Arboriculture & Urban Forestry 50(1): January 2024 109

Arboriculture & Urban Forestry 2024. 50(1):109–130

https://doi.org/10.48044/jauf.2023.029
Scientific Journal of the
International Society of Arboriculture

Green Urban Scenarios: A Framework for Digital

Twin Representation and Simulation for Urban
Forests and Their Impact Analysis

By Bulent Ozel and Marko Petrovic

Abstract. Background: Trees are a critical part of urban infrastructure. Cities worldwide are pledging afforestation objectives due to net-zero
targets; however, their realisation requires a comprehensive framework that combines science, policy, and practice. Methods: The paper pres-
ents the Green Urban Scenarios (GUS) framework for designing and monitoring green infrastructures. GUS considers weather, maintenance,
tree species, diseases, and spatial distributions of trees to forecast their impacts. The framework uses agent-based modelling (ABM) and simu-
lation paradigm to integrate green infrastructure into a city’s ecological, spatial, economic, and social context. ABM enables the creation of dig-
ital twins for urban ecosystems at any level of granularity, including individual trees, to accurately predict their future trajectories. Digital
representation of trees is created using a combination of datasets such as earth observations from space, street view images, field surveys, and
qualitative descriptions of typologies within existing and future projects. Machine learning and statistical models calibrate biomass growth pat-
terns and carbon release schemes. Results: The paper examines various green area typologies, simulating several hypothetical scenarios based
on Glasgow’s urban forests. It exhibits the emergence of heterogeneity features of the forests due to interactions among trees. The growth tra-
jectory of trees has a non-linear transition phase toward stable growth in its maturity. Reduced maintenance deteriorates the health of trees lead-
ing to lower survival rate and increased CO2 emissions, while the stormwater alleviation capacity may differ among species. Conclusions: The
paper demonstrates how GUS can facilitate policies and maintenance of urban forests with environmental, social, and economic benefits.
Keywords. Agent-Based Modelling; Digital Twins; Nature-Based Solutions; Scenario Analysis; Urban Forest.

INTRODUCTION researchers. Besides, in the growth process of an

Trees are a critical part of urban infrastructure along- urban forest, trees are not the sole actors. Other actors,
side bridges, roads, and rails, yet the lack of under- such as the built environment and humans, through
standing around the ecosystem benefits they provide their policies and interventions, can all affect the
stands in the way of them being conceived and imple- growth of an urban forest. The scale of the problem
mented as such. Even though cities are pledging chal- requires approaching practices of planning and main-
lenging targets—Prague has made a commitment to taining urban forests as well as estimating their impacts
plant 1 million trees (Dimitrova 2021), Sydney to plant through a complex systems lens (Zhang and DeAngelis
5 million trees by 2030 (New South Wales Government 2020).
2020)—the realisation of such targets is questionable The work presented in this paper takes a holistic
given the current practices, tools, and methodologies. approach combining science, policy, and technology.
Models and software tools evaluating the ecosystem It proposes a theoretical and computational frame-
benefits of trees fall short because they are too general: work that other researchers and practitioners can
they fail to include location and species-specific data adopt and extend. It contributes to the state of the art
(Zhang and DeAngelis 2020; Grueters et al. 2021). In in a number of complementary ways. On the scien-
addition, they lack modularity and extensibility and tific side, it adopts and extends previous work on
cannot be fully interrogated as they lack sufficient agent-based forest growth models. On the policy
transparency to be adopted by other practitioners and side, it adopts the Cynefin framework (Snowden

©2024 International Society of Arboriculture

110 Ozel and Petrovic: Green Urban Scenarios

2000), a conceptual framework for decision-making. inherent feature of complex systems (Newman 2011).
On the technical side, it creates an end-to-end open- Third, a tree constantly interacts with its environment
source and documented software framework. A detailed and other trees (Coates et al. 2003). Due to specificity
discussion of previous work and the gap in the state and heterogeneity, its growth process is dynamic,
of the art is provided below. depending on its size, age, and reach to resources
The overall objective is to enable cities to monitor, such as light, soil, water, or other factors (Berger and
(re)design, and forecast green infrastructure portfolios Hildenbrandt 2000). The collection of individual tree-
and their long-term impacts under varying weather level interactions leads to the emergence of a forest-level
conditions, maintenance regimes, species composi- growth pattern that might differ from an individual
tions, spatial distributions and their exposure to dis- tree. This characteristic is highlighted in Figure 1,
eases. This paper presents a Green Urban Scenarios comparing the growth trajectory of a tree’s trunk size
(GUS) framework and its application under a set of during its life cycle versus the growth trajectory of
simplified urban forest planning schemes. For more the whole tree population, where individual tree ver-
complex and granular use cases, as well as to enable sus forest growth generally exhibits varying patterns
replicability, integration with other tools and models, (Dale et al. 1985). Last, an urban forest is an open
and further extensions by other practitioners and system and is exposed to external shocks such as
researchers, a detailed model specification and code invasive insects, fires, and frequent human interven-
documentation is provided at its public source code tions. Such externalities are essential and integral to
distribution channel as a live document (Ozel and the analysis of complex systems.
Petrovic 2022). An agent-based paradigm is adopted to model an
This manuscript discusses relevant works within urban forest as a complex system with its compo-
the field on complex systems, presenting the novel- nents and interactions within a geophysical context.
ties introduced in this work; outlines the methodol- An agent is an abstraction for an autonomous, reactive,
ogy for applications of the framework; provides key and proactive information-processing entity (Epstein
model specifications and the current calibration of its and Axtell 1996). Thus, trees and other entities such
parameters; demonstrates an exemplary geo-localised as birds, bees, insects, or sensors that collect data on
scenario design set-up that would enable a policy- soil health could all be construed as autonomous agents.
maker or a project owner to gain insights into plan- The physical system can then be mimicked digitally
ning, monitoring, and measuring the impacts of green by configuring agents and their interactions within
infrastructure, presenting the results from the simula- the surrounding ecosystem to capture the context and
tions; and discusses the modelling approach and points conduct very granular computational experiments.
to potential use cases and extensions of the framework. The agent-based modelling (ABM) approach of
the framework provides a complex systems design
BACKGROUND and analysis methodology, a modelling paradigm, a
In this study, an urban forest is considered as a com- software programming style, and a simulation frame-
plex system with 4 underlying characteristics, adopted work (Savaglio et al. 2020). There are many purposes
from Newman (2011). Figure 1 depicts the complex for using simulation modelling (Epstein 2008). Sim-
systems features relevant to urban forests. First, each ulations can serve as decision-support tools to create
urban forest is located at a physical location where would-be worlds, explore alternative counterfactuals,
the density of its trees, exposure to the sun, access to and perform what-if scenarios. As a prime example of
water, and other location-specific conditions can vary this method, the Decision Theatre at Arizona State
drastically, comparing one city to another and within University uses wall-sized 260° screens and surround
the same city. Capturing such location-specific condi- sound to create an immersive experience, helping
tions is referred to as a specificity characteristic. Sec- participants make informed policy decisions. In addi-
ond, each urban forest may have different species tion, Dubbelboer et al. (2017) employ ABM to simu-
composition. Even 2 trees within the same species late the dynamical evolution of flood risk and economic
often have different shapes and physiology and vulnerability to facilitate London’s more informed
hence, may respond to the same environment differ- insurance mechanism. Other common uses of simula-
ently (Bittebiere et al. 2012). Such heterogeneity is an tion include using it for new theory formation (Schelling

©2024 International Society of Arboriculture

Arboriculture & Urban Forestry 50(1): January 2024 111

Figure 1. Modelling features of the Green Urban Scenarios (GUS) framework representing an urban forest as a complex system.
Adopted from a complex systems science perspective (Newman 2011). The figure highlights the role of location-specific conditions
on the growth of trees, the impact of externalities such as pollutants from neighbouring industries, the heterogeneity among trees, and
the variance in growth trajectories of an individual tree versus the tree population as a whole.

1978), for testing new market mechanisms (Gode and into the role of competition on the growth of plants,
Sunder 1993), for the role of green housing (Raberto while Grimm et al. (2017) argue that ABM is the best
et al. 2019), or as a purely pedagogical device (Kir- fit to study biodiversity in ecosystems. Zhang and
man 1993). DeAngelis (2020) provide an in-depth literature review
This paper develops a simulator to explore aspects on the use of ABM in plant biology and ecology.
such as the emergence of forest growth patterns at the This paper advances existing works on ecosystem
site and city scale through event-based scenarios, growth models (e.g., Canham et al. 1994; Berger and
leading to more efficient design and implementation Hildenbrandt 2000; Bittebiere et al. 2012; Zhang and
of green infrastructures. This approach has been used DeAngelis 2020; Grueters et al. 2021), by introducing
in urban planning. For example, Karnouskos and de a generic and modular framework that can be applied
Holanda (2009) created an agent-based environment to various urban forests at different levels of granular-
to simulate a smart city where different smart objects ity as well as further extended and integrated with
could act autonomously, collaborate and dynamically other tools in the literature. It explicitly shares the
use or produce energy. In addition, ABMs have also position of Grueters et al. (2021) that growth models
been employed in scientific research in ecology. Berger should pay attention to spatial aspects and move from
and Hildenbrandt (2000) discuss the need for spa- mean-field models to the granularity of individual
tially explicit modelling of forest dynamics, looking trees. However, unlike Grueters et al. (2021), which
into the growth of individual trees and their competi- is specifically designed to study the growth dynamics
tion for space and other resources. Bittebiere et al. of Mangrove forests, the model presented in this
(2012) used an agent-based approach to look deeper work can be used for any forest type with single or

©2024 International Society of Arboriculture

112 Ozel and Petrovic: Green Urban Scenarios

multiple tree species. This novelty is introduced by its METHODOLOGY

computational model, where species-specific allome- The methodology is inspired by the Cynefin frame-
tric equations and their location-specific parameters work (Snowden 2000), a framework for decision-
are dynamically loaded during the initialisation phase making. The Cynefin framework initially proposes
of the simulations. This computational modularity is context-aware decision-making schemes. This work
further extended to capture location-specific weather adopts the set of actions that the Cynefin framework
and landscape conditions. Ozel and Petrovic (2022) proposes for complex systems, where actions to make
provide detailed documentation for adopting the model a qualitative change in the system are based on proba-
to study urban or non-urban forest growth dynamics. tion and making sense of alternative pathways.
JABOWA, a software with proprietary and closed- While working in complexity, Snowden (2000)
source code (Botkin 1993), can be regarded as one of proposes to start with invoking new experiments or
the pioneering software applications where the growth collecting information on emergent features (probe)
of individual trees serves as the foundation for simu- by a careful reflection to make sense of how the sys-
lating forest growth, while SORTIE (Coates et al. tem reacts to new experiments (sense), followed by
2003) is a computational model that predicts forest calibrated actions that may lead to a new state
dynamics by simulating changes in tree populations (respond). Figure 2 depicts the adoption of the Cyne-
over time. In a similar vein, GUS considers interac- fin framework as an overarching methodology, and
tions among nearby individual trees, much like SOR- Figure 3 presents technical infrastructure on how the
TIE. However, SORTIE is specifically designed to methodology is implemented as software and data
offer growth predictions for individual trees in processing components. The adoption aims to make a
multi-species forests. On the other hand, mesoFON systemic change by combining policy intervention,
(Grueters et al. 2019), a newly developed agent-based planning, impact forecasting and monitoring.
model, focuses on mangrove forest dynamics and Urban forests are in constant interaction with their
aims to identify forest management scenarios that environments; therefore, GUS requires a continuous
maximise timber yield. GUS encompasses modelling and iterative process in monitoring, sense-making,
features found in both SORTIE and mesoFON, and it and interventions with feedback loops. Figure 2 visu-
adopts a generic and modular framework that is appli- alises how a linear probe-sense-respond process is
cable to urban forests. Its modular nature allows for made iterative with feedback loops.
integration with other tools, including iTree (Nowak As the figure suggests, a typical process starts with
2020), a widely used software for assessing the value data curation and processing to create a digital repre-
of ecosystem services provided by urban forests. sentation of urban forests at the desired level of gran-
In addition to combining existing computational ularity and fidelity. The upper and lower panels are
and scientific features in a modular and interoperable contrasted in the figure to point out potential simplifi-
manner, the current version extends the state-of-the- cations while moving from real physical trees into
art studies by implementing a dynamic carbon release their digital twin representations. For this paper, the
process as part of the growth dynamics. For details, database from the iTree survey on tree canopy in
see the Methodology section. Individual tree growth Glasgow (Rumble et al. 2015) is used to create a sim-
dynamics not only consider competition between plified and generic representation of tree populations
neighbouring trees (Coates et al. 2003) but also the in Glasgow. However, an automated monitoring com-
current health of a tree and the health of other sur- ponent is being developed within ReTreevAIable Proj-
rounding trees. This, in turn, enables GUS to study ect (2022) where remote sensing data is used to detect
disease contagion without a need to develop a conta- and recognise individual trees while creating their
gion model from scratch (see the Discussion section digital twin representations.
for details). Figure 3 details the overall architecture, presenting
Lastly, as presented in the subsequent section, the how the methodology is implemented. It depicts con-
forest growth model and simulator are designed and tinuous feedback loops between 3 main pillars: policy
implemented as an integral component of an urban interventions (module 1); scenario analysis, including
policy-making scheme and a novel decision-making inputs and scenario design (module 2); population ini-
methodology. tialisation (module 3); population dynamics (module 4);

©2024 International Society of Arboriculture

Arboriculture & Urban Forestry 50(1): January 2024 113

Figure 2. Change model in complexity that combines science, practice, and policy adopted from Cynefin framework (Snowden 2000)
as an overarching methodology for Green Urban Scenarios framework. The figure depicts how data collection is connected to digital
twin representation and scenario analysis with designated parameters, which in return may inform practitioners or policymakers to
allocate financial and other resources for creating and maintaining urban forests.

Figure 3. Green Urban Scenarios (GUS) framework architecture diagram.

©2024 International Society of Arboriculture

114 Ozel and Petrovic: Green Urban Scenarios

impact analysis (module 5), and data curation and of the existing data or to enter information about new
monitoring (modules 6 and 7). Each of these elements sites, species, allometric equations, and experiments.
is designed as an independent module; when com- The site configuration is a database with non-
bined, they form the GUS framework. A typical use exhaustive information about specific sites. For instance,
case starts by proposing policy intervention (module it stores the data about the exact location of a site, its
1), for example, building a new urban park in a city. boundaries, the total area of the site, the data about
Different park typologies and scenarios are created the surface, and any other relevant data that may
for examination using the experiment setup (module describe a specific location. The list of variables is
2). Typologies can differ according to the species expandable; therefore, users may use the database to
composition, tree distribution, different maintenance improve the quality of descriptions of the existing
strategies, etc. Once determined, scenarios are passed sites or enter information about new sites. This paper
to the population initialisation module (3), where creates a hypothetical site that matches species and
hypothetical tree populations are created for each sce- climate conditions in Glasgow for demonstration pur-
nario. Each hypothetical population is then forwarded poses. For this purpose, an i-Tree survey on the urban
to module 4 (population dynamics) to simulate the area is used (Rumble et al. 2015).
population growth and development over N years. The The population configuration is a database describ-
simulation results can be either sent back to module 2 ing the tree population at a specific location. It can
for a redesign of the scenario, sent back to the policy- consist of the existing trees at a given location, the
makers, stored in the data curation module (6), or for- planned tree population, or any hypothetical tree pop-
warded to module 5 to calculate ecosystem services ulation that could be studied at a particular site. The
(carbon sequestration, water retention, and air pollu- granularity of the data in the database is flexible,
tion removal) over N years in the future. Each of these meaning that one can describe each tree in the popu-
modules is further explained in the following sections. lation using a non-exhaustive list of variables in
detail. The current paper employs a hypothetical tree
Policy Intervention population of newly planted young trees at the site
Analysis in the GUS starts by entering the description using the following variables to describe it: species,
of a nature-based solutions (NbS) project. The descrip- diameter at breast height (DBH), tree height, canopy
tion is given in terms of input parameters to the model height, canopy width, leaf area index (LAI), bark area
describing the population’s size and characteristics index (BAI), plant area index (PAI), tree dieback
(e.g., the number of trees, species, typology, density), ratio, tree age, canopy overlap, and the exact coordi-
planting/maintaining activities, etc. If the typology nate of each tree in the site.
already exists, the population is initialised based on On the other hand, if a real tree population is under
actual observations; otherwise, the population is ini- study and some data points are missing, one can use
tialised using the rest of the input parameters given in less granular data, such as more generic species infor-
the NbS description. mation or DBH and other tree size measures drawn
from a particular distribution. If missing, some of the
Scenario Analysis variables can also be estimated, such as tree height,
The second pillar, scenario analysis, includes 4 modules: canopy height, canopy width, LAI, and BAI. Other
inputs, population initialisation, population dynamics, variables, such as tree dieback ratio, tree age, canopy
and impact analysis. overlap, and coordinates, can be assumed.
The allometric database consists of allometric
Input Data Types equations for specific species at specific locations.
The framework is flexible and applicable to any for- Each species at each site is described with an allome-
est typology at any location. It divides the inputs into tric equation. If some allometric equations are miss-
4 different segments: (i) site configuration, (ii) popu- ing, the researcher can define a representative or mean
lation configuration, (iii) allometric equations, and allometric for the specific area, considering a more
(iv) experiment configuration, where the user can generic family of species. In the most generic case, there
provide an independent (stand-alone) database for is one mean allometric for deciduous and one for
each segment. All databases are easily expandable, conifers at the specific location. This work considers 2
allowing users to improve the quality of the description species: European ash (Fraxinus excelsior) and Port

©2024 International Society of Arboriculture

Arboriculture & Urban Forestry 50(1): January 2024 115

Orford cedar (Chamaecyparis lawsoniana). Thus, which influences their growth. They would be exposed
species-specific allometric equations to calculate bio- to various diseases which can affect their diebacks
mass, tree height, grown height, and crown are iden- and probability of dying. If a tree dies, humans can
tified from Nowak (2020). proceed with site maintaining actions which could
The experiment configuration database describes result in removing or replacing the dead tree or could
the experiments to be performed on the particular tree result in no action. One simulation of GUS over N
population at the given location. For now, experiments years provides one unique realisation of the world.
are available related to the different maintenance Nevertheless, by repeating the simulation process,
strategies. This paper uses three levels (M0, M1, M2) one can obtain another possible/unique outcome. By
of tree maintenance which is related to planting, removal simulating the model M times over N years, GUS
and disposal, and tree replacement. Maintaining level provides the outputs and M unique possible realisa-
M0 disregards any maintenance activities, while the tions. Hence, the outputs can be analysed either by
activities increase with maintaining levels M1 and M2. observing each unique realisation (so-called “run”,
“simulation”, or “seed”), or by aggregating the outputs
Population Initialisation and looking at the average outcome and its variance
Based on inputs (described in the previous section), over M seeds, which allows the researchers to measure
the model initialises the digital ecosystem, creating a the confidence interval and the risk of the estimation.
digital representation of each tree in the population.
Each tree can be described by a list of variables such Data Curation and Monitoring
as DBH, tree height, canopy height, canopy width, The data on population growth and ecosystem ser-
tree health, and percentage crown loss from the pop- vices can be stored in the data curation module and
ulation configuration database. In addition, each tree reutilised at any time during the simulations. Finally,
is assigned an allometric equation from the allometric monitoring collects the data from the field, real data,
database based on species and location. The digital during the project time span. The methodology devel-
trees are then distributed on a digital grid. The model oped in this paper has fully or semi-automated com-
uses a special configuration method that creates a dig- ponents for data collection, curation, integration, and
ital site based on the site configuration database. It generation processes. The use of actual data spans the
allocates each digital tree on a grid, where each tree phases from model calibration and validation to veri-
has its own x and y coordinate. Therefore, each tree fication of the outcomes.
has a unique place in the digital site, surrounded by
neighbouring trees. The spatial configuration trans- MODEL SPECIFICATION AND
lates all other site characteristics from the actual site CALIBRATION
into digital space, such as site size, the distance The tree growth occurs within each iteration/simula-
among trees, and the sun exposure. The initialised tion step, mimicking a calendar year. At each step, the
tree population, including all parameters and weather growth model predicts DBH, height, crown width,
data, is passed to the core simulation engine in mod- and crown height growths taking into account envi-
ule 4 (Figure 3), where the tree population growth is ronmental and tree health conditions (Figure 4). It
simulated over N years. also calculates tree biomass and net carbon sequestra-
tion (NCS). NCS is a difference between gross car-
Population Dynamics bon sequestration (GCS) and decomposition. GCS
The main component of the population dynamics represents the amount of sequestrated carbon dioxide
component is a flexible agent-based model that simu- in one year, while decomposition denotes the emis-
lates a spectrum of interactions among the trees. Each sion/release of carbon dioxide due to the tree decom-
simulation consists of N iterations (steps), where each position process (Nowak et al. 2002; Nowak et al.
iteration represents a period of one year. Within each 2008). In the following section, each segment of the
iteration, each agent performs tasks, imitating what a tree growth process is presented.
real agent (e.g., a tree) would do during one year. For At the initial state, t = 0, the entire digital ecosys-
instance, trees would be exposed to the weather con- tem is initialised, all trees’ variables are set to their
ditions that would trigger their growth. They would initial values, and the system is ready to be simulated.
be competing with other trees for sun exposure, When the simulation runs from time t = 0 to time t = 1

©2024 International Society of Arboriculture

 116 Ozel and Petrovic: Green Urban Scenarios

Figure 4. ABM diagram: the state graph showing the agents’ interactions, decision-making, and adaptation processes.

(i.e., one iteration takes place), trees perform the fol- overlap in meters between trees i and j rescaled for a
lowing tasks. multiplier of 0.5 to correct for the square-shaped
Each tree reads information about weather condi- assumption of the tree crown. Based on the crown
tions for the current year. In particular, each tree gets overlap size, the overlap ratio is calculated as:
the information about frost-free days ffd(t) at year t. ‫݌݈ܽݎ݁ݒ݋‬௜௝
Each tree at each iteration also checks the state of its ‫݋݅ݐܽݎ݌݈ܽݎ݁ݒ݋‬௜௝ ൌ ͲǤʹͷ ൈ ݉݅݊ ቆͳǡ ቇ(2)
ܿ‫݄݅ݐ݀݅ݓ݊ݓ݋ݎ‬
neighbouring trees and determines its current crown
light exposure (CLE). Trees compute the crown over- where 0.25 multiplier accounts for 1 of the 4 sides of
lap at time t concerning their neighbours: the grid cell. In addition, the crown overlap ratio is
adjusted for the neighbouring relative height, assum-
‫݌݈ܽݎ݁ݒ݋‬௜௝ ൌ ݉ܽ‫ݔ‬൫Ͳǡ ͲǤͷ ൈ ൣܿ‫݄ݐ݀݅ݓ݊ݓ݋ݎ‬௜ ൅ ܿ‫݄ݐ݀݅ݓ݊ݓ݋ݎ‬௝ing െ ܾܽ‫݁ܿ݊ܽݐݏ݅݀݇ݎ‬ ௜௝ ൧൯ሺ݉ሻ
that a taller tree creates more shading. Thus, the
(1)
ൈ ൣܿ‫݄ݐ݀݅ݓ݊ݓ݋ݎ‬௜ ൅ ܿ‫݄ݐ݀݅ݓ݊ݓ݋ݎ‬௝ െ ܾܽ‫݁ܿ݊ܽݐݏ݅݀݇ݎ‬௜௝ ൧൯ሺ݉ሻ
adjusted crown overlap ratio is given as:
where index j denotes a neighbouring tree, while ௞
݄݄݁݅݃‫ݐ‬௝
index i stands for the reference tree. Note that index t ൌ ෍ ‫݋݅ݐܽݎ݌݈ܽݎ݁ݒ݋‬௜௝ ൈ (3) 
‫݋݅ݐܽݎ݌݈ܽݎ݁ݒ݋‬௜ ݄݄݁݅݃‫ݐ‬௝ ൅ ݄݄݁݅݃‫ݐ‬௜
is omitted for the sake of better readability. Therefore, ௞
௝ୀଵ

bark distance(ij) is the distance between tree i and its ݄݄݁݅݃‫ݐ‬௝

ൌ ෍ ‫݋݅ݐܽݎ݌݈ܽݎ݁ݒ݋‬௜௝ ൈ 
neighbouring tree j, while overlap(ij) is‫݋݅ݐܽݎ݌݈ܽݎ݁ݒ݋‬
the crown ௜ ݄݄݁݅݃‫ݐ‬௝ ൅ ݄݄݁݅݃‫ݐ‬௜
௝ୀଵ

©2024 International Society of Arboriculture

 Arboriculture & Urban Forestry 50(1): January 2024 117

where k is the number of neighbours, while height(i) increasing in dieback ratio (i.e., the worse the condi-
and height(j) are the heights of the reference and tion, the higher the probability of dying), as well as
neighbouring trees, respectively. The crown light the function is decreasing in maintenance scope (i.g.,
exposure is then calculated for each tree i at time t as: the higher the maintenance scope, the lower the prob-
ability of dying).
݈ܿ݁௜ ൌ ݉ܽ‫ ݔ‬ቆͲǡ ͳ െ ݈݄݅݃‫ ݎ݈݁݅݌݅ݐ݈ݑ݉ݏݏ݋݈ݐ‬ൈ (4) ቇ Another component of the core module is the tree
‫݋݅ݐܽݎ݌݈ܽݎ݁ݒ݋‬௜
ǡ ͳ െ ݈݄݅݃‫ ݎ݈݁݅݌݅ݐ݈ݑ݉ݏݏ݋݈ݐ‬ൈ ቇ growth model developed following the work of Smith
‫݋݅ݐܽݎ݌݈ܽݎ݁ݒ݋‬௜ and Shifley (1984), Nowak (1994), and Nowak
where the light loss multiplier is a constant set to (2020). These growth models are extended by con-
0.75. sidering the health of a tree and the health of its neigh-
bours, as well as exposure to light. The resilience of a
Dieback tree against crown loss or diseases is heterogeneous
The dieback ratio indicates the percentage of the tree and correlates with its current age or DBH. Combin-
that is dying, and/or it indicates the health condition ing the input data on tree species, site configuration,
of trees. For instance: dieback < 0.01 ⇒ excellent allometric equations, and experiments (e.g., main-
condition; 0.01 ≤ dieback ≤ 0.1 ⇒ good condition; taining activities), the core module simulates the tree
0.1 < dieback ≤ 0.25 ⇒ fair condition; 0.25 < dieback population’s growth and development. In particular,
≤ 0.5 ⇒ poor condition; 0.5 < dieback ≤ 0.75 ⇒ critical the model calculates DBH growth (cm) as:
condition; 0.75 < dieback ≤ 0.99 ⇒ dying condition;
οܾ݄݀௜௧ ൌ ݀݅ܽ݉݁‫݄ݐݓ݋ݎ݃ݎ݁ݐ‬௜௧ ൈ ݃‫ݐ݊݁݉ݐݏݑ݆݄݀ܽݐݓ݋ݎ‬௜௧ 
dieback = 1 ⇒ dead tree. The modelled dieback ratio (9)
οܾ݄݀ ൌ
depends on (i) the latest condition of the tree, (ii) ௜௧the ݀݅ܽ݉݁‫݄ݐݓ݋ݎ݃ݎ݁ݐ‬ ௜௧ ൈ ݃‫ݐ݊݁݉ݐݏݑ݆݄݀ܽݐݓ݋ݎ‬௜௧ 

age via DBH, and (iii) the health of neighbouring where diameter growth (cm) is given as:
trees. The health of neighbouring trees determines the ݂݂݀௧
݀݅ܽ݉݁‫݄ݐݓ݋ݎ݃ݎ݁ݐ‬௜௧ ൌ ‫݄ݐݓ݋ݎ݃݀ݎܽ݀݊ܽݐݏ‬௧ ൈ ൈ ݈ܿ݁ ൈ ሺͳ െ
contagion risk, which is calculated as: ݂݂݀
(10)
݂݂݀௧
௞
݀݅ܽ݉݁‫݄ݐݓ݋ݎ݃ݎ݁ݐ‬௜௧ ൌ ‫݄ݐݓ݋ݎ݃݀ݎܽ݀݊ܽݐݏ‬௧ ൈ ൈ ݈ܿ݁ ൈ ሺͳ െ ܾ݀݅݁ܽܿ݇௜௧ ሻ
σ௝௥ୀଵ ݇ ௥ିଵ
ൈ ܾ݀݅݁ܽܿ݇௥ ݂݂݀
ܿ‫݇ݏ݅ݎ݊݋݅݃ܽݐ݊݋‬௜ ൌ ͲǤͻ ൈ ෍  (5)
݇௝
௝ୀଵ Standard growth is a location-species specific param-
where k is the total number of neighbours, j, and 0.9 eter which is provided in the allometric database for
is an adjustment parameter. The new dieback rate is each species and location, and it is constant during
drawn from a uniform distribution between the heal- the simulation. If missing, the moderate annual
ing range and dying range, calculated as growth rate of 0.8382 cm is applied (see the literature
ξܾ݄݀ on diameter growth for details, e.g., Smith and Shif-
݄݈݁ܽ݅݊݃‫݁݃݊ܽݎ‬௜ ൌ െͳ ൈ ሺͳ െ ܿ‫݇ݏ݅ݎ݊݋݅݃ܽݐ݊݋‬௜ ሻ ൈ
‫ܯ‬ ley 1984;௜ DeVries 1987; Lorimer et al. 1992; Nowak
ൈ ݄݈݁ܽ݅݊݃‫݁ݐܽݎ‬
ξܾ݄݀ (6) 1994; Nowak 2020). is the number of frost-free days
ൈ ሺͳ െ ܿ‫݇ݏ݅ݎ݊݋݅݃ܽݐ݊݋‬௜ ሻ ൈ ൈ ݄݈݁ܽ݅݊݃‫݁ݐܽݎ‬௜ 
‫ܯ‬ in the year t, while ffd is the average number of frost-
where M is a parameter indicating maintenance free days at the given location. The growth adjust-
scope, such that M(M0) = 5, M(M1) = 4 and M(M2) = ment is a parameter calculated as a ratio of tree height
1; while healing rate(i) is a parameter set to 0.005. to tree height at maturity (Ozel and Petrovic 2022). In
Furthermore, the dying range is calculated as: addition, tree height, crown height, and crown width
are calculated as functions of DBH using the species-
‫݁ݐܽݎ݇ݏ݅ݎ‬௜ ൈ ‫ܯ‬
݀‫݁݃݊ܽݎ݃݊݅ݕ‬௜ ൌ 
(7) specific and location-specific equations provided in
ሺͳ െ ܿ‫݇ݏ݅ݎ݊݋݅݃ܽݐ݊݋‬௜ ሻ ൈ ξܾ݄݀ the allometric database.
where risk rate is a parameter set to 0.005. Therefore, Based on the predicted tree growth, the model esti-
the new dieback of tree i at time t is calculated as: mates the biomass, carbon stock, yearly sequestra-
tion, and CO2 release for each tree in the population
ܾ݀݅݁ܽܿ݇௜௧ ൌ ܾ݀݅݁ܽܿ݇௜ǡ௧ିଵ ൅ ܷ̱ሺ݄݈݁ܽ݅݊݃‫݁݃݊ܽݎ‬௜௧ ǡ ݀‫݁݃݊ܽݎ݃݊݅ݕ‬௜௧ ሻ
(8) over N years. The descriptions of all calculations can
௜௧ ൌ ܾ݀݅݁ܽܿ݇௜ǡ௧ିଵ ൅ ܷ̱ሺ݄݈݁ܽ݅݊݃‫݁݃݊ܽݎ‬௜௧ ǡ ݀‫݁݃݊ܽݎ݃݊݅ݕ‬௜௧ ሻ be found in the model specification (Ozel and Petro-
At every period t, there is a probability that a tree vic 2022).
dies. It is modelled as a step function, and it is

©2024 International Society of Arboriculture

 118 Ozel and Petrovic: Green Urban Scenarios

Decomposition where root to shoot ratio(i) is a percentage of a tree’s

Decomposition is a process of carbon dioxide release biomass under the ground. There is a 50% probability
from dead or alive trees. If a tree is still alive, decom- that a dead tree will be removed from the site.
position may occur on its dead branches due to die- Given that it is removed, there is a 70% probability
backs or crown loss. The amount of carbon to be that it will be burnt, in which case the release of CO2
released due to dead branches of alive trees is calcu- is immediate and given as:
lated as: ‫ܱܥ‬ଶ ‫݈݂݁݅݋ݎ݌݁ݏ݈ܽ݁݁ݎ‬௜ఛ ൌ ͲǤ͹ ൈ ‫ܱܥ‬ଶ ‫ݎ݃݁ݒ݋ܾܽ݁ݏ݈ܽ݁݁ݎ‬
(16)
‫ܾܱ݇ܿܽ݁݅݀ܥ‬
‫ܱܥ‬ଶ ‫݁ݒ݈݅ܽ݁ݏ݈ܽ݁݁ݎ‬௜௧ ൌ ܿ‫݋݅ݐܽݎ݇݊ݑݎݐ݋ݐ݊ݓ݋ݎ‬௜ ൈ ଶ ‫݈݂݁݅݋ݎ݌݁ݏ݈ܽ݁݁ݎ‬
௜௧ ൈ ܿܽ‫݁݃ܽݎ݋ݐݏ݊݋ܾݎ‬
௜ఛ ൌ ͲǤ͹ ൈ ௜௧  ଶ ‫݀݊ݑ݋ݎ݃݁ݒ݋ܾܽ݁ݏ݈ܽ݁݁ݎ‬௜ఛ 
‫ܱܥ‬
(11)
‫݋݅ݐܽݎ݇݊ݑݎݐ݋ݐ݊ݓ݋‬௜ ൈ ܾ݀݅݁ܽܿ݇௜௧ ൈ ܿܽ‫݁݃ܽݎ݋ݐݏ݊݋ܾݎ‬௜௧ 
The remaining 30% is assumed to be converted into
where crown to trunk ratio(i) is a parameter set to sustainable products. If the tree is not removed from
0.05. the site, there is a 40% chance that it will stay untouched
The decomposition rate of dead branches depends (standing) and a 60% probability that it will be
mainly on human actions, such as maintenance activ- mulched. In the first case, the slow decomposition
ities. For now, the model distinguishes 2 scenarios: (i) rate is applied (Equation 12) for both root and above-
if the dead branches are mulched and left to be decom- ground biomass, with parameter k = 5; while in the
posed in nature, or (ii) if the dead branches are imme- second case, the root is still decomposed by the slow
diately decomposed (e.g., burnt). The first scenario rate, while the biomass above ground is decomposed
happens with 90% probability, and in this case, the by the fast rate setting parameter k = 2. Therefore, the
decomposition rate is given as: CO2 release profile of a standing tree is computed as:
ͳ ௦௟௢௪
‫ܱܥ‬ଶ ‫݈݂݁݅݋ݎ݌݁ݏ݈ܽ݁݁ݎ‬௜ǡ௧ିఛ ൌ ݀݁ܿ‫݁ݐܽݎ݊݋݅ݐ݅ݏ݋݌݉݋‬௜ǡ௧ିఛ ൈ
݀݁ܿ‫݁ݐܽݎ݊݋݅ݐ݅ݏ݋݌݉݋‬௜ǡ௧ିఛ ൌ ௧ିఛ  (12) (17)
௦௟௢௪
݇ ൈ‫ܱܥ‬
݁ ଶ௞‫݈݂݁݅݋ݎ݌݁ݏ݈ܽ݁݁ݎ‬௜ǡ௧ିఛ ൌ ݀݁ܿ‫݁ݐܽݎ݊݋݅ݐ݅ݏ݋݌݉݋‬௜ǡ௧ିఛ ൈ ܿܽ‫݁݃ܽݎ݋ݐݏ݊݋ܾݎ‬௜ఛ 
where parameter k determines the speed of the
decomposition. It is set to 2 for mulched dead ‫ܱܥ‬ଶ ‫݈݂݁݅݋ݎ݌݁ݏ݈ܽ݁݁ݎ‬௜ǡ௧ିఛ ൌ ݀݁ܿ‫݁ݐܽݎ݊݋݅ݐ݅ݏ݋݌݉݋‬௜ǡ௧ିఛ ௦௟௢௪
ൈ‫ܥ‬
branches, indicating the fast decomposition observed ‫ܱܥ‬ ‫݈݂݁݅݋ݎ݌݁ݏ݈ܽ݁݁ݎ‬௦௟௢௪ ൌ ݀݁ܿ‫݁ݐܽݎ݊݋݅ݐ݅ݏ݋݌݉݋‬ ௦௟௢௪
ൈ ‫ܱܥ‬ ‫݈݁݁ݎ‬
‫ܱܥ‬ଶ ‫݈݂݁݅݋ݎ݌݁ݏ݈ܽ݁݁ݎ‬௜ǡ௧ିఛ ൌ ݀݁ܿ‫݁ݐܽݎ݊݋݅ݐ݅ݏ݋݌݉݋‬௜ǡ௧ିఛ ൈ ‫ܱܥ‬ଶ ‫ݐ݋݋ݎ݁ݏ݈ܽ݁݁ݎ‬
ଶ ௜ǡ௧ିఛ ௙௔௦௧ ௜ఛ  ௜ǡ௧ିఛ ଶ
elsewhere (Nowak and Crane 2002), where it is seen
‫ܱܥ‬ଶ ‫݈݂݁݅݋ݎ݌݁ݏ݈ܽ݁݁ݎ‬ ൅݀݁ܿ‫݁ݐܽݎ݊݋݅ݐ݅ݏ݋݌݉݋‬
௦௟௢௪ ൈ ‫ܱܥ‬ ଶ ‫݁ݒ݋ܾܽ݁ݏ݈ܽ݁݁ݎ‬
(18)
௜ǡ௧ିఛ ൌ ݀݁ܿ‫݁ݐܽݎ݊݋݅ݐ݅ݏ݋݌݉݋‬ ௜ǡ௧ିఛ ൈ ‫ܱܥ‬ଶ ‫ݐ݋݋ݎ݁ݏ݈ܽ݁݁ݎ‬
௙௔௦௧ ௜ǡ௧ିఛ ௜ఛ
that in the first 3 years, around 60% of carbon is
൅݀݁ܿ‫݁ݐܽݎ݊݋݅ݐ݅ݏ݋݌݉݋‬
௙௔௦௧ ൅݀݁ܿ‫݁ݐܽݎ݊݋݅ݐ݅ݏ݋݌݉݋‬
ൈ ‫ܱܥ‬ଶ ‫݀݊ݑ݋ݎ݃݁ݒ݋ܾܽ݁ݏ݈ܽ݁݁ݎ‬ ൈ
௜ǡ௧ିఛ ௜ఛ ‫ܱܥ‬ ଶ ‫݊ݑ݋ݎ݃݁ݒ݋ܾܽ݁ݏ݈ܽ݁݁ݎ‬
௜ǡ௧ିఛ
௙௔௦௧
released, corresponding to empirical findings ൅݀݁ܿ‫݁ݐܽݎ݊݋݅ݐ݅ݏ݋݌݉݋‬
for ௜ǡ௧ିఛ ൈ ‫ܱܥ‬ଶ ‫݀݊ݑ݋ݎ݃݁ݒ݋ܾܽ݁ݏ݈ܽ݁݁ݎ‬௜ఛ 
mulched biomass above ground. The index τ indi-
cates the year when the branches died. The release
profile is then given as: The dead tree can also be replaced with a new tree
which is planted at the location where the tree is dead.
‫ܱܥ‬ଶ ‫݈݂݁݅݋ݎ݌݁ݏ݈ܽ݁݁ݎ‬௜ǡ௧ିఛ ൌ ݀݁ܿ‫݁ݐܽݎ݊݋݅ݐ݅ݏ݋݌݉݋‬௜ǡ௧ିఛ ൈ ‫ܱܥ‬ଶ ‫݁ݒ݈݅ܽ݁ݏ݈ܽ݁݁ݎ‬௜ఛ 
(13) The replacement will depend on the maintenance
‫݈݂݁݅݋ݎ‬௜ǡ௧ିఛ ൌ ݀݁ܿ‫݁ݐܽݎ݊݋݅ݐ݅ݏ݋݌݉݋‬௜ǡ௧ିఛ ൈ ‫ܱܥ‬ଶ ‫݁ݒ݈݅ܽ݁ݏ݈ܽ݁݁ݎ‬௜ఛ 
project in place. For instance, if the maintenance
When the dead branches are burnt, the amount of CO2 scope is M0, there will not be any replacement; if it is
release alive is immediately released. When a tree M1 the tree is replaced with 30% probability, and in
dies, the entire sequestrated carbon dioxide is released the case of maintenance scope M2, any dead tree will
in the years ahead. The decomposition of the tree’s be replaced.
root and the biomass above ground are considered
separately since the decomposition of the root is inde- Water Retention
pendent of human actions. Therefore, the CO2 release The population dynamic module provides growth
from the 2 parts is given as: projection data as output, which are passed to the
impact analysis module (Figure 3) along with the
‫ܱܥ‬ଶ ‫ݐ݋݋ݎ݁ݏ݈ܽ݁݁ݎ‬௜௧ ൌ ‫݋݅ݐܽݎݐ݋݋݄ݏ݋ݐݐ݋݋ݎ‬௜ ൈ ܿܽ‫݁݃ܽݎ݋ݐݏ݊݋ܾݎ‬௜௧ 
hourly weather data to estimate the ecosystem services.
݈݁݁ܽ‫ݐ݋݋ݎ݁ݏ‬௜௧ ൌ ‫݋݅ݐܽݎݐ݋݋݄ݏ݋ݐݐ݋݋ݎ‬௜ ൈ ܿܽ‫݁݃ܽݎ݋ݐݏ݊݋ܾݎ‬௜௧  (14)
Currently, the module incorporates computations
and based on existing literature to address water retention,
‫ܱܥ‬ଶ ‫݀݊ݑ݋ݎ݃݁ݒ݋ܾܽ݁ݏ݈ܽ݁݁ݎ‬௜௧ ൌ ሺͳ െ ‫݋݅ݐܽݎݐ݋݋݄ݏ݋ݐݐ݋݋ݎ‬carbon sequestration, and
௜ ሻ ൈ ܿܽ‫݁݃ܽݎ݋ݐݏ݊݋ܾݎ‬ ௜௧  air pollution removal (Wang
(15) et al. 2008; Hirabayashi 2013; Castro and Maidment
݃‫݀݊ݑ݋ݎ‬௜௧ ൌ ሺͳ െ ‫݋݅ݐܽݎݐ݋݋݄ݏ݋ݐݐ݋݋ݎ‬௜ ሻ ൈ ܿܽ‫݁݃ܽݎ݋ݐݏ݊݋ܾݎ‬௜௧ 

©2024 International Society of Arboriculture

Arboriculture & Urban Forestry 50(1): January 2024 119

2020; Lin et al. 2020) as well as i-Tree models such conditions and different probabilities of agents’ inter-
as i-Tree Eco, and i-Tree Hydro (United States action. The results are then aggregated over 2 dimen-
Department of Agriculture Forest Service, Washing- sions, time and simulation runs, and presented
ton DC, USA)(Ozel and Petrovic 2022). The output coherently.
from the impact analysis modules on ecosystem ser- Figure 5 summarises the 3 stages of water reten-
vices can be further used to update population data tion through panels 5a to 5c. Figure 5a presents the
and/or NbS design and rerun simulations and recalcu- first stage, where the precipitation starts and the tree
late the ecosystem services. Since the nature of the canopy (vegetation storage [Sv]) still has the capacity
model is stochastic, the simulation can be performed to retain the water. A part of the water from the can-
M times over N years, considering different weather opy evaporates (Ev) while a small part of the

Figure 5. Precipitation-evaporation dynamic. Panels (a) to (c) depict stages, and panel (d) presents a hypothetical baseline on the process
of water run-off in the absence of trees.

©2024 International Society of Arboriculture

120 Ozel and Petrovic: Green Urban Scenarios

precipitation reaches the previous and/or impervious Hourly meteorological data from several official
ground under the canopy due to precipitation drip weather stations are combined to create controlled
through the canopy (Pd). Figure 5b presents the sec- weather projections for long-term impact analysis.
ond stage, where the vegetation storage reaches its The weather data is also used to characterise the
maximum capacity Sv = Sv(max), and the entire pre- length of a stochastic tree growth season and to tune
cipitation hits the ground. If the rain hits the previous the parameters of the evapotranspiration mechanisms.
cover, it infiltrates the ground; if it hits the impervious i-Tree survey Rumble et al. (2015) has provided
cover, it is considered runoff. The evaporation from the tree canopy, species composition and the city-scale
the canopy (Ev) continues. Figure 5c shows the third tree inventory used in the analysis. High-resolution
stage when the precipitation stops, and the evapora- satellite images are used to identify flood risk zones and
tion from vegetation (Ev) continues until the canopy potentially effective locations for green infrastructures.
dries up (vegetation storage Sv = 0) or until new pre- To understand the role of species in providing eco-
cipitation begins. system services, 2 species are selected: European ash
Figure 5d represents the hypothetical case of the (Fraxinus excelsior) and Port Orford cedar (Chamae-
considered site without trees. To trace out the addi- cyparis lawsoniana). Ash is Glasgow’s most com-
tionality of trees, the following equation calculates mon deciduous tree, while cedar is the most common
the stormwater retention and runoffs for the hypothet- evergreen tree. Following the urban tree categorisa-
ical case without trees, along with the regular sce- tion guidelines (Nowak 1996), Woodland, Street, and
nario under the study and calculates the water-related Park typologies are examined. Each typology is char-
benefits of trees as acterised by the density of trees and the percentage of
improvement in water retention = impervious and pervious ground cover. In addition, 3
water retention − water retention(hypothetical) (19) tiers of maintenance and care for each scenario are
investigated. At maintenance Scope-0, there is no
and human care, while at maintenance Scope-2, state-of-art
avoided runoff = runoff(hypothetical) − runoff (20) maintenance is assumed.
Each scenario is identified by maintenance, typology,
Where improvement in water retention and avoided and species compositions. Nevertheless, the com-
runoff are net water-related benefits of the trees under the plexity and variability are further reduced to compare
study. The detailed procedure to calculate water reten- the impact of each experiment parameter. In the
tion and runoff are given in (Ozel and Petrovic 2022). reported experiments for this paper, the initial hypothet-
ical population of trees are composed of young trees
APPLICATION of the same species; they have the same girth (DBH)
Scenario Design size and are in good health condition. Besides, within
This section develops several controlled experiments a given typology, trees are planted at equal distances
demonstrating how different portfolios of urban forests from each other, and they all have the same initial sun
can aid long-term flood risk management. The objec- and weather exposure. There is no shading due to build-
tive of this section is twofold: to present the granular- ings, hills, or other taller trees. However, one should
ity and flexibility of GUS through several simplified note that these assumptions can be relaxed to match
experiments, and to point out how controlled compu- the heterogeneity and specificity features of a more
tational experiments within a geo-context can lead to complex and realistic green infrastructure within the
efficient and informed decision-making. GUS framework. Due to dynamic interactions between
The model is calibrated to match the geo-context trees and their exposure to externalities, initially, homo-
of a hypothetical site in the City of Glasgow. There- geneous tree populations gain heterogeneity through-
fore, using the Glasgow weather data over the last 15 out their life cycle. They compete for canopy space and
years, data on tree species that are native or common, sun exposure (Coates et al. 2003). Each tree’s dieback
and urban forest typologies observable or applicable rate and recovery from a poor or critical health condi-
in the area, this paper examines the role of species, tion depend on its unique history and location.
maintenance scopes, and urban forest typology in the The configuration described above translates into an
provision of ecosystem services. experiment set-up composed of 18 different scenarios:

©2024 International Society of Arboriculture

Arboriculture & Urban Forestry 50(1): January 2024 121

2 (species) × 3 (typology) × 3 (maintenance). Since light, and the under-canopy ground surface type. Next,
GUS is a stochastic framework that captures the prob- each scenario is simulated 100 times. This number
abilistic nature of weather and the interaction of trees can be reduced or increased depending on the vari-
with each other and the environment, each scenario is ability and required size of data points for statistical
run multiple times. Probabilistic variables, such as analysis. Finally, the data generated through scenario
the recovery chance of a dying tree, are drawn from simulation are fed into impact models. Figures 6 and
the underlying probability distribution according to 7 below present exemplary outputs relevant to carbon
the Monte Carlo method (Metropolis and Ulam 1949), sequestration and flood avoidance ecosystem services.
using different seeds for random number generators. The plots in Figure 8 illustrate how the GUS model
Each tree has a unique identifier and designated demonstrates an urban forest’s growth and emergence
spatial coordination depending on its host typology. features at the individual tree agent and population
This hypothetical experiment set-up has 6 different level, revealing the characteristics of a complex system.
initial tree populations: 2 (species) × 3 (typology).
Result 1. Heterogeneity Features of the Urban
However, one should note that each scenario and its
Forest as a Complex System
tree population can be created using a physical site.
The size distribution plot in Figure 8a is a tree popu-
Then each unique tree in the digital space can be cre-
lation at year 50 from one selected simulation run out
ated using monitored features such as girth, height,
of 100 runs. The histogram gives distribution in tree
canopy dimension, health condition, and exact geolo-
diameters measured at breast height, typically 1.35 m
cation. In such cases, the population initialisation mod-
from the ground. Each of the 100 runs starts with the
ule of the GUS framework (Figure 3) would enable a
same population. However, each initially identical
practitioner to create a one-to-one representation of
population evolves into a distinct urban forest over
an actual site and its digital twin version.
time. At the very inception of each project (i.e., each
In the following section, this paper presents a demon-
run), they all start with a composition of identical
strative subset of results from the experiment set-ups
trees: a healthy young ash (Fraxinus spp.) tree with a
relevant to growth patterns, carbon sequestration and
diameter of 4.77 cm. The initial uniform distribution
release, water retention capacity, and avoided water
transforms its shape over the lifetime of the forest.
runoff.
The size distribution by the end of year 50 reveals the
emergence of heterogeneity features of the urban for-
Growth and Impacts
est as a complex system.
As Figure 3 above suggests, a typical analysis cycle
comprises multiple phases. A project planning or pol- Result 2. The Validity of the Observed Growth
icy intervention initiates a scenario design with rele- Trajectory Over 100 Runs
vant intervention or control parameters. For instance, Figure 8b displays the mean net sequestration of 100
as laid out previously, maintenance as an intervention runs and the variance. The validity of the observed
instrument is parameterised with 3 distinctive scopes, growth trajectory is examined by running statistical
while street trees, woodland, and parks characterise tests in R version 0.1.3 (Marwick and Krishnamoor-
types of green infrastructures. A combination of design thy 2019). For instance, in the scenario in Figure 8, a
parameters leads to the identification of a scenario. woodland of ash trees with Scope-2 maintenance, the
Then, an initial tree population and site-specific con- average coefficient of variation (CV) of the seques-
figuration files are then prepared and linked for each tration is 5.4% for 50 years; it is 4.1% for the first 10
scenario. As part of this initialisation process, each years and 6.8% for the last 10 years. The solid line is
tree agent automatically fetches its locality and species- the average sequestration over the total number of
specific base biomass and canopy growth functions planted trees, while the grey band around the line
from a respective database. In the current version of shows its variance over 100 Monte Carlo simula-
the model, an individual tree agent holds and keeps tions. Halfway through the simulated time, a healthy
track of information on its state of health, girth, can- part of the tree population reaches maturity and keeps
opy height, crown size, the ratio of its crown dieback, per-tree sequestration performance stable, even
rate of recovery from the latest dieback, distance and though young and under-performing trees are added
position from neighbouring trees, its exposure to to populations to replace the dead ones.

©2024 International Society of Arboriculture

122 Ozel and Petrovic: Green Urban Scenarios

Figure 6. Results from maintenance scenarios.

Figure 7. Results from species-related experiments.

©2024 International Society of Arboriculture

Arboriculture & Urban Forestry 50(1): January 2024 123

Figure 8. Model dynamics and emergence.

Result 3. The Biomass Function of a Single Ash The water-related ecosystem services are calcu-
Tree lated hourly and derived from the tree population and
The biomass function of a single ash tree in the model, weather conditions. In particular, stormwater reten-
plotted in Figure 8c, is exponential. tion, improvement in stormwater retention, and
avoided runoff are calculated for different typologies
Result 4. Growth Trajectory of European Ash
of European ash and Port Orford cedar based on the
Within an Urban Park Has a Non-Linear
50 years of forecasted weather conditions in Glasgow.
Transition Phase Toward a Stable Growth in
Figure 6 displays several representative analyses
Its Maturity
where the role of maintenance on the health of the
Figure 8d shows a typical growth trajectory of ash
urban forests and their water-related ecosystem ser-
trees within an urban park that has grown into matu-
vices is explored.
rity and stayed healthy. The growth pattern, observed
through its biomass within the simulations, expresses Result 5. Good Maintenance Improves Carbon
a linear growth during its youth, followed by an expo- Sequestration and Prevents Storm-Water Run-Off
nential-like transition into steady-state sequestration Figure 6a compares the number of live trees per year
during its maturity. Empirical studies confirm this under different maintenance scopes. Lower or no
non-linear growth pattern of trees (Muukkonen and maintenance scope leads to a decrease in the number
Mäkipää 2006). of surviving trees and an increase in the likelihood of

©2024 International Society of Arboriculture

124 Ozel and Petrovic: Green Urban Scenarios

the release of stored carbon stocks. Figures 6b, 6c, light background show i-Tree reported data, while
and 6d highlight the difference good maintenance can columns in the dark background indicate the results
make in avoiding carbon release, improving carbon from the computational experiments of GUS. This
sequestration, and preventing storm-water run-off. validation exercise selects the most common species
The current version of GUS estimates carbon in Glasgow, European ash (Fraxinus excelsior), and
release at an individual tree and site level. The model calculates (i) under canopy area using the crown
tracks the time-dependent carbon release process tak- width, (ii) leaf area index (LAI) dividing leaf area
ing into account differing release trajectories of dead (LA) by under-canopy area, (iii) bark area (BA) using
roots, standing dead trees, mulched parts, and parts trees’ diameter at breast height (DBH), and (iv) plant
that are removed from the site immediately. area index (PAI) taking into account leaf-off and
leaf-on seasons (Wang et al. 2008). The tree popula-
Result 6. Poorly Maintained Trees Release
tion sample is matched with hourly weather data from
Significant Amounts of Carbon Back into the
Glasgow and potential evapotranspiration is calcu-
Atmosphere
lated for each tree in each hour. Finally, the yearly
Controlled experiments show that poorly maintained
potential evapotranspiration (PET) for each tree is
trees or immediately removed dead tree trunks and
provided in the last column of the table, showing that
roots release significant amounts of carbon back into
the results match the outputs of i-Tree in the penulti-
the atmosphere. Although some of this carbon release
mate column.
is inevitable and necessary for soil regeneration and
For certain ecosystem services, such as flood pre-
biodiversity, good maintenance practices can avoid a
vention or heat island reduction, the location-specific
significant amount of unnecessary release, as shown
effect of urban forests can be optimised further by
in plot b of Figure 6.
combining insights from GUS with additional geo-
Result 7. As Cedar Trees Reach Maturity, They spatial data analysis.
Surpass European Ash in Their Ability to Alleviate The following section discusses current limitations,
Stormwater, Proving More Effective after a Span ongoing development efforts, and future pathways.
of 25 Years
Figure 7 compares the role of tree species on ecosys- DISCUSSION
tem services. The figure shows that European ash, a The work presented in this paper aims to serve as a
deciduous tree, can capture more carbon than cedar, framework and a tool for the research community,
an evergreen tree. However, as a cedar tree grows and as a software product for practitioners both in the
into maturity in the longer term, it becomes more public and private sectors. Given its overarching
effective in stormwater alleviation. In short, this sub- objective, its development process follows good sci-
set of controlled experiments indicates that species ence and software engineering practices.
composition significantly impacts the outcomes of eco-
system services. Therefore, the results suggest that Validation and Reusability
given the location and urgency, the decision-makers The model development process follows a recursive
should consider the species composition of urban for- cycle of verification, validation, and calibration steps
ests as a design parameter to optimise their impact. (Law and Kelton 1991). Model verification should
Good practices in complex system design and assure that the programming implementation of the
analysis require conducting systematic internal and conceptual model is correct. This process follows typ-
external validation exercises. The internal validation ical software debugging and testing exercises. Model
examines how the model responds to changes in validation checks whether the simulation model can
weather conditions and other model parameters. For accurately mimic real-world tree growth patterns.
instance, an increase in solar radiation and/or vapour Model calibration is an iterative process that adjusts
pressure deficit raises the potential evapotranspiration. model parameters to match the location and scenario-
In addition, the external validation compares results specific requirements.
with the outputs of other models in the literature. Biomass growth, carbon release, sequestration, leaf
Table 1 summarises water-related outcomes from the and bulk area estimation, evapotranspiration, and other
i-Tree report on Glasgow and the results from GUS impact assessment models are based on peer-reviewed
considering the same set of trees. Columns in the scientific literature. The source code and model

©2024 International Society of Arboriculture

Arboriculture & Urban Forestry 50(1): January 2024 125

Table 1. Validation: water impact of European ash (Fraxinus excelsior). The table shows the output for several selected trees,
while the entire table includes 109 trees provided by the i-Tree report. In addition, Δ(PET)i = PET(i-Tree)i − PET(GUS)i is calculated,
taking into account all 109 trees, and the following results are obtained: Δ(PET)mean = −0.195, Δ(PET)median = 0.181, along with
t-test statistics = −1.301 and P-value = 0.196, not rejecting the null hypothesis Δ(PET) = 0. PET (potential evapotranspiration);
DBH (diameter at breast height); BAI (bark area index); LAI (leaf area index); PAI (plant area index).

Leaf area DBH Height Crown Under-canopy PET (m3/ PET (m3/
Plot ID BAI LAI PAI
(m2) (cm) (m) width (m) area (m2) yr) – i-Tree yr) − GUS

131 36 2.4 8.5 10.0 1.0 0.8 2.7 3.1 5.8 0.1 0.1
198 26 2.0 7.0 11.5 1.0 0.8 2.4 2.5 4.9 0.1 0.1
37 1 9.1 7.5 6.0 2.0 3.1 0.2 2.9 3.0 0.3 0.4

… … … … … … … … … … … …

… … … … … … … … … … … …

198 20 131.8 19.5 14.0 6.0 28.3 0.1 4.7 4.8 4.1 4.0
131 74 12.2 8.8 10.0 2.0 3.1 0.4 3.9 4.2 0.4 0.4
11 3 116.4 16.6 10.0 6.0 28.3 0.1 4.1 4.2 3.7 3.7

specification documents are made open to enable transparency about these sources of uncertainty and
transparency, replicability of the analysis, and ease of how they affect measurements. Figure 9 depicts
adoption of the framework. potential sources of errors in the design, calibration,
The GUS framework modules are architected to and interpretation of modelling outputs. These uncer-
complement or extend the functionality of existing tainties have further implications on outcome-driven
tools, such as the i-Tree species tool. When the data valuations and financing strategies in planning and
from a granular tree detection and recognition facility maintaining green infrastructures. There are 3 poten-
feeds GUS, it functions as a digital twin version of the tial sources of divergence from actuality:
actual site. In such a use case, GUS serves as a virtual
Modelling: The models diverge from reality. As dis-
laboratory.
cussed earlier in this paper, continuity in verification,
validation, and calibration cycles may reduce the
Uncertainty and Transparency
divergence or help make the divergence explicit and
Limitations and errors are inherent to any modelling
quantifiable.
exercise on complex systems, which mandate

Figure 9. Potential sources of errors or uncertainties.

©2024 International Society of Arboriculture

126 Ozel and Petrovic: Green Urban Scenarios

Monitoring: The data points may not represent real- In addition to carbon and water-related modules
ity accurately enough. For example, human monitor- that are presented in this paper, impact estimation and
ing can have lower accuracy than a remote sensor. In prediction models on the other ecosystem services
some cases, remote sensor devices fail to capture will be made available as application programming
accurate enough estimates or inject additional errors. interface (API) endpoints. The access will enable
Recurrent data validation practices need to be inte- other researchers and practitioners to configure or
grated into impact assessment processes. monitor location-specific urban forest typologies and
gain insights into their effect on air pollution removal,
Association: An urban forest’s contribution to remov-
heat island reduction, energy-saving, and mental
ing carbon dioxide from the atmosphere can be fully
health improvements.
quantified. However, its contribution to other out-
Future extensions of GUS will keep following the-
come types, such as mental health, may not be effec-
oretical and practical guidelines of the augmented
tively isolated from other socio-economic factors.
collective intelligence (ACI) that have been devel-
For instance, to what extent fewer mental health cases
oped elsewhere for civic initiatives (Ozel and van der
can be only associated with residential proximity to
Hoog 2020). ACI accommodates trees, humans, and
the parks? In this case, decoupling the additionality of
other non-human actors and their interactions using
urban forests from the contribution of other underly-
artificial intelligence (AI). As containers of AI, agent-
ing sociodemographic factors requires extensive and
based models explore counterfactual scenarios, tak-
detailed empirical studies. Future extensions of GUS
ing data from the physical environment to simulate
aim to reduce this gap by serving as a computational
potential outcomes, which can be used to help decision-
laboratory to experiment and observe other sociode-
making in the real world. Thus, augmented collective
mographic effects. Its underlying agent-based para-
intelligence emerges from interacting with people,
digm enables it to scale its complexity by adding
technology, and nature in the physical world and the
human agents, considering mobility and housing
digital environment.
choices concerning their sociodemographic profiles.
The model implementation is open-sourced, along
CONCLUSION
with detailed model specifications (Ozel and Petrovic
This paper presented GUS, an innovative and practical
2022) and code documentation. It enables interroga-
scenario analysis framework for green infrastructures.
tion and replicability by other practitioners and
The paper analyses several urban forest typologies
researchers and hence, continuous improvement in
within Glasgow’s context, examining how different
calibration and interpretation, which in return, closes
maintenance strategies may affect tree population
the gap between the model and the reality.
dynamics as well as the provision of ecosystem
services.
Further Directions
The results show emergent behaviours can be dis-
Further extension of the GUS framework will main-
covered without explicitly being coded into the simu-
tain its modular architecture while adding new com-
lation. The analysis confirms that building and using
ponents and features. Modularity provides flexibility
agent-based models is an essential tool for understand-
in turning system components on and off depending
ing complex systems aspects of biological, ecological,
on the context and its required level of granularity
and urban planning issues.
and complexity.
A subset of the computational experiments pro-
A generic contagion model of infectious tree dis-
vided in the paper shows that poorly maintained trees
eases is under development. Once validated, it will be
or immediately removed dead tree trunks and roots
activated and calibrated for the urban forests where
release significant amounts of captured carbon.
there is a likelihood of ash dieback cases, a fungal
Although some of this carbon release is inevitable
disease of ash trees in Europe characterised by leaf
and is necessary for soil regeneration and biodiver-
loss and crown dieback in infected trees (Musolin et
sity, a large part of unnecessary release can be avoided
al. 2017). This feature should help tree officers to
by improving maintenance practices. Another subset
gain insight into contagion patterns and restore spa-
of the computational experiments indicates that spe-
tial compositions of urban trees with more resistant
cies composition significantly impacts the outcomes
species.

©2024 International Society of Arboriculture

Arboriculture & Urban Forestry 50(1): January 2024 127

of ecosystem services and that impact measurements Castro CV, Maidment DR. 2020. GIS preprocessing for Rapid
for each species need to be considered. For example, initialization of HEC-HMS hydrological basin models using
web-based data services. Environmental Modelling & Software.
as cedar trees reach maturity, they have been found to
130:104732. https://doi.org/10.1016/j.envsoft.2020.104732
surpass European ash in their ability to alleviate Coates KD, Canham CD, Beaudet M, Sachs DL, Messier C. 2003.
stormwater, proving more effective after a span of 25 Use of a spatially explicit individual-tree model (SORTIE/
years. Depending on the location and the urgency of BC) to explore the implications of patchiness in structurally
particular ecosystem services, this paper suggests that complex forests. Forest Ecology and Management. 186(1-3):
decision-makers should consider the species compo- 297-310. https://doi.org/10.1016/S0378-1127(03)00301-3
Dale VH, Doyle TW, Shugart HH. 1985. A comparison of tree
sition of urban forests as a design parameter. growth models. Ecological Modelling. 29(1-4):145-169.
Using the current version of GUS, practitioners https://doi.org/10.1016/0304-3800(85)90051-1
can carefully design a new urban forest or explore the DeVries R. 1987. A preliminary investigation of the growth and
impact of an existing one. The urban forest can encom- longevity of trees in Central Park [dissertation]. New Brunswick
pass various typologies, comprising a diverse range (NJ, USA): Rutgers University. 95 p.
Dimitrova A. 2021. Prague with a major tree-planting programme.
of species, while its growth and benefits can be simu-
Sofia (Bulgaria): TheMayor. [Updated 2021 January 4;
lated over any desired timeframe. The growth of the Accessed 2021 May 17]. https://www.themayor.eu/en/a/
urban forest considers the health of individual trees view/prague-with-a-major-tree-planting-programme-2337
and their influence on neighbouring trees, resulting in Dubbelboer J, Nikolic I, Jenkins K, Hall J. 2017. An agent-based
population growth through interactions between them. model of flood risk and insurance. Journal of Artificial Soci-
Practitioners can delve into each ecosystem service, eties and Social Simulation. 20(1):6. https://doi.org/10.18564/
jasss.3135
tailoring the level of granularity and complexity on a Epstein JM. 2008. Why model? Journal of Artificial Societies
project-by-project basis. Additionally, the modular and Social Simulation. 11(4):12. https://www.jasss.org/11/4/
nature of the framework facilitates expansion of eco- 12.html
system services and integration with other existing Epstein JM, Axtell R. 1996. Growing artificial societies: Social
tools in the field. science from the bottom up. Washington, DC (USA): The
Brookings Institution. 228 p. https://doi.org/10.7551/
The science-based scenario design and impact analy-
mitpress/3374.001.0001
sis models presented in this paper can also be extended Grimm V, Ayllon D, Railsback SF. 2017. Next-generation
as an open-source software toolset for researchers, individual-based models integrate biodiversity and ecosystems:
practitioners, and policymakers. Its use also puts forth Yes we can, and yes we must. Ecosystems. 20(2):229-236.
an economic case where the benefits of urban forests https://doi.org/10.1007/s10021-016-0071-2
far exceed the costs, making a compelling argument Gode DK, Sunder S. 1993. Allocative efficiency of markets with
zero intelligence traders: Market as a partial substitute for
for maintaining a green infrastructure with carbon, individual rationality. Journal of Political Economy. 101(1):119-
water, health, energy, and economic and social bene- 137. https://doi.org/10.1086/261868
fits as a basis for co-investments. Grueters U, Ibrahim MR, Satyanarayana B, Dahdouh-Guebas F.
2019. Individual-based modeling of mangrove forest growth:
LITERATURE CITED MesoFON—Recent calibration and future direction. Estuarine
Berger U, Hildenbrandt H. 2000. A new approach to spatially Coastal and Shelf Science. 227:106302. https://doi.org/10
explicit modelling of forest dynamics: Spacing, ageing and .1016/j.ecss.2019.106302
neighbourhood competition of mangrove trees. Ecological Grueters U, Ibrahim MR, Schmidt H, Tiebel K, Horn H, Pranchai
Modelling. 132(3):287-302. https://doi.org/10.1016/S0304 A, Vovides AG, Vogt J, Otero V, Satyanarayana B, Dahdouh-
-3800(00)00298-2 Guebas F. 2021. Stable coexistence in a field-calibrated
Bittebiere AK, Mony C, Clement B, and Garbey M. 2012. Mod- individual-based model of mangrove forest dynamics caused
eling competition between plants using an Individual Based by inter-specific crown plasticity. Forests. 12(7):955. https://
Model: Methods and effects on the growth of two species with doi.org/10.3390/f12070955
contrasted growth forms. Ecological Modelling. 234:38-50. Hirabayashi S. 2013. i-Tree eco precipitation interception model
https://doi.org/10.1016/j.ecolmodel.2011.05.028 descriptions. 21 p. https://www.itreetools.org/documents/61/
Botkin DB. 1993. Forest dynamics: An ecological model. New iTree_Eco_Precipitation_Interception_Model_Descriptions.pdf
York (NY, USA): Oxford University Press, Inc. 309 p. Karnouskos S, de Holanda TN. 2009. Simulation of a smart grid
Canham CD, Finzi AC, Pacala SW, Burbank DH. 1994. Causes city with software agents. In: Al-Dabass D, Katsikas S, Koukos
and consequences of resource heterogeneity in forests: Inter- I, Abraham A, Zobel R, editors. EMS ’09: Proceedings of the
specific variation in light transmission by canopy trees. 2009 third UKSim European symposium on computer model-
Canadian Journal of Forest Research. 24(2):337-349. https:// ing and simulation; 2009 November 25–27; Athens (Greece).
doi.org/10.1139/x94-046 Washington, DC (USA): IEEE Computer Society. p. 424-429.
https://doi.org/10.1109/EMS.2009.53

©2024 International Society of Arboriculture

128 Ozel and Petrovic: Green Urban Scenarios

Kirman A. 1993. Ants, rationality, and recruitment. Quarterly Ozel B, Petrovic M. 2022. Green urban scenarios: Model speci-
Journal of Economics. 108(1):137-156. https://doi.org/10 fication document. [Updated 2022 October 28]. https://github
.2307/2118498 .com/lucidmindsai/gus/tree/main/docs
Law AM, Kelton WD. 1991. Simulation modeling and analysis. Ozel B, van der Hoog S. 2020. Augmented collective intelligence
2nd Ed. New York (USA): McGraw-Hill. 672 p. a framework for self-organizing community actions in
Lin J, Kroll CN, Nowak DJ. 2020. Ecosystem service-based response to climate change. [Accessed 2022 May 28]. https://
sensitivity analyses of i-Tree Eco. Arboriculture & Urban lucidminds.ai
Forestry. 46(4):287-306. https://doi.org/10.48044/jauf.2020 Raberto M, Ozel B, Ponta L, Teglio A, Cincotti S. 2019. From
.021 financial instability to green finance: The role of banking and
Lorimer CG, Frelich LE, Nordheim EV. 1992. Forest-tree growth credit market regulation in the Eurace model. Journal of
rates and probability of gap origin: A reply to clark. Ecology. Evolutionary Economics. 29(1):429-465. https://doi.org/10
73(3):1124-1128. https://doi.org/10.2307/1940185 .1007/s00191-018-0568-2
Marwick B, Krishnamoorthy K. 2019. Cvequality: Tests for the ReTreevAIable Project. 2022. Research and Development coop-
equality of coefficients of variation from multiple groups. R eration project co-funded by the Dutch Enterprise Agency
software package version 0.1.3. https://cran.r-project.org/ (RVO). https://www.rvo.nl
web/packages/cvequality/readme/README.html Rumble H, Rogers K, Doick K, Albertini A, Hutchings T. 2015.
Metropolis N, Ulam S. 1949. The Monte Carlo method. Journal Valuing urban trees in Glasgow: Assessing the ecosystem
of the American Statistical Association. 44(247):335-341. services of Glasgow’s urban forest: A technical report. Dumfries
https://doi.org/10.1080/01621459.1949.10483310 (Scotland): The Research Agency of the Forestry Commission.
Musolin DL, Selikhovkin AV, Shabunin DA, Zviagintsev VB, http://www.urbantreecover.org/wp-content/uploads/2015/01/
Baranchikov YN. 2017. Between ash dieback and emerald FR_Doick_GlasgowItreereportFINAL.pdf
ash borer: Two Asian invaders in Russia and the future of ash Savaglio C, Ganzha M, Paprzycki M, Bădică C, Ivanović M,
in Europe. Baltic Forestry. 23(1):316-333. Fortino G. 2020. Agent-based internet of things: State-of-the
Muukkonen P, Mäkipää R. 2006. Biomass equations for European -art and research challenges. Future Generation Computer
trees: Addendum. Silva Fennica. 40(4):475. https://doi.org/ Systems. 102:1038-1053. https://doi.org/10.1016/j.future.2019
10.14214/sf.475 .09.016
Newman MEJ. 2011. Complex systems: A survey. American Schelling TC. 1978. Micromotives and macrobehavior. New York
Journal of Physics. 79(8):800-810. https://doi.org/10.1119/1 (NY, USA): W.W. Norton and Co. 252 p.
.3590372 Smith W, Shifley S. 1984. Diameter growth, survival, and volume
Nowak DJ. 1994. Atmospheric carbon dioxide reduction by Chi- estimates for trees in Indiana and Illinois. St. Paul (MN, USA):
cago’s urban forest. In: McPherson EG, Nowak DJ, Rowntree USDA, Forest Service, North Central Forest Experiment Sta-
RA, editors. Chicago’s urban forest ecosystem: Results of the tion. Research Paper NC-257. https://doi.org/10.2737/NC-RP
Chicago urban forest climate project. Madison (WI, USA): -257
USDA, Forest Service, Northeastern Forest Experimental Snowden D. 2000. The social ecology of knowledge management.
Station. General Technical Report NE-186. p. 83-94. In: Despres C, Chauvel D, editors. Knowledge horizons.
Nowak DJ. 1996. Estimating leaf area and leaf biomass of open- Boston (MA, USA): Butterworth-Heinemann. p. 237-265.
grown deciduous urban trees. Forest Science. 42(4):504-507. https://doi.org/10.1016/B978-0-7506-7247-4.50015-X
https://doi.org/10.1093/forestscience/42.4.504 Wang J, Endreny TA, Nowak DJ. 2008. Mechanistic simulation
Nowak DJ. 2020. Understanding i-Tree: Summary of programs of tree effects in an urban water balance model1. Journal of
and methods. Madison (WI, USA): USDA Forest Service, the American Water Resources Association. 44(1):75-85.
Northern Research Station. General Technical Report NRS-200. https://doi.org/10.1111/j.1752-1688.2007.00139.x
100 p. https://doi.org/10.2737/NRS-GTR-200 Zhang B, DeAngelis DL. 2020. An overview of agent-based
Nowak DJ, Crane DE. 2002. Carbon storage and sequestration by models in plant biology and ecology. Annals of Botany.
urban trees in the USA. Environmental Pollution. 116(3):381- 126(4):539-57. https://doi.org/10.1093/aob/mcaa043
89. https://doi.org/10.1016/S0269-7491(01)00214-7
Nowak DJ, Crane DE, Stevens JC, Hoehn RE, Walton JT. 2008. ACKNOWLEDGEMENTS
A ground-based method of assessing urban forest structure This work has received partial support from the Google.org
and ecosystem services. Arboriculture & Urban Forestry. Impact Challenge on Climate through the TreesAI project and the
34(6):347-358. https://doi.org/10.48044/jauf.2008.048 Netherlands Enterprise Agency (RVO). We would like to extend
Nowak DJ, Crane DE, Stevens JC, Ibarra M. 2002. Brooklyn’s our gratitude to our dedicated team members at Lucidminds AI
urban forest. Newtown Square (PA, USA): USDA, Forest and Dark Matter Labs. Oguzhan Yayla made significant contribu-
Service, Northeastern Research Station. General Technical tions to the software architecture and implementation of GUS,
Report NE-290. 107 p. https://doi.org/10 .2737/NE-GTR-290 open-source package distribution, and maintenance. Jake Doran
New South Wales Government. 2020. Greening our city premier’s has been contributing to refactoring existing modules and new
priority. [Updated 2020 May 17; Accessed 2022 May 17]. extensions. Cynthia Mergel and Gurden Batra played vital roles
https://www.dpie.nsw.gov.au/premiers-priorities/greening-our in UI design and implementation. Axel Nilsson contributed to
-city data curation and preparation. Seowoo Nam provided illustra-
tions for Figures 1, 2, and 9, while Cynthia Mergel contributed

©2024 International Society of Arboriculture

Arboriculture & Urban Forestry 50(1): January 2024 129

illustrations for Figures 3 and 4. Arianna Smaron contributed Zusammenfassung. Hintergrund: Bäume sind ein wichtiger
illustrations for Figure 5. Special thanks go to Anne Mul for her Bestandteil der städtischen Infrastruktur. Städte auf der ganzen
copywriting support. Welt verpflichten sich aufgrund von Netto-Null-Zielen zur Auf-
forstung; ihre Verwirklichung erfordert jedoch einen umfassen-
den Rahmen, der Wissenschaft, Politik und Praxis miteinander
Bulent Ozel (corresponding author) verbindet. Methoden: In diesem Beitrag wird das Rahmenwerk
Lucidminds AI Green Urban Scenarios (GUS) für die Gestaltung und Überwa-
Amsterdam, Netherlands chung grüner Infrastrukturen vorgestellt. GUS berücksichtigt
[email protected] Wetter, Pflege, Baumarten, Krankheiten und die räumliche Ver-
Marko Petrovic teilung von Bäumen, um deren Auswirkungen zu prognostizieren.
Lucidminds AI Der Rahmen nutzt agentenbasierte Modellierung (ABM) und ein
Amsterdam, Netherlands Simulationsparadigma, um grüne Infrastrukturen in den ökologi-
Department of Economic Analysis schen, räumlichen, wirtschaftlichen und sozialen Kontext einer
University of Valencia Stadt zu integrieren. ABM ermöglicht die Erstellung digitaler
Valencia, Spain Zwillinge für städtische Ökosysteme auf jeder Granularitätsebene,
[email protected] einschließlich einzelner Bäume, um deren zukünftige Entwick-
lung genau vorherzusagen. Die digitale Darstellung von Bäumen
wird anhand einer Kombination von Datensätzen wie Erdbeob-
Conflicts of Interest: achtungen aus dem Weltraum, Straßenansichten, Feldstudien und
The authors reported no conflicts of interest. qualitativen Beschreibungen von Typologien innerhalb bestehen-
der und zukünftiger Projekte erstellt. Durch maschinelles Lernen
und statistische Modelle werden Biomassewachstumsmuster und
Résumé. Contexte: Les arbres sont un élément essentiel de l’in-
Kohlenstofffreisetzungspläne kalibriert. Ergebnisse: In dem Auf-
frastructure urbaine. Les villes du monde entier s’engagent à
satz werden verschiedene Grünflächentypologien untersucht und
atteindre des objectifs de boisement en raison des cibles de zéro
mehrere hypothetische Szenarien auf der Grundlage der städti-
émission nette de carbone. Toutefois, la réalisation de ces objec-
schen Wälder von Glasgow simuliert. Es zeigt die Entstehung von
tifs nécessite un cadre global combinant la science, la politique et
Heterogenitätsmerkmalen der Wälder aufgrund von Interaktionen
la pratique. Méthodes: L’article présente le cadre de Scénarios
zwischen den Bäumen. Der Wachstumsverlauf von Bäumen weist
Urbains Verts (SUV) pour la conception et le suivi des infrastruc-
eine nichtlineare Übergangsphase hin zu einem stabilen Wachs-
tures vertes. Les SUV tient compte des conditions météorolo-
tum im Reifestadium auf. Eine geringere Pflege verschlechtert
giques, de l’entretien, des espèces d’arbres, des maladies et de la
den Gesundheitszustand der Bäume, was zu einer geringeren
répartition spatiale des arbres afin de prévoir leur impact. Le cadre
Überlebensrate und erhöhten CO2-Emissionen führt, während die
utilise une modélisation basée sur les agents (MBA) et un para-
Fähigkeit zur Minderung des Regenwassers je nach Baumart
digme de simulation pour intégrer l’infrastructure verte dans le
unterschiedlich sein kann. Schlussfolgerungen: Der Beitrag zeigt,
contexte écologique, spatial, économique et social d’une ville. La
wie GUS die Politik und die Pflege von städtischen Wäldern mit
MBA permet de créer des jumeaux numériques pour les écosys-
ökologischen, sozialen und wirtschaftlichen Vorteilen erleichtern
tèmes urbains à tous les niveaux de granularité, incluant les arbres
kann.
individuels, afin de prédire avec précision leurs développements
Resumen. Antecedentes: Los árboles son una parte funda-
futurs. La représentation numérique des arbres est créée à l’aide
mental de la infraestructura urbana. Las ciudades de todo el mundo
d’une combinaison d’ensembles de données tels que des observa-
se están comprometiendo a objetivos de forestación debido a los
tions terrestres depuis l’espace, des images de l’outil Street View,
objetivos de cero emisiones netas. Sin embargo, su realización
des relevés de terrain et des descriptions qualitatives de typolo-
requiere un marco integral que combine la ciencia, la política y la
gies au sein de projets actuels et futurs. L’apprentissage automa-
práctica. Métodos: El artículo presenta el marco de los Escenarios
tique et les modèles statistiques permettent de calibrer des
Urbanos Verdes (GUS) para el diseño y seguimiento de infrae-
modèles de croissance de la biomasse et des schémas de libéra-
structuras verdes. GUS considera el clima, el mantenimiento, las
tion du carbone. Résultats: L’article examine diverses typologies
especies arbóreas, las enfermedades y la distribución espacial de
d’espaces verts, en simulant plusieurs scénarios hypothétiques
los árboles para pronosticar sus impactos. El marco utiliza el
basés sur les forêts urbaines de Glasgow. Elle montre l’émer-
modelado basado en agentes (ABM) y el paradigma de simu-
gence des caractéristiques d’hétérogénéité des forêts dues aux
lación para integrar la infraestructura verde en el contexto
interactions entre les arbres. La projection de croissance des
ecológico, espacial, económico y social de una ciudad. ABM per-
arbres présente une phase de transition non linéaire vers une
mite la creación de gemelos digitales para ecosistemas urbanos
croissance stable à maturité. La réduction des soins d’entretien
en cualquier nivel de granularidad, incluidos árboles individuales,
détériore la santé des arbres, ce qui entraîne une baisse du taux de
para predecir con precisión sus trayectorias futuras. La represent-
survie et une augmentation des émissions de CO2, tandis que la
ación digital de los árboles se crea utilizando una combinación de
capacité d’atténuation des eaux pluviales peut varier selon les
conjuntos de datos, como observaciones de la Tierra desde el
espèces. Conclusions: L’article démontre comment les SUV peuvent
espacio, imágenes de Street View, estudios de campo y descrip-
favoriser les politiques et l’entretien des forêts urbaines en lien
ciones cualitativas de tipologías dentro de proyectos existentes y
avec les avantages environnementaux, sociaux et économiques.
futuros. El aprendizaje automático y los modelos estadísticos

©2024 International Society of Arboriculture

130 Ozel and Petrovic: Green Urban Scenarios

calibran los patrones de crecimiento de la biomasa y los esquemas

de liberación de carbono. Resultados: El artículo examina varias
tipologías de áreas verdes, simulando varios escenarios hipotéti-
cos basados en los bosques urbanos de Glasgow. Presenta la apa-
rición de características de heterogeneidad de los bosques debido
a las interacciones entre árboles. La trayectoria de crecimiento de
los árboles tiene una fase de transición no lineal hacia un crec-
imiento estable en su madurez. La reducción del mantenimiento
deteriora la salud de los árboles, lo que conduce a una menor tasa
de supervivencia y a un aumento de las emisiones de CO2, mien-
tras que la capacidad de alivio de las aguas pluviales puede diferir
entre especies. Conclusiones: El artículo demuestra cómo GUS
puede facilitar las políticas y el mantenimiento de los bosques
urbanos con beneficios ambientales, sociales y económicos.

©2024 International Society of Arboriculture