Common Dispersion Modelling Errors

Design Hazard Review for New Processing Facility

Common Dispersion Modelling Errors are a recurring theme whenever industrial teams review toxic release scenarios, emergency planning zones, or environmental impact predictions. In high-hazard UAE sectors—oil and gas processing, petrochemical clusters, power generation, logistics terminals, and large construction sites that handle hazardous substances—dispersion outputs often shape decisions that carry operational, regulatory, and reputational consequences. When models are built on shaky inputs or mismatched methods, the result can be understated risks, misplaced emergency resources, or over-designed controls that inflate cost without improving safety. This article examines where dispersion studies most often go wrong, how those missteps manifest in practice, and what HSE and process safety leaders can do to improve reliability without turning modelling into an academic exercise detached from site realities.

Although dispersion modelling spans environmental compliance and process safety, the focus here is on releases of toxic or asphyxiant gases and vapours, and on the real-world factors that influence predicted concentrations at worker locations, site boundaries, and neighbouring communities. We will reference common toolchains (Gaussian, dense-gas, and computational fluid dynamics), typical inputs (source terms, meteorology, building data), and the decision frameworks that rely on model output: QRA contours, emergency response plans, shelter-in-place triggers, and land-use planning inputs. Along the way, we will connect these elements to prevalent standards and practices used across the UAE, such as OSHAD requirements, Dubai Municipality approvals, Civil Defence expectations for emergency planning, ADNOC HSE codes of practice, and management systems under ISO 45001 and ISO 14001.

Who Does This Affect?

Common Dispersion Modelling Errors

Dispersion modelling touches many functions, which is partly why errors arise—assumptions cross boundaries between engineering, HSE, and operations. Those who should pay particular attention include:

  • Process Safety and HSE Managers responsible for QRA contours, hazard studies (HAZOP/HAZID), LOPA/SIL selections (IEC 61511 context), and emergency planning.
  • Environmental Specialists preparing impact assessments and permit applications, especially when modelling odours or toxic air pollutants near sensitive receptors.
  • Operations and Maintenance Leaders who own isolation timings, relief system behaviour, and gas detection strategies that feed the source term.
  • Project Engineering teams and EPC contractors who set the physical layout, building heights, and ventilation intakes that drive downwash and exposure.
  • Regulatory and Compliance Officers coordinating with ADNOC group requirements, OSHAD frameworks, Dubai Municipality approvals, Emirates Civil Defence codes, and MOHRE occupational exposure expectations.
  • Emergency Response and Business Continuity teams who translate model outputs into shelter-in-place instructions, evacuation routes, and muster point locations.

Facility types where the stakes are especially high include refineries and gas processing plants, petrochemical crackers and tank farms, offshore platforms and onshore terminals, power stations with ammonia-based SCR systems, bulk liquid storage and distribution terminals, chemical warehouses in free zones, and major construction projects that store or use hazardous agents (e.g., tunnel ventilation systems with fuel handling, high-rise sites with refrigerant or solvent storage). In each, dispersion predictions underpin controls that must work on the worst day, not just on paper.

Why This Topic Matters

Common Dispersion Modelling Errors

It is not just about producing a contour map. Dispersion outputs frame arguments about what is reasonably practicable. In ALARP demonstrations and LOPA calibrations, toxic frequency-consequence curves rely on predicted dose and exposed populations. If those predictions are built on simplistic source terms or generic meteorology, decision-makers may select the wrong layers of protection or misjudge emergency response needs. For example, a contour that underestimates ammonia concentrations at an air intake could encourage an evacuation when shelter-in-place would be safer—or, conversely, lead to sheltering where plume infiltration is likely. Either outcome introduces avoidable harm.

Regulatory expectations heighten the stakes. ADNOC HSE codes of practice and facility interface guidelines commonly expect dispersion to be coherent with QRA and with credible process scenarios. OSHAD and Dubai Municipality reviews look for appropriate use of models, receptor definitions, and averaging times, especially for public exposures. Civil Defence authorities examine emergency planning assumptions, challenging optimistic gas detection claims or isolation times that lack operational realism. MOHRE’s focus on worker health further raises the bar for occupational exposure criteria and re-entry timing after a release.

Finally, cost and credibility are tied to getting this right. Overly conservative modelling can drive unnecessary retrofits, large fixed water systems, or redundant detection networks; over-optimistic modelling risks actual harm and regulatory non-compliance. Striking the right balance requires clarity about where errors most often creep in, and how to guard against them.

Key Points to Understand

Common Dispersion Modelling Errors

Because dispersion studies draw from multiple disciplines, the most common issues fall into a few recurring domains: source term definition, meteorology, model selection, geometry/terrain, receptors and endpoints, and model configuration. Addressing these domains systematically prevents many downstream headaches.

1) Source term realism is the foundation

Common Dispersion Modelling Errors

Almost every flawed dispersion study can be traced back to a weak source term. A credible source term captures composition, temperature, pressure, phase behaviour, and time-varying release rate. Frequent missteps include:

  • Treating the entire inventory as instantaneously released, or conversely assuming unrealistic isolation/closure times without validating with operations.
  • Ignoring two-phase flashing for high-pressure liquefied gases (e.g., chlorine, ammonia, LPG components), which affects jet momentum, droplet rainout, and pool evaporation.
  • Overlooking heat effects (auto-refrigeration, dissolution heat in aqueous releases), which influence density and vapour generation.
  • Assuming a constant leak rate rather than time-dependent relief valve dynamics, choked flow transitions, or operator-driven mitigations.

Good practice is to couple process modelling or validated engineering calculations (e.g., relief scenarios, hole size distribution from HAZOP/LOPA) with release physics consistent with the selected dispersion model. That connection must include rainout estimates, pool spread and depth, and the effect of mitigation (scrubbers, water curtains, enclosures) on effective release characteristics.

2) Meteorology must match the modelling objective

Common Dispersion Modelling Errors

Using airport data for a coastal terminal or desert site may be convenient, but it invites error. Sea breezes, shamal winds, nocturnal inversions, and frequent calm conditions common in the UAE materially alter plume behaviour, especially for dense gases hugging the ground. Typical pitfalls include:

  • Substituting calm winds with arbitrary low speeds, which artificially disperses dense plumes.
  • Applying annual average datasets to acute toxic scenarios where stability class and nighttime conditions dominate worst-case exposure.
  • Mismatched averaging times (e.g., using 1-hour met statistics to evaluate 10-minute AEGL endpoints) that dilute peak concentrations.
  • Ignoring surface roughness and mixing height appropriate to refineries versus open desert, skewing vertical dispersion.

On-site meteorological monitoring provides the best fidelity. Where that is not available, regional datasets should be adjusted with surface characteristics and stability analyses that reflect the site’s true conditions, acknowledging uncertainty ranges in the final interpretation.

3) Choose the model for the physics you face

Common Dispersion Modelling Errors

Neutrally buoyant Gaussian models (e.g., AERMOD) are useful for many air quality problems but can under-predict for dense gas releases. Conversely, dense-gas models (e.g., SLAB-type algorithms within tools like PHAST/SAFETI) can over-predict for light, hot jets or for releases that quickly transition to buoyancy. Common mistakes include:

  • Using a single model for all scenarios irrespective of density, momentum, and thermal buoyancy.
  • Invoking CFD to “solve everything” without adequate boundary conditions, mesh independence checks, or validation, producing visually convincing but unverified detail.
  • Not accounting for building downwash or wake effects that dominate dispersion in congested process units and pipe racks.

Select models based on expected regimes: dense-gas for cold, heavier-than-air clouds; Gaussian or puff models for neutrally buoyant releases; CFD for localized, building-dominated flows where detailed recirculation governs exposure. The model’s limits must be stated up front.

4) Geometry and terrain matter more than you think

Common Dispersion Modelling Errors

Flat-plate assumptions break down near tall equipment, tank bunds, warehouses, and control buildings. Wake recirculation can trap a plume over an air intake or route it around obstacles to unexpected hotspots. Terrain steps, quay walls at marine terminals, or dunes near desert pipelines also alter flow patterns. If building data are rough or outdated, even the best source term and meteorology will disappoint. Capturing key dimensions, porosity (e.g., pipe racks vs. solid walls), and intake/exhaust positions is essential, even for screening models that parameterize downwash.

5) Receptors and toxic endpoints must match the decision

Common Dispersion Modelling Errors

One of the easiest errors to make is to choose toxic criteria that do not match the question. AEGL and ERPG values vary by exposure duration; IDLH and STELs serve worker safety but not necessarily public protection. Endpoints for decision-making (e.g., emergency planning zones, shelter-in-place triggers) must align with the likely exposure time and the vulnerability of the receptor population:

  • Workers near source: short durations, higher tolerance thresholds; relevant to MOHRE occupational health expectations.
  • Control room occupants: potential infiltration; relevant to HVAC design and shelter-in-place integrity testing.
  • Public receptors at site boundaries: longer durations, more conservative endpoints; relevant to Dubai Municipality and Civil Defence reviews.

Consistent units are non-negotiable: mixing mg/m³ and ppm without temperature and pressure normalization is a classic cause of erroneous contour interpretation.

6) Configuration choices can bias results

Grid size, domain extent, time step, and run duration materially affect peak capture, especially for puffs or time-varying releases. If the receptor grid is too coarse, peak concentrations are averaged away; if the domain is too small, reflected boundaries clip the plume. Downwash algorithms require correct building heights and proximity; deposition or reactivity flags must be switched with care. These “knobs” are easy to mis-set and can be hidden defaults in software tools.

Typical Modelling Error Likely Consequence Quick Diagnostic Check
Assuming steady leak rate for a flashing release Underestimation of early peak; wrong shelter/evacuation timing Plot time-history of mass flux and temperature; verify choked/unchoked transitions
Using airport met data without stability adjustment Under-predicted nighttime concentrations for dense gases Compare with site roughness and nocturnal stability; test sensitivity to Class F
Gaussian model applied to heavy vapor cloud Excess lateral spreading; lower near-field peak than credible Check density ratio and Froude number; screen with a dense-gas model
Ignoring building downwash near intake Missed hot spot on leeward side; intake contamination underestimated Run with and without downwash; compare plume centreline deflection
Mismatched exposure endpoint duration Contours too small/large; poor emergency planning decisions Confirm AEGL/ERPG duration equals model averaging window
Coarse receptor grid Peak concentrations smeared; false pass against criteria Refine grid near source and intakes; check convergence
Unit conversion errors (ppm vs mg/m³) Order-of-magnitude error in interpretation Document T/P assumptions; replicate with alternate unit set

Robust governance around assumptions and documentation helps prevent these pitfalls. Teams that maintain version-controlled inputs, justification memos for each modelling choice, and peer reviews catch most issues before decisions are made. Curated technical resources—such as CABINET NPM—can be useful for preserving checklists, calculation templates, and lessons learned so that institutional memory survives staff rotations.

How Does This Work in Practice?

Bringing the above into daily engineering and HSE workflows requires connecting process knowledge, site realities, and regulatory context. A typical chain starts upstream in hazard studies and runs through to emergency planning.

From HAZOP/HAZID to scenarios: Hazard identification flags credible loss-of-containment modes—small-bore leaks, pump seal failures, PSV lifts, loading arm ruptures. LOPA layers assign frequencies, and in IEC 61511 contexts, SIL targets may depend on consequence severity informed by dispersion. If the hazard study captures isolation timings, ESD logic, and operator response windows, these become source-term inputs rather than vague “assumptions.”

Source term development: Engineering calculations define hole sizes, relief rates, and phase splits. Where utilities interact (e.g., water deluge on ammonia), the combined effect on evaporation and heat transfer is quantified. Operations validate that isolation times and detection setpoints reflect real practice, not idealised intent—an important bridge between design and reality.

Meteorology and site data: If on-site met towers exist, data are analysed for stability class distribution, calm occurrence, and sea-breeze cycles. Without them, a hybrid dataset is constructed from regional data adjusted for surface roughness representative of refineries versus open desert corridors like pipelines. Geometry models compile building dimensions, intake/exhaust elevations, and porous structures like pipe racks. These details feed downwash algorithms or CFD domains.

Model selection and configuration: For dense-gas releases (e.g., refrigerated ammonia), a dense-gas algorithm is chosen for near-field predictions; if far-field public impacts are also relevant, a staged approach compares results with Gaussian puffs at downwind distances where buoyancy transition is expected. Time steps, receptor grids, and domain extents are tested for convergence. Where building wakes dominate exposures—such as control building intakes shielded by process units—CFD may be used with strict verification: mesh independence, turbulence model sensitivity, and validation against wind tunnel benchmarks if available.

Decision translation: Outputs are mapped to decisions. Emergency planning zones consider AEGL/ERPG durations; shelter-in-place advisories reference likely infiltration; HVAC shutdown times for control rooms are cross-checked with predicted arrival times; muster points are evaluated against plume paths. For projects, air intake placement, duct routing, and stack heights are adjusted where contours indicate risk. In QRAs, event trees use dose-response or probit functions aligned with endpoints to estimate individual and societal risk compatible with ADNOC HSE acceptance criteria.

Regulatory coordination: During submissions to Dubai Municipality or in OSHAD-aligned management system updates, the documentation highlights modelling rationales, data sources, and conservative choices. Civil Defence reviews of emergency plans are supported with clear maps showing decision thresholds and the basis for them. Auditors under ISO 45001/14001 appreciate traceable logic from hazard identification to risk control and monitoring.

Benefits and Practical Use Cases

When dispersion modelling avoids the common traps, its value becomes tangible across operations, projects, and regulatory interfaces. A few concrete examples illustrate the range:

  • Emergency response tuning: By calibrating models with realistic isolation times and meteorology, a gas processing plant refines its shelter-in-place protocol, reducing unnecessary evacuations and the risk of exposure during movement. Muster points move out of recirculation zones predicted near tall fractionation towers.
  • HVAC and control building design: A refinery upgrades air intakes after modelling shows downwash-driven hotspots at the original location. Filtration and shutdown logic are aligned with predicted arrival times and AEGL-2 thresholds.
  • Detection and ESD optimisation: LOPA decisions benefit from dose calculations, leading to revised gas detector locations and voting logic that demonstrably cut early-peak exposure from flashing releases, supporting SIL verification under IEC 61511.
  • Land-use and stakeholder engagement: A coastal terminal demonstrates, using dense-gas modelling under sea-breeze conditions, that planned housing set-backs remain outside AEGL-2 contours for credible events—strengthening its case in municipal planning consultations.
  • Turnaround planning: Construction teams near operating units use modelling to determine time-of-day work windows that avoid stable nighttime conditions, reducing risk to temporary workforces and aligning with MOHRE occupational exposure vigilance.
  • Environmental compliance confidence: An ammonia SCR system at a power plant quantifies potential releases from storage and dosing equipment, informing spill containment and ventilation design to remain compliant with facility permits and ISO 14001 commitments.

Across these scenarios, the benefit is not a prettier contour; it is a crisper decision with documented logic that stands up to internal challenge and external review.

Best Practices and Points of Caution

Because many errors are preventable with disciplined practice, the following recommendations help teams build credibility into their studies from the outset:

  • Start with the scenario logic: Trace each modelled case back to HAZOP/HAZID nodes, include enabling conditions, and capture LOPA assumptions for isolation and detection. Consistency with QRA is essential.
  • Codify the source term: Use engineering calculations that reflect phase behaviour, rainout, and pool dynamics. Document operator actions and ESD timings with operations sign-off.
  • Use meteorology fit for purpose: Prefer site data; if not available, justify substitutes, capture stability classes, and test sensitivity to calm and nighttime conditions.
  • Match model to physics: Apply dense-gas models where appropriate; reserve CFD for building-dominated flows and perform verification (mesh independence, boundary sensitivity).
  • Represent key geometry: Include buildings, tanks, berms, and air intakes with correct heights and porosity; use downwash algorithms or CFD accordingly.
  • Choose endpoints that fit decisions: Align AEGL/ERPG durations with exposure windows; avoid mixing worker and public criteria without clarity.
  • Control configuration drift: Standardize grids, domains, and run settings; run convergence checks and preserve configuration files under version control.
  • Peer review and back-calculation: Have an independent reviewer reproduce at least one scenario end-to-end, checking units and conversions explicitly.
  • Keep the story auditable: Maintain an assumptions log, change log, and a one-page rationale for each scenario, ready for regulator or internal audit.

There are also cautions worth repeating:

  • Avoid “worst of everything” stacking that combines mutually exclusive conditions; instead, bracket plausible bounds and explain choices.
  • Do not equate visual detail with accuracy; CFD that is not validated can mislead more than a simple, well-parameterized dense-gas run.
  • Be careful with mitigation credits; proof of performance (test data, vendor curves, or conservative derating) is necessary before taking large risk reductions.
  • Remember that internal consistency across QRA, SIL/LOPA, and emergency plans matters as much as any single contour.

Frequently Asked Questions About Common Dispersion Modelling Errors

How do I decide between a dense-gas model and a Gaussian model for a given release?

Check the density ratio of the cloud to air, expected temperature after flashing, and jet momentum. Cold, heavier-than-air clouds from liquefied gases (e.g., ammonia, chlorine) typically require a dense-gas model at least in the near field. Hot, buoyant jets or quickly mixing vapours often suit Gaussian or puff models. If the cloud transitions (dense near field, neutral far field), use a staged approach or compare both methods where appropriate.

What meteorological data are acceptable for acute toxic modelling in the UAE?

On-site data are preferred, ideally with stability classification and sufficient record to capture diurnal and seasonal patterns. If on-site data are unavailable, regional stations can be used with adjustments for surface roughness, stability class distribution, and site topography. Many reviewers will scrutinize the treatment of calm conditions and nighttime inversions because they drive worst-case toxic exposures for dense gases.

Which exposure averaging time should I use for toxic endpoints?

Use the duration associated with the endpoint (e.g., AEGL-1/2/3 at 10 minutes versus 60 minutes) matched to the exposure window relevant to the decision. Worker protection near source may warrant shorter durations; public protection at boundaries often aligns with 10- or 60-minute endpoints. The averaging window in the model must align with the selected endpoint to avoid under- or over-predicting risk.

When is CFD warranted, and what are the common pitfalls?

CFD is warranted when building and terrain effects dominate (e.g., complex process units, control building intakes, street-canyon effects in dense facilities) and a simpler model cannot represent recirculation accurately. Pitfalls include inadequate boundary conditions, coarse meshes that smear gradients, failure to perform mesh independence studies, and a lack of validation. Without verification, CFD can give a false sense of precision.

How should I represent mitigation measures like water curtains, enclosures, or scrubbers?

Mitigation should be credited only with documented efficacy. For water curtains, use vendor data and conservative derating; for enclosures, quantify leakage and ventilation rates; for scrubbers, apply removal efficiencies supported by performance curves at relevant flow rates. Then modify the source term (e.g., reduced effective release rate or altered temperature) accordingly, and be explicit about the assumptions in documentation.

What does “building downwash” mean, and why does it matter for control room intakes?

Building downwash is the deflection and recirculation of airflow around structures, which can concentrate a plume on the leeward side, sometimes at lower elevations. For control room intakes, this can turn a seemingly upwind, safe location into a hotspot during certain wind conditions. Models that include downwash effects—or CFD—are essential to evaluate intake placement and to design shutdown or filtration responses.

How do dispersion results tie into LOPA and SIL determinations under IEC 61511?

Dispersion affects consequence severity in LOPA by informing dose at target receptors, which in turn shapes risk reduction requirements. Where exposure severity is lower than initially assumed (e.g., due to rapid isolation or effective mitigation), SIL targets may drop; conversely, more severe predicted dose can justify higher SIL or additional layers. Documentation must show the modelling basis that underpins these determinations to satisfy functional safety audits.

What are typical acceptance criteria for toxic risk in QRAs reviewed by UAE stakeholders?

While specifics vary by operator and regulator, many ADNOC-group reviews consider individual and societal risk criteria that incorporate toxic endpoints alongside flammable hazards. The key is not a single numeric value but the coherence of assumptions across the QRA, the plausibility of scenarios, and the credible use of toxic endpoints. Explaining alignment with recognised practices (API RPs, NFPA guidance where relevant) helps establish confidence.

Conclusion

Dispersion modelling sits at the junction of process safety, environmental stewardship, and emergency preparedness. As we have seen, the most consequential errors stem from ungrounded source terms, misapplied meteorology, mismatched model physics, and geometry oversights. These then cascade into poor endpoint choices or configuration shortcuts that distort decisions. By tying scenarios back to hazard studies, validating operational assumptions, matching models to physics, and documenting choices, organisations produce outputs that hold up under audit and, more importantly, guide safer, more defensible actions on the ground—from shelter-in-place calls to HVAC design and land-use consultations. Teams that apply these principles consistently will notice fewer surprises in reviews, clearer ALARP arguments in LOPA/SIL contexts, and emergency plans that reflect the real behaviour of releases in their specific facilities.

To go further on this topic, NPM’s broader Toxic Dispersion Modelling Services UAE page covers this in more depth.

Explore NPM’s References

NPM’s work across complex industrial settings has consistently reinforced one lesson: context matters as much as calculation in dispersion modelling. Our references demonstrate how careful source-term development, rigorous meteorological analysis, and fit-for-purpose modelling have improved decision quality in refineries, petrochemical complexes, offshore platforms, bulk terminals, power generation assets, and high-density logistics hubs. In many of these projects, dispersion analysis was not an isolated exercise; it informed QRA contours, supported LOPA and SIL verifications, and helped emergency response teams shape shelter-in-place versus evacuation strategies that genuinely matched predicted plume behaviour. Reviewers appreciate transparency, and our case histories document the full rationale—from HAZOP-derived scenarios through to receptor definitions and endpoint alignment—so that stakeholders, including regulators and internal auditors under ISO 45001 and ISO 14001 management systems, can trace how each choice was made. You will also find examples where building aerodynamics near air intakes required CFD-supported assessments, as well as instances where dense-gas models were the more reliable choice for cold, heavy releases. Across all of them, the through-line is disciplined governance: assumption logs, verification checks, and independent review. These references illustrate how robust modelling translates into practical actions, whether repositioning HVAC intakes, refining gas detection networks, updating muster points, or supporting land-use dialogues with authorities. If you are exploring how to strengthen the credibility of your modelling or seeking insight into how others have navigated similar regulatory and operational questions, our portfolio offers concrete, technically grounded illustrations rather than generic claims.

For more information about our references, visit Our References.

Start a Conversation

If this article raised questions about your facility’s current assumptions, or if you would value a second set of eyes on how dispersion informs your emergency planning, we are happy to discuss practical next steps. NPM engages with engineering, HSE, and operations teams to align modelling choices with real operating constraints and regulatory expectations. We welcome technical questions and brief scoping discussions to point you in the right direction.

For any further questions or to speak with our team, Contact Us.