Colmet-Daage et al. (2018) Evaluation of uncertainties in mean and extreme precipitation under climate change for northwestern Mediterranean watersheds from high-resolution Med and Euro-CORDEX ensembles

Identification

• Journal: Hydrology and earth system sciences
• Year: 2018
• Authors: Antoine Colmet-Daage, Emilia Sánchez-Gómez, Sophie Ricci, Cécile Llovel, Valérie Borrell Estupina, Pere Quintana Seguí, María Carmen Llasat, Éric Servat
• DOI: 10.5194/hess-22-673-2018

Research Groups

• CECI, CERFACS – CNRS TOULOUSE, Toulouse, France
• WSP France, Toulouse, France
• Hydrosciences Montpellier, Univ. Montpellier, Montpellier, France
• Observatori de l'Ebre Fundació, Tarragona, Spain
• Universitat de Barcelona, Barcelona, Spain
• Institut Montpelliérain de l'Eau et de l'Environnement – IRD, Montpellier, France

Short Summary

This study evaluates the uncertainties in mean and extreme precipitation projections for northwestern Mediterranean watersheds using high-resolution Med and Euro-CORDEX (EMCORDEX) ensembles. The core finding is that while mean precipitation is projected to decrease under the RCP8.5 scenario, extreme precipitation events (above the 90th quantile) are intensified across all studied catchments, particularly toward the end of the 21st century.

Objective

• To assess the skills of Regional Climate Models (RCMs) from the EMCORDEX multi-model ensemble regarding mean and extreme precipitation over past periods.
• To assess the influence of Global Climate Models (GCMs)' lateral boundary conditions on the RCMs' skills.
• To evaluate future changes in precipitation extremes using a "futurization" approach (past/future change coefficients of quantile-ranked RCM precipitation outputs) for subsequent flash flood simulations.

Study Configuration

• Spatial Scale: Three mesoscale northwestern Mediterranean watersheds: Lez (114 km²), Aude (over 5000 km²), and Muga (854 km²). The RCM simulations are at approximately 12 km resolution.
• Temporal Scale: Historical period (1976–2005 or 1981–2010 for evaluation) and future periods (2011–2040, 2041–2070, and 2071–2100), analyzed in 30-year time slices.

Methodology and Data

• Models used: Five Regional Climate Models (RCMs) from the EMCORDEX initiative: ALADIN52, ALADIN53, RACMO22E, RCA4, and HIRHAM5. These RCMs were driven by four different Global Climate Models (GCMs) from CMIP5 (CNRM-CM5, ICHEC, MOHC, MPI) resulting in eight GCM–RCM pairs for historical and future scenarios (RCP4.5 and RCP8.5).
• Data sources: SAFRAN (Système d'Analyse Fournissant des Renseignements Atmosphériques à la Neige) reanalysis product (8 km resolution for France, 5 km for Spain) was used as the reference observational dataset. ERA-Interim reanalysis was used to drive the RCM evaluation simulations (EVAL).
• Key Method: A "futurization" approach was used, calculating past/future change coefficients ($A_{qi}$) for quantile-ranked precipitation outputs to estimate future changes in extreme event intensity. Quantiles analyzed ranged from the 90th to the 99.9th percentile of daily precipitation.

Main Results

• Historical Bias (1981–2010): RCMs generally overestimate cumulative precipitation over mountainous regions (up to +30%) and underestimate it over coastal regions (down to -30%) in autumn. Extreme precipitation events (quantiles above the 95th) are systematically underestimated by most RCMs compared to the SAFRAN reference data.
• GCM Influence: GCM forcing tends to enhance the RCMs' intrinsic underestimation of extreme precipitation values. CNRM-CM5 forcing systematically led to overestimation of summer precipitation.
• Mean Precipitation Change (2071–2100): Under the RCP4.5 scenario, the annual cycle of mean precipitation remains largely unchanged compared to the historical period. Under the RCP8.5 scenario, a general decrease in mean precipitation is projected from April to October across all three catchments.
• Extreme Precipitation Change (2071–2100): Extreme precipitation events (beyond the 90th quantile) are projected to intensify for both RCP4.5 and RCP8.5 scenarios. The ensemble mean change coefficient for the 99.5th quantile over the Aude catchment reaches 1.15 (a 15% increase) for both RCPs.
• Uncertainty: The ensemble spread (uncertainty) is smaller under the RCP8.5 scenario, indicating a higher level of certainty regarding the intensification of extreme precipitation compared to RCP4.5, especially toward the end of the 21st century. For the 99.9th quantile under RCP8.5 over the Lez catchment, the mean change coefficient is 1.35, representing an increase from 140 mm/day to 189 mm/day.

Contributions

• Provides a high-resolution, multi-model assessment of climate change impacts on mean and extreme precipitation specifically focused on small, flash-flood-prone northwestern Mediterranean watersheds (Lez, Aude, Muga).
• Quantifies the uncertainties associated with both RCM intrinsic biases and GCM lateral boundary conditions on precipitation projections.
• Introduces and applies a "futurization" approach using quantile-ranked change coefficients, which is highly relevant for subsequent event-based hydrological modeling of future flash floods.
• Confirms with high confidence, using the EMCORDEX ensemble, that extreme precipitation events are projected to increase in intensity in this climate change hotspot region, despite a projected decrease in total monthly precipitation.

Funding

• WSP-France (under the CIFRE contract 2015/005)
• International HYMEX project
• Spanish HOPE project (CGL2014-52571-R)

Citation

@article{ColmetDaage2018Evaluation,
  author = {Colmet-Daage, Antoine and Sánchez-Gómez, Emilia and Ricci, Sophie and Llovel, Cécile and Estupina, Valérie Borrell and Quintana‐Seguí, Pere and Llasat, María Carmen and Servat, Éric},
  title = {Evaluation of uncertainties in mean and extreme precipitation under climate change for northwestern Mediterranean watersheds from high-resolution Med and Euro-CORDEX ensembles},
  journal = {Hydrology and earth system sciences},
  year = {2018},
  doi = {10.5194/hess-22-673-2018},
  url = {https://doi.org/10.5194/hess-22-673-2018}
}

Generated by BiblioAssistant using gemini-flash-latest (Google API)