Acero et al. (2010) Peaks-over-Threshold Study of Trends in Extreme Rainfall over the Iberian Peninsula

Identification

• Journal: Journal of Climate
• Year: 2010
• Date: 2010-10-26
• Authors: Francisco Javier Acero, J. A. García, M. C. Gallego
• DOI: 10.1175/2010jcli3627.1

Research Groups

• School of Civil Engineering and Geosciences, Newcastle University, Newcastle upon Tyne, United Kingdom
• Met Office Hadley Centre, Exeter, United Kingdom
• Met Office Reading, Reading, United Kingdom
• College of Engineering, Mathematics and Physical Sciences, University of Exeter, Exeter, United Kingdom

Short Summary

This study evaluates two Met Office regional climate models (RCMs) at 12-kilometer (parameterized convection) and 1.5-kilometer (explicit convection) resolutions for simulating multihourly precipitation extremes over the southern United Kingdom. It concludes that the 1.5-kilometer RCM provides a more realistic representation of summer subdaily extreme events, despite a positive bias, while the 12-kilometer RCM suffers from unrealistic "gridpoint storms" leading to overly intense long-return-period extremes.

Objective

• To diagnose the differences in extreme precipitation behavior between the 1.5-kilometer and 12-kilometer RCMs.
• To assess the reliability of models for hourly extremes given limited United Kingdom hourly observations and determine if observations are adequate for extreme value analysis and model evaluation.
• To evaluate the physical realism of 1.5-kilometer and 12-kilometer RCM precipitation extremes compared to observations.
• To evaluate the value of the 1.5-kilometer RCM for providing future projections of extreme precipitation.

Study Configuration

• Spatial Scale: Southern United Kingdom (SUK) domain. Model grid resolutions of 12 kilometers and 1.5 kilometers. Observational data at 5-kilometer (daily gridded) and point (hourly gauge) resolutions, with radar data gridded. All gridded data re-gridded to a 12-kilometer grid for analysis.
• Temporal Scale:
  • Simulation Period: 1990–2008 (19 years) for both RCMs.
  • Observation Periods:
    • UKMO daily gridded (UK5): 1990–2008.
    • UKMO hourly gridded radar (Radarnet): 2003–2010 (8 years).
    • UKMO hourly station gauges (MIDAS): 1992–2010 (19 years).
  • Accumulation Periods Analyzed: 1 hour, 3 hours, 6 hours, 12 hours, 1 day, and 5 days.
  • Seasons Analyzed: June–August (JJA) and December–February (DJF).

Methodology and Data

• Models used:
  • Met Office Regional Climate Model (RCM) at 12-kilometer resolution, employing parameterized convection, driven by ERA-Interim reanalysis.
  • Met Office Regional Climate Model (RCM) at 1.5-kilometer resolution, employing explicit convection, driven by the 12-kilometer RCM.
• Data sources:
  • Observation (Gauge-based):
    • UKMO daily gridded precipitation (UK5) at 5-kilometer resolution.
    • UKMO hourly station (tipping-bucket rain gauge) measurements from the Integrated Data Archive System (MIDAS) for 28 stations in the SUK.
  • Radar: UKMO hourly gridded radar (Radarnet) precipitation measurements.
  • Reanalysis: ERA-Interim reanalysis (used to drive the 12-kilometer RCM).
  • Statistical Methods: Peaks-over-threshold (PoT) extreme value theory (EVT) to characterize extreme recurrence rates and probability distributions of excesses above a threshold using the Generalized Pareto (GP) distribution. L-moments were used for GP parameter estimation, with standard errors estimated via 1000-time bootstrapping. Goodness of fit was tested using the Anderson–Darling test. Declustering was applied with a minimum 1-day time. Growth curves G(n) = z(n)/z(2) were used for normalized comparisons.

Main Results

• JJA Extremes: The 1.5-kilometer RCM more realistically represents (multi)hourly JJA very extreme events compared to the 12-kilometer RCM. The 12-kilometer RCM exhibits overly large high return levels and a "heavy-tailed" distribution (high shape parameter), leading to unrealistically rapid increases in return levels with rarity. This is attributed to "gridpoint storms" – localized, intense, and stationary events where convective parameterization assumptions are violated. The 1.5-kilometer RCM, while having a consistent positive bias in return levels, shows shape parameters closer to observations ("thinner-tailed") and accurately captures the observed insensitivity of growth rates to accumulation period changes. As accumulation periods lengthen towards daily scales, the 12-kilometer RCM's biases diminish.
• DJF Extremes: Both RCMs simulate DJF extremes comparably, and these are generally weaker than JJA extremes. The 12-kilometer RCM's JJA-specific biases (e.g., high shape parameters) are not evident in DJF, suggesting the convective parameterization's role in summer biases. The 1.5-kilometer RCM shows consistently higher DJF return levels than the 12-kilometer RCM and observations for accumulation periods longer than 6 hours.
• Observational Data Utility: Radar-estimated return levels and growth curves are found to be comparable with gauge and daily gridded data, supporting the use of spatially averaged radar data for subdaily extreme precipitation studies despite known limitations.
• Growth Curves: Observed JJA growth rates are found to be insensitive to accumulation periods of 1 hour or longer, a feature well-captured by the 1.5-kilometer RCM but not the 12-kilometer RCM. This contradicts some previous findings (e.g., Flood Estimation Handbook).

Contributions

• Provides a comprehensive diagnostic evaluation of the performance of convective-permitting (1.5-kilometer) versus parameterized convection (12-kilometer) RCMs in simulating multihourly precipitation extremes over the United Kingdom.
• Demonstrates the significant added value of the 1.5-kilometer RCM in realistically representing summer subdaily extreme precipitation, despite its overall positive intensity bias, by accurately capturing the shape of the extreme value distribution and the insensitivity of growth rates to accumulation period.
• Identifies "gridpoint storms" as a key physical mechanism leading to unrealistic very extreme events in the 12-kilometer RCM, linking this to the limitations of convective parameterization in the "gray zone" resolution.
• Validates the utility of spatially averaged radar data for extreme value analysis of subdaily precipitation, providing crucial information where gauge observations are sparse.
• Challenges existing hydrological guidance (e.g., Flood Estimation Handbook) regarding the accumulation period sensitivity of extreme precipitation growth rates, suggesting an invariance for periods of 1 hour or longer.

Funding

• CONVEX project
• United Kingdom NERC CWC programme (Grant NE/I006680/1)
• UK Met Office (for model simulations)
• Newcastle University (for lead author's financial support)

Citation

@article{Acero2010PeaksoverThreshold,
  author = {Acero, Francisco Javier and García, J. A. and Gallego, M. C.},
  title = {Peaks-over-Threshold Study of Trends in Extreme Rainfall over the Iberian Peninsula},
  journal = {Journal of Climate},
  year = {2010},
  doi = {10.1175/2010jcli3627.1},
  url = {https://doi.org/10.1175/2010jcli3627.1}
}