MODELLING OF EXTREME RAINFALL IN PUNJAB: PAKISTAN USING BAYESIAN AND FREQUENTIST APPROACH

Touqeer AhmadApplied Ecology and Environmental Research 2019 · DOI

Abstract

In this study we compared the efficiency of frequentist approach with Bayesian approach by carrying out extreme value analysis of Annual Maximum Daily Rainfall (AMDR). For frequentist frequency analysis of AMDR, we used the data of one station i.e. Lahore in Punjab province, Pakistan while for Bayesian analysis we used the data of three other neighboring stations as prior information. During frequentist approach, Generalized Extreme Value (GEV) was found to be a best-fit distribution on the data. In frequentist method, the parameters of GEV distribution were estimated using Maximum Likelihood Estimation (MLE), while in the Bayesian framework the Markov Chain Monte Carlo (MCMC) simulation technique along with Metropolis-Hasting algorithm and Gibbs sampler were implemented. Findings of this study indicate that despite the asymptotic requirements of the MLE, its performance can be enhanced by adopting the Bayesian MCMC paradigm. In order to acquire the posterior densities of GEV parameters, non-informative and informative priors based on the historical data of surrounding weather stations were employed. The result of Bayesian MCMC might be affected by the choice of priors. In addition, the performance of the parameters estimation methods was verified by employing several robustness measures. Robustness measures results proved that the Bayesian MCMC method performed better than MLE in estimating GEV parameters and future return levels. Therefore, the findings of these analyses could be helpful in adopting proper flood protection measures and designing infrastructures of culverts, buildings, bridges and dams in the region.

Introduction

Predicting "once-in-a-century" storms is a critical challenge in hydrology, particularly in regions like Pakistan where monsoon-driven floods have caused over $38 billion in losses over the last seven decades. Traditional statistical methods often struggle with extreme events due to the scarcity of data points. This study evaluates whether a Bayesian framework, which allows for the integration of regional knowledge, can provide more reliable predictions than the standard Frequentist approach for infrastructure design.

Methodology

The research utilizes Extreme Value Analysis (EVA) centered on the Generalized Extreme Value (GEV) distribution. The GEV distribution is the mathematical standard for modeling "block maxima"—the single highest rainfall day recorded each year.

Two primary estimation techniques were compared:

  1. Frequentist (Maximum Likelihood Estimation - MLE): A standard method that identifies parameter values maximizing the probability of observed data.
  2. Bayesian MCMC: This approach treats parameters as random variables. Using Metropolis-Hastings and Gibbs sampling, the researchers simulated posterior distributions. A key innovation was the use of Informative Priors, where historical data from neighboring stations informed the model for the target site.

Study Area and Data

The study focused on Lahore as the primary target station, with data spanning 1985–2014. To inform the Bayesian model, data from three neighboring stations—Faisalabad, Sialkot, and Jhelum—were used as priors. These stations were selected based on their proximity (ranging from 105km to 164km) and data quality.

Map of study locations in Punjab, Pakistan, showing distances between Lahore and neighboring stations.

The descriptive statistics for these locations highlight the variability in rainfall across the region, with Sialkot showing the highest mean rainfall and Jhelum recording the highest maximum daily rainfall.

Descriptive statistics of Annual Maximum Daily Rainfall (AMDR) for the four study locations.

Key Results and Discussion

The Bayesian MCMC method consistently outperformed MLE across several robustness measures, including Relative Root Mean Square Error (RRMSE) and the Probability Plot Correlation Coefficient (PPCC).

  • "Borrowing Strength": By using informative priors from neighboring stations, the Bayesian model reduced uncertainty (standard deviation) in its estimates.
  • Geographical Influence: The effectiveness of priors was linked to proximity; Faisalabad, the closest station to the south-west, provided the most significant improvement in model fit (PPCC of 0.9721 compared to 0.8858 for MLE).
  • Risk of Underestimation: The study found that standard MLE might underestimate extreme return levels. For a 100-year event, MLE estimated ~214mm, whereas the Bayesian approach suggested ~234mm. This 20mm discrepancy represents a critical margin for engineering safety.

Conclusion

The findings demonstrate that the Bayesian MCMC approach, particularly when utilizing informative priors, offers a superior framework for modeling extreme rainfall in data-scarce environments. While the study assumes regional uniformity, future work should explore full spatial modeling to account for topographical variations. These more precise estimates are essential for designing resilient infrastructure such as dams, bridges, and drainage systems in flood-prone regions of Pakistan.

Summary updated

Abstract

In this study we compared the efficiency of frequentist approach with Bayesian approach by carrying out extreme value analysis of Annual Maximum Daily Rainfall (AMDR). For frequentist frequency analysis of AMDR, we used the data of one station i.e. Lahore in Punjab province, Pakistan while for Bayesian analysis we used the data of three other neighboring stations as prior information. During frequentist approach, Generalized Extreme Value (GEV) was found to be a best-fit distribution on the data. In frequentist method, the parameters of GEV distribution were estimated using Maximum Likelihood Estimation (MLE), while in the Bayesian framework the Markov Chain Monte Carlo (MCMC) simulation technique along with Metropolis-Hasting algorithm and Gibbs sampler were implemented. Findings of this study indicate that despite the asymptotic requirements of the MLE, its performance can be enhanced by adopting the Bayesian MCMC paradigm. In order to acquire the posterior densities of GEV parameters, non-informative and informative priors based on the historical data of surrounding weather stations were employed. The result of Bayesian MCMC might be affected by the choice of priors. In addition, the performance of the parameters estimation methods was verified by employing several robustness measures. Robustness measures results proved that the Bayesian MCMC method performed better than MLE in estimating GEV parameters and future return levels. Therefore, the findings of these analyses could be helpful in adopting proper flood protection measures and designing infrastructures of culverts, buildings, bridges and dams in the region.
BibTeX
@inproceedings{ahmad2019modelling,
  author = {Touqeer Ahmad},
  title = {{MODELLING OF EXTREME RAINFALL IN PUNJAB: PAKISTAN USING BAYESIAN AND FREQUENTIST APPROACH}},
  booktitle = {Applied Ecology and Environmental Research},
  year = {2019},
  doi = {10.15666/aeer/1706\_1372913748},
}
Figure