Geri Dön

Çoklu afetlerde yapay zeka yaklaşımı

Artificial intelligence approach in multi-disasters

PDF İndir

  1. Tez No: 997344
  2. Yazar: SANEM ÖZTÜRK
  3. Danışmanlar: DOÇ. DR. TİMUR TEZEL
  4. Tez Türü: Yüksek Lisans
  5. Konular: Meteoroloji, Meteorology
  6. Anahtar Kelimeler: Afet yönetimi, İklim değişikliği, Disaster management, Climate change
  7. Yıl: 2026
  8. Dil: Türkçe
  9. Üniversite: Sakarya Üniversitesi
  10. Enstitü: Fen Bilimleri Enstitüsü
  11. Ana Bilim Dalı: Afet Yönetimi Ana Bilim Dalı
  12. Bilim Dalı: Belirtilmemiş.
  13. Sayfa Sayısı: Belirtilmemiş.

Özet

Bu tez çalışması, iklim değişikliğinin atmosferik süreçler üzerindeki yapısal etkilerini“çoklu tehlike”(multi-hazard) perspektifinden incelemekte ve Türkiye'nin bu süreçteki uyum kapasitesini yapay zekâ tabanlı erken uyarı sistemleri (AI-EWS) ekseninde analiz etmektedir. Çalışma kapsamında, 1980-2025 dönemini kapsayan küresel, doğa kaynaklı meteorolojik afet eğilimleri EM-DAT ve IPCC verileri ışığında ele alınarak, afetlerin yalnızca frekansının değil, aynı zamanda fiziksel karakterinin de evrildiği saptanmıştır. 193 sayfalık bu araştırma, teorik çerçeve, küresel ülke karşılaştırmaları ve Türkiye'nin kurumsal altyapısına yönelik saha analizlerinden oluşmaktadır. Araştırmanın kuramsal bölümünde, meteorolojik afetlerin geleneksel kriz yönetiminden bütünleşik afet risk yönetimine geçiş süreci değerlendirilmiş,“bileşik olaylar”(compound events) ve“zincirleme tehlikeler”(cascading hazards) gibi modern afet kavramları tezin merkezine yerleştirilmiştir. Karşılaştırmalı analiz aşamasında, ABD, Japonya, Hollanda ve Avustralya gibi ileri teknoloji kullanan ülkeler ile Çin ve Hırvatistan gibi gelişmekte olan ekonomilerin yanı sıra Bangladeş ve Pakistan gibi iklim hassasiyeti yüksek bölgelerin erken uyarı modelleri incelenmiştir. Bu analizler, teknik tahmin doğruluğunun yüksek olmasının afet riskini azaltmak için tek başına yeterli olmadığını, sistemin başarısının kurumlar arası veri entegrasyonu ve yönetişim kapasitesine bağlı olduğunu kanıtlamaktadır. Türkiye özelinde yapılan incelemeler, ülkenin Meteoroloji Genel Müdürlüğü (MGM) nezdinde güçlü bir veri üretim altyapısına sahip olduğunu, ancak bu verilerin yapay zekâ destekli, insan denetimli karar destek mekanizmalarıyla bütünleşmesinde yapısal kopukluklar yaşandığını ortaya koymaktadır. Tezin özgün bulguları, Türkiye'deki erken uyarı pratiklerinin ağırlıklı olarak“tehlike bildirimi”aşamasında kaldığını; bu durumun risk azaltma ve sakınım (mitigation) süreçleriyle olan bağının güçlendirilmesi gerektiğini göstermektedir. Ayrıca, Türkiye'nin uyum kapasitesi SWOT analizi ile değerlendirilerek, TAMP, TARAP ve TASİP gibi ulusal planların sahadaki uygulama ve yerel yönetim düzeyindeki yetkinlik farkları analiz edilmiştir. Sonuç olarak, çalışma, yapay zekâ teknolojilerinin büyük veri işleme ve tahmin sürelerini uzatma konusundaki teknik avantajlarını kabul etmekle birlikte; Türkiye için mevcut aktörler (AFAD, MGM, DSİ vb.) arasındaki veri akışını ve teknik entegrasyonu merkeze alan bir uyum stratejisi önerisi sunmaktadır. Tez, bu sistemlerin“insan denetimli karar destek mekanızmaları”(human-in-the-loop) esas alan, etik ve şeffaf bir yönetişim çerçevesinde kurgulanmasının, Türkiye'nin iklim değişikliğine bağlı meteorolojik afetler karşısındaki direncini artıracağını savunmaktadır.

Özet (Çeviri)

SUMMARY The global climate crisis has shifted from a distant threat to an immediate and systemic danger that impacts environmental, social, and institutional systems worldwide. This change requires a fundamental overhaul of disaster mitigation and adaptation strategies. According to the Intergovernmental Panel on Climate Change (IPCC) Sixth Assessment Report, the Mediterranean Basin is a climate-sensitive hotspot, warming faster than the global average. This accelerated warming considerably heightens regional vulnerability to extreme weather events. Türkiye, with its varied climate zones, faces growing exposure to meteorological hazards such as flash floods, prolonged droughts, and intense heatwaves. These events display complex spatial interactions that amplify their combined impacts, often triggering cascading disasters that exceed local response capabilities. Traditional disaster management — centred on response after events — is increasingly insufficient. This thesis assesses Türkiye's adaptation capacity through a Multi-Hazard Risk Management framework. By analysing global trends from 1980 to 2025, the study uncovers structural gaps in national strategies such as Türkiye Disaster Risk Reduction Plan (TARAP) and Türkiye Disaster Response Plan (TAMP), and explores the potential of Artificial Intelligence–based Early Warning Systems (AI-EWS) as a strategic tool for risk reduction and proactive decision-making. Ultimately, the research aims to support the development of a more resilient and technologically advanced model of disaster governance. In light of these objectives, the research is based on social vulnerability and resilience theories, which state that disasters arise not only from physical events but also from the interaction between natural hazards and vulnerable social environments. This view highlights that disaster risk depends on both environmental triggers and existing societal weaknesses. Meteorological disasters are understood as socially constructed risks influenced by governance and spatial planning. At the core of this thesis is the shift from reactive crisis management to proactive risk governance. This change advances beyond simple emergency response towards a more integrated and anticipatory approach. Special focus is given to compound hazards and cascading impacts. Understanding these complex interactions is crucial, as a single event can often set off a chain of secondary disasters. Compound hazards involve simultaneous or sequential events, such as heatwaves causing wildfires and subsequent erosion, which increases flood vulnerability. While global casualty rates have decreased thanks to better warnings, rising economic losses highlight the need for mitigation strategies that safeguard urban systems and critical infrastructure. Therefore, strengthening both structural and non-structural resilience is vital to reducing the long-term socio-economic impacts of these multifaceted disasters. To operationalise this inquiry, the thesis adopts a qualitative comparative research design, incorporating a multi-dimensional approach to evaluate disaster governance. Primary data sources include the The International Disaster Database (EM-DAT) database, World Meteorological Organization (WMO) reports, and the IPCC Sixth Assessment Report (AR6), which provides the scientific basis for future climate projections. These are supplemented by national documents from Disaster and Emergency Management Authority of Türkiye (AFAD) and Türkiye State Meteorological Service (TSMS). The integration of these diverse datasets ensures that the analysis is grounded in both high-level global scientific findings and localised operational data. The methodological framework consists of three components: (1) Thematic analysis classifying disasters into hydrological, climatological, and meteorological categories in alignment with international climate standards; (2) Comparative policy analysis benchmarking Türkiye against international models such as the United States, Japan, the Netherlands, and Bangladesh; and (3) A SWOT analysis assessing Türkiye's institutional strengths and weaknesses in multi-hazard risk management. By synthesising IPCC climate scenarios with national disaster data, this methodology allows for the identification of best practices and the formulation of adaptable strategies for the Turkish context. Within this comparative framework, global analysis reveals a shift towards Impact-Based Forecasting (IBF), where systems prioritise potential consequences over merely meteorological thresholds. This paradigm shift emphasises the necessity of translating technical weather data into specific socio-economic risk assessments. Japan demonstrates advanced lead-time management, converting satellite data into actionable warnings, setting a global benchmark for rapid institutional response. The Netherlands offers a distinctive model through its“Room for the River”programme, combining engineering with ecological adaptation to manage long-term hydrological risks. The United States highlights the role of public–private partnerships in data sharing and risk financing, showcasing how multi-sector collaboration can enhance fiscal resilience. Compared to these models, Türkiye exhibits strong response capacity but faces limitations in proactive risk reduction, high-resolution spatial modelling, and cross-sectoral data integration, especially regarding cascading hazards. Bridging these gaps requires a move beyond traditional structural measures towards more integrated, AI-driven predictive frameworks that can account for the complexity of multi-disaster interactions. Furthermore, detailed findings on Türkiye's adaptation capacity reveal that while the country possesses substantial data capabilities, TSMS's (MGM) extensive observation data remains underused in municipal-level risk management. This underuse creates a gap between high-level data collection and local implementation. Applying a multi-hazard approach across ministries remains limited. The findings of this thesis show that current practices in Türkiye mainly stay at the“hazard notification”stage, which results in weak integration with long-term risk reduction processes. Significant disparities are present at the local level, where smaller municipalities lack the technical expertise to turn climate data into local mitigation measures. Additionally, the lack of a unified digital infrastructure prevents real-time sharing of vulnerability assessments between central and local authorities, further complicating responses to multi-disaster scenarios. To address these challenges, the main contribution of this study is a proposed national AI-EWS framework tailored to Türkiye's unique geographical and institutional context. The model highlights three key components: (1) Deep learning architectures, such as Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM) networks, for predicting rapid-onset events like pluvial floods (seller) with greater accuracy and longer lead times; (2) Big data analytics for real-time situational awareness, combining diverse data streams from meteorological sensors, satellite imagery, and social dynamics; and (3) A human-in-the-loop mechanism. This hybrid approach ensures that AI functions as a reliable decision-support tool under human supervision, maintaining ethical accountability and transparent governance. By integrating predictive technology with expert judgment, the proposed model seeks to reduce false alarms and improve the distribution of emergency resources during complex multi-disaster events. In conclusion, Türkiye must shift from a traditional, crisis-focused management approach to a comprehensive and proactive risk governance model. This strategic change is crucial to reduce the systemic weaknesses in current disaster response systems. Key suggestions include updating TAMP to include AI-driven protocols that enable real-time, data-based coordination among various institutional stakeholders. Furthermore, the study advocates investing in green infrastructure as a nature-based solution to lessen the increasing impacts of urban heat islands and flash floods in rapidly expanding cities. Improving digital disaster literacy at both administrative and community levels is also vital to ensure effective use of advanced early warning technologies. Aligning technological innovation with multi-layered governance is necessary to boost national disaster resilience in an ever more unpredictable and volatile climate. Ultimately, Türkiye's ability to adapt relies on connecting technological innovation with adaptable, risk-informed institutional frameworks capable of evolving alongside emerging global threats. This research concludes that implementing an AI-EWS framework is not just a technical upgrade, but a fundamental strategic requirement for the sustainable future and long-term stability of Türkiye's disaster management system.

Benzer Tezler

  1. Tez No

    952951

    Doğal afetlerde çevrimiçi davranış verilerine dayalı ihtiyaç tahmini ve kaynak tahsis optimizasyonu: 2023 kahramanmaraş depremleri uygulaması

    Demand prediction and resource allocation optimization based on online behavioral data in natural disasters: An application to the 2023 Kahramanmaraş earthquakes

    MEHMET DEMİR

    Yüksek Lisans

    Türkçe

    Türkçe

    2025

    Endüstri ve Endüstri MühendisliğiHacettepe Üniversitesi

    Endüstri Mühendisliği Ana Bilim Dalı

    DR. ÖĞR. ÜYESİ ERDİ DAŞDEMİR

  2. Tez No

    886463

    Orthogonality based feature selection for ai applications

    Yapay zeka uygulamaları için ortogonalite tabanlı öznitelik seçimi

    MEHMET SELAHADDİN ŞENTOP

    Yüksek Lisans

    İngilizce

    İngilizce

    2024

    Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrolİstanbul Teknik Üniversitesi

    Bilgisayar Mühendisliği Ana Bilim Dalı

    PROF. DR. BURAK BERK ÜSTÜNDAĞ

  3. Tez No

    1004660

    Minimal invaziv adrenal cerrahide yapay zekâ destekli cerrahi kopilot: Anatomik yapılar, cerrahi aletler ve prosedür adımlarının entegre tanınması

    Artificial intelligence-powered surgical copilot in minimally invasive adrenal surgery: Integrated recognition of anatomy, tools, and surgical phases

    BERKE ŞENGÜN

    Tıpta Uzmanlık

    Türkçe

    Türkçe

    2026

    Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrolİstanbul Üniversitesi

    Genel Cerrahi Ana Bilim Dalı

    DOÇ. DR. AHMET YALIN İŞCAN

  4. Tez No

    986783

    AI applıcatıons ın dısaster rısk management: lessons from the 2023 turkıye-syrıa earthquake

    Afet risk yönetiminde yapay zeka uygulamaları: 2023 Iürkiye-Suriye depreminden dersler

    MOHAMMED ABDULBASIT SHAREEF AL-TAWEEL

    Yüksek Lisans

    İngilizce

    İngilizce

    2025

    Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolAnkara Yıldırım Beyazıt Üniversitesi

    Afet ve Savaş Sonrası İmar ve Rehabilitasyon Ana Bilim Dalı

    PROF. DR. SALAH HAJISMAIL

  5. Tez No

    942814

    Lise öğrencilerinin iş sağlığı ve güvenliği eğitimlerinde dijital oyun tasarımı, uygulanması ve geliştirilmesi

    Design, development, implementation, and evaluation of digital game-based tools for occupational health and safety education in high schools

    HÜSEYİN ENİS KARA

    Doktora

    Türkçe

    Türkçe

    2025

    Eğitim ve Öğretimİstanbul Medeniyet Üniversitesi

    İş Sağlığı ve Güvenliği Ana Bilim Dalı

    PROF. DR. EMİNE CAN

Geri Dön