Convolutional neural network and long short-term memory algorithms for groundwater potential mapping in Anseong, South Korea

Journal of Hydrology: Regional Studies

Mahdi Panahi¹  Arip Syaripudin Nur²  Chang-Wook Lee²  Saro Lee²  Wahyu Luqmanul Hakim²  Fatemeh Rezaie³ 

[1] Department of Geophysical Exploration, Korea University of Science and Technology, 217 Gajeong-ro, Yuseong-gu, Daejeon 305-350, Republic of Korea;Division of Smart Regional Innovation, Kangwon National University, 1 Gangwondaehak-gil, Chuncheon-si, Gangwon-do 24341, Republic of Korea;Geoscience Platform Division, Korea Institute of Geoscience and Mineral Resources (KIGAM), 124, Gwahak-ro, Yuseong-gu, Daejeon 34132, Republic of Korea;

关键词: Groundwater;    Potential mapping;    Convolutional neural network (CNN);    Long short-term memory (LSTM);   

DOI  :  

来源: DOAJ

【 摘 要 】

Study region: The study area was the Anseong-si area that located in the southernmost part of Gyeonggi-do Province at 127°19′ E, 36°82′ N. Anseong has a transitional climate between that the north and the south regions. Its climate is characterized by the geographical conditions of forming expansive plains that stretch from the Charyeong Range. The entire area of the city is surrounded by many high and low mountains to the south and the west, late-middle age and old age hills are spread, while there are many plains due to the development of rivers. Study focus: In this study, machine learning algorithms were used based on convolutional neural network (CNN) and long short-term memory (LSTM) to generate a groundwater potential map in Anseong, South Korea. A total of 295 wells locations were divided by the median value of transmissivity data (T) and produced “1” point with high groundwater productivity data and “0” as the point with the low groundwater productivity and divided into 70:30 for training and validation the model. 14 groundwater potential related factors were used in this study such as topo-hydrological and geo-environmental factors to define the spatial correlation with the high groundwater productivity data. The validation of the model is evaluated using the receiver operating characteristics (ROC) curve analysis method. The area under the ROC curve (AUC) is calculated to test the model. New hydrological insight for the region: The results shows a good accuracy, as the AUC values were all higher than 0.8. Finally, the groundwater potential maps generated by CNN and LSTM can be used to analyze the areas that potentially provided groundwater in Anseong, South Korea. This study could help the local environment to manage groundwater resources and can be used to assist development planners and decision-makers in groundwater sustainability planning.

【 授权许可】

Unknown