Agent-based modelling of interactions between air pollutants and greenery using a case study of Yerevan, Armenia

Highlights

• We developed an original agent-based model of the interactions between air pollutants and greenery.

• A decision-making system for greenery planning was developed for Yerevan, Armenia.

• Two bi-objective optimisation problems were suggested and solved for reducing the air pollution concentration.

• The Pareto optimal solutions for urban greenery were computed through a genetic optimisation algorithm.

Abstract

Urban greenery such as trees can effectively reduce air pollution in a natural and eco-friendly way. However, how to spatially locate and arrange greenery in an optimal way remains as a challenging task. We developed an agent-based model of air pollution dynamics to support the optimal allocation and configuration of tree clusters in a city. The Pareto optimal solutions for greenery in the city were computed using the suggested heuristic optimisation algorithm, considering the complex absorptive-diffusive interactions between agent-trees (tree clusters) and air pollutants produced by agent-enterprises (factories) and agent-vehicles (car clusters) located in the city. We applied and tested the model with empirical data in Yerevan, Armenia, and successfully found the optimal strategy under the budget constraint: planting various types of trees around kindergartens and emission sources.

Introduction

As is known, many countries and cities currently face the problem of increasing air pollution caused by industrial activities and the increasing number of vehicles. Previous research (e.g., Brunekreef and Holgate, 2002) has proven the strictly negative influence of air pollutants (organic and inorganic dust, carbon monoxide (CO), sulphur dioxide (SO2), nitrogen oxides (NOx), volatile organic compounds (VOCs), heavy metals, etc.) on human health. Air pollution mainly affects respiratory diseases such as asthma, allergy, chronical obstructive pulmonary diseases, lung cancer, cardiovascular diseases and skin diseases. In addition, infants and children are especially vulnerable to the influence of air pollution.

There are known approaches to reducing air pollution using natural-based solutions (NSB); in particular, urban greenery (e.g., Nesshöver et al., 2017). Among research concerning this topic can be highlighted a work (Fuiii et al., 2005) which investigated air pollutant removal by plant absorption and adsorption in Japan. This research yielded the important conclusion that the efficiency of plant absorption depends on the photosynthesis actively performed in the spring and summer. Thus, the contribution of NBS to air pollution removal can be significantly improved in countries having better climate characteristics—e.g., a higher number of sunny days during the year, as in the Republic of Armenia (having approximately 300 sunny days per year in the city of Yerevan).

Thus, the internal chemical processes of absorption and adsorption of different kinds of air pollutants by different kinds of plants have been very well investigated on the micro level (e.g., Fuiii et al., 2005; Omasa et al., 2002; Bell and Treshow, 2002). The results raise the flowing questions: what is the positive impact of urban greenery in reducing air pollution; how should the best kinds of trees be chosen and allocated in a city to protect the human population; and which greenery strategies are better, taking into account limited urban budgets?

Solving such problems are impossible without modelling air pollution dynamics and forecasting the movement of air pollutants in the urban atmosphere. There are three main groups of models that can be used to model air pollution dynamics.

The first is based on a statistical approach (e.g., Bolzern et al., 1982; Harnandez et al., 1992) and intended mostly for analysis of historical data on pollutant concentrations and short-term forecasting. The second group consists of deterministic and deterministic-statistical models (e.g., Lamb and Seinfeld, 1973; Genikhovich, 2004). Pure statistical approaches to the investigation of air pollution dynamics are mainly simplified, and their applications are limited. The main reason for the limitation is that they ignore many factors with important influences on the dynamics of air pollution concentration, such as plants located on paths of pollutant masses. Such plants are natural barriers implementing absorption effects. Some environmental factors, e.g., changing wind directions and collusions between trees and landscape objects, cannot be taken into account using statistical methods.

The last group of methods is based on modelling the dynamics of emission plumes using Gaussian dispersion models (e.g., Turner, 1994; Beychok, 2005) and are intended for the investigation of air diffusion processes, taking into account different climate factors. Such models are the most realistic, but they are characterised by significant computation complexity.

However, such models are not always capable of representing the roles of different sources and sinks of air pollution, such as cars, factories and trees, along with the multiple interactions between them. Specifically, agent-based models can provide a better understanding of the role of trees as sinks of air pollution than provided by diffusion reaction models and, thus, support the implementation of informed public policies to reduce pollution effects.

There are other well-known and simpler models of air pollution dynamics such as Lagrangian models, Eulerian models, chemically reactive compound models and analytical models (e.g., Seinfeld, 1975; Zak, 1983; Seinfeld, 1986; Barbulescu and Barbes, 2017; Ridley, 2017; Corani and Scanagatta, 2016). Such models focus on forecasting trajectories of moving air pollutants without taking into account the impact of interaction with other landscape objects (e.g., trees, buildings). Nevertheless, some studies have been devoted to researching air pollution dynamics taking into account some elements of the urban landscape (e.g., Turner, 1964; Finzi and Tebaldi, 1982; Santiago et al., 2017).

In contrast, the agent-based simulation approach suggested here for modelling air pollution dynamics involving interaction with landscape objects, particularly with tree clusters, allows the dimensionality of the considered problem to be reduced significantly. This ensures that the advantages of deterministic models are maintained.

The idea of combining computational air fluid dynamics (Abbot and Basco, 1989) and agent-based modelling has already been suggested elsewhere (Epstein et al., 2011). However, the agent-based model of air pollution dynamics using interactions with tree clusters having individual characteristics, such as the geometry of planting (e.g., simple circle, arithmetic spiral, double circle), the type of tree (e.g., popular, oak, maple), and the distance between the nearest tree clusters is suggested herein for the first time. This method aims to reduce the air pollution concentration in cities.

Many studies have confirmed the positive impact of greenery in reducing air pollution in urban areas (e.g., Jim and Chen, 2008; Wong et al., 2009). However, there are not any systems that determine the optimal number of tree clusters, their location coordinates and planting geometry, the best kind of trees and other parameters that could be computed to minimise the average daily concentration of air pollution if the urban greenery budget is limited. This omission has been due to the overly high computational complexity needed for such tasks, which involve large-scale multi-objective optimisation problems.

To achieve this goal, we have combined computational air pollution dynamics with simulations of the ecological behaviour of different urban agents, such as agent-enterprises (factories), agent-vehicles (car clusters) and agent-trees (tree clusters). Moreover, we have used a special multi-objective genetic algorithm (Akopov and Hevencev, 2013) aggregated with the developed simulation through objective functions to determine the Pareto optimal solutions for greenery in a city.

There is a line of research to develop agent-based models and multi-objective systems to enhance the management of complex environmental systems (Hadka and Reed, 2015; Sun et al., 2016; Tesfatsion et al., 2017; West et al., 2018).

In the suggested model, trees and air pollutants are considered interactional agents with individual characteristics. This means that different tree clusters interacting with heterogeneous air pollutants have dissimilar absorptive-diffusive characteristics and different influences on the daily air pollution concentration.

Agent-based modelling approaches in combination with other simulation methods have been described elsewhere (Papaleonidas and Iliadis, 2012; Letcher et al., 2013; Zenonos et al., 2015; Vallejo et al., 2015; Ridley, 2017; Akopov et al., 2017). These studies have identified certain associated advantages for ecological modelling, the most important of which is the possibility of modelling the dynamics of agent interactions without the need to develop complex analytical models.

This study is focused on designing an agent-based model to determine the best ecological trade-offs for urban greenery. The case study of the city of Yerevan, Armenia will be presented. The main purpose behind the development of the original decision-making system was to analyse effective scenarios for greenery allocation under the budget constraint in the city of Yerevan, Armenia to reduce the average daily pollution concentration.

The suggested system can be applied to identify optimal greenery strategies to reduce air pollution concentrations in other urban areas and regions if appropriate data are available. In addition, the developed approach improves the precision of the model by including additional characteristics, such as the prevailing wind direction, wind velocity, air pollution intensities generated by agent-enterprises and agent-vehicles.