Başkesen et al. (2025) Comparative analysis of interpolation methods for missing daily precipitation data by suggesting an alternative inverse distance weighted model

Identification

• Journal: Theoretical and Applied Climatology
• Year: 2025
• Date: 2025-11-18
• Authors: Kevser Merkür Başkesen, Yavuz Selim Güçlü, Sena Ecem Yakut Şevik, Ahmet Duran Şahin
• DOI: 10.1007/s00704-025-05907-z

Research Groups

• Basin Management, Monitoring and Allocations Branch, State Hydraulic Works (DSI), 11th Regional Directorate, Edirne, Türkiye
• Department of Civil Engineering, Hydraulic Division, Istanbul Technical University, Istanbul, Türkiye
• Department of Climate Science and Meteorological Engineering, Istanbul Technical University, Istanbul, Türkiye

Short Summary

This study compares various interpolation methods for estimating missing daily precipitation data in Istanbul, proposing an Alternative Inverse Distance Weighting (AIDW) model. The AIDW model consistently performs comparably to or slightly better than established methods like IDW and MNR-T, offering a reliable solution for data infilling.

Objective

• To compare the performance of various interpolation methods, including a newly proposed Alternative Inverse Distance Weighting (AIDW) model, for estimating missing daily precipitation data in Istanbul.
• To evaluate if the AIDW method, which normalizes station distances relative to the total sum of neighboring distances, can offer improved accuracy and adaptability compared to classical interpolation methods, particularly in areas with irregular station coverage.
• To analyze the impact of different input scenarios (number of neighboring stations and training-test data ratios) on model performance to provide an effective solution for data gaps, crucial for flood/drought analysis and water resource management.

Study Configuration

• Spatial Scale: ITU Maslak meteorological station (target) in Istanbul, Türkiye, with five nearby reference stations (Sarıyer, Eyüp, Beykoz, Şişli, and Üsküdar). The region exhibits characteristics of both Mediterranean and humid subtropical climates.
• Temporal Scale: Daily precipitation data for the period 2014–2024 for reference stations and 2017–2024 for the target station. Missing data for the ITU Maslak station were estimated for the period 2014–2017.

Methodology and Data

• Models used:
  • Alternative Inverse Distance Weighting (AIDW) - proposed
  • Inverse Distance Weighting (IDW)
  • Modified Inverse Distance Weighting 1 (MIDW1)
  • Modified Inverse Distance Weighting 2 (MIDW2)
  • Inverse Exponential Weighting Method (IEWM)
  • Modified Normal Ratio Based On Square Root Distance (MNR-T)
  • Model performance was evaluated using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Coefficient of Determination (R²), and Nash-Sutcliffe Efficiency (NSE).
  • Scenarios included varying numbers of input stations (3, 4, or 5) and training-test data ratios (60%-40%, 70%-30%, 80%-20%).
• Data sources: Daily precipitation data obtained from the General Directorate of Meteorology (MGM) of Türkiye.

Main Results

• The AIDW, classic IDW, and MNR-T methods consistently demonstrated the lowest error rates and highest performance among the six tested interpolation techniques.
• The proposed AIDW model achieved MAE = 0.001027 m, RMSE = 0.002051 m, R² = 0.930, and NSE = 0.930, showing performance comparable to or slightly better than IDW (MAE = 0.001006 m, RMSE = 0.002014 m, R² = 0.932, NSE = 0.935) and MNR-T (MAE = 0.001039 m, RMSE = 0.002059 m, R² = 0.929, NSE = 0.932).
• MIDW1 and MIDW2 consistently exhibited the weakest performance, with RMSE values often exceeding 0.0035 m and R²/NSE values frequently below 0.85. IEWM showed intermediate performance.
• The optimal scenario for estimating missing data involved using 4 reference stations (Eyüp, Şişli, Beykoz, and Üsküdar) with a 70% training – 30% test data ratio, which generally yielded the lowest MAE and RMSE values.
• Increasing the number of input stations from three to five generally improved the accuracy of the best-performing models.
• The 70% training – 30% test split often resulted in the lowest MAE and RMSE, while the 80% training – 20% test split generally produced the best R² and NSE values. The 60% training – 40% test scenario consistently showed the weakest performance.

Contributions

• Introduction and validation of an Alternative Inverse Distance Weighting (AIDW) model that enhances weight calculation by normalizing each station's distance relative to the total sum of all neighboring distances, leading to more balanced and accurate estimations, particularly in areas with irregular station networks.
• A comprehensive comparative analysis of AIDW against five established deterministic interpolation methods, providing robust evidence of its competitive performance for daily precipitation data infilling.
• Identification of optimal configurations (number of reference stations and training-test data splits) for missing daily precipitation data estimation in a climatically complex urban environment like Istanbul.
• Provision of an effective and reliable solution for hydrological data completion, which is critical for informed decision-making in water resource management, flood forecasting, and climate change impact assessment.

Funding

• Istanbul Technical University Scientific Research Projects Coordination Unit.
• Project Grant Number: MGA-2024-45366.

Citation

@article{Başkesen2025Comparative,
  author = {Başkesen, Kevser Merkür and Güçlü, Yavuz Selim and Şevik, Sena Ecem Yakut and Şahin, Ahmet Duran},
  title = {Comparative analysis of interpolation methods for missing daily precipitation data by suggesting an alternative inverse distance weighted model},
  journal = {Theoretical and Applied Climatology},
  year = {2025},
  doi = {10.1007/s00704-025-05907-z},
  url = {https://doi.org/10.1007/s00704-025-05907-z}
}