Comparative Analysis of Transfer Function Method with Advanced Flood Prediction Techniques

In this paper, the evaluation of the performance of five flood prediction models in the Simineh-Rood River, Lake Urmia basin, Iran, is discussed in detail. To this purpose, the performance of Transfer Function, Saint-Venant equations, Artificial Neural Network, Adaptive Neuro-Fuzzy Inference System,...

Full description

Saved in:
Bibliographic Details
Main Author: Jafar Chabokpour
Format: Article
Language:English
Published: University of Birjand 2024-09-01
Series:Water Harvesting Research
Subjects:
Online Access:https://jwhr.birjand.ac.ir/article_3050_96af733f3a28d6923ddd58e66d18cfeb.pdf
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849723020130648064
author Jafar Chabokpour
author_facet Jafar Chabokpour
author_sort Jafar Chabokpour
collection DOAJ
description In this paper, the evaluation of the performance of five flood prediction models in the Simineh-Rood River, Lake Urmia basin, Iran, is discussed in detail. To this purpose, the performance of Transfer Function, Saint-Venant equations, Artificial Neural Network, Adaptive Neuro-Fuzzy Inference System, and Support Vector Machine models are evaluated for 2018 and 2019 flood data. Specifically, the models are rated according to their accuracy, computational efficiency, and robustness under different flow regimes and at various forecast times. This now leads to a maximum Nash-Sutcliffe Efficiency of 0.91 for the Saint-Venant equations during the 2019 flood event, followed by ANN with 0.89, ANFIS with 0.87, SVM with 0.85, and lastly, Transfer Function with 0.78. The same is the case for peak flow discharge, which was best predicted by the Saint-Venant model to be 193.80 m³/s while the observed value was 200.83 m³/s. This model maintained its consistency with respect to low, medium, and high flows, where the values of NSE were 0.89, 0.92, and 0.91, respectively. However, compared to the other models, which took 0.5–8 s, it had a much larger computational time, 120 s for a 72-h simulation. The sensitivity analysis returned variable model responses to the quality of the input data; an input variation of 20% reduced the NSE of the Saint-Venant model to 0.73 and that of the Transfer Function to 0.44. This study provides quantitative insight into the choice of flood prediction methods in a semi-arid region, with respect to required accuracy, computational resources, and forecast lead-time.
format Article
id doaj-art-a84a8cf7ff484e8bba15f0fc8369b606
institution DOAJ
issn 2476-6976
2476-7603
language English
publishDate 2024-09-01
publisher University of Birjand
record_format Article
series Water Harvesting Research
spelling doaj-art-a84a8cf7ff484e8bba15f0fc8369b6062025-08-20T03:11:10ZengUniversity of BirjandWater Harvesting Research2476-69762476-76032024-09-017219420910.22077/jwhr.2024.7966.11473050Comparative Analysis of Transfer Function Method with Advanced Flood Prediction TechniquesJafar Chabokpour0Associate Professor, Department of Civil Engineering, University of Maragheh, Maragheh, Iran.In this paper, the evaluation of the performance of five flood prediction models in the Simineh-Rood River, Lake Urmia basin, Iran, is discussed in detail. To this purpose, the performance of Transfer Function, Saint-Venant equations, Artificial Neural Network, Adaptive Neuro-Fuzzy Inference System, and Support Vector Machine models are evaluated for 2018 and 2019 flood data. Specifically, the models are rated according to their accuracy, computational efficiency, and robustness under different flow regimes and at various forecast times. This now leads to a maximum Nash-Sutcliffe Efficiency of 0.91 for the Saint-Venant equations during the 2019 flood event, followed by ANN with 0.89, ANFIS with 0.87, SVM with 0.85, and lastly, Transfer Function with 0.78. The same is the case for peak flow discharge, which was best predicted by the Saint-Venant model to be 193.80 m³/s while the observed value was 200.83 m³/s. This model maintained its consistency with respect to low, medium, and high flows, where the values of NSE were 0.89, 0.92, and 0.91, respectively. However, compared to the other models, which took 0.5–8 s, it had a much larger computational time, 120 s for a 72-h simulation. The sensitivity analysis returned variable model responses to the quality of the input data; an input variation of 20% reduced the NSE of the Saint-Venant model to 0.73 and that of the Transfer Function to 0.44. This study provides quantitative insight into the choice of flood prediction methods in a semi-arid region, with respect to required accuracy, computational resources, and forecast lead-time.https://jwhr.birjand.ac.ir/article_3050_96af733f3a28d6923ddd58e66d18cfeb.pdfflood predictionlake urmiasoft computing methodstransfer function
spellingShingle Jafar Chabokpour
Comparative Analysis of Transfer Function Method with Advanced Flood Prediction Techniques
Water Harvesting Research
flood prediction
lake urmia
soft computing methods
transfer function
title Comparative Analysis of Transfer Function Method with Advanced Flood Prediction Techniques
title_full Comparative Analysis of Transfer Function Method with Advanced Flood Prediction Techniques
title_fullStr Comparative Analysis of Transfer Function Method with Advanced Flood Prediction Techniques
title_full_unstemmed Comparative Analysis of Transfer Function Method with Advanced Flood Prediction Techniques
title_short Comparative Analysis of Transfer Function Method with Advanced Flood Prediction Techniques
title_sort comparative analysis of transfer function method with advanced flood prediction techniques
topic flood prediction
lake urmia
soft computing methods
transfer function
url https://jwhr.birjand.ac.ir/article_3050_96af733f3a28d6923ddd58e66d18cfeb.pdf
work_keys_str_mv AT jafarchabokpour comparativeanalysisoftransferfunctionmethodwithadvancedfloodpredictiontechniques