|
Prediction of bathymetry Using Neuro-Fuzzy Models: A Comparative Study with Geodetic Data fusion and Optimization Based on Local Data
|
Mohammad Ali Mohammad   , Iraj Jazireeyan * , Mahmoud Pirooznia |
| K. N. Toosi University of Technology |
|
|
Abstract: (35 Views) |
Accurate knowledge of seabed depth plays a crucial role in understanding oceanic processes, physical oceanography, marine biology, ecohydraulics, and marine geology. Conventional depth modeling methods are commonly based on satellite altimetry, gravity models or marine gravimetry data, which often lack sufficient accuracy or spatial resolution. In this study, a comparative analysis of neuro-fuzzy models was conducted for regional bathymetry modeling in the Persian Gulf and the Oman Sea, with results further optimized using local datasets. For this purpose, Adaptive Neuro-Fuzzy Inference System (ANFIS) and Local Linear Model Tree (LOLIMOT) were applied as nonlinear models to identify the relationship between inputs and outputs. The model inputs consisted of geodetic data, including geoid height, gravity gradient, and gravity anomaly, while the output was the GEBCO bathymetric dataset. The results indicated that the LOLIMOT model showed better agreement with the test data. Consequently, the baseline model was refined by assimilating in-situ depth observations using the three-dimensional variational (3DVAR) optimization method, leading to the development of the final bathymetric model. The proposed approach was validated along control track soundings from Chabahar, Genaveh, and Alamshah regions, demonstrating high accuracy with RMSE values of approximately 0.4 m, 0.8 m, and 0.9 m, respectively. This modeling framework provides a robust and accurate method for seabed depth analysis and prediction in both scientific and applied studies.
|
|
| Keywords: Bathymetry Modeling, Satellite Altimetry, Marine Gravity, Machine Learning, Neuro-Fuzzy Model |
|
|
|
Type of Study: Research |
Subject:
Geodesy
Received: 2025/09/30 | Accepted: 2025/12/20 | ePublished ahead of print: 2026/01/31
|
|
|
|
|
|
|
| Send email to the article author |
|
|
