remote sensing
Article
Use of Bi-Temporal ALS Point Clouds for Tree Removal
Detection on Private Property in Racibórz, Poland
2 ,
Patrycja Przewoźna 1,2, * , Paweł Hawryło 3 , Karolina Zi˛eba-Kulawik 3 , Adam Inglot 4 , Krzysztof Maczka
˛
2
3
and Piotr Matczak
Piotr W˛eżyk
1
2
3
4
*
Citation: Przewoźna, P.; Hawryło, P.;
Zi˛eba-Kulawik, K.; Inglot, A.;
Maczka,
˛
K.; W˛eżyk, P.; Matczak, P.
Use of Bi-Temporal ALS Point Clouds
for Tree Removal Detection on
Private Property in Racibórz, Poland.
Department of Geoinformation, Faculty of Geographical and Geological Sciences,
Adam Mickiewicz University, ul. Wieniawskiego 1, 61-712 Poznan, Poland
Department of the Study of Social Dynamics, Faculty of Sociology, Adam Mickiewicz University,
ul. Szamarzewskiego 89c, 60-568 Poznan, Poland; krzysztof.maczka@amu.edu.pl (K.M.);
matczak@amu.edu.pl (P.M.)
Department of Forest Resources Management, Faculty of Forestry, University of Agriculture in Krakow,
al. 29 Listopada 46, 31-425 Kraków, Poland; p.hawrylo@ur.krakow.pl (P.H.);
karolina.zieba@urk.edu.pl (K.Z.-K.); p.wezyk@ur.krakow.pl (P.W.)
Department of Geodesy, Faculty of Civil and Environmental Engineering, Gdansk University of Technology,
ul. Narutowicza 11/12, 80-233 Gdansk, Poland; adam.inglot@pg.edu.pl
Correspondence: pwysocka@amu.edu.pl
Abstract: Trees growing on private property have become an essential part of urban green policies.
In many places, restrictions are imposed on tree removal on private property. However, monitoring
compliance of these regulations appears difficult due to a lack of reference data and public administration capacity. We assessed the impact of the temporary suspension of mandatory permits on
tree removal, which was in force in 2017 in Poland, on the change in urban tree cover (UTC) in the
case of the municipality of Racibórz. The bi-temporal airborne laser scanning (ALS) point clouds
(2011 and 2017) and administrative records on tree removal permits were used for analyzing the
changes of UTC in the period of 2011–2017. The results show increased tree removal at a time when
the mandatory permit was suspended. Moreover, it appeared that most trees on private properties
were removed without obtaining permission when it was obligatory. The method based on LiDAR
we proposed allows for monitoring green areas, including private properties.
Remote Sens. 2021, 13, 767. https://
doi.org/10.3390/rs13040767
Keywords: urban tree cover; environmental management; tree removal permits; airborne laser
scanning; private property
Academic Editor: Ramón Alberto
Díaz-Varela
Received: 6 January 2021
Accepted: 13 February 2021
Published: 19 February 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affiliations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
1. Introduction
Urban forests give residents many benefits known as “ecosystem services” (ES) [1–4].
They have vital importance for sustainable development and inhabitants’ quality of life
and include all types of services: provisioning (e.g., wood and fruits), regulation and
maintenance of conditions (e.g., reducing the formation of urban heat islands and habitats),
resources for services (e.g., biomass production) and cultural importance (e.g., beauty) [4–7].
However, in recent decades, urban areas have been significantly affected by tree removal,
giving space to urban services such as buildings, roads and others. According to Nowak
and Greenfield [8], in the United States, a country with fast population growth in cities,
urban tree cover (UTC) fell from 40.4% to 39.4% in 2009–2014, while impervious cover
(mainly roads and buildings) increased by 1%. In Melbourne, Australia, UTC declined
in 2008–2016 by 0.7% [9], with the most significant losses being observed in public parks
and private properties (coverage reduction in these areas was 2.9% and 3.5%, respectively).
A similar situation is happening elsewhere in the world [10,11]. The increasing density
of buildings, including single-family housing, means that tree space decreases [10,12],
reducing the essential ES that trees provide, undermining life quality and contributing to
social inequalities.
Remote Sens. 2021, 13, 767. https://doi.org/10.3390/rs13040767
https://www.mdpi.com/journal/remotesensing
Remote Sens. 2021, 13, 767
2 of 21
1.1. Regulations Concerning Tree Removal on Private Property
Effective policies regarding the protection of trees and shrubs in cities are very much
needed, but the increasing urban development makes public lands insufficient in this
respect. As a result, urban green space management needs to take into account private
landowners. The inclusion of the private sector in environmental policies is not new. Such
attempts are observed in flood risk management [13] and nature conservation [14]. For trees,
private property often constitutes the majority of the city area grown by trees [12,15,16],
and private land yards and gardens comprise a substantial part of the urban tree canopy.
Trees on residents’ land are essential for supporting biodiversity [17,18] and the structural
connectivity of urban forest networks, which is crucial for many species [19,20].
However, incorporating private owners into tree management appears problematic.
In general, individuals’ contributions to the public good involve the collective action
problem [21]. Individuals are inclined to free-ride, e.g., refrain from contributing to the
public good while counting on contributions from others, assuming that they will be
available freely anyway, regardless of one’s input [22]. The collective action problem is
widely recognized in environmental management, as most environmental goods have the
common good’s essential characteristics. If the collective action problem is not resolved,
the underprovision of the public good can be expected. Besides the collective action
problem, there are several specific difficulties. Firstly, ecosystem disservices are functions
of ecosystems that are perceived as unfavorable for human well-being [23]. Speak [24] has
noticed significant differences between ES and ecosystem disservices provided by trees
on private and public land. It may explain why residents generally do not support the
rules treating a tree in a private area as a public good and restricting their freedom of tree
removal [25,26]. Secondly, a tree on a neighbor’s property can be a source of conflict due to
the shadow cast on the neighbor’s property, problems with falling leaves, or safety issues.
Pandit et al. [27] noticed that trees on nearby plots can significantly reduce the estate price.
The urban fabric also matters in this respect. According to Saphores and Li [28]), trees
are more valuable to people who live in denser areas. They would like new trees in their
neighborhood, but they are still unwilling to pay for the maintenance costs of new trees on
their properties. Moreover, urban trees are a natural habitat for many species, which might
also cause some nuisance [29], influencing the decision to remove a tree.
This situation necessitates the protection of trees on private property. Indeed, regulations are applied, aiming at securing the public good [30–32]. In the field of tree protection
and management, there are regulatory mechanisms established at the state and the municipality levels concerning tree removal on private property in many countries. The
regulations that control or even prohibit it on this type of property are one of the possible
solutions. For example, in most cities in Germany or Australia, a private owner has to
submit an appropriate, detailed application to the nature protection authority for permission to cut a tree in his/her own garden. The application needs to provide arguments to
remove the tree, detailed photos, a sketch of the property with the location of tree species,
the tree’s parameters (height and trunk circumference), and the numbers of all trees on the
property. The property owner should also deliver information on a planned replacement
of trees and if it is not possible, he must pay an appropriate fee for each tree removed. The
fees are spent to maintain and plant new trees in public places. Usually, though there are
exceptions, the owner receives approval from the office if there are appropriate arguments.
There are also large fines for illegal tree removals [33].
Nevertheless, in many cities, tree removal ordinances are not present, or regulations,
even if they include tree removal permit obligations, have numerous exceptions [34]. This
makes urban tree policies inefficient. Therefore, it is particularly important to monitor
spatio-temporal variation in UTC on public land and in private areas to determine the scale
of changes taking place and their causes. Tracking the number of removed trees is critical
for the management of green areas in urban space. For a proper assessment of the changes,
such factors as the current legal status or ecosystem disservices related to trees’ presence
(e.g., shading or maintenance problems) need also be considered. Although regulations on
Remote Sens. 2021, 13, 767
3 of 21
tree removal play a significant role, data that assess the effectiveness of these regulations
are scarce. In many cases, the scale of UTC changes is hard to estimate, as many cities
have little or no information on private green space [35]. Little research has focused on
tree removal on private properties [10,15,36], and most of the research focused on studying
the residents’ attitudes toward their own trees and the importance of legal regulations in
their protection [24–26]. Tools for supporting the management available in many cities
also focus on public urban forests—for example, the GreenSpaces (previously known as R3
TREES, R3GIS) application used by municipal administrations in 129 places in Europe [37].
Since it is very comprehensive, it allows the management of all information about public
greenery and helps schedule maintenance and care activities on all managed objects, but
trees on private properties are not included in this database.
In Poland, the municipal administration focuses only on urban greenery growing
on the municipality property. As a result, private areas are beyond any tree or shrub
monitoring database [38]. No private greenery inventories are made, which means that
there is no assessment of the actual impact of the existing policies on tree protection found
in these areas. At the same time, monitoring is essential when regulations exist with
numerous exceptions regarding tree removal apply, and this is the case in Poland. A
specific Polish enactment regulates this issue and the legal procedure of application for
cutting down trees has undergone development. Since April 16th, 2004, there have been
established regulations on tree removal permits that forced private owners to apply for
permission to cut trees on private land. The mandatory permission does not apply to fruit
trees and small trees less with trunk circumferences less than 25/35 cm (depending on
the species) and twelve other exceptions, mostly for security reasons [39]. Local checks
controlling the situation on private land occurred occasionally at that time, and information
from neighbors appeared to be the main control of compliance. The number of applications
for removal was steadily growing, which could be attributed to increasing awareness of
the obligation. However, the number of applications resulted in a considerable burden
for the administration. At the same time, the vast majority of applications were granted.
According to research conducted in provincial cities in 2011–2014, 96.4% of applicants were
granted permission to cut down their trees [39]. Eventually, a change in the Act on Nature
Conservation [40] of December 2016 abolished the requirement to obtain a permit to cut
down trees on private property. It went into force on 1 January 2017. It caused controversy,
as there were cases of large clearings and the media reported that there was tree removal
on a dramatic scale. On 1 March, the protection period started due to bird nesting. It
suspended the tree removal, and in June 2017, the mandatory permits were restored.
The regulation change that took place in Poland in 2017 constitutes an interesting
case from the urban green area management perspective. Assessing the impact of the
regulation’s change on the actual number of trees cut is difficult, as there are no effective
monitoring mechanisms in place. In Poland, the administrative process is monitored, e.g.,
monitoring replacement plantings or the number of permits for tree removal issued [39].
The outcome of the procedure, in terms of the number of removed trees, is not monitored
directly by the regulation. Moreover, only public urban greenery is subject to inventory,
and there is a growing need to have in place smart and integrated GIS systems that can
provide tree management.
1.2. Remote Sensing Data in UTC Change Analysis
This paper deals with the monitoring of tree removal on private properties in the
municipality of Racibórz (South of Poland) using the LiDAR (light detection and ranging)
3D point clouds approach and an administrative record to estimate the UTC loss caused
by the temporary suspension of mandatory permits to tree removal on private property
introduced in Poland in 2017. LiDAR is an active remote sensing technology that uses its
own energy source, i.e., a laser, to map objects. The laser energy (pulses in the near infra-red
range, e.g., 1064 nm) is sent from the transmitter which is the laser diode. The pulses are
scattered by rotating elements, e.g., mirrors, and they reach the object at the speed of light,
Remote Sens. 2021, 13, 767
4 of 21
are reflected from it, and some of the photons return to the detector as signal echoes [41].
Since each light pulse is numbered, the system uses the time it travels to determine the
path to and from the object. This way one can calculate the distance of the object from
the device, taking into account the angle of impulses sent (horizontal and vertical). In
the case of airborne laser scanning (ALS) technology, which was used in this research, a
laser altimeter is mounted on an aircraft and the precise location of the platform (plane or
helicopter) is determined thanks to an INS (inertial measurement system) which consists
of a GNSS (global navigation satellite system) and an inertial measurement unit (IMU).
The IMU records with high precision the platform inclination angles in three dimensions,
and directions and acceleration values. System integration allows one in the case of ALS to
achieve measurement accuracy of 20–30 cm horizontally (XY) and <10 cm vertically (Z)
depending on the scanning density and the technology used and geodetic transformations
(projections). Modern aerial scanners are able to generate several hundred thousand pulses
per second and provide a scanning density of 10–40 points/m2 (one strip, depending on the
flight altitude and speed). For trees, the ALS technology allows for very precise registration
of objects thanks to the penetration of the laser through vertical structure of vegetation.
The individual echoes (beam parts of light) reflect off the crown, trunks, and branches
and reach the bushes or the ground. Depending on the phenological period (leaves on
or off), the degree of tree crown damage, and the treatments performed, the horizontal
structure (range of the crown; areas of trees and shrubs 2D/3D) and the vertical structure
(tree height, crown length, crown base, percentiles, penetration, cover, etc.) can be precisely
described [42].
ALS point cloud analyses are a promising approach to monitor the protection of trees
and shrubs in private properties. However, its potential has not been not sufficiently
utilized till today. First of all, that is due to the relatively high costs of regularly repeated
ALS campaigns in a short period. Existing research in terms of using remote sensed data in
urban forestry is based mostly on aerial orthophotos or satellite (high resolution or very
high resolution) imagery and their derivatives, e.g., the normalized difference vegetation
index (NDVI) [43,44] or the Landsat-5 Thematic Mapper (TM) [45]. Another usually
applied solution is to use LiDAR data of single time points [46]. Both approaches give
ambiguous results. Landry and Pu [44] used IKONOS-2 scenes to verify the introduction
of tree permit regulations. Sung applied LiDAR data (2012) and Landsat TM images for
the same purpose (2013). The results of the studies suggest that the law can at least partly
influence the UTC. Nevertheless, the regulation change is not the only possible explanation.
Conway and Urbani’s analysis of satellite images (2007) revealed no significant differences
between residential areas in which the regulations on tree removal were and were not
implemented. Thus, the results obtained in previous studies are not conclusive in the
context of the effectiveness of legal rules. Moreover, all of the above research focused
on the comparison of tree cover between neighboring municipalities with and without
tree removal regulations, which does not enable taking into account the temporal lag.
Meanwhile, past planning decisions and behavior of citizens may significantly influence
the present UTC [47].
The second difficulty related to applied remote sensing methods is the spatial scale of
data. Very often, the availability of UTC data only allows comparison of changes across the
city that do not take into account district or parcel-level processes, such as tree removal
on private property [43–46]. More detailed spatial analyses relate to comparisons between
types of land cover or administrative units [48]. However, these may be misleading, as
the spatial variability in such cases is closely related to the size of the analyzed areas [12].
Therefore, an impact evaluation of policies should be based on both a detailed analysis
carried out at the parcel level and data from at least two-time points. Such a study
was performed in Christchurch (New Zealand) in terms of redevelopment’s influence on
UTC [36]. However, it concerned only selected plots of land, not the entire city, as only the
manual method of marking removed trees was used.
Remote Sens. 2021, 13, 767
5 of 21
Moreover, analyzing the effectiveness of tree management policies on private property
should not be limited to comparing UTC changes between private and public property.
If UTC is the only variable analyzed in the study, many non-legal factors may affect it,
such as population density, social stratification, luxury effect, and ecology of prestige [47].
Therefore, explicit efficiency of the implemented solutions requires analysis of UTC changes
and comparison with actual permits for tree removal. Such a comparison would allow
not only determining the changes in UTC but also checking the extent to which the
legal provisions are respected and whether they really influence changes regarding trees.
Effective regulations in this respect should assume a high consistency in the number of trees
removed on the basis of administrative data and resulting from the actual change in UTC.
Only then can it be concluded that the solutions introduced allow local authorities to have
real control over the number of trees cut down. That, in turn, requires the development of
an appropriate methodology to convert the surface of canopy cover into individual trees.
Such an analysis could provide more reliable information on the effectiveness of existing
regulations in tree protection in private areas.
This study responded to that challenge. It had two main goals: (1) to propose a new
method of analyzing changes in UTC using various data sources, and to apply it for (2)
evaluation of the regulations’ effectiveness concerning tree protection on private properties.
We tested a specific approach in which two ALS point cloud data sets (2011 and 2017) and
administrative decisions were compared to assess the effectiveness of legal regulations
concerning tree removal. The analysis was performed for the Polish municipality of
Racibórz. The study used bi-temporal ALS point clouds for assessment of the effectiveness
of the tree preservation legislation in force in Poland between 2011 and 2016. The work
describes how these data can be supplemented by administrative decision records that
authorize the trees’ removal to produce a more precise analysis of a policy output. The
results include the analysis of the impact of the regulation change, i.e., the suspension of
mandatory tree removal permits for private purposes that was introduced in Poland at the
beginning of 2017.
2. Materials and Methods
2.1. The City of Racibórz Case Study
The paper investigated the case study of Racibórz, a medium-sized city with approximately 55,000 inhabitants, located in Silesian Province in southern Poland, near the border
with the Czech Republic (Figure 1). The choice was constrained by the availability of
ALS data from the beginning of 2017, allowing for the analysis of the regulation change’s
impact. The data were available for the city of Racibórz, which did not differ from a typical
Polish town of this size regarding the types of green areas and their share in covering the
city. It is an administrative center, being the seat of Racibórz County. Its area is about
75 km2 . Agricultural land accounts for over 68.9% of the total area; forests and forest
land 3.7%; built-up land 22.9% [49]. The spatial structure of Racibórz is characteristic of
other medium-sized Polish cities. A significant portion is constituted by open areas located
mainly on the outskirts of the town on the left side of the Odra River. This is mostly arable
land and meadows. Urbanized areas with the highest density is concentrated in the central
part of the city, along the river. The vision of the city of “Racibórz 2020” takes the strategy
of openness to cooperation, for a vibrant, clean and green city [49].
Remote Sens. 2021, 13, 767
6 of 21
Figure 1. Localization of study area. Left top: border of Poland; left bottom: Silesian provincial borders; right: orthophotomap of Racibórz city.
2.2. Input Data
Two types of data were taken for the analysis: remote sensing data and information
from the administrative record. They are presented below.
2.2.1. Remote Sensing Data
Two types of remotely sensed data were used in the study. The primary data set
includes two acquisitions of ALS point clouds collected on 2011 (6 June 2011) and 2017
(5 March 2017). The ALS data set from 2011 was acquired within the framework of the
country-wide ISOK project dedicated to the country’s protection against extraordinary
threats [50]. The second acquisition in 2017 was performed as part of the CAPAP project [51].
ALS data were obtained by two standards: (a) standard I—applied mainly for forest and
rural areas outside cities, where point cloud density was at least 4 points per square meter
(ppsqm); (b) standard II—applied in 94 cities in Poland [50], with a density at least 12
ppsqm. The ALS point clouds were previously classified by data providers according to the
specifications of the American Society for Photogrammetry and Remote Sensing (ASPRS)
for LiDAR Exchange Format (LAS 1.2). The second type of remotely sensed data were
digital aerial orthophotomaps available in natural color composites (RGB) through the
Google Earth application (Table 1). The orthophotomaps were used as an additional data
source to make the UTC layer up-to-date at the beginning of the period of no tree removal
permit regulations for private purposes (Figure 2).
Remote Sens. 2021, 13, 767
7 of 21
Table 1. Specifications of airborne laser scanning (ALS) and aerial orthophotos used.
Data Type
Data Acquisition Date
Resolution
Flight Altitude
ALS point clouds
ALS point clouds
Aerial orthophoto RGB
(Google Earth)
11 June 2011
5 March 2017
≥12 ppsqm
≥12 ppsqm
∼1000 m AGL
∼1000 m AGL
1 September 2016
0.15 m
—
Figure 2. The timeline shows changes in tree removal regulations (above the horizontal line) and
remote sensing data (below the horizontal line). The red segment represents the time period when
the mandatory tree removal permission was suspended.
It should be emphasized that 5 March 2017 (when the ALS data set was available for
Racibórz) was the perfect time to analyze the impact of the change in legislation on tree
felling. The period from 1 March to 16 September is considered in Poland as the “bird
breeding period,” which introduces additional restrictions related to tree removal due to
protection of nesting birds. Thus, for many areas, the appropriate period during which the
removal of trees on private properties was beyond control was January–February 2017.
2.2.2. Administrative Data
Local authorities provided two types of spatial information to conduct the study:
cadastral data about land ownership and permits for tree removal. The first type of data,
obtained from the County Office in Racibórz, contained information on the type of the
real estate ownership assigned to registered plots according to 44 different sub-groups.
These data covered the studied area, and the accuracy of the border points of the land
plot does not exceed 0.3 m. For the purpose of this study, the plots were divided into
four categories based on the dominant sub-group registration codes of the registration
entities [52]. The first two (accounting for 54% of the city area) are related to private
owners located in (a) residential areas and (b) farms (in the outskirts of the city). These two
categories were related to the changes in the law that entered into force at the beginning of
2017. The remaining areas are mainly various types of non-private land, which constitute
the third distinguished category and 45% of the city’s area. The housing cooperatives (not
covered by the regulation change from the beginning of 2017) and the individual plots
for which the type of registration group was not specified, were classified in the fourth
group—others—and will not be discussed in detail in the results.
The data concerning the locations of trees with administrative permission for tree
removal were obtained from the Racibórz City Council for decisions issued between 2011
and 2017 (by the City Council and the County Office). The information contained 1763
records, each including the permit date, address, plot number, number of trees and species
name. The plot number was the basis for combining the removal data with the cadastral
data. Unfortunately, the system of collecting this data leaves much to be desired, and
therefore the correct allocation of individual decisions was not always possible. Therefore,
we assessed the policy effects on the city level. We assumed that the proportional shares of
Remote Sens. 2021, 13, 767
8 of 21
particular types of land ownership regarding tree removal, calculated for decisions whose
relevant locations were provided, represent the whole city’s situation.
2.3. Creating the Urban Tree Cover Loss Layers
In this section, the consecutive steps of the analysis are presented.
2.3.1. Generating Normalized Digital Surface Models and Canopy Height Models
In the first step of analyzing the remote sensing data, a Digital Terrain Model (DTM)
and a normalized Digital Surface Model (nDSM) were generated based on the ALS LiDAR point clouds. The models were created with a spatial resolution of 1.0 m using the
FUSION software [53]. The nDSMs were generated in two variants—in the first variant
(nDSM_build) using points classified (according to ASPRS) into classes (2) ground and (6)
buildings, and in the second variant (nDSM) using points classified into classes (2) ground,
(3) low vegetation, (4) medium vegetation, (5) high vegetation and (6) buildings. Height
models were generated separately for data obtained in 2011 and 2017. In the first step of
analyzing the ALS data, Digital Terrain Models (DTM) with 1.0 m spatial resolution were
created separately for both ALS datasets using the “GridSurfaceCreate” algorithm from the
FUSION software [53]. ALS point clouds were then normalized by subtracting the DTM
elevation from the corresponding ALS points. The normalized point clouds were then
used for creating normalized Digital Surface Models describing the elevation of buldings
(nDSM_buildings). For that purpose, only the points classified (according to ASPRS) as
classes (2) ground and (6) buildings were used during genearation of nDSM. In the next
step, the Canopy Height Models (011–201s) were generated using points classified as
classes (2) ground, (3) low vegetation, (4) medium vegetation and (5) high vegetation. The
nDSMs and CHMs were genrated using the “CanopyModel” algorithm from the FUSION
software [53]. Height models were generated separately for point clouds acquired in 2011
and 2017 airborne campaigns. Results of generating a DTM, an nDSM_buildings and a
CHM for the selected subset of the study area are presented in Figure 3.
Figure 3. Results of processing airborne laser scanning point clouds acquired in 2017 for the subset of the study area: Digital
Terrain Model (DTM); (A) normalized Digital Surface Model for buildings (nDSM_buildings); (B) Canopy Height Model
(CHM) (C).
2.3.2. Initial Classification of Areas Covered by Trees
The nDSMs and CHMs were imported into the eCognition Developer 9.3 software for
the preliminary classification of areas covered by high vegetation. It was performed using
the object-based classification method with the use of the rule-based approach. The objectbased classification method known under the acronym GEOBIA (Geographic Object-Based
Image Analysis) is an image analysis method that enables the integration of multi-source
data [54]. This method has many advantages over pixel classification, especially in the
Remote Sens. 2021, 13, 767
9 of 21
context of analyzing data with a high spatial resolution [55,56]. A similar approach was
used by Guo et al. [36] for creating an initial tree land cover class for further tree crown
delineation. In the first step, the buildings were classified using a nDSM_buildings layer.
Objects were considered to be buildings with an area of at least 4.0 m2 and a height of
at least 1.0 m. In the next step, based on CHMs generated from classes of points 2–5,
vegetation areas were classified, which could constitute potential tree crowns. It was
assumed that the crown of a single tree should have an area of at least 3.0 m2 , and the
minimum top height should be at least 3.0 m. The assumption about the minimum size
of the tree crown was adopted to eliminate minor errors in the classification of ALS point
clouds. These errors mainly consisted of an incorrect classification of power lines into the
high vegetation class. The adoption of the minimum height of the tree top at 3.0 m was also
aimed at eliminating from the analysis areas covered with shrubs, for which it is difficult
to precisely segment individual shrubs based on the nDSM derived from ALS. Finally, the
areas classified as potential trees were used in the subsequent analysis for individual tree
segmentation. The described analyses were performed separately for ALS from the years
2011 and 2017.
2.3.3. Segmentation of Individual Trees Based on Canopy Height Models
Using the previously created CHMs, single tree segmentation was performed. For
this purpose, the algorithm of inverse watershed implemented in eCognition Developer
software was used. It is an image segmentation algorithm using the local minima of the
indicated raster layer (in this case inverted from a CHM), which are treated as the so-called
seeds (treetops) from which segments expand into adjacent pixels with higher values [57].
The expansion of a given segment (tree crown) occurs until it comes into contact with the
border of the next segment growing from the adjacent seed (in this case, the top of the
tree). This algorithm works well in conditions of managed forest stands where the trees
growing next to each other have comparable sizes. In the case of urban green areas, small
and big trees often grow next to each other. Thus, it is difficult to properly parameterize
the segmentation algorithm to correctly separate all trees [58]. Increasing the minimum
area of segments representing tree crowns causes incorrect merging of neighboring crowns
of small trees. On the other hand, the use of too small a minimum size for segments results
in generating several segments within the crown of a single tree. In order to mitigate the
described issues, the segmentation algorithm was parameterized using data collected on 41
reference plots established in June 2019 in Racibórz city. In June 2019, fieldwork was carried
out within 41 reference areas of an average area of 0.004 km2 . The plots were established
in places where tree removal had not been carried out in recent years so that the number
of trees measured in the field corresponded to the number of trees identified on the CHM
from 2017.
The first step of segmentation of tree crowns was performed based on CHMs from 2017
using the same segmentation parameters for all trees. Then, the differences between the
number of trees obtained via the segmentation and the actual number of trees determined
on 41 reference plots were analyzed. Based on the analysis of errors, in the next step,
the segmentation was performed in two stages—using different parameters for trees
lower and higher than 16 m. The segmentation parameters were tuned in both groups to
minimize the systematic percentage error (percentage bias) calculated based on reference
data. Comparison of number of trees detected through segmentation of CHMs with
the data from 41 reference plots showed relatively high agreement (R2 = 0.86; mean
absolute percentage error = 21.1%; mean percentage error (percentage bias) = 0.3%). The
segmentation parameters tuned based on the field data and CHM from 2017 were also used
to perform the CHM segmentation from 2011. The results of tree segmentation based on
CHMs were saved as two vector layers in ESRI Shapefile format: UTC 2011 and UTC 2017.
Remote Sens. 2021, 13, 767
10 of 21
2.3.4. Generating UTC Loss Layers
Based on the two vector layers, e.g., UTC 2011 and UTC 2017, the UTC loss vector
layer was generated (UTCL 2011–2017). The GIS analysis was performed to compare
the UTC 2011 and UTC 2017 layers by calculating the overlapping areas between the
layers. The tree loss was identified for tree crowns segmented in 2011 if the tree crown area
decreased by 75% or more. In the subsequent analysis, the number of trees identified in
the UTCL 2011–2017 layer was compared with the number of trees removed according to
administrative decisions within the same period.
The separate UTC loss layer was created for the period from 1 January 2017 to 28
February 2017 when the permit for tree removal in Poland was not required (UTCL 2017).
The first step of the analysis was to identify the tree crowns segmented in 2011 for which
the crown area decreased by 40% or more between two ALS acquisitions (2011–2017).
The threshold of 40% was used in this case to make the analysis more robust and to
minimize the potential error of missing the trees removed within the mentioned period.
Such detected trees were then subject to visual verification when the operator analyzed
each tree on the computer monitor and decided whether this tree could be removed in the
period 1 January 2017–28 February 2017. The visual verification was based on the aerial
orthophotomap from 1 September 2016 available in the Google Earth application. If the
tree crown did not exist on the orthophotomap (2016), it was removed, since it means
that it disappeared before the period of interest. To obtain the most accurate estimation
of the number of trees removed within the targeted period (1 January 2017–28 February
2017), the information from administrative decisions authorizing tree removal was also
analyzed. Based on the information about the number of trees removed with permission
within the period 1 September 2016–31 December 2016 for an individual cadastral parcel,
the corresponding number of trees were removed from the UTCL 2017 layer within borders
of that parcel.
2.4. Comparing Administrative Decisions Regarding the Number of Removed Trees with
ALS-Derived UTC Loss Layers
To compare the data from administrative decisions with ALS-derived UTCL layers,
the polygons representing removed tree crowns (UTCL layers) were converted to points
by finding the centroid of each polygon. In the next step, the obtained points were
spatially assigned to the appropriate cadastral parcel. For each parcel, the decision date
and the number of trees removed according to administrative decisions were available. The
comparison of two data sets consisted of calculating the sum of removed trees according
to ALS-derived UTCL layers and the sum of trees from the removal permits for a single
registered parcel, taking into account the same period. That allowed for the estimation of
the percentage of trees that were removed without permission:
treeremoval LiDAR − treeremovaldecisions
· 100%
treeremoval LiDAR
(1)
The flowchart summarizing the steps of performed analysis is presented in Figure 4.
Remote Sens. 2021, 13, 767
11 of 21
Figure 4. Flowchart of the analysis performed.
3. Results
An analysis of UTC changes in Racibórz based on LiDAR data alone shows that
in the period of 2011–2017 a total of 113.5 ha of tree crowns disappeared, representing
1.5% of the city area, and 10.8% of the area of all trees growing in 2011. It means the loss
of almost 48,000 trees, according to estimates based on the used segmentation method.
More than half of them were removed from non-private areas. In the case of private
properties 34% disappeared from farms, and 13% from residential areas. Results of tree
crown segmentation for the selected reference plot is presented in Figure 5.
Figure 5. Results of tree crown segmentation obtained for the selected reference plot: (A) Canopy
Height Model (CHM) created from ALS data acquired in 2017; (B) results of tree crown segmentation
based on CHM 2017.
3.1. Number of Trees Removed from 2011 to 2016
In order to estimate for what proportion of removed trees the permissions were issued
we analyze the period of 2011–2016 when the permissions were mandatory. In accordance
with the adopted segmentation method, a total of 42,839 trees were cut in 2011–2016 while
permits issued at that time concerned 8473 trees. It means that 80% of trees disappeared at
that time without obtaining formal consent. According to the administrative records, for
Remote Sens. 2021, 13, 767
12 of 21
the whole city 7.2% of trees for which permits for removing were issued were in private
residential areas, 35.4% in farms, and 55.6% in non-private land (Figure 6). Thus it can
be estimated that in 2011–2016 a permit was issued for the removal of 593 trees grown in
private residential areas and 2966 growing on private farms.
In order to estimate for what proportion of removed trees the permissions were issued
we analyze the period of 2011–2016 when the permissions were mandatory. In accordance
with the adopted segmentation method, a total of 42,839 trees were cut in 2011–2016
while permits issued at that time concerned 8473 trees only. It means that 80% of trees
disappeared at that time without obtaining formal consent. According to the administrative
records, for the whole city 7.2% of trees for which permits for removing were issued were
in private residential areas, 35.4% in farms, and 55.6% in non-private land (Figure 6). Thus
it can be estimated that in 2011–2016 a permit was issued for the removal of 593 trees grown
in private residential areas and 2966 growing on private farms.
Figure 6. The shares of land ownership types for the total number of trees removed according to
administrative decisions from 2011 to 2016.
According to the analysis of ALS data, 44% of trees were cut on private properties. The
estimated number of trees that disappeared from the two categories of private areas was
4622 and 14,224 trees, respectively. Thus comparing the number of trees removed in the
period 2011–2016 according to ALS data and administration decisions, 87.2% of trees that
disappeared from private residential land were cut down without obtaining permission.
The same situation applies to 79.1% of trees owned by farmers.
The number of applications for permission to remove trees increased annually. We
cannot determine whether this increase was related to residents’ growing awareness
regarding applicable legal provisions or the increasing demand for tree removal. Assuming
the constant number of trees cut down annually by the inhabitants, the number of tree
removal applications in 2016 was lower than the actual number of trees removed by 7.6%
(Figure 7).
Remote Sens. 2021, 13, 767
13 of 21
Figure 7. Comparison of the number of tree removal permits issued annually with the average
number of trees removed annually estimated based on ALS data.
3.2. Number of Trees Removed in the Period of January–February 2017
Concerning the removal of trees at the beginning of 2017, when the regulations were
changed, it can be estimated that 5075 trees were cut at that time, 78% of which were in
both types of private areas: 34% in residential areas and 44% on private farms. It is a
substantial increase in the share of trees removed on these land ownership compared to
the previous period (Figure 8). In the period of 2011–2016, average 70 trees were removed
monthly on private residential areas and 216 on private farms. During the first two months
of 2017, when the mandatory permits for tree removal were lifted, the estimated numbers
of removed monthly trees in these areas were 874 and 1125, respectively while the rest
concerned non-private lands. Noteworthy, a significant contribution to this number was
made by large-area tree removal taking place on 14 large parcels located on the outskirts
of the city. Almost 2000 trees, which covered 22% to 35.8% of the plot’s area, disappeared
from those parcels. In other locations, mostly a few trees were cut down (average two per
property).
Figure 8. The shares of types of land ownership in the total tree removal in two time periods: 2011–2016 (tree permit
obligation) and the beginning of 2017 (the tree permit obligation on private properties lifted).
Remote Sens. 2021, 13, 767
14 of 21
4. Discussion
4.1. Influence of Tree Removal Policy on the UTC Change on Private Properties
This research shows that the Racibórz case follows the general trend observed around
the world in relation to decreasing UTC in cities [8–10,12], the area covered by trees shrank
by 10% during the studied period of 2011–2017. It resulted in the loss of 48,000 trees, among
which almost 47% disappeared on private properties. Thus, the regulations aiming at the
protection of trees in these area, appeared unsatisfactory.
The change of the regulation in 2017 that allowed tree removal without permits
increased tree removal. It is in line of the previous observations [44–46]. However, it cannot
be determined whether this change would be long-lasting or temporary, as Conway and
Urbani have already pointed out [43]. In this context, it is essential to analyze not only the
UTC change but also the reasons and aggregated dynamics of the removal of trees. The
increased removal in the time of releasing the restrictions could stem from a need to get
rid of individual trees that were problematic for the owners. After a while, the situation
would stabilize. Just as opportunity makes the thief, a change in the law could have only
accelerated the decision, which would happen in the long run anyway.
There are many reasons why people choose to fell trees [59]. In the case of Racibórz it
appeared that in the period when the permits were lifted, it regarded mainly large-scale
tree removal in a few plots. Considering the whole period 2011–2017 (Figure 9A), the most
significant UTC loss is related to the regulation of the banks of the Odra River and related
floodplains. The decision to remove trees in such areas belonged to the relevant water
management bodies and did not require obtaining a felling permit. The large share of
tree removal also concerns the city center, while when considering only the trees removed
in early 2017 (Figure 9B), it is barely noticeable. Instead, the share of housing estates
located on the outskirts of the city becomes clear. Removal in 14 plots (one of them is
shown on the Figure 9D) constituted 65% of trees cut at the beginning of 2017. That
allows us to suppose that the main reasons inhabitants cut down trees were not security
or lack of knowledge but rather redevelopment activities. A similar situation took place
in Christchurch, New Zealand [11]. However, it is difficult to interpret to what extent the
cutting resulted from private landowners’ needs and to what extent, in the long run, it was
aimed at commercial, developmental purposes. In the latter case, it would mean abuse of
the introduced legal changes. A more straightforward assessment of this situation would
require further observation in subsequent years, as the amendment to the 2016 Act [40]
prohibits the sale of a plot in the next five years after trees were removed. In any case, if
reconstruction or development activities were the reason for large-scale tree removal, it
would be a clear signal to national and local law-making institutions that this issue requires
attention. The regulations present in Poland till the end of 2016 were criticized as imposing
the unnecessary burden for both landowners and the administration as the same procedure
applied for a single tree and a large number of trees removal. However, liquidating the
regulation involved additional tree removal. Thus, there is a need for urban tree protection
policies sensitive to the types of land in terms of number of trees.
The study underlines the importance of regulations compliance. As the changes to
the law introduced at the beginning of 2017 were short-lived, assessing the effectiveness
of previously binding legal regulations is critical. Meanwhile, the results from 2011–2016
indicate that a relatively large number of trees were cut down without permission in 2011–
2016. Two explanations can be posed in this context. First, the tree preservation regulations
on private properties were ineffective due to a lack of proper monitoring systems. The
second explanation is that the majority of trees in private areas did not require permission
for their removal (e.g., being fruit trees or small trees in diameter). Thus, the exceptions
in existing regulation might have influenced UTC significantly. The letter explanation
suggests that tree protection in private areas need to be even more restricted to protect
against decreasing UTC. That supports empirically the results of Lavy and Hegelman’s [34]
analysis on law in Texas. Both reasons confirm the importance of regular monitoring of
tree protection policies’ effects, and remote sensed data has a vital role in this subject.
Remote Sens. 2021, 13, 767
15 of 21
Figure 9. The intensity of UTC loss visualized in the form of heat maps created for two time periods: 2011–2017 (A) and the
beginning of 2017 (B), compared with the location of trees for which a felling permit was issued in 2011–2017 (C); and the
zoom-in showing exemplary large-scale tree removal (D) also indicated with Google Earth images from 1 September 2016
(E) and 30 July 2017 (F), not indicated on the Racibórz map to preserve the anonymity of the landowners; Basemap: the land
cover based on open source data available on [51].
Remote Sens. 2021, 13, 767
16 of 21
ALS point clouds allow not only to analyze the intensity of tree removal in selected
time periods but also enable checking the effectiveness of restrictions in this regard, if they
exist. In conjunction with a reliable spatial database on the granted permits for felling
(Figure 9C), they enable one to indicate the locations for which such a decision was not
issued and whether the situation was legal. Local authorities can also analyze the affected
land cover types and draw conclusions for future protection strategies. It is especially
important when regulations assume certain exceptions. In the case study discussed in this
paper, thanks to the analysis of bi-temporal ALS point clouds, it was possible to identify
specific locations that may raise doubts about whether the tree removal without permission
was following the regulations that entered into force in Poland in 2017 (i.e., Figure 9D). It is
also important that at this stage, future monitoring can be realized even with the use of
satellite images only (Figure 9E,F). Therefore, the knowledge that we gain by using remote
sensed data (ASL and aerial orthophotos) appears crucial for assessing environmental
policies and the management of urban greenery on private properties.
However, residents’ support for the principles of tree protection involves acceptance
of treating their properties as a public good [25,26]. It is not always the case. Residents’
behavior concerning greenery in private areas may be influenced by their neighbors [47].
It is related to the willingness to maintain nearby households’ prestige and expresses
belonging to a specific lifestyle group. Exploring these issues requires both detailed
spatial data, in terms of issued administrative decisions, and data on residents’ knowledge,
awareness, and attitudes. This would allow for the development of a more elaborate
analysis informative for effective regulations that could propose not only the penalty
system but also the incentives.
4.2. Combing Information of UTC Change with Permits Issued for Tree Removal
We claim that monitoring of tree cover is crucial in green areas management. Nevertheless, ALS data from various time points are so far rarely used to analyze factors affecting
the number of trees cut down in private areas. Usually, such analyzes are subject to many
limitations, such as data availability at selected time points or their accuracy or spatial
extent. The methodology presented in the article allows supplementing incomplete ALS
data with administrative information regarding the permits issued to remove trees, thus
limiting the interpretation uncertainty of results. The observation presented above would
not be possible if the change in UTC based on ALS data and the administrative data were
analyzed separately.
A reliable assessment requires the ability to compare administrative data with an
analysis of actual changes in UTC. Our work presents a workable methodological approach
in this regard. However, tree crown segmentation in the city environment is a challenging
task. To avoid big over- or underestimating the real number of trees, we parameterized the
segmentation algorithm based on the reference field plots. The comparison of the number
of trees derived from segmentation and the actual number of trees showed relatively high
agreement and low systematic error. However, it has to be kept in mind that the applied
segmentation algorithm was not validated based on “unseen data” [60] because all of
the 41 field plots were used for parameterization of the algorithm. The analysis of UTC
loss for the beginning of the year 2017 can be treated as more accurate. In that case, the
results of automatic segmentation were also checked through manual photo interpretation.
We decided to apply this approach since the manual delineation of all trees, as done for
example by Guo et al. [36], although it is time-consuming and not error-free. Thus, it is not
feasible for large-scale cases. One of the aims of our study was to combine ALS data with
administrative decisions in the context of tree removal monitoring rather than providing a
ready to use tree segmentation approach for urban areas. In the case of using ALS data for
UTC monitoring, it is usually more common to base on the area of tree crowns instead of
the number of trees [9]. A tree crown area can be easier derived from ALS point clouds and
provide a better proxy to calculate ES [61]. However, for monitoring a tree protection policy,
tree segmentation was necessary. It enabled the comparison of permits’ administrative
Remote Sens. 2021, 13, 767
17 of 21
records, where information about tree crown area was not available with the independent
data set to assess different law regulations’ effectiveness. Future studies should put an
effort in development of robust tree segmentation methods, specifically dedicated for urban
areas. Recent advancements in methods of data analysis, such as deep learning approaches
dedicated for analysis of 3D point clouds [62], together with possibilities of acquiring very
high density point clouds [63], might provide big contribution to research and practical
applications of LiDAR technology in context of UTC monitoring.
4.3. Limitations of the Study
Our study has several limitations. The first one is the quality and availability of administrative data mentioned in the chapter concerning input data, especially those relating
to removal permits. The accuracy of the estimation depends on the quality efficiency of
tree location data collected by offices. Cadastral data having accurate georeferencing of
every single property is also important. The database we used was often inconsistent and
incomplete. There were missing or erroneous data in the records resulting in the impossibility of correct allocation of individual decision detection. This problem concerned 11% of
the analyzed decisions. That is why a detailed analysis indicating land plots that had not
obtained consent for tree removal was not possible. Records often lacked the exact locations
of tree removal, or the addresses provided do not exist. In many cases, it was impossible to
geolocate removed trees and make a more elaborated spatial analysis. The cadastral data
were also incomplete, although to a much lesser extent. The lack of information about the
type of ownership concerned only 1% of the plots. Thus, the methodology we propose
does not allow detailed indications of plots of land where abuses occurred.
The approach we propose would also provide more opportunities if ALS data were
collected more frequently, which is still a long way off, at least in Poland. It would help
analyze UTC changes in the city’s specific areas in more detail and investigate their causes.
Additionally, it would enable us to compare UTC changes in subsequent years with the
administrative decisions issued. Without this information, we cannot determine yearly
dynamics of the number of trees being cut down between 2011–2016. Nevertheless, the
proposed method makes it possible to assess the effectiveness of introduced regulations in
general. Of course, frequently collecting ALS data requires additional financial resources,
so it is worth considering in the future also comparative analysis of accuracy and costs
using alternative methods allowing for detailed monitoring of UTC. Automatic urban
tree coverage mapping with LiDAR data can be also performed, e.g., using artificial
intelligence models, to ensure efficiency and quality. Timilsina et al. [64] demonstrated
that a combination of Convolution Neural Network (CNN) with OBIA based on very highresolution (VHR) aerial ortophotomap and LiDAR data results in 93.2% overall accuracy
(OA) of classification for mapping urban trees within cadastral parcels. Object-Based CNN
(OB-CNN) was also used to mapping UTC changes using Google Earth imageries and
LiDAR datasets with 96% and 98% OA [65]. Deeper CNN methods such as Region-based
CNN (R-CNN) with manually generated training samples were used by Wang et al. [58]
with 86% OA in identifying flowering trees. One of the most promising techniques is
generating image-derived point clouds from images acquired from aircraft or Remotely
Piloted Aircraft Systems (RPAS), commonly known as Unmanned Aerial Vehicles (UAVs) or
just as drones (Dandois i Ellis, 2013). One should also not forget about the new possibilities
of the Earth Observation (EO) satellites which can provide multispectral images (0.15 to 3
m ground sampling distance; GSD), even daily. The plans of the new constellations show
that the acquisition of images will be possible even several dozen times a day (SkySat
Planet Labs., MAXAR), which will give decision-makers strong geoinformatics tools to
monitor the state of greenery in the city.
5. Conclusions
The study presented a new adapted method of using LiDAR point clouds for UTC
change analyses. Its application may be of particular importance when the ALS datasets
Remote Sens. 2021, 13, 767
18 of 21
are insufficient. The obtained results point out a few insights essential for the effective
assessment of tree protection regulations on private property.
First of all, it shows that the necessity to obtain consent for tree removal on private
property does not guarantee local authorities’ full control in this regard. In the case of
Racibórz, it turned out that in 2011–2016 most of the trees were cut down without obtaining
permission. No matter the actual reason behind that issue, the uncontrolled loss of almost
19,000 trees in practice shows such regulations’ ineffectiveness.
Secondly, lifting the obligatory permission for tree removal on private property caused
even more intense UTC loss in this type of land ownership. This indicates that restrictions
concerning the owners of trees growing on private land are not without significance in
this respect. Nevertheless, to ensure satisfactory policy outcomes’ results, they must be
introduced together with properly planned monitoring of UTC changes in these areas.
Thirdly, the presented results highlight the role of ALS point clouds in this scope.
Their wider use by the public administration is essential to create environmental policies
not only based on theoretical assumptions but also real data. These, if collected regularly,
can provide new insights into the behavior of residents in the field of tree removal. In
a situation where there are gaps in geodata, in periods when LiDAR point clouds are
not obtained due to high costs, it is necessary to use aerial photo matching techniques
or unmanned aircraft systems (generating even denser point clouds but with limited
penetration through vegetation). Taking them into account may help construct better legal
solutions and incentives to plant and keep urban trees and shrubs in the entire city also on
private properties.
Fourthly, the presented results show the very necessity to regularly monitor the
functioning of implemented policies based on spatial data. The state’s role should be to
facilitate setting appropriate standards in this regard than on creating specific regulations
concerning removing trees on private properties on the local level. We agree with the
recent claim in a study of the effective implementation of land-use planning regulations
in Australia that evaluation guidelines are crucial to achieving the intended results [66].
Otherwise, top-down goals are often not included in local documents, not to mention
monitoring their implementation. Similar conclusions appeared in the review of legal
regulations regarding urban sustainability plans in selected US cities [35]. It showed that in
many cases, UTC growth goals do not take into account the monitoring plan. At the same
time, only an adequately designed valuation of existing regulation enables one to create
practical solutions in this respect.
Author Contributions: Conceptualization, P.P., P.H., K.Z.-K., A.I., K.M., P.W., P.M.; methodology,
P.P., P.H., K.Z.-K., A.I., P.W.; validation, P.H., A.I.; formal analysis, P.P., P.H.; investigation, P.P.,
P.H., K.Z.-K., A.I.; resources, P.P., P.H., K.Z.-K., P.W., A.I., K.M.; data curation, P.P., P.H., A.I.; P.W.;
writing—original draft preparation, P.P., P.H., K.Z.-K., A.I., K.M., P.W., P.M.; writing—review and
editing, P.P., P.H., K.Z.-K., A.I., K.M., P.W., P.M.; visualization, K.Z.-K., P.P.; P.H.; supervision, P.M,
P.W.; project administration, P.M., K.M; funding acquisition, P.P., K.M., P.W., P.M. All authors have
read and agreed to the published version of the manuscript.
Funding: This research was funded by the National Science Centre, Poland, under grant number
2017/25/B/HS6/00954; and the Foundation for Polish Science (FNP) (scholarship START edition
2020).
Acknowledgments: We would like to thank the Central Office of Geodesy and Cartography in
Poland (GUGiK) for providing open LiDAR data and aerial orthophotography data and the City Hall
in Racibórz for providing the database of permission for felling trees. Research assistance by Marcin
Mielewczyk is gratefully acknowledged.
Conflicts of Interest: The authors declare no conflict of interest.
Remote Sens. 2021, 13, 767
19 of 21
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
Costanza, R.; d’Arge, R.; de Groot, R.; Farber, S.; Grasso, M.; Hannon, B.; Limburg, K.; Naeem, S.; O’Neill, R.V.; Paruelo, J.; et al.
The value of the world’s ecosystem services and natural capital. Nature 1997, 387, 253–260. [CrossRef]
Jennings, V.; Larson, L.; Yun, J. Advancing Sustainability through Urban Green Space: Cultural Ecosystem Services, Equity, and
Social Determinants of Health. Int. J. Environ. Res. Public Health 2016, 13, 196. [CrossRef]
Biernacka, M.; Kronenberg, J. Urban green space availability, accessibility and attractiveness, and the delivery of ecosystem
services. Cities Env. 2019, 12, 5.
Reid, W.; Mooney, H.; Cropper, A.; Capistrano, D.; Carpenter, S.; Chopra, K. Millennium Ecosystem Assessment. Ecosystems and
Human Well-Being: Synthesis. 2005. Available online: www.islandpress.org/bookstore/details.php?prod_id=1119 (accessed on
16 February 2021).
Haines-Young, R.; Potschin, M. Common International Classification of Ecosystem Services (CICES, Version 4.1); European Environment
Agency: Copenhagen, Denmark, 2012.
Takács, Á.; Kiss, M.; Hof, A.; Tanács, E.; Gulyás, Á.; Kántor, N. Microclimate Modification by Urban Shade Trees—An Integrated
Approach to Aid Ecosystem Service Based Decision-making. Procedia Environ. Sci. 2016, 32, 97–109. [CrossRef]
Hasselström, L.; Visch, W.; Gröndahl, F.; Nylund, G.M.; Pavia, H. The impact of seaweed cultivation on ecosystem services—A
case study from the west coast of Sweden. Mar. Pollut. Bull. 2018, 133, 53–64. [CrossRef] [PubMed]
Nowak, D.J.; Greenfield, E.J. Declining urban and community tree cover in the United States. Urban For. Urban Green. 2018,
32, 32–55. [CrossRef]
Croeser, T.; Ordóñez, C.; Threlfall, C.; Kendal, D.; van der Ree, R.; Callow, D.; Livesley, S.J. Patterns of tree removal and canopy
change on public and private land in the City of Melbourne. Sustain. Cities Soc. 2020, 56, 102096. [CrossRef]
Daniel, C.; Morrison, T.H.; Phinn, S. The governance of private residential land in cities and spatial effects on tree cover. Environ.
Sci. Policy 2016, 62, 79–89. [CrossRef]
Guo, T.; Morgenroth, J.; Conway, T.; Xu, C. City-wide canopy cover decline due to residential property redevelopment in
Christchurch, New Zealand. Sci. Total Environ. 2019, 681, 202–210. [CrossRef]
Mincey, S.K.; Schmitt-Harsh, M.; Thurau, R. Zoning, land use, and urban tree canopy cover: The importance of scale. Urban For.
Urban Green. 2013, 12, 191–199. [CrossRef]
Hartmann, T.; Slavíková, L.; McCarthy, S. (Eds.) Nature-Based Flood Risk Management on Private Land; Springer International
Publishing: Cham, Switzerland, 2019. [CrossRef]
Kamal, S.; Grodzińska-Jurczak, M.; Brown, G. Conservation on private land: A review of global strategies with a proposed
classification system. J. Environ. Plan. Manag. 2014, 58, 576–597. [CrossRef]
Kaspar, J.; Kendal, D.; Sore, R.; Livesley, S. Random point sampling to detect gain and loss in tree canopy cover in response to
urban densification. Urban For. Urban Green. 2017, 24, 26–34. [CrossRef]
Heynen, N.; Perkins, H.A.; Roy, P. The Political Ecology of Uneven Urban Green Space. Urban Affairs Rev. 2006, 42, 3–25.
[CrossRef]
Goddard, M.A.; Dougill, A.J.; Benton, T.G. Scaling up from gardens: Biodiversity conservation in urban environments. Trends
Ecol. Evol. 2010, 25, 90–98. [CrossRef] [PubMed]
Braaker, S.; Moretti, M.; Boesch, R.; Ghazoul, J.; Obrist, M.K.; Bontadina, F. Assessing habitat connectivity for ground-dwelling
animals in an urban environment. Ecol. Appl. 2014, 24, 1583–1595. [CrossRef]
Hale, J.D.; Fairbrass, A.J.; Matthews, T.J.; Sadler, J.P. Habitat Composition and Connectivity Predicts Bat Presence and Activity at
Foraging Sites in a Large UK Conurbation. PLoS ONE 2012, 7, e33300. [CrossRef]
Ossola, A.; Locke, D.; Lin, B.; Minor, E. Yards increase forest connectivity in urban landscapes. Landsc. Ecol. 2019, 34, 2935–2948.
[CrossRef]
Mincey, S.K.; Vogt, J.M. Watering strategy, collective action, and neighborhood-planted trees: A case study of Indianapolis,
Indiana, US. Arboric. Urban For. 2014, 40, 84–95.
Marschall, M.J. Citizen Participation and the Neighborhood Context: A New Look at the Coproduction of Local Public Goods.
Political Res. Q. 2004, 57, 231–244. [CrossRef]
Lyytimäki, J.; Sipilä, M. Hopping on one leg—The challenge of ecosystem disservices for urban green management. Urban For.
Urban Green. 2009, 8, 309–315. [CrossRef]
Speak, A.; Escobedo, F.J.; Russo, A.; Zerbe, S. An ecosystem service-disservice ratio: Using composite indicators to assess the net
benefits of urban trees. Ecol. Indic. 2018, 95, 544–553. [CrossRef]
Kirkpatrick, J.; Davison, A.; Daniels, G. Resident attitudes towards trees influence the planting and removal of different types of
trees in eastern Australian cities. Landsc. Urban Plan. 2012, 107, 147–158. [CrossRef]
Conway, T.M.; Bang, E. Willing partners? Residential support for municipal urban forestry policies. Urban For. Urban Green. 2014,
13, 234–243. [CrossRef]
Pandit, R.; Polyakov, M.; Sadler, R. Valuing public and private urban tree canopy cover. Aust. J. Agric. Resour. Econ. 2013,
58, 453–470. [CrossRef]
Saphores, J.D.; Li, W. Estimating the value of urban green areas: A hedonic pricing analysis of the single family housing market
in Los Angeles, CA. Landsc. Urban Plan. 2012, 104, 373–387. [CrossRef]
Remote Sens. 2021, 13, 767
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
20 of 21
Adams, L.W.; Van Druff, L.W.; Luniak, M. Managing urban habitats and wildlife. In Techniques for Wildlife Investigations and
Management, 6th ed.; The Wildlife Society: Bethesda, MD, USA, 2005; pp. 714–739.
Narducci, J.; Quintas-Soriano, C.; Castro, A.; Som-Castellano, R.; Brandt, J.S. Implications of urban growth and farmland loss for
ecosystem services in the western United States. Land Use Policy 2019, 86, 1–11. [CrossRef]
Rimal, B.; Sharma, R.; Kunwar, R.; Keshtkar, H.; Stork, N.E.; Rijal, S.; Rahman, S.A.; Baral, H. Effects of land use and land cover
change on ecosystem services in the Koshi River Basin, Eastern Nepal. Ecosyst. Serv. 2019, 38, 100963. [CrossRef]
Wang, W.; Wu, T.; Li, Y.; Xie, S.; Han, B.; Zheng, H.; Ouyang, Z. Urbanization Impacts on Natural Habitat and Ecosystem Services
in the Guangdong-Hong Kong-Macao “Megacity”. Sustainability 2020, 12, 6675. [CrossRef]
Nasemann, A. Fällen Verboten. Available online: https://www.sueddeutsche.de/geld/baeume-faellen-verboten-1.3585349
(accessed on 30 September 2020).
Lavy, B.L.; Hagelman, R.R. Protecting the urban forest: Variations in standards and sustainability dimensions of municipal tree
preservation ordinances. Urban For. Urban Green. 2019, 44, 126394. [CrossRef]
Benton-Short, L.; Keeley, M.; Rowland, J. Green infrastructure, green space, and sustainable urbanism: Geography’s important
role. Urban Geogr. 2017, 40, 330–351. [CrossRef]
Guo, T.; Morgenroth, J.; Conway, T. Redeveloping the urban forest: The effect of redevelopment and property-scale variables on
tree removal and retention. Urban For. Urban Green. 2018, 35, 192–201. [CrossRef]
Green Spaces. Available online: https://www.r3-trees.com/ (accessed on 30 September 2020).
Feltynowski, M.; Kronenberg, J.; Bergier, T.; Kabisch, N.; Łaszkiewicz, E.; Strohbach, M.W. Challenges of urban green space
management in the face of using inadequate data. Urban For. Urban Green. 2018, 31, 56–66. [CrossRef]
Krynicki, M.; Witkoś Gnach, K. Monitoring standardów w zarzadzaniu
˛
zielenia˛ wysoka˛ w najwi˛ekszych miastach Polski.
Technical Report; 2016. Available online: http://drzewa.org.pl/publikacja/1075-2/ (accessed on 30 September 2020).
The Act of December 16, 2016 Amending the Act on Nature Protection and the Act on Forests. 2016. Available online:
http://isap.sejm.gov.pl/isap.nsf/DocDetails.xsp?id=WDU20160002249 (accessed on 30 September 2020).
Wagner, W.; Ullrich, A.; Melzer, T.; Briese, C.; Kraus, K. From Single-Pulse to Full-Waveform Airborne Laser Scanners: Potential
and Practical Challenges. 2004. Available online: https://www.isprs.org/proceedings/XXXV/congress/comm3/papers/267.pdf
(accessed on 16 February 2021).
Wezyk, P. The integration of the terrestrial and airborne laser scanning technologies in the semi-automated process of retrieving
selected trees and forest stand parameters Integração das tecnologias terrestre e aerotransportada de scanner laser no processo
semi. Ambiencia 2012, 8, 533–548.
Conway, T.M.; Urbani, L. Variations in municipal urban forestry policies: A case study of Toronto, Canada. Urban For. Urban
Green. 2007, 6, 181–192. [CrossRef]
Landry, S.; Pu, R. The impact of land development regulation on residential tree cover: An empirical evaluation using highresolution IKONOS imagery. Landsc. Urban Plan. 2010, 94, 94–104. [CrossRef]
Sung, C.Y. Mitigating surface urban heat island by a tree protection policy: A case study of The Woodland, Texas, USA. Urban
For. Urban Green. 2013, 12, 474–480. [CrossRef]
Sung, C.Y. Evaluating the efficacy of a local tree protection policy using LiDAR remote sensing data. Landsc. Urban Plan. 2012,
104, 19–25. [CrossRef]
Grove, J.M.; Locke, D.H.; O’Neil-Dunne, J.P.M. An Ecology of Prestige in New York City: Examining the Relationships Among
Population Density, Socio-economic Status, Group Identity, and Residential Canopy Cover. Environ. Manag. 2014, 54, 402–419.
[CrossRef] [PubMed]
Lee, S.J.; Longcore, T.; Rich, C.; Wilson, J.P. Increased home size and hardscape decreases urban forest cover in Los Angeles
County’s single-family residential neighborhoods. Urban For. Urban Green. 2017, 24, 222–235. [CrossRef]
Biuletyn Informacji Publicznej. Available online: https://www.bipraciborz.pl (accessed on 30 September 2020).
IT System of the Country Protection. Available online: https://isok.gov.pl/index.html (accessed on 30 September 2020).
Head Office of Geodesy and Cartography. Available online: http://www.gugik.gov.pl/ (accessed on 30 September 2020).
Ordinance of the Minister of Regional Development and Construction of March 29, 2001 on Land and Building Records. Available
online http://isap.sejm.gov.pl/isap.nsf/DocDetails.xsp?id=wdu20010380454 (accessed on 5 January 2021).
McGaughey, R.J. FUSION/LDV: Software for LIDAR Data Analysis and Visualization; US Department of Agriculture, Forest Service,
Pacific Northwest Research Station: Seattle, WA, USA, 2009; p. 123.
Castilla, G.; Hay, G. Image objects and geographic objects. In Object-Based Image Analysis; Springer: Berlin/Heidelberg, Germany,
2008; pp. 91–110.
Shojanoori, R.; Shafri, H.Z. Review on the use of remote sensing for urban forest monitoring. Arboric. Urban For. 2016, 42, 400–417.
Blaschke, T. Object based image analysis for remote sensing. ISPRS J. Photogramm. Remote Sens. 2010, 65, 2–16. [CrossRef]
Vincent, L.; Soille, P. Watersheds in digital spaces: An efficient algorithm based on immersion simulations. IEEE Trans. Pattern
Anal. Mach. Intell. 1991, 13, 583–598. [CrossRef]
Wang, K.; Wang, T.; Liu, X. A Review: Individual Tree Species Classification Using Integrated Airborne LiDAR and Optical
Imagery with a Focus on the Urban Environment. Forests 2018, 10, 1. [CrossRef]
Conway, T.M. Tending their urban forest: Residents’ motivations for tree planting and removal. Urban For. Urban Green. 2016,
17, 23–32. [CrossRef]
Remote Sens. 2021, 13, 767
60.
61.
62.
63.
64.
65.
66.
21 of 21
Redyuk, S.; Schelter, S.; Rukat, T.; Markl, V.; Biessmann, F. Learning to Validate the Predictions of Black Box Machine Learning
Models on Unseen Data. In Proceedings of the Workshop on Human-In-the-Loop Data Analytics; ACM Press: New York, NY, USA,
2019. [CrossRef]
Nowak, D.J.; Hirabayashi, S.; Bodine, A.; Greenfield, E. Tree and forest effects on air quality and human health in the United
States. Environ. Pollut. 2014, 193, 119–129. [CrossRef] [PubMed]
Kuželka, K.; Slavík, M.; Surový, P. Very High Density Point Clouds from UAV Laser Scanning for Automatic Tree Stem Detection
and Direct Diameter Measurement. Remote Sens. 2020, 12, 1236. [CrossRef]
Bello, S.A.; Yu, S.; Wang, C.; Adam, J.M.; Li, J. Review: Deep Learning on 3D Point Clouds. Remote Sens. 2020, 12, 1729. [CrossRef]
Timilsina, S.; Sharma, S.K.; Aryal, J. Mapping Urban Trees Within Cadastral Parcels Using An Object-Based Convolutional Neural
Network. ISPRS Ann. Photogramm. Remote Sens. Spat. Inf. Sci. 2019, IV-5/W2, 111–117. [CrossRef]
Timilsina, S.; Aryal, J.; Kirkpatrick, J.B. Mapping Urban Tree Cover Changes Using Object-Based Convolution Neural Network
(OB-CNN). Remote Sens. 2020, 12, 3017. [CrossRef]
Phelan, K.; Hurley, J.; Bush, J. Land-Use Planning’s Role in Urban Forest Strategies: Recent Local Government Approaches in
Australia. Urban Policy Res. 2018, 37, 215–226. [CrossRef]