An observation-constrained multi-physics RCM ensemble for simulating European mega-heatwaves

Annemiek I. Stegehuis, Robert Vautard, Philippe Ciais, Adriaan J. Teuling, Diego G. Miralles, Martin Wild2014 · DOI

TL;DR

This paper tackles the challenge of accurately simulating extreme heatwaves with regional climate models by creating a large ensemble of simulations with varying physics schemes. They identified that convection schemes are the most critical factor for heatwave simulation and developed a reduced, observation-constrained ensemble of five configurations that better capture these events, offering a more reliable tool for future climate change impact studies.

What Problem Does This Paper Address?

Climate models are crucial for predicting future climate scenarios, but their ability to accurately simulate extreme events like heatwaves is often limited. This is partly because models are not always rigorously evaluated or calibrated against observed extreme conditions. Heatwaves, in particular, pose significant threats to society and ecosystems, and understanding their dynamics and future projections requires models that can reliably reproduce past events. The authors highlight that even when models are validated for general climate conditions, their performance during extreme events can be uncertain. This study aims to address this by systematically evaluating a large number of different physical process configurations within a regional climate model to identify those best suited for simulating European mega-heatwaves.

Methods

The researchers employed the Weather Research and Forecasting (WRF) regional climate model and created a large "multi-physics" ensemble. This involved systematically combining different parameterization schemes for various physical processes within WRF. Specifically, they explored combinations of schemes for:

  • Microphysics (MP): 3 schemes
  • Planetary Boundary Layer and Surface Layer (PBL-SF): 6 schemes
  • Radiation (RA): 3 schemes
  • Convection (CU): 4 schemes

This resulted in a total of 216 unique model configurations. These simulations were then compared against observational data for temperature, precipitation, and radiation during specific heatwave events (2003 in Western Europe and 2010 in Russia) and a regular summer (2007). A rigorous selection process was used to identify the best-performing configurations, focusing on those that met specific temperature accuracy criteria and then ranking them based on multiple variables. This led to the creation of a reduced ensemble of five diverse and well-performing configurations.

Datasets

The study utilized the following observational datasets for evaluation and validation:

  • E-OBS (Haylock et al., 2008): Gridded daily temperature and precipitation data for Europe.
  • Global Energy Balance Archive (GEBA) network (Wild et al., 2009): Station data of monthly global radiation.
  • GLEAM (Miralles et al., 2011): Satellite observation-driven estimates of daily latent heat fluxes.

The study focused on simulating the summers of 2003, 2007, and 2010. The regional domain used was the EURO-CORDEX domain.

Key Insights and Ideas

  • Convection is King (for Heatwaves): The study found that the choice of convection scheme in WRF has the most significant impact on simulating heatwave temperatures. This is a crucial finding for model development, suggesting that improvements in this area are paramount for better heatwave prediction.
  • Interactions Matter: The authors observed that the impact of convection schemes on temperature is largely mediated through their influence on radiation. This highlights the complex interplay between different physical processes and how errors can propagate.
  • Reduced Ensembles Can Be Powerful: By carefully selecting and constraining simulations based on observations, the researchers were able to create a smaller, more reliable ensemble that still captures a good range of physical variability relevant to heatwaves. This offers a more efficient approach than using extremely large ensembles.
  • Systematic Biases Persist: Despite the large number of simulations, most still exhibited systematic biases, such as underestimating temperature and overestimating precipitation during heatwaves, a pattern consistent with previous studies.

Key Results

  • Out of 216 simulations, only a subset could reproduce extreme temperatures during heatwaves, with most showing a cold bias of 2-3°C.
  • Precipitation was generally overestimated before heatwaves, and shortwave radiation was slightly underestimated.
  • The study identified a reduced ensemble of five WRF configurations that best simulate European heatwaves, showing improved performance across temperature, precipitation, and radiation compared to the full ensemble.
  • The selected five configurations performed well in capturing temperature dynamics and were consistent with latent heat flux data, indicating a robust selection.

Limitations and Future Work

  • Single Land Surface Scheme: A significant limitation acknowledged by the authors is the use of only one land surface scheme (NOAH). They suggest that incorporating different land surface models and permuting them with various atmospheric schemes would likely increase the ensemble spread and provide a more comprehensive understanding of uncertainties.
  • Future Studies: The authors propose that the developed constrained WRF ensemble can be used for future climate change impact studies, as each member may exhibit different sensitivities to future climate conditions, allowing for a more constrained exploration of uncertainty. They also emphasize the need to carefully design or adjust convection schemes for accurate representation of summer climate during heatwaves.

Summary updated

Abstract

Climate models are often not evaluated or calibrated against observations of past climate extremes, resulting in poor performance during for instance heatwave conditions. Here we use the Weather Research and Forecasting (WRF) regional climate model with a large combination of different atmospheric physics schemes, with the goal of detecting the most sensitive ones and identifying those that appear most suitable for simulating the heatwave events of 2003 in Western Europe and 2010 in Russia. 55 Out of 216 simulations combining different atmospheric physical schemes can reproduce the extreme temperatures observed during heatwaves, the majority of simulations showing a cold bias of on average 2–3 °C. Conversely, precipitation is mostly overestimated prior to heatwaves, and short wave radiation is slightly underestimated. Convection is found to be the most sensitive process impacting simulated heatwave temperature, across 4 different convection schemes in the simulation ensemble. Based on these comparisons, we design a reduced ensemble of five well performing and diverse scheme combinations, which may be used in the future to perform heatwave analysis and to investigate the impact of climate change in summer in Europe. Future studies could include the use of different land surface models together with varied physics scheme.
BibTeX
@misc{stegehuis2014observation,
  author = {Annemiek I. Stegehuis and Robert Vautard and Philippe Ciais and Adriaan J. Teuling and Diego G. Miralles and Martin Wild},
  title = {{An observation-constrained multi-physics RCM ensemble for simulating European mega-heatwaves}},
  year = {2014},
  doi = {10.5194/gmdd-7-7861-2014},
}

Loading related content...

Figure