Machine learning and geographic information systems-based framework for multidimensional analysis of cascading drought impacts using remote sensing and in-situ data

Serkendiz H., TATLI H., ÖZELKAN E., ÇETİN M.

Science of the Total Environment, cilt.1001, 2025 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 1001
  • Basım Tarihi: 2025
  • Doi Numarası: 10.1016/j.scitotenv.2025.180504
  • Dergi Adı: Science of the Total Environment
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, PASCAL, Aerospace Database, Analytical Abstracts, Aqualine, Aquatic Science & Fisheries Abstracts (ASFA), BIOSIS, Biotechnology Research Abstracts, CAB Abstracts, Chemical Abstracts Core, Chimica, Communication Abstracts, Compendex, Environment Index, Food Science & Technology Abstracts, Geobase, Greenfile, Metadex, Pollution Abstracts, Veterinary Science Database, Civil Engineering Abstracts
  • Anahtar Kelimeler: Agricultural production, Cascading Hazard, Cascading risk, Change detection, Drought, Groundwater, Land use change, Machine learning
  • Çukurova Üniversitesi Adresli: Evet

Özet

This study proposes a multidimensional conceptual framework to assess the cascading impacts of drought on the agricultural sector. The framework consists of four interconnected components: the triggering hazard, biophysical drivers, socio-ecological impacts, and socio-economic outcomes. To demonstrate its applicability, the framework was applied to the Konya Closed Basin, a drought-sensitive agricultural region in central Türkiye. The study integrates remote sensing indicators (NDVI, NDWI, LST, and land cover), ground-based observations (precipitation, temperature, groundwater levels), and statistical trend analyses (Mann-Kendall) to characterize drought dynamics and land use transitions. Machine learning algorithms were used to model land use change: drought-related indicators such as NDVI, NDWI, LST, and Palmer Drought Severity Index (PDSI) served as input variables, while Support Vector Machines (SVM), Artificial Neural Networks (ANN), and Random Forests (RF) were applied as classification tools to predict land cover changes over time. Between 1990 and 2018, approximately 510,000 ha of irrigated land, including rice fields, were converted into non-irrigated areas. Despite this trend, the production of water-intensive crops such as maize and sugar beet continued to rise, indicating a maladaptive trajectory in agricultural practices. This mismatch between environmental constraints and production patterns highlights unsustainable water use and signals potential long-term risks to both water and food security. The proposed framework not only enhances understanding of cascading drought impacts but also offers critical insights for adaptive agricultural and water governance, supporting evidence-based policymaking in climate-vulnerable regions.