Observation policy and model selection in integrated flood frequency assessment
Year: 2026
Type: article
Abstract: When do observation and model choices support reliable flood-frequency inference, and when does dependence improve a design decision? We examine this question using Najaf–Kufa, five regional Australian gauges, 12 additional public catchments and controlled experiments. Matched-year observation restrictions change regional L-moment Q100 by 20.8–53.6%; additional-catchment effects span −3.34% to +76.20%. A new 2,160-record model-selection test uses observed values only and estimates peak-loss probability. It reduces combined-mechanism errors from 236 to 167/432, while worsening complete-record and missed-peak cases; short-record selection does not guarantee latent-tail recovery. Expanding-origin comparison supplies 607 observed annual forecasts at three probabilities, providing predictive evidence beyond estimator disagreement without identifying true field Q100. Extended daily magnetic tests provide 10–22 nonempty test years at supported sites and 137 Iraqi high-flow days; no supported incremental comparison has a positive simultaneous lower bound across calendar-block sensitivities. Genuine trivariate field copulas alter joint probabilities without established general hazard benefit. A separate physically coherent storage benchmark uses 216 training samples and independent validation under three dependence structures. With 200 training events and Student-t dependence, independence yields median overflow risk 3.661% against a 1% target, versus 0.913% for selected dependence; the independent control does not favor unnecessary complexity. These findings distinguish observation sensitivity, predictive performance and conditional hydraulic utility. Browser execution, LLM validation and public analysis deposition remain incomplete. Keywords: flood frequency; observation quality; Q100; geomagnetic activity; copula; reproducibility; uncertainty.
Author: Alkhafaji, Ayad K. H.
Institution: University of Kufa
Language: English
Subfield: Global and Planetary Change
Field: Environmental Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7225336388