Using GIS, Remote Sensing, and Machine Learning to Highlight the Correlation between the Land-Use/Land-Cover Changes and Flash-Flood Potential

• Published in: Remote sensing (Basel, Switzerland) Vol. 12; no. 9; p. 1422
• Main Authors: Costache, Romulus, Bao Pham, Quoc, Corodescu-Roșca, Ema, Cîmpianu, Cătălin, Hong, Haoyuan, Thi Thuy Linh, Nguyen, Ming Fai, Chow, Najah Ahmed, Ali, Vojtek, Matej, Muhammed Pandhiani, Siraj, Minea, Gabriel, Ciobotaru, Nicu, Cristian Popa, Mihnea, Diaconu, Daniel Constantin, Thai Pham, Binh
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01-05-2020
• Subjects: Algorithms; Climate change; Correlation analysis; Evolution; Flash floods; flash-flood potential index; Flood frequency; Flood predictions; Flooding; Floods; Geographic information systems; geographically weighted regression; Hydrology; Image processing; Land cover; Land use; Landsat; Landsat images; Landsat satellites; Learning algorithms; Lithology; Machine learning; multilayer perceptron; Multilayer perceptrons; Multiplication; Neural networks; Remote sensing; Researchers; River catchments; Rivers; Runoff; Satellite imagery; Studies; Surface runoff; total relative difference-synthetic dynamic land-use index; Vegetation; Zăbala; Romania
• ISSN: 2072-4292; 2072-4292
• DOI: 10.3390/rs12091422

The aim of the present study was to explore the correlation between the land-use/land cover change and the flash-flood potential changes in Zăbala catchment (Romania) between 1989 and 2019. In this regard, the efficiency of GIS, remote sensing and machine learning techniques in detecting spatial patterns of the relationship between the two variables was tested. The paper elaborated upon an answer to the increase in flash flooding frequency across the study area and across the earth due to the occurred land-use/land-cover changes, as well as due to the present climate change, which determined the multiplication of extreme meteorological phenomena. In order to reach the above-mentioned purpose, two land-uses/land-covers (for 1989 and 2019) were obtained using Landsat image processing and were included in a relative evolution indicator (total relative difference-synthetic dynamic land-use index), aggregated at a grid-cell level of 1 km2. The assessment of runoff potential was made with a multilayer perceptron (MLP) neural network, which was trained for 1989 and 2019 with the help of 10 flash-flood predictors, 127 flash-flood locations, and 127 non-flash-flood locations. For the year 1989, the high and very high surface runoff potential covered around 34% of the study area, while for 2019, the same values accounted for approximately 46%. The MLP models performed very well, the area under curve (AUC) values being higher than 0.837. Finally, the land-use/land-cover change indicator, as well as the relative evolution of the flash flood potential index, was included in a geographically weighted regression (GWR). The results of the GWR highlights that high values of the Pearson coefficient (r) occupied around 17.4% of the study area. Therefore, in these areas of the Zăbala river catchment, the land-use/land-cover changes were highly correlated with the changes that occurred in flash-flood potential.