Accurate prediction of salinity in Chott Djerid shallow aquifers, southern Tunisia: Machine learning model development

A backpropagation neural network (BPNN) was used to predict salinity levels in the Chott Djerid shallow aquifers. A set of 51 water samples was collected from the Chott Djerid plio-quaternary aquifers for geochemical analysis. Major elements and nitrates were ascertained by using high performance li...

Full description

Saved in:
Bibliographic Details
Main Authors: Zohra Kraiem, Kamel Zouari, Najiba Chkir
Format: Article
Language:English
Published: Taylor & Francis Group 2024-12-01
Series:Water Science
Subjects:
Online Access:https://www.tandfonline.com/doi/10.1080/23570008.2023.2294535
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850247736478138368
author Zohra Kraiem
Kamel Zouari
Najiba Chkir
author_facet Zohra Kraiem
Kamel Zouari
Najiba Chkir
author_sort Zohra Kraiem
collection DOAJ
description A backpropagation neural network (BPNN) was used to predict salinity levels in the Chott Djerid shallow aquifers. A set of 51 water samples was collected from the Chott Djerid plio-quaternary aquifers for geochemical analysis. Major elements and nitrates were ascertained by using high performance liquid-ion chromatography. The BPNN was trained on a dataset of 51 water samples with variable geochemical parameters. Our results indicated a high accuracy when applying a model with 13 inputs, 1 hidden layer (6 neurons) and 1 output (TDS in mg/L). The collected data were split into 80% for training the model and 20% for testing and cross validation. The result was evaluated using various statistical performance criteria (i.e., MSE, RMSE, R2, SSE, SD, Accuracy, Sensitivity, specificity, and Kappa test); it showed that BPNN model properly predicted the salinity of the Chott Djerid plio-quaterny water samples (RMSE = 0.0402; R2 = 0.9721 and SSE = 0.0146). The BPNN was able to capture the complex relationship between salinity levels and other aquifer parameters. The potential application of BPNNs for predicting salinity levels in shallow aquifers was crucial in supporting decision-makers for water management; it provided valuable insights into the salinity fluctuation of the studied shallow aquifers.
format Article
id doaj-art-415728af58ee4f55a8bed74f0536c011
institution OA Journals
issn 2357-0008
language English
publishDate 2024-12-01
publisher Taylor & Francis Group
record_format Article
series Water Science
spelling doaj-art-415728af58ee4f55a8bed74f0536c0112025-08-20T01:58:51ZengTaylor & Francis GroupWater Science2357-00082024-12-01381334710.1080/23570008.2023.2294535Accurate prediction of salinity in Chott Djerid shallow aquifers, southern Tunisia: Machine learning model developmentZohra Kraiem0Kamel Zouari1Najiba Chkir2National Engineering School, Laboratory of Radio-Analyses and Environment, University of Sfax, Sfax, TunisiaNational Engineering School, Laboratory of Radio-Analyses and Environment, University of Sfax, Sfax, TunisiaDepartment of Geography, Faculty of Letters and Humanities, Laboratory of Radio-Analyses and Environment, University of Sfax, Sfax, TunisiaA backpropagation neural network (BPNN) was used to predict salinity levels in the Chott Djerid shallow aquifers. A set of 51 water samples was collected from the Chott Djerid plio-quaternary aquifers for geochemical analysis. Major elements and nitrates were ascertained by using high performance liquid-ion chromatography. The BPNN was trained on a dataset of 51 water samples with variable geochemical parameters. Our results indicated a high accuracy when applying a model with 13 inputs, 1 hidden layer (6 neurons) and 1 output (TDS in mg/L). The collected data were split into 80% for training the model and 20% for testing and cross validation. The result was evaluated using various statistical performance criteria (i.e., MSE, RMSE, R2, SSE, SD, Accuracy, Sensitivity, specificity, and Kappa test); it showed that BPNN model properly predicted the salinity of the Chott Djerid plio-quaterny water samples (RMSE = 0.0402; R2 = 0.9721 and SSE = 0.0146). The BPNN was able to capture the complex relationship between salinity levels and other aquifer parameters. The potential application of BPNNs for predicting salinity levels in shallow aquifers was crucial in supporting decision-makers for water management; it provided valuable insights into the salinity fluctuation of the studied shallow aquifers.https://www.tandfonline.com/doi/10.1080/23570008.2023.2294535Artificial neural networksalinity predictiongroundwater managementChott Djerid shallow aquifersgeochemical assessment
spellingShingle Zohra Kraiem
Kamel Zouari
Najiba Chkir
Accurate prediction of salinity in Chott Djerid shallow aquifers, southern Tunisia: Machine learning model development
Water Science
Artificial neural network
salinity prediction
groundwater management
Chott Djerid shallow aquifers
geochemical assessment
title Accurate prediction of salinity in Chott Djerid shallow aquifers, southern Tunisia: Machine learning model development
title_full Accurate prediction of salinity in Chott Djerid shallow aquifers, southern Tunisia: Machine learning model development
title_fullStr Accurate prediction of salinity in Chott Djerid shallow aquifers, southern Tunisia: Machine learning model development
title_full_unstemmed Accurate prediction of salinity in Chott Djerid shallow aquifers, southern Tunisia: Machine learning model development
title_short Accurate prediction of salinity in Chott Djerid shallow aquifers, southern Tunisia: Machine learning model development
title_sort accurate prediction of salinity in chott djerid shallow aquifers southern tunisia machine learning model development
topic Artificial neural network
salinity prediction
groundwater management
Chott Djerid shallow aquifers
geochemical assessment
url https://www.tandfonline.com/doi/10.1080/23570008.2023.2294535
work_keys_str_mv AT zohrakraiem accuratepredictionofsalinityinchottdjeridshallowaquiferssoutherntunisiamachinelearningmodeldevelopment
AT kamelzouari accuratepredictionofsalinityinchottdjeridshallowaquiferssoutherntunisiamachinelearningmodeldevelopment
AT najibachkir accuratepredictionofsalinityinchottdjeridshallowaquiferssoutherntunisiamachinelearningmodeldevelopment