Skroufouta et al. (2025) Rainfall Disaggregation in Data-Scarce Regions Using the Random Bartlett-Lewis Rectangular Pulse Model

Identification

Journal: Climate
Year: 2025
Date: 2025-11-27
Authors: Sofia Skroufouta, Evangelos Baltas
DOI: 10.3390/cli13120242

Research Groups

Water Resources and Environmental Engineering, National Technical University of Athens, Athens, Greece

Short Summary

This study evaluates the Random Bartlett-Lewis Rectangular Pulse Model (RBLRPM) for rainfall disaggregation in data-scarce Mediterranean regions, comparing it against a machine learning benchmark and assessing pulse intensity distributions and parameter uncertainty. It finds that RBLRPM effectively reproduces essential rainfall properties, with the Gamma distribution generally outperforming the Exponential, offering a robust stochastic approach for hydrological applications.

Objective

To evaluate the performance of the RBLRPM for rainfall disaggregation in data-scarce Mediterranean regions.
To compare the stochastic RBLRPM with a deterministic machine learning (ML) benchmark method, highlighting their respective strengths and limitations.
To assess the influence of alternative pulse intensity distributions (Gamma and Exponential) on the model’s ability to reproduce rainfall statistics.
To quantify parameter uncertainty through a two-tier sensitivity and record-length analysis, testing the robustness of calibration under short datasets.
To identify the implications of these findings for hydrological modeling, flood risk assessment, and infrastructure design in regions with limited rainfall data.

Study Configuration

Spatial Scale: Four regions in Greece: Corfu, Rhodes, Alexandroupolis, and Herakleion.
Temporal Scale: Daily precipitation records (1995–2004) used for calibration (1995–1999) and hourly data (2000–2004) for validation. Disaggregation from daily to hourly resolution.

Methodology and Data

Models used: Random Bartlett-Lewis Rectangular Pulse Model (RBLRPM), Evolutionary Annealing-Simplex (EAS) optimization, Machine Learning (ML) based linear regression.
Data sources: Observed daily and hourly precipitation data from meteorological stations in Corfu, Rhodes, Alexandroupolis, and Herakleion (Hellenic National Meteorological Service).

Main Results

The RBLRPM successfully reproduces essential rainfall properties (variance, autocorrelation, skewness, and dry spell probabilities) even when calibrated with as little as three years of data.
The Gamma distribution for pulse intensity generally provides a closer match to observed rainfall statistics (autocorrelation, variability) compared to the Exponential distribution, though both remain applicable depending on the month and region.
The ML approach ensures perfect conservation of daily totals and computational efficiency but tends to smooth temporal variability and underestimate rainfall extremes.
The stochastic RBLRPM captures rainfall clustering, intermittency, and heavy tails more realistically, which is crucial for hydrological design and flood risk analysis.
Uncertainty analysis revealed that parameters controlling rainfall structure (e.g., pulse intensity, storm duration) are more sensitive to input perturbations, while storm arrival frequency (λ) shows strong stability. Longer records (5 years vs. 3 years) moderately increased parameter stability, reducing the confidence interval range for λ by approximately 2.3%.

Contributions

First systematic evaluation of the Random Bartlett-Lewis Rectangular Pulse Model (RBLRPM) under data-scarce Mediterranean conditions.
Direct comparison of the stochastic RBLRPM with a deterministic machine learning (ML) disaggregation benchmark.
Comprehensive assessment of the influence of alternative pulse intensity distributions (Gamma and Exponential) on model performance.
Quantification of parameter uncertainty through a two-tier sensitivity and record-length analysis, demonstrating robustness under limited data.
Provides clear guidance on the trade-offs between stochastic realism and computational efficiency for rainfall disaggregation in data-scarce environments.

Funding

This research received no external funding.

Citation

@article{Skroufouta2025Rainfall,
  author = {Skroufouta, Sofia and Baltas, Evangelos},
  title = {Rainfall Disaggregation in Data-Scarce Regions Using the Random Bartlett-Lewis Rectangular Pulse Model},
  journal = {Climate},
  year = {2025},
  doi = {10.3390/cli13120242},
  url = {https://doi.org/10.3390/cli13120242}
}

Original Source: https://doi.org/10.3390/cli13120242