⚡ electrical engineering

Bayesian–Deterministic Seismic-Demand Scenarios for Baish Dam, Southwestern Saudi Arabia: An Engineering Hazard-Screening Assessment

This study develops a Bayesian–deterministic stochastic framework to generate physically traceable seismic-demand scenarios for the Baish Dam in Saudi Arabia, utilizing the 2014 Jazan earthquake source to establish conditional engineering screening envelopes while identifying rupture-length constraints and uncertainty drivers for future finite-fault evaluations.

Original authors: Hashim Al Aoun¹, Ali Abdelfattah¹, Abdullah Al-Amri¹, Hassan Alzahrani¹

Published 2026-10-23

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Hashim Al Aoun¹, Ali Abdelfattah¹, Abdullah Al-Amri¹, Hassan Alzahrani¹

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Earthquakes do not strike with a single, predictable rhythm; they arrive as a complex mix of shaking that depends on where the ground breaks, how the energy travels through the earth, and what the ground feels like right beneath a building. For engineers designing critical structures like dams, knowing the exact nature of this shaking is a matter of public safety. They must imagine the worst possible earthquake that could reasonably happen nearby and calculate how the structure would react. This process, known as seismic hazard assessment, involves creating realistic scenarios of ground motion. It requires combining the physics of how an earthquake source radiates energy with the way that energy fades as it travels and how local soil conditions amplify or dampen the vibrations. The goal is not just to predict a single number, but to understand the range of possible shaking intensities, accounting for the fact that scientists cannot know every detail of the earth's interior with perfect precision.

In the southwestern corner of Saudi Arabia, a team of researchers has applied this rigorous thinking to the Baish Dam, a vital water resource located just 36 kilometers from the epicenter of a significant earthquake that struck the nearby Jazan region in 2014. The team, led by geologists and geophysicists at King Saud University, set out to create a detailed set of "what-if" scenarios for the dam. Instead of relying on a single, simplified estimate of danger, they built a sophisticated computer model that simulates nine different earthquake sizes, ranging from a moderate tremor to a massive, catastrophic event. They used the specific characteristics of the 2014 Jazan earthquake—its location, depth, and the direction the fault moved—as the foundation for all these simulations. By keeping the distance and location fixed, they could isolate exactly how the shaking would change as the magnitude of the earthquake increased, providing a clear picture of the forces the dam might face.

The researchers did not just run one calculation; they ran hundreds of variations to account for the unknowns in the earth's crust. They treated key factors, such as the energy released by the fault and how the ground absorbs that energy, as variables with a range of likely values rather than fixed numbers. This approach, known as Bayesian analysis, allowed them to produce not just a single prediction, but a spectrum of probable outcomes with upper and lower bounds. They tested two different types of ground conditions at the dam site: one representing hard, competent rock and another representing slightly softer, stiffer soil. This distinction proved critical, as the simulations showed that the softer soil condition could amplify the shaking significantly more than the hard rock, particularly for the shorter, faster vibrations that are most dangerous to tall structures.

The results of these simulations revealed a clear and escalating pattern of risk. For the smaller, more frequent earthquakes in their scenarios, the shaking at the dam site remained relatively low. However, as the simulated magnitude increased, the intensity of the shaking grew rapidly. When they modeled the largest possible scenario, a magnitude 8.0 earthquake, the predicted shaking became severe. Under the softer soil conditions, the peak shaking reached levels that could challenge the safety of the dam, with the most extreme estimates suggesting ground acceleration could exceed 1.1 times the force of gravity. This is a substantial force, capable of causing significant stress on concrete and earth structures. The study also highlighted that the uncertainty in these predictions is driven mostly by how much energy the fault releases and how quickly the ground absorbs that energy as the waves travel, rather than by small variations in the fault's orientation.

A crucial part of the team's work was to check the physical reality of their largest scenario. While they included the magnitude 8.0 case to test the absolute upper limits of safety, they found that such an earthquake would require a fault rupture nearly 230 kilometers long. The mapped faults in the Jazan region are not that long. This means that while the magnitude 8.0 scenario is a useful tool for stress-testing the dam's design limits, it is not a realistic prediction of what the local geology can actually produce. The researchers concluded that for earthquakes of magnitude 6.5 and larger, the simple point-source model they used is no longer sufficient, and more complex simulations that account for the length and direction of the breaking fault will be necessary before final design decisions are made.

The study provides a clear, conditional set of guidelines for engineers and safety officials. It confirms that the Baish Dam is sensitive to the specific type of ground it sits on, with softer soil leading to significantly higher shaking forces. It also establishes a range of likely shaking intensities for various earthquake sizes, complete with the margins of error that come from not knowing every detail of the earth's structure. The work serves as a vital screening tool, helping to prioritize where further investigation is needed, such as more precise measurements of the soil under the dam and more detailed modeling of larger earthquakes. By mapping out these scenarios, the researchers have given dam operators a better understanding of the forces they must prepare for, ensuring that the safety of this critical infrastructure is built on a foundation of realistic, physics-based scenarios rather than guesswork.

Technical Summary: Bayesian–Deterministic Seismic-Demand Scenarios for Baish Dam

Problem Statement
Dam safety assessments require physically traceable earthquake scenarios that explicitly account for epistemic uncertainty. While Probabilistic Seismic Hazard Analysis (PSHA) is the standard for code-based design, Deterministic Seismic Hazard Assessment (DSHA) is essential for critical facilities to evaluate ground-motion consequences of credible, controlling scenarios. However, conventional DSHA often relies on a single "Maximum Credible Earthquake" (MCE) at a closest-approach geometry, potentially overlooking magnitude sensitivity and the propagation of uncertainty in source, path, and site parameters. This study addresses the need for a scenario-based engineering extension of DSHA for Baish Dam in southwestern Saudi Arabia, located approximately 36 km from the tectonically active 2014 Jazan earthquake source. The goal is to quantify seismic demand across a range of magnitudes while explicitly treating epistemic uncertainty, rather than defining a single fault-specific MCE.

Methodology
The study employs a deterministic Bayesian point-source stochastic framework to simulate seismic demand for nine scenario magnitudes ( $M_w$ 4.0–8.0) originating from the fixed epicenter of the January 23, 2014, Jazan earthquake (hypocentral distance = 37.0 km).

Source Characterization: The analysis utilizes the well-constrained focal mechanism of the 2014 Jazan mainshock (strike 259°, dip 87°, rake −171°), explicitly incorporating these parameters into the double-couple radiation term. Unlike conventional DSHA, the study does not select a single controlling magnitude but sweeps through nine scenarios. The $M_w$ 8.0 case is treated strictly as a conservative upper engineering-screening bound, not a demonstrated fault-specific maximum, as its estimated rupture length (~230 km) exceeds mapped local fault dimensions.
Stochastic Simulation: Ground motions are generated using a Brune omega-square source spectrum. The model accounts for geometrical spreading, anelastic path attenuation (using regional coda-Q relations), site amplification based on $V_{S30}$ , and high-frequency diminution ( $\kappa$ ).
Site Conditions: Two engineering site classes are evaluated: NEHRP B/C ( $V_{S30}$ = 760 m/s) and NEHRP C ( $V_{S30}$ = 560 m/s).
Bayesian Uncertainty Propagation: Epistemic uncertainty in source (strike, dip, rake, depth), path (stress drop, $Q_0$ , $\eta$ ), and site ( $\kappa_0$ , amplification scale) parameters is modeled using truncated Gaussian distributions. For each scenario, 500 posterior realizations are sampled and propagated through the stochastic model to generate response spectra with credible intervals (5th–95th percentiles).
Validation and Screening: Results are benchmarked against an independent regional ground-motion prediction equation (Kiuchi et al., 2019). Additionally, a rupture-length screening based on Wells and Coppersmith (1994) is applied to assess the physical plausibility of the point-source approximation for larger magnitudes.

Key Results

Magnitude Scaling and Site Effects: Seismic demand increases systematically with magnitude. For the $M_w$ 8.0 screening scenario, median Peak Ground Acceleration (PGA) values are 0.202 g (NEHRP B/C) and 0.321 g (NEHRP C). The NEHRP C condition yields significantly higher demands (approx. 1.6 times higher for spectral acceleration) due to lower $V_{S30}$ and frequency-dependent amplification.
Uncertainty Quantification: The Bayesian analysis reveals substantial epistemic uncertainty. For $M_w$ 8.0 under NEHRP C conditions, the 95th percentile of the maximum spectral acceleration ( $SA_{max}$ ) reaches 1.134 g, compared to a median of 0.704 g. The 5th–95th percentile credible bounds for $SA_{max}$ are 0.228–0.759 g (NEHRP B/C) and 0.380–1.134 g (NEHRP C).
Sensitivity Analysis: Stress drop and regional attenuation parameters ( $Q_0$ and its frequency dependence) are identified as the dominant contributors to uncertainty in the predicted spectral demand, particularly for the $M_w$ 8.0 scenario.
Benchmark Discrepancy: Comparison with the regional empirical benchmark shows a systematic positive offset in the stochastic PGA predictions. The stochastic model predicts PGA values approximately 28% higher than the benchmark at $M_w$ 4.0, increasing to a factor of ~2.7 at $M_w$ 8.0. The authors attribute this to differences in faulting mechanisms (strike-slip vs. normal-faulting swarms) and stress-drop assumptions, noting the benchmark is extrapolated beyond its primary empirical range for high magnitudes.
Applicability Limits: Rupture-length screening indicates the point-source approximation is defensible for $M_w$ 4.0–6.0. For $M_w \ge 6.5$ , the estimated rupture length approaches or exceeds half the source-to-site distance, suggesting that $M_w \ge 7.0$ scenarios require finite-fault simulations before design-basis use.

Significance and Claims
The paper positions its contribution as a scenario-based engineering extension of DSHA rather than a strict application of the conventional single-controlling-earthquake procedure. Its primary significance lies in:

Explicit Mechanism Coupling: Integrating the specific focal mechanism of the 2014 Jazan event directly into the radiation pattern, providing a mechanism-specific representation not available in standard empirical models.
Uncertainty Transparency: Propagating epistemic uncertainty through a Bayesian framework to provide credible intervals for seismic demand, moving beyond single-value deterministic estimates.
Engineering Screening Utility: Providing conditional demand envelopes (median and 95th percentile) suitable for preliminary dam-safety screening, instrumentation planning, and emergency scenario development.

The authors explicitly state that these results are not a formal design-basis seismic hazard assessment or a certified worst-case earthquake. The $M_w$ 8.0 scenario is a conservative screening bound, and the study identifies the need for future work to establish fault-specific maximum magnitudes, resolve the empirical benchmark offset, and transition to finite-fault simulations for final design criteria. The study serves as a tool for prioritizing dynamic analyses and site investigations rather than replacing facility-specific probabilistic hazard analyses.

More like this