Extreme Value Analysis of the Coupled Coastal Hazard Prediction System (CCHaPS) Hindcast for Australia

Extreme value analysis of CCHaPS, the Coupled Coastal Hazard Prediction System, which is a 2D hydrodynamic and wave model (Semi-implicit Cross-scale Hydroscience Integrated System Model and Wind Wave Model III, SCHISM-WWMIII), configured around the Australian coastline including the Great Barrier Reef, and extending out to deep offshore waters or neighbouring landmasses (https://data.csiro.au/collection/csiro:65669).\n\nThe model simulates water levels due to astronomical tides, weather, waves and aspects of wave-flow interaction over multiple decades, allowing for full consideration of the dynamics of extreme sea levels and waves in the Australian region, at high spatial resolution from ~7 km offshore down to ~250 m at the coast, and ~100 m in major river mouths. The model is run on an unstructured (triangular) grid comprising over 1.4 million nodes, which extends overland up to a 12 m elevation contour, to enable modelling of inundation events and sea level rise.\n\nThe CCHaPS hindcast covers the 40-year period 1981 to 2020.\n\nExtreme value analysis has been performed on\n1) annual maximum{yr_max} significant wave heights {hs}, and\n2) detrended annual maximum {yr_max-detrended} water levels {zos}, with the linear trend in mean water levels removed.\n\nThree types of extreme value distributions (EVDs) are fitted to the annual maxima:\n- two-parameter Gumbel\n- three-parameter Generalised Extreme Value (GEV)\n- four-parameter mixed Gumbel distribution (https://doi.org/10.1038/s41598-022-08382-y)\n\nDatasets are provided for\n1) the commonly used GEV {fgev}, and\n2) the best of the three EVD {bestEVD} types (with positive shape parameters), selected using the Akaike Information Criterion (AIC). Lower values of AIC indicate better models, in the sense that the model fit is better relative to the number of model parameters.\nNote: GEV fits with a negative shape parameter (i.e. bounded upper tail) can yield lower estimated extremes and are therefore not always preferred (e.g. Haigh et al., 2014; https://doi.org/10.1007/s00382-012-1652-1). \n\nThe dataset comprises 1, 2, 5, 10, 20, and 63% Annual Exceedance Probabilities (AEPs) with {upper} and {lower} 95% confidence intervals, stored as netCDF data on an unstructured grid, compliant with CF, ACDD and UGRID metadata conventions. Additional processing is also applied to the confidence intervals. A notebook demonstrating how to interact with the data will be attached under Supporting Documentation, and also available via GitHub (see Related Links, LINK TBA).\n\nThe netCDF files can be viewed in QGIS by importing them as a mesh through the Data Source Manager. (https://docs.qgis.org/3.40/en/docs/user_manual/working_with_mesh/mesh.html?utm_source=chatgpt.com)\n\nThe data is also stored on CSIRO infrastructure in the {ev-acs-wp3-cchaps} volume, and at NCI in /g/data/ia39/WP3/release/CCHaPS.\n\nA manuscript describing this dataset is in preparation and will be linked to this metadata record in due course.\nLineage: Extreme value analysis of hourly CCHaPS data (https://data.csiro.au/collection/csiro:65669) was performed using the R CRAN package loopevd (https://doi.org/10.32614/CRAN.package.loopevd), June 2025. File converted to metadata standards compliant netCDF via Python notebook, June 2025. Final dataset copied from NCI to CSIRO Bowen storage in preparation for publication via CSIRO DAP.