Monitoring of Inland Excess Water Inundations Using Machine Learning Algorithms

被引:6
作者
Kajari, Balazs [1 ,2 ]
Bozan, Csaba [2 ]
Van Leeuwen, Boudewijn [1 ]
机构
[1] Univ Szeged, Dept Geoinformat Phys & Environm Geog, Egyet U 2-6, H-6722 Szeged, Hungary
[2] Hungarian Univ Agr & Life Sci, Inst Environm Sci, Ctr Irrigat & Water Management, Res, Annaliget 35, H-5540 Szarvas, Hungary
关键词
inland excess water; water logging; water classification; machine learning; convolutional neural network; deep learning; Sentinel-2; RANDOM FOREST; INDEX NDWI; FEATURES; IMPACTS;
D O I
10.3390/land12010036
中图分类号
X [环境科学、安全科学];
学科分类号
08 ; 0830 ;
摘要
Nowadays, climate change not only leads to riverine floods and flash floods but also to inland excess water (IEW) inundations and drought due to extreme hydrological processes. The Carpathian Basin is extremely affected by fast-changing weather conditions during the year. IEW (sometimes referred to as water logging) is formed when, due to limited runoff, infiltration, and evaporation, surplus water remains on the surface or in places where groundwater flowing to lower areas appears on the surface by leaking through porous soil. In this study, eight different machine learning approaches were applied to derive IEW inundations on three different dates in 2021 (23 February, 7 March, 20 March). Index-based approaches are simple and provide relatively good results, but they need to be adapted to specific circumstances for each area and date. With an overall accuracy of 0.98, a Kappa of 0.65, and a QADI score of 0.020, the deep learning method Convolutional Neural Network (CNN) gave the best results, compared to the more traditional machine learning approaches Maximum Likelihood (ML), Random Forest (RF), Support Vector Machine (SVM) and artificial neural network (ANN) that were evaluated. The CNN-based IEW maps can be used in operational inland excess water control by water management authorities.
引用
收藏
页数:22
相关论文
共 69 条
[1]  
Abadi M., 2016, TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems, DOI DOI 10.48550/ARXIV.1603.04467
[2]   FLOODPLAIN SEDIMENTATION - QUANTITIES, PATTERNS AND PROCESSES [J].
ASSELMAN, NEM ;
MIDDELKOOP, H .
EARTH SURFACE PROCESSES AND LANDFORMS, 1995, 20 (06) :481-499
[3]   Extracting water-related features using reflectance data and principal component analysis of Landsat images [J].
Balazs, Boglarka ;
Biro, Tibor ;
Dyke, Gareth ;
Singh, Sudhir Kumar ;
Szabo, Szilard .
HYDROLOGICAL SCIENCES JOURNAL-JOURNAL DES SCIENCES HYDROLOGIQUES, 2018, 63 (02) :269-284
[4]   Random forest in remote sensing: A review of applications and future directions [J].
Belgiu, Mariana ;
Dragut, Lucian .
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING, 2016, 114 :24-31
[5]   Integrated spatial assessment of inland excess water hazard on the Great Hungarian Plain [J].
Bozan, Csaba ;
Takacs, Katalin ;
Korosparti, Janos ;
Laborczi, Annamaria ;
Turi, Norbert ;
Pasztor, Laszlo .
LAND DEGRADATION & DEVELOPMENT, 2018, 29 (12) :4373-4386
[6]   Landslide Susceptibility Mapping of Landslides with Artificial Neural Networks: Multi-Approach Analysis of Backpropagation Algorithm Applying the Neuralnet Package in Cuenca, Ecuador [J].
Bravo-Lopez, Esteban ;
Fernandez Del Castillo, Tomas ;
Sellers, Chester ;
Delgado-Garcia, Jorge .
REMOTE SENSING, 2022, 14 (14)
[7]   Random forests [J].
Breiman, L .
MACHINE LEARNING, 2001, 45 (01) :5-32
[8]   Random forests: Finding quasars [J].
Breiman, L ;
Last, M ;
Rice, J .
STATISTICAL CHALLENGES IN ASTRONOMY, 2003, :243-254
[9]  
Chollet F., 2015, KERAS
[10]   Convolutional neural networks for global human settlements mapping from Sentinel-2 satellite imagery [J].
Corbane, Christina ;
Syrris, Vasileios ;
Sabo, Filip ;
Politis, Panagiotis ;
Melchiorri, Michele ;
Pesaresi, Martino ;
Soille, Pierre ;
Kemper, Thomas .
NEURAL COMPUTING & APPLICATIONS, 2021, 33 (12) :6697-6720