Back to results

Democritus University of Thrace (DUTH)

Ανάπτυξη πληροφοριακού συστήματος μη κατευθυνόμενων αλγορίθμων και υβριδικών προτύπων στην εκτίμηση περιβαλλοντικών κινδύνων

Abstract

dc:description

The beginning of the 21st century is accompanied by enormous environmental challenges that already affect and have a catalytic effect not only on biodiversity but also on everyday life of modern man. Aim of this PhD research: The aim of this Ph.D. thesis is to model and predict the potential impacts of climate change on various areas such as: Atmospheric Pollution, increase of extreme weather incidents, extensive forest fires, introduction and installation of Invasive Species in new environments, increase of Cardiovascular and Respiratory incidents in the urban area of Thessaloniki, Greece. This research employs Hybrid Soft Computing methodologies and Unsupervised Machine Learning algorithms. Climate change: One of the most serious impacts of climate change is the dramatic degradation of air quality in urban centers. The degradation of air quality is interwoven not only with the modern way of life, but also with climate change. In particular, temperature, humidity and hours of sunshine as meteorological parameters are determinants contributing to the composition of the smog in the winter and to the Photochemical cloud in the summer, as opposed to the rainfall and wind that contribute to the diffusion and dispersion of pollutants. People living in air polluted areas, are facing cardiovascular and respiratory problems, and moreover they are at serious risk of developing myocardial infarction and stroke. At the same time, long-term exposure to high concentrations of pollutants can cause various forms of cancer. Air Pollution of urban and industrial areas in Greece is a major environmental problem, as apart from Climate Change it is also associated with urban population growth and increase of industrial production trends. The recent economic crisis in the country has led to an increase of the PM10 and PM2.5 particulates concentration, due to the increased combustion of solid fuels. Natural Disasters-Extreme weather phenomena Forest fires are one of the most important natural phenomena and are indissolubly linked to the Greek ecosystems during the summer months. Each factor has its own effect, while risk of forest fires is under the cumulative effect of all the involved parameters that contribute to their initiation and more over to their expantion. The most important factors are the humidity of the matter (how dry is the herbaceous vegetation) the wind speed, the density of the matter and the topography (Kailidis and Karanikola, 2004). Forest fires destroy properties, threaten human lives, degrade the environment, disrupt the ecosystems’ balance, erode the soil which results in more floods, affect the microclimate and destroy the flora and fauna ecosystems. Hail as an extreme seasonal phenomenon causes serious damage to plant production. The possibility of reporting the occurrence of hail in a rural area has multiple benefits. Farmers will be able to get informed and notified in time to protect their crops, and Weather Modification Services will receive valid information to directly bomb the clouds with silver iodine. Also, immediate notification of the occurrence of hail in a rural area plays a key role in cases where weather radar is out of order. In particular, this PhD research focuses on the development of an innovative Hybrid Model and a corresponding Computational Intelligence Information System, based exclusively on Hybrid approaches of Computational Intelligence and Unsupervised Machine Learning Algorithms (UMLA). The produced system consists of several subsystems that model multifactor environmental problems. The proposed information subsystems use Statistical or Hybrid Artificial Intelligence methods to study the current state of the conditions and interrelationships of the parameters constituting an environmental problem. Review of the methodology. The statistical part comprises of Fuzzy Chi-Square Test (FCST) and Correlation Analysis (CA), whereas projection of the environmental problem to the future was achieved by using Fuzzy Cognitive Maps (FCM) and by simulating the Greek climate model till 2100. The climate model simulation was done by employing Coupled Atmospheric-Ocean General Circulation Models, which model general circulation through the whole planet (Mearns et al., 2001). These models accept different data as an input, such as carbon dioxide (CO2), solar radiation and evapotranspiration. These models run on computers, simulating climate change, and their output is mainly in the medium to long-term evolution of temperature and rainfall, as well as other variables in any part of the globe (Denman et al., 2007; Καψωμενάκης et al., 2011).In order to simulate smaller spatial scale phenomena at a high resolution in a prefecture or in a country, the Regional Climate Models are used in which climate information is input either by Global Atmospheric General circulation Models, or by Statistical subscaling methods, or by dynamic subscaling. The proposed subsystems have modeled various environmental hazards namely: (Air Pollution, Forest Fires, Extreme Weather Phenomena, Introduction and Invasion of Invasive Species in New Regions). These problems are relevant to the Greek space and are expected to have severe impacts based on climate change in the coming decades. Modeling of the above environmental risks by the individual subsystems was achieved by considering simulations of the most timely Climate Models and Climate Change scenarios that were produced under the framework of the program Coupled Model Intercomparison Project, Phase 5 (CMIP5) till 2100.(http://sdwebx.worldbank.org/climateportal/index.cfm?page=country_future_climate&ThisRegion=Europe&ThisCcode=GRC). Indices production Through the simulations, it is possible to find the medium- and long-term forecast of the values of all the analyzed parameters that constitute an environmental problem, by changing the values of the meteorological parameters (temperature and precipitation). Fuzzy Inference Systems (FIS) Mamdani & Takagi-Sugeno-Kang (TSK) were implemented in a MATLAB environment. Their advantage is that they provide the ability to incorporate a number of Fuzzy Rules using weightings of significance for the synthesis of individual indices in order to model an environmental problem. Sub-indices can produce a Unified Index through the Fuzzy T-Norms Conjunctions (FTNC) procedure of Fuzzy T-Norms.The Unified index may represent the overall risk of the problem under consideration. The intelligence of the individual subsystems is derived from the MLM2 algorithms and specifically from: a) Self Organizing Maps-SOM, b) the Expectation Maximization-EM method, c) the method of the Sequential Information Bottleneck-SIB, d) the Simple K-Means Algorithm-SKMA algorithm for estimating the risk of Environmental Problems. The purpose of Unsupervised machine learning algorithms (UMLA) is to group and rank clusters by evaluating parameter values that determine to a greater or lesser extent to an environmental problem. The cluster analysis extracts important information about how the values of the parameters in each cluster affect the problem. The proposed innovative subsystems use UMLA and hybrid patterns, contributing to the production of new knowledge to further improve and extend the proposed methodologies. Soft Computing (SF) approaches are particularly innovative and powerful, very important and have been used by many researchers internationally. Innovation - Gaps covered by this PhD researchThe gap in this literature lies in the lack of an integrated approach and investigation of environmental risks rather than fragmented. In the literature, there is not yet an integrated approach to developing robust and reliable Intelligent Hybrid Patterns, which will work by combining Methods of Fuzzy Logic (FL) and Artificial Neural Networks (ANN) (such as Fuzzy Cognitive Maps and Fuzzy Chi-Square Test) with the Unsupervised machine learning algorithms for the purpose of modeling and projecting in the future the evolution and impact of environmental risks. The innovation of this PhD research is based on the design, development and integration under a common umbrella, innovative Hybrid Intelligent Patterns, while developing a corresponding integrated information system that will provide assessment, analysis and projection in the future of the evolution and impact of major environmental risks-challenges, quality of life and biodiversity. Aims-benefits. This research utilizes Computational Intelligence (CI) methods, in particular Unsupervised Machine Learning algorithms and Soft Computing Approaches to optimize the reliability and performance of models that are required to assess the impact of climate change on public health and the environment. In addition, the system is a predictive real-time information tool in depth of a few days, with the ultimate goal of informing local authorities and strengthening civil protection on the basis of its assessments. This timely update can lead to valid decision-making protection of the inhabitants of an area and especially of vulnerable groups. In particular, it can be applied to problems that affect our post-industrial societies, such as Atmospheric Pollution, increasing the incidence of extreme weather events, forest fires, introducing and installing Invasive Species in new areas requiring immediate readiness and intervention of statemechanisms.Along with estimates of the evolution of environmental risks based on climate scenarios, it is possible to design a long-term plan to reduce the impact of climate change on public health, forest ecosystems and the environment. In this dissertation, the application of Computational Intelligence methods demonstrates that they can assess an environmental problem of a wider region or a county and project its evolution in the future after having studied the existing state of correlation of the parameters that make up the problem.

Degree

thesis:*
Grantor dc:publisher
Democritus University of Thrace (DUTH)
Year dc:date
2018

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Anezakis, Vardis-Dimitrios
  • Ανεζάκης, Βαρδής-Δημήτριος

Subjects

dc:subject × 18

Rights

dc:rights
Statement dc:rights
  • BY_NC_ND
Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/44421

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Anezakis, Vardis-Dimitrios; Ανεζάκης, Βαρδής-Δημήτριος. Ανάπτυξη πληροφοριακού συστήματος μη κατευθυνόμενων αλγορίθμων και υβριδικών προτύπων στην εκτίμηση περιβαλλοντικών κινδύνων. Democritus University of Thrace (DUTH), 2018. http://hdl.handle.net/10442/hedi/44421