This study evaluates the agreement and structural consistency of three atmospheric dispersion modeling approaches applied across three U.S. urban environments for two chemically distinct hazardous materials. Areal Locations of Hazardous Atmospheres (ALOHA), an analytic Gaussian plume model, and OpenFOAM computational fluid dynamics (CFD) passive scalar transport are each applied to continuous release scenarios for chlorine and ammonia at Colorado Springs (Colorado), Newark (New Jersey; serving as the proxy location for the New York City area due to ALOHA software constraints), and Pittsburgh (Pennsylvania). Chlorine is modeled as a heavy gas in ALOHA and ammonia is modeled using ALOHA's Gaussian module, reflecting the distinct dispersion physics of each material. Scenario parameters—including release rate, source height, wind speed, and stability class—are held constant within each chemical scenario across all three cities. Cross-model evaluation is performed at a canonical set of eleven receptor offsets, defined as the intersection of downwind and off-centerline positions present across all six ALOHA exports. Concentrations or tracer values at each receptor are compared using Spearman rank correlation to accommodate both missing values and the non-calibrated nature of the OpenFOAM passive scalar field. Results demonstrate that ALOHA and the analytic Gaussian model maintain consistent rank ordering across all cities and both chemicals (ρ = 0.786–0.821), confirming their shared simplified-physics basis. OpenFOAM agreement with the two reduced-order models is city-dependent: strong structural concordance is observed for Colorado Springs (Gaussian–OpenFOAM ρ = 0.857–0.905), weak or negligible agreement for Pittsburgh (ρ ≈ −0.10), and mixed results for Newark (ρ = 0.297). Pittsburgh's limited OpenFOAM receptor coverage—five of eleven canonical receptors fall outside the mesh boundary—further constrains interpretation for that site. These findings suggest that while reduced-order models produce reliable relative rankings across diverse urban settings, CFD solutions introduce urban-geometry-driven redistributions that are site-specific and cannot be captured by simplified parameterizations. Methodological contributions include a reproducible, locked-artifact workflow verified by SHA-256 checksum and a transparent policy for handling missing values without ad hoc receptor reassignment.
Devin Peters (Tue,) studied this question.