Anıl Çelik
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Anıl Çelik has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
Abdüsselam Altunkaynak
Istanbul Technical Univ
12 shared publications- Abdüsselam Altunkaynak
Hydraulics and Water Resource Engineering Division, Faculty of Civil Engineering, Istanbul Technical University, Maslak, 34469 Istanbul, Türkiye
2 shared publications - Mehmet Özger
Hydraulics and Water Resource Engineering Division, Faculty of Civil Engineering, Istanbul Technical University, Maslak, Istanbul 34469, Turkey
2 shared publications Murat Barış Mandev
2 shared publications- Murat Barış Mandev
Istanbul Technical Univ
2 shared publications Elif Kartal
1 shared publication
Latest Publications
- 1 December 2026
- Published by American Society of Civil Engineers (ASCE) in Journal of Hydrologic Engineering
- Vol. 31 (6) , 04026054
Advanced Approaches for Imputing Missing Rainfall Data in Time Series: Dynamic Adaptive Wavelet-Based Compressed Sensing
Discrete wavelet (DW) analysis is commonly employed to decompose raw observed data into stochastic and deterministic components, aiming to improve model accuracy. Nonetheless, the DW technique is recognized for its significant limitations, often leading to inaccurate predictions of daily rainfall. In this study, a dynamic adaptive wavelet algorithm is introduced to decompose the raw observed data into components, aiming to address the drawbacks associated with DW analysis. The ultimate goal is to enable the estimation of daily rainfall with enhanced accuracy and precision in missing value imputation. The decomposed stochastic and deterministic components derived from the maximum overlap discrete wavelet (MODW) and DW techniques are incorporated as inputs in compressed sensing (CS) framework. This procedure facilitates the creation of novel methodologies MODW-based compressed sensing (MODW-CS) and DW-based compressed sensing (DW-CS) for daily rainfall prediction. These methodologies are deployed across scenarios utilizing 50%, 70%, and 90% of the available data. Twenty-two years of validated daily rainfall data were collected from five stations situated within the Euphrates-Tigris Basin. The novel methodologies were assessed against the CS model, with root-mean squared error (RMSE), mean absolute error (MAE), and coefficient of efficiency (CE) being taken into account. Findings revealed that both the MODW-CS and DW-CS models outperformed the CS model, with enhanced prediction accuracy observed through the utilization of decomposed signals. However, the MODW-CS model exhibited superior performance compared with the DW-CS model for all scenarios. This suggests that the MODW algorithm possesses a dynamic adaptive structure enabling it to capture the natural structure of observed data more effectively than DW analysis. The findings of this study have significant implications across various domains, including hydrology, agriculture, flood control, drought management, and weather forecasting. Accurate predictions of missing data points in time series are essential for better understanding and forecasting rainfall event behavior.
- 1 December 2026
- Published by American Society of Civil Engineers (ASCE) in Journal of Hydrologic Engineering
- Vol. 31 (6) , 04026054
Advanced Approaches for Imputing Missing Rainfall Data in Time Series: Dynamic Adaptive Wavelet-Based Compressed Sensing
Discrete wavelet (DW) analysis is commonly employed to decompose raw observed data into stochastic and deterministic components, aiming to improve model accuracy. Nonetheless, the DW technique is recognized for its significant limitations, often leading to inaccurate predictions of daily rainfall. In this study, a dynamic adaptive wavelet algorithm is introduced to decompose the raw observed data into components, aiming to address the drawbacks associated with DW analysis. The ultimate goal is to enable the estimation of daily rainfall with enhanced accuracy and precision in missing value imputation. The decomposed stochastic and deterministic components derived from the maximum overlap discrete wavelet (MODW) and DW techniques are incorporated as inputs in compressed sensing (CS) framework. This procedure facilitates the creation of novel methodologies MODW-based compressed sensing (MODW-CS) and DW-based compressed sensing (DW-CS) for daily rainfall prediction. These methodologies are deployed across scenarios utilizing 50%, 70%, and 90% of the available data. Twenty-two years of validated daily rainfall data were collected from five stations situated within the Euphrates-Tigris Basin. The novel methodologies were assessed against the CS model, with root-mean squared error (RMSE), mean absolute error (MAE), and coefficient of efficiency (CE) being taken into account. Findings revealed that both the MODW-CS and DW-CS models outperformed the CS model, with enhanced prediction accuracy observed through the utilization of decomposed signals. However, the MODW-CS model exhibited superior performance compared with the DW-CS model for all scenarios. This suggests that the MODW algorithm possesses a dynamic adaptive structure enabling it to capture the natural structure of observed data more effectively than DW analysis. The findings of this study have significant implications across various domains, including hydrology, agriculture, flood control, drought management, and weather forecasting. Accurate predictions of missing data points in time series are essential for better understanding and forecasting rainfall event behavior.
- Open Access
High-Fidelity Wind Speed Network Reconstruction from Sparse Monitoring Stations Using QR Pivoting, Shallow Decoder Networks, and Shallow Recurrent Decoders
Accurate reconstruction of spatially distributed wind speed is important for wind-resource assessment and monitoring when station coverage is sparse. This study compares three data-driven approaches for spatiotemporal wind-field reconstruction: a QR-pivoting linear model, the Shallow Decoder Network (SDN), and the Shallow Recurrent Decoder (SHRED). Monthly mean wind-speed observations from 26 stations in Türkiye’s Marmara region were evaluated across input configurations ranging from one station to full-network coverage. Under sparse conditions, SHRED showed the strongest and most stable performance, achieving a mean relative error (MRE) of 0.120 and a mean coefficient of efficiency (MCE) of 0.878 with a single input station. Its recurrent structure enabled effective use of temporal information and reduced sensitivity to the selected station across the tested realizations. As sensor density increased, SDN and QR improved substantially. QR outperformed SHRED on both MRE and MCE from 20 stations onward and slightly surpassed SDN at 25 stations. However, SDN achieved the highest full-network accuracy with all 26 stations (MRE = 0.034; MCE = 0.990). These findings demonstrate that reconstruction performance depends on monitoring-network density, with SHRED being particularly effective under sparse conditions and SDN and QR becoming increasingly competitive as spatial coverage increases.
Open AccessHigh-Fidelity Wind Speed Network Reconstruction from Sparse Monitoring Stations Using QR Pivoting, Shallow Decoder Networks, and Shallow Recurrent Decoders
Accurate reconstruction of spatially distributed wind speed is important for wind-resource assessment and monitoring when station coverage is sparse. This study compares three data-driven approaches for spatiotemporal wind-field reconstruction: a QR-pivoting linear model, the Shallow Decoder Network (SDN), and the Shallow Recurrent Decoder (SHRED). Monthly mean wind-speed observations from 26 stations in Türkiye’s Marmara region were evaluated across input configurations ranging from one station to full-network coverage. Under sparse conditions, SHRED showed the strongest and most stable performance, achieving a mean relative error (MRE) of 0.120 and a mean coefficient of efficiency (MCE) of 0.878 with a single input station. Its recurrent structure enabled effective use of temporal information and reduced sensitivity to the selected station across the tested realizations. As sensor density increased, SDN and QR improved substantially. QR outperformed SHRED on both MRE and MCE from 20 stations onward and slightly surpassed SDN at 25 stations. However, SDN achieved the highest full-network accuracy with all 26 stations (MRE = 0.034; MCE = 0.990). These findings demonstrate that reconstruction performance depends on monitoring-network density, with SHRED being particularly effective under sparse conditions and SDN and QR becoming increasingly competitive as spatial coverage increases.
- Open Access
Combining Multivariate Singular Spectrum Analysis and Dynamic Mode Decomposition for Daily Sea Surface Temperature Forecasting up to 60 Days Ahead in the Marmara Sea
Accurate long-range forecasting of sea surface temperature (SST) fields remains challenging because regional SST variability reflects the combined influence of recurrent seasonal behavior and shorter-term oceanographic processes. This study develops a forecasting framework combining Multivariate Singular Spectrum Analysis (MSSA) and Dynamic Mode Decomposition (DMD), hereafter referred to as MSSA-DMD, for daily SST forecasting in the Marmara Sea. The framework first represents the SST field in a reduced empirical orthogonal function (EOF) space, separates the temporal dynamics into reconstructed MSSA components, and independently propagates these components using DMD before reconstructing the full SST field. Daily Copernicus Marine Service SST data were evaluated for forecast horizons ranging from 1 to 60 days. The proposed approach was compared with Standard DMD, multivariate empirical mode decomposition (MEMD)–DMD, multivariate variational mode decomposition (MVMD)–DMD, direct inverted Transformer (iTransformer), MSSA-iTransformer, persistence, and climatology. MSSA-DMD provided the most consistent overall forecasting performance, with its advantage becoming particularly evident at extended lead times while maintaining stable spatial and temporal error characteristics. The improvement of MSSA-iTransformer over direct iTransformer further indicates that MSSA decomposition itself contributes substantially to predictive skill. Computational analysis additionally showed that MSSA-DMD achieved a considerably lower end-to-end execution cost than the iTransformer-based configurations. These findings demonstrate that decomposition-assisted reduced-order forecasting, particularly through MSSA-DMD, provides an effective accuracy–cost balance for long-range regional SST prediction.
Open AccessCombining Multivariate Singular Spectrum Analysis and Dynamic Mode Decomposition for Daily Sea Surface Temperature Forecasting up to 60 Days Ahead in the Marmara Sea
Accurate long-range forecasting of sea surface temperature (SST) fields remains challenging because regional SST variability reflects the combined influence of recurrent seasonal behavior and shorter-term oceanographic processes. This study develops a forecasting framework combining Multivariate Singular Spectrum Analysis (MSSA) and Dynamic Mode Decomposition (DMD), hereafter referred to as MSSA-DMD, for daily SST forecasting in the Marmara Sea. The framework first represents the SST field in a reduced empirical orthogonal function (EOF) space, separates the temporal dynamics into reconstructed MSSA components, and independently propagates these components using DMD before reconstructing the full SST field. Daily Copernicus Marine Service SST data were evaluated for forecast horizons ranging from 1 to 60 days. The proposed approach was compared with Standard DMD, multivariate empirical mode decomposition (MEMD)–DMD, multivariate variational mode decomposition (MVMD)–DMD, direct inverted Transformer (iTransformer), MSSA-iTransformer, persistence, and climatology. MSSA-DMD provided the most consistent overall forecasting performance, with its advantage becoming particularly evident at extended lead times while maintaining stable spatial and temporal error characteristics. The improvement of MSSA-iTransformer over direct iTransformer further indicates that MSSA decomposition itself contributes substantially to predictive skill. Computational analysis additionally showed that MSSA-DMD achieved a considerably lower end-to-end execution cost than the iTransformer-based configurations. These findings demonstrate that decomposition-assisted reduced-order forecasting, particularly through MSSA-DMD, provides an effective accuracy–cost balance for long-range regional SST prediction.
- Open Access
- 29 June 2026
- Published by Springer Nature in Discover Applied Sciences
- Vol. 8 (7) , 733
Hybrid principal component analysis and singular value decomposition based fuzzy models for multi-step daily rainfall forecasting
- 29 June 2026
- Published by Springer Nature in Discover Applied Sciences
- Vol. 8 (7) , 733
Open AccessHybrid principal component analysis and singular value decomposition based fuzzy models for multi-step daily rainfall forecasting
- 16 May 2026
- Published by Elsevier in Estuarine, Coastal and Shelf Science
- Vol. 339
Robust reconstruction of sea level fields from sparse observations using shallow recurrent decoders
- 16 May 2026
- Published by Elsevier in Estuarine, Coastal and Shelf Science
- Vol. 339
Robust reconstruction of sea level fields from sparse observations using shallow recurrent decoders
- Open Access
Dynamic Mode Decomposition for Forecasting Flood-Driven Sedimentation at a River Mouth: A Data-Driven Coastal Modelling
Accurate forecasting of sediment accumulation under extreme hydrodynamic forcing is essential for coastal engineering design and harbor management. This study evaluates the performance of Dynamic Mode Decomposition (DMD), optimized DMD (optDMD), and optimized DMD with stability constraints (optDMDs) for reconstructing and forecasting sediment accumulation height fields at the Dilderesi River mouth under a 50-year return period flood scenario. Sediment height fields generated using Delft3D are represented through reduced-order modal decompositions and the truncation rank is determined based on reconstruction-error analysis. Although all formulations reproduce the training data with negligible error, their predictive behavior differs during temporal extrapolation. Standard DMD exhibits rapid error growth at longer lead times. The optDMD formulation improves short- and intermediate-horizon performance but shows gradual degradation at extended lead times. Optimized DMD with stability constraints provides the most consistent long-horizon forecasts, maintaining high Nash–Sutcliffe efficiency and low RMSE across the full 9 h prediction interval. Examination of the continuous-time eigenvalue distributions and modal dynamics indicates that spectral characteristics of the reduced-order representation govern forecast robustness. The results demonstrate that enforcing spectral stability within reduced-order frameworks substantially enhances morphodynamic forecasting reliability under extreme flood conditions. The proposed approach provides a computationally efficient and physically consistent tool for sediment dynamics prediction in coastal engineering applications.
Open AccessDynamic Mode Decomposition for Forecasting Flood-Driven Sedimentation at a River Mouth: A Data-Driven Coastal Modelling
Accurate forecasting of sediment accumulation under extreme hydrodynamic forcing is essential for coastal engineering design and harbor management. This study evaluates the performance of Dynamic Mode Decomposition (DMD), optimized DMD (optDMD), and optimized DMD with stability constraints (optDMDs) for reconstructing and forecasting sediment accumulation height fields at the Dilderesi River mouth under a 50-year return period flood scenario. Sediment height fields generated using Delft3D are represented through reduced-order modal decompositions and the truncation rank is determined based on reconstruction-error analysis. Although all formulations reproduce the training data with negligible error, their predictive behavior differs during temporal extrapolation. Standard DMD exhibits rapid error growth at longer lead times. The optDMD formulation improves short- and intermediate-horizon performance but shows gradual degradation at extended lead times. Optimized DMD with stability constraints provides the most consistent long-horizon forecasts, maintaining high Nash–Sutcliffe efficiency and low RMSE across the full 9 h prediction interval. Examination of the continuous-time eigenvalue distributions and modal dynamics indicates that spectral characteristics of the reduced-order representation govern forecast robustness. The results demonstrate that enforcing spectral stability within reduced-order frameworks substantially enhances morphodynamic forecasting reliability under extreme flood conditions. The proposed approach provides a computationally efficient and physically consistent tool for sediment dynamics prediction in coastal engineering applications.
- 4 February 2026
- Published by Elsevier in Wave Motion
- Vol. 143
Improving missing data imputation in significant wave height time series data using Fourier-wavelet-based compressed sensing
- 4 February 2026
- Published by Elsevier in Wave Motion
- Vol. 143
Improving missing data imputation in significant wave height time series data using Fourier-wavelet-based compressed sensing
- 2 January 2026
- Published by Elsevier in Ocean Engineering
- Vol. 349
Sparse-to-full reconstruction of ocean wave height fields using lightweight recurrent neural models
- 2 January 2026
- Published by Elsevier in Ocean Engineering
- Vol. 349
Sparse-to-full reconstruction of ocean wave height fields using lightweight recurrent neural models
- 5 June 2025
- Published by Elsevier in Ocean Engineering
- Vol. 335
Eigen-ensemble-transformer model: An exceptionally accurate approach to spatiotemporal sea water level prediction
- 5 June 2025
- Published by Elsevier in Ocean Engineering
- Vol. 335
Eigen-ensemble-transformer model: An exceptionally accurate approach to spatiotemporal sea water level prediction