Assessing structural changes at the forest edge using kernel density estimation

https://doi.org/10.1016/j.foreco.2019.117639Get rights and content

Highlights

  • Automatic quantification and detection of forest edge changes for a large area.

  • Using kernel density estimation (KDE) to describe ALS data.

  • 3 indicators describe the KDE curves and reflected the complexity.

  • Promising tool for using ALS data to analyze the change of forest edges.

Abstract

Due to the strong impact of the composition and structure of the forest edge on forest ecosystems, the value of investigating changes at the forest edge has been recognized for a few decades. Automatic quantification and detection of forest edge changes has advantages over fieldwork with respect to the time and cost required. Based on the structural features extracted from Airborne Laser Scanning (ALS) data, in this study an automatic method using Kernel Density Estimation (KDE) was developed to describe ALS point data related to the forest edge in terms of a finite number of interpretable indicators. The method was applied to identify patterns in data sets at the forest edge between two given dates (2001 and 2014) in European temperate mixed forests. The study area has an extent of 1403 km2 and the total length of forest boundary is approx. 5000 km.

The calculated structural features explained the vegetation’s height distribution and the structural variance at the forest edge. The kernel density curves represented the general structural diversity of the study area. Three extracted indicators described the shape of the kernel density curves and reflected the dynamics of the studied forest edge area. Overall, results from this study show that the proposed approach is a promising tool for using remote sensing data to analyze whether or not the vertical structure of the forest edge for the entire study area has changed and thus to assess automatically the progress of forest management at forest edge.

Introduction

The forest edge, defined as the transition zone from forest to open land (Smith, 1974), plays a key role in forest ecosystem functions (Murcia, 1995, Cadenasso et al., 2003, Esseen et al., 2006). It serves as a filter, corridor and habitat for numerous plant and animal species and is related to forest structure and species composition (Williams-Linera, 1990, de Casenave et al., 1995).

Human impacts, such as expansion of settlements, agriculture and deforestation, have resulted in the fragmentation of forest ecosystems everywhere, which has in turn led to significant microclimatic and structural changes – in particular at forest edges (Keenan et al., 2015). Owing to an increased intensification of land use, forest edges have become more abrupt and the gradual transition zone has become either lost completely or poorly developed (Harper et al., 2005).

Ideally, the forest edge is a gradual transition zone, ranging from young trees of intermediate height to a shrub belt and low-growing plants (Keenan et al., 2015) and the forest edge area is characterized by high tree cover variance (Dantas de Paula et al., 2016). Thus, the ecological enhancement (e.g. different light regimes, different species etc. lead to more niches) of forest edges has enormous potential to preserve biodiversity in forest ecosystems. Moreover, the structure and extent of forest edges have a strong impact on the stability of the adjacent forest, by providing protection against strong winds. Therefore, a more distinct gradual transition zone in the forest edge area has a greater protective function (Yang et al., 2006, Dupont and Brunet, 2008).

Up to now, forest edge research has mainly been conducted to explore interactions of the forest edge with the forest or the landscape; according to Remy et al. (2017), it has received relatively little attention. Furthermore, existing studies have been based on time-consuming fieldwork. For example, Kleinn et al. (2010) estimated forest edge length based on forest inventory data. Rodrigues (1998) examined edge effects on the regeneration of forest fragments in southern Brazil. Aragon et al. (2015) analyzed the effect of edge type on forest edge habitats, and Reinmann and Hutyra, 2017, Smith et al., 2018 among others, examined forest edge effects to improve the assessment of carbon dynamics in temperate forests.

In comparison to field work, remote sensing technologies enable the required information to be acquired over large areas at a relatively low cost (Franklin et al., 2002, White et al., 2016). In the past, remote sensing techniques have mainly been used to estimate forest- or tree-specific parameters, although not explicitly at the forest edge, using the whole spectrum of sensors of two primary types – active and passive, and platforms, i.e., airborne, spaceborne or a combination of the two systems. The most popular parameters have been: forest area or extent (Waser et al., 2017a), tree crown (Zhen et al., 2016), above-ground biomass (Kumar and Mutanga, 2017, Price et al., 2017), tree species distribution (Fassnacht et al., 2017), forest structure (Latifi, 2012), and canopy gap (Rehush and Waser, 2017). Changes in the tree cover fraction in relation to distance from the forest edge were assessed by Dantas de Paula et al. (2016) in the context of the role of fragmented landscapes in retaining biodiversity.

Remote sensing has frequently been used to quantify and map the forest area or its change by assessing the tree line or forest border. Bouvet et al. (2018) used Sentinel-1 Synthetic Aperture Radar (SAR) data to detect borders of deforestation. Zhang et al. (2009) used Landsat imagery to explore tree line dynamics in order to assess increases or decreases in forest area, with a specific focus on alpine regions. Eysn et al. (2012) used Airborne Laser Scanning (ALS) to detect borders of mixed forest patches. However, in all these previously mentioned studies forest edge area as a whole was only included as a side topic. Consequently, the evolution of forest edge changes is not yet fully understood. Only recently, identifying and monitoring changes at the forest edge area and taking appropriate measures to preserve or restore this edge have become recognized as important issues for forest management (Spörri et al., 2014). Currently, monitoring the progress of forest edge management is mostly based on fieldwork (Buckley et al., 1997, Koch, 1997, Bolliger, 2009) rather than remote sensing techniques.

In recent years, ALS data in particular has emerged as one of the most promising remote sensing technologies to provide data for research and the management of forest ecosystems (Ackermann, 1999, Corona et al., 2012, Hyyppä et al., 2012, Bolton et al., 2018). As an active 3-D remote sensing technology, ALS penetrates the canopy and provides detailed information on vertical forest structure (Lefsky et al., 1999, Vastaranta et al., 2013, Stepper et al., 2017). Furthermore, ALS data is applicable in inaccessible areas of dense canopy and provides the required information, whereas traditional vegetation structure surveys in the field may lack precision and spatial extent.

In Switzerland, the ecological enhancement of forest edges has been recognized as an important measure in recent years, and it is now considered essential in the context of biodiversity, as well as forest and natural hazard management. Tools to monitor the effectiveness of ecological improvement are required by the national government and individual cantons (BAFU, 2008, Spörri et al., 2009, BVU, 2012, Imesch et al., 2015). With a total forest edge length of around 111,000 km in Switzerland (Abegg et al., 2014), the Federal Office of the Environment (FOEN) has realized the value of forest edges (BAFU, 2008, Rigling and Schaffer, 2015) based on the data of the third Swiss National Forest Inventory (NFI3) (Brändli, 2010) and has asked local authorities to increase their ecological value (Imesch et al., 2015). Up to now, research has been sparse and few local authorities have developed a concept to identify long-term forest edge areas with a high potential for ecological improvement. For example, the canton of St. Gallen developed a GIS model (Babbi, 2017) based on ecological and economic criteria in order to simplify the selection procedure for forest edges with potential for improvement. The structure of the forest edge and its composition of woody plants were also measured in the field according to the evaluation criteria developed by Krüsi. (2013). Although this measurement is appropriate to determine the ecological value of a forest edge, and also its changes, through repeated measurements, it is time consuming and thus limited for large areas.

Two questions arise in this context: first, how can we generally assess changes at forest edges for the entire study area between two given dates, i.e. the vertical structure at forest edge area has positively changed or not from an ecological point of view? And second, how can a progress in vertical variability at managed forest edges for large areas be assessed automatically? Vertical structure information from ALS point-cloud data might be useful to extract forest edge structure features cost effectively over large areas based on the evaluation criteria developed by Krüsi, e.g. forest edge depth and length and shrub belt depth and length. These parameters are measured manually in the field. Vertical structural information from ALS data acquired on two different dates can give insight into forest edge change and might serve as a way to assess the success of canton-wide forest edge management. However, to the best of our knowledge, up to now only Wehrli (2015) and Bühler et al. (2017) have carried out research related to forest edge areas based on ALS data.

In the present study we developed a novel approach to quantify the spatial and temporal structures of forest edge areas, in terms of a finite number of interpretable indicators based ALS data from two different dates. We applied Kernel Density Estimation (KDE), a non-parametric statistical method for estimating probability densities (Silverman, 1986), to explore the different patterns from two time steps. KDE is advantageous for estimating unknown distributions, and it accounts for irregular structures without requiring an understanding of the underlying processes (Bowman et al., 1997). KDE enables calculation of the density of any shape and is not influenced by grid size or location effects, and it has thus been widely used in research, e.g. for detecting and analyzing wild-land risk zones (Koutsias et al., 2014, Guo et al., 2015), housing and urban development (Fleming et al., 2017) and ecological networks (Li et al., 2016) based on the spatial distribution of landscape elements. In this study, KDE was used to determine the distribution patterns of the histograms that are required to assess dynamics at forest edges.

We developed an automatic approach to assess the structural change at forest edges over the last 13 years for the whole of the canton of Aargau, which has an area of 1403 km2 and a total length of forest boundary of approx. 5000 km. With our approach the following three questions are addressed: Is ALS point cloud data from two different dates suitable to quantify the change of vertical structure at the forest edge for a large area? Can differences in the structure of managed forest edges and all other forest edges be automatically detected? Is the proposed method helpful to evaluate the effectiveness of management activities?

Section snippets

Materials and methods

The following procedure was applied. First, a sample plot area is generated by assigning each sample point to a specified local study area. The corresponding ALS points from two different dates (ALS_time1 and ALS_time2) are then extracted. Second, for each sample plot area, a forest edge structure feature is calculated that represents the scale between the intermediate-height vegetation area and the entire woody area. The forest edge structure features are generated for both dates. Third,

KDE bandwidth selection

According to the workflow illustrated in Section 2, the forest structure feature w should always be calculated as a first step. Then, the normalized histogram of w is generated individually for each time step with the width 0.01. Therefore, we got two histograms, one generated from ALS_time1 data and the other from ALS_time2 data. Fig. 10 shows a histogram of the w values of all managed sample plot centers from ALS_time1 data, where the lines denote the density estimation functions with

General aspects of the proposed method

Assessing changes in structure at forest edges is considered an important task in that provides information relevant for forest edge management. This management aims to enhance the ecological value of the forest edge by increasing the structural diversity of the vegetation. In the present study, we introduced a novel approach to assessing structural changes at the forest edge between two different time steps based on ALS point-cloud data, the extracted forest structure feature w and the KDE.

Conclusions

The ecological enhancement of forest edges has recently become an important topic in the context of biodiversity and natural hazard management, and it is an essential part of sustainable forest management. Up to now, evaluation indicators have been used in the management field to describe the ecological quality and protection status of an examined forest edge, which is related to the vertical structure.

In this study, we developed a novel and highly automated approach to assess changes in the

Acknowledgments

This study was carried out in the framework of the Swiss National Forest Inventory (NFI), a cooperative effort between the Swiss Federal Institute for Forest, Snow and Landscape Research (WSL) and the Swiss Federal Office for the Environment (FOEN). We thank the department of forest of the canton Aargau for providing the field management information, Isabelle Livebardon for the field pictures and Melissa Dawes for professional language editing.

References (73)

  • P. Marchand et al.

    Spatial patterns of plant species richness along a forest edge: what are their determinants?

    For. Ecol. Manage.

    (2006)
  • C. Murcia

    Edge effects in fragmented forests: implications for conservation

    Trends Ecol. Evol.

    (1995)
  • E. Remy et al.

    Altered microbial communities and nitrogen availability in temperate forest edges

    Soil Biol. Biochem.

    (2018)
  • C. Wekesa et al.

    Interactions between native tree species and environmental variables along forest edge-interior gradient in fragmented forest patches of Taita Hills, Kenya

    For. Ecol. Manage.

    (2018)
  • B. Yang et al.

    Wind loading on trees across a forest edge: a large eddy simulation

    Agric. For. Meteorol.

    (2006)
  • Abegg, M., Brändli, U.-B., Cioldi, F., Fischer, C., Herold-Bonardi, A., M., H., Keller, M., Meile, R., Rösler, E.,...
  • G. Aragón et al.

    Edge type determines the intensity of forest edge effect on epiphytic communities

    Eur. J. For. Res.

    (2015)
  • Babbi, M., 2017. GIS-basierte Priorisierung der Waldränder im Kanton St.Gallen hinsichtlich ihres öko-logischen...
  • BAFU, 2008. Handbuch NFA im Umweltbereich. Mitteilung des BAFU als Vollzugsbehörde an Gesuchsteller. In: BAFU, B.f.U....
  • Bolliger, M., 2009. Waldrandpflege: Eine Chance für die Natur! Mitteilungsblatt BirdLife Aargau...
  • A. Bouvet et al.

    Use of the SAR shadowing effect for deforestation detection with sentinel-1 time series

    Remote Sens.

    (2018)
  • A.W. Bowman et al.

    Applied Smoothing Techniques for Data Analysis: The Kernel Approach with S-Plus Illustrations

    (1997)
  • Brändli, U.-B. (Eds.), 2010. Schweizerisches Landesforstinventar. Ergebnisse der dritten Erhebung 2004–2006....
  • C. Bühler et al.

    Pflege von Waldrändern –Erfolgskontrolle mit Laserscanning (LIDAR)

    Zürcher Wald

    (2017)
  • BVU, 2012. Naturschutzprogramm Wald. Zwischenbericht 2013, Ziele und Handlungsbedarf vierte Etappe (2014–2019). In:...
  • M.L. Cadenasso et al.

    A framework for a theory of ecological boundaries

    Bioscience

    (2003)
  • P. Corona et al.

    Airborne Laser Scanning to support forest resource management under alpine, temperate and Mediterranean environments in Italy

    Eur. J. Remote Sens.

    (2012)
  • S. Dupont et al.

    Impact of forest edge shape on tree stability: a large-eddy simulation study

    Forestry: Int. J. For. Res.

    (2008)
  • L. Eysn et al.

    Forest Delineation Based on Airborne LIDAR Data

    Remote Sens.

    (2012)
  • F.E. Fassnacht et al.

    Estimating stand density, biomass and tree species from very high resolution stereo-imagery – towards an all-in-one sensor for forestry applications?

    For.: Int. J. For. Res.

    (2017)
  • C.H. Fleming et al.

    A new kernel density estimator for accurate home-range and species-range area estimation

    Meth. Ecol. Evol.

    (2017)
  • S.E. Franklin et al.

    Change detection and landscape structure mapping using remote sensing

    For. Chronicle

    (2002)
  • F. Guo et al.

    Historic distribution and driving factors of human-caused fires in the Chinese boreal forest between 1972 and 2005

    J. Plant Ecol.

    (2015)
  • K.A. Harper et al.

    Edge influence on forest structure and composition in fragmented landscapes

    Conserv. Biol.

    (2005)
  • J. Hyyppä et al.

    Advances in forest inventory using airborne laser scanning

    Remote Sens.

    (2012)
  • Imesch, N., Stadler, B., Bolliger, M., Schneider, O., 2015. Biodiversität im Wald: Ziele und Massnahmen. Vollzugshilfe...
  • Cited by (0)

    View full text