages are going well and I already have some results to report. Discussion and conclusion sections I will start in May and continue through until the end of the project in August, alongside a further review of literature in July and August. The final month of the study – August – is left relatively clear except for finalisation of the written aspects of the project. This allows extra time for any overrunning of other project objectives, such as the modelling and analysis.
Budget
Datasets required for the project are either openly available (AHN3, WorldClim, De Bodemkaart van Nederland) or provided for free by Sovon (bearded reedling observation data). LGN7 land use data was purchased previously by the University of Amsterdam and requires no additional purchases for the purpose of this study. Data analysis for the project will be carried out in R software, which is open source and free to use, and ArcGIS Pro, which is provided through the University of Amsterdam as a student licence.
Equipment and Insurance
No specialised equipment is required for the fulfilment of the project. The use of a personal computer will be required and potentially remote access to a computer with a higher processing power, dependent on the performance of the personal computer. This remote access would be made available through the University of Amsterdam or provided in the GIS lab when corona restrictions are eased. R software is open source and ArcGIS software is provided by the University of Amsterdam.
10
References
Ackers, S. H., Davis, R. J., Olsen, K. A., and Dugger, K. M. (2015) The evolution of mapping habitat for northern spotted owls (Strix occidentalis caurina): A comparison of photo-interpreted, Landsat-based, and lidar-based habitat maps. Remote Sensing of Environment 156: 361–373.
AHN: The making of | AHN (/2021). Accessed: 21st March 2021 <https://www.ahn.nl/ahn-making. >. Araújo, M. B., and Guisan, A. (2006) Five (or so) challenges for species distribution modelling. Journal of Biogeography 33(10): 1677–1688.
Araújo, M. B., and New, M. (2007) Ensemble forecasting of species distributions. Trends in ecology & evolution 22(1): 42–47.
Austin, M. P. (2002) Spatial prediction of species distribution: an interface between ecological theory and statistical modelling. Ecological modelling 157(2–3): 101–118.
Bakx, T. R. M., Koma, Z., Seijmonsbergen, A. C., and Kissling, W. D. (2019) Use and categorization of Light Detection and Ranging vegetation metrics in avian diversity and species distribution research. Diversity and Distributions 25(7): 1045–1059.
Barbet‐Massin, M., Jiguet, F., Albert, C. H., and Thuiller, W. (2012) Selecting pseudo-absences for species distribution models: how, where and how many? Methods in Ecology and Evolution 3(2): 327–338. Beemster, N., Troost, E., and Platteeuw, M. (2010) Early successional stages of Reed Phragmites australis vegetations and its importance for the Bearded Reedling Panurus biarmicus in Oostvaardersplassen, The Netherlands. Ardea 98(3): 339–354.
Davies, A. B., and Asner, G. P. (2014) Advances in animal ecology from 3D-LiDAR ecosystem mapping. Trends in ecology & evolution 29(12): 681–691.
de Vries, J. P. R., Koma, Z., WallisDeVries, M. F., and Kissling, W. D. (2021) Identifying fine‐scale habitat preferences of threatened butterflies using airborne laser scanning. Diversity and Distributions.
Elith, J., and Graham, C. H. (2009) Do They? How Do They? Why Do They Differ? On Finding Reasons for Differing Performances of Species Distribution Models. Ecography 32(1): 66–77.
11 Elith, J., and Leathwick, J. R. (2009) Species Distribution Models: Ecological Explanation and Prediction Across Space and Time. Annual Review of Ecology, Evolution, and Systematics 40(1): 677–697.
Hewitt, J. E., Thrush, S. F., Halliday, J., and Duffy, C. (2005) The importance of small‐scale habitat structure for maintaining beta diversity. Ecology 86(6): 1619–1626.
Koma, Z., Grootes, M. W., Meijer, C. W., Nattino, F., Seijmonsbergen, A. C., Sierdsema, H., Foppen, R., and Kissling, W. D. (2021) Niche separation of wetland birds revealed from airborne laser scanning. Ecography n/a(n/a).
Koma, Z., Seijmonsbergen, A. C., and Kissling, W. D. (2020) Classifying wetland‐related land cover types and habitats using fine‐scale lidar metrics derived from country‐wide Airborne Laser Scanning. Remote Sensing in Ecology and Conservation.
Kwaliteitsbeschrijving | AHN (/2021). Accessed: 21st March 2021 <https://www.ahn.nl/kwaliteitsbeschrijving. >.
Lohani, B., and Ghosh, S. (2017) Airborne LiDAR technology: a review of data collection and processing systems. Proceedings of the National Academy of Sciences, India Section A: Physical Sciences 87(4): 567–579. MacArthur, R. H., and MacArthur, J. W. (1961) On bird species diversity. Ecology 42(3): 594–598.
Maltamo, M., Næsset, E., and Vauhkonen, J. (2014) Forestry applications of airborne laser scanning. Concepts and case studies. Manag For Ecosys 27: 2014.
Malzer, I. (2017) Patterns in the space use of the Bearded Reedling, Panurus biarmicus, on the Tay Reedbeds, Scotland. University of Glasgow.
Malzer, I., and Hansell, M. (2017) Nest timing, nest site selection and nest structure in a high latitude population of Bearded Reedlings Panurus biarmicus. Bird Study 64(1): 51–61.
Müller, J., Stadler, J., and Brandl, R. (2010) Composition versus physiognomy of vegetation as predictors of bird assemblages: The role of lidar. Remote Sensing of Environment 114(3): 490–495.
Naimi, B., and Araújo, M. B. (2016) sdm: a reproducible and extensible R platform for species distribution modelling. Ecography 39(4): 368–375.
12 Onojeghuo, A. O., and Blackburn, G. A. (2011) Optimising the use of hyperspectral and LiDAR data for
mapping reedbed habitats. Remote Sensing of Environment 115(8): 2025–2034.
Pearson, R. G., and Dawson, T. P. (2003) Predicting the impacts of climate change on the distribution of species: are bioclimate envelope models useful? Global ecology and biogeography 12(5): 361–371. Robinson, S. K., and Holmes, R. T. (1984) Effects of plant species and foliage structure on the foraging behavior of forest birds. The Auk 101(4): 672–684.
Roll, U., Geffen, E., and Yom‐Tov, Y. (2015) Linking vertebrate species richness to tree canopy height on a global scale. Global Ecology and Biogeography 24(7): 814–825.
Roussel, J.-R., documentation), D. A. (Reviews the, features), F. D. B. (Fixed bugs and improved catalog, segment_snags()), A. S. M. (Implemented wing2015() for, track_sensor()), B. J.-F. (Contributed to R. for, and track_sensor()), G. D. (Implemented G. for (2021) lidR: Airborne LiDAR Data Manipulation and Visualization for Forestry Applications.
Shabani, F., Kumar, L., and Ahmadi, M. (2018) Assessing accuracy methods of species distribution models: AUC, specificity, sensitivity and the true skill statistic. Glob. J. Hum. Soc. Sci 18(1).
Smart, L. S., Swenson, J. J., Christensen, N. L., and Sexton, J. O. (2012) Three-dimensional characterization of pine forest type and red-cockaded woodpecker habitat by small-footprint, discrete-return lidar. Forest Ecology and Management 281: 100–110.
Trnka, A., and Prokop, P. (2006) Reedbed structure and habitat preference of reed passerines during the post-breeding period. Biologia 61(2): 225–230.
Vaughan, I. P., and Ormerod, S. J. (2005) The continuing challenges of testing species distribution models. Journal of Applied Ecology 42(4): 720–730.
Vierling, K. T., Vierling, L. A., Gould, W. A., Martinuzzi, S., and Clawges, R. M. (2008) Lidar: shedding new light on habitat characterization and modeling. Frontiers in Ecology and the Environment 6(2): 90–98.
Zurell, D., Franklin, J., König, C., Bouchet, P. J., Dormann, C. F., Elith, J., et al. (2020) A standard protocol for reporting species distribution models. Ecography 43(9): 1261–1277.
