Sensitivity analysis of mesoscale simulations to physics parameterizations: a case study of storm Ciara over the Belgian North Sea using WRF-ARW

The Weather, Research and Forecasting (WRF) model includes a multitude of physics parameterizations to account for atmospheric dynamics and interactions such as turbulent fluxes within the planetary boundary layer (PBL), long and short wave radiation, hydrometeor representation in microphysics, cloud ensemble representation in cumulus, amongst others. A sensitivity analysis is conducted in order to identify the optimal WRF-physics set-up and impact of temporal resolution of re-analysis dataset for the event of sudden changes in wind direction that can become challenging for reliable wind energy operations. In this context, Storm Ciara has been selected as a case study to investigate the influence of a broad combination of different interacting physics-schemes on quantities of interest that are relevant for energy yield assessment. Of particular relevance to fast transient weather events, two different temporal resolutions (1-hourly and 3-hourly) of the lateral boundary condition's re-analysis dataset, ERA5, are considered. Physics parameterizations considered in this study include: two PBL schemes (MYNN2.5 and scale-aware Shin Hong PBL), four cumulus schemes (Kain-Fritsch, Grell-Devenyi, and scale-aware Grell-Freitas and multi-scale Kain-Fritsch,) and three microphysics schemes (WSM5, Thompson and Morrison) coupled with two geospatial configurations for WRF simulation domains. The resulting WRF predictions are assessed by comparison to observational RADAR reflectivity data on precipitation. In addition, SCADA data on wind direction and wind speed from an offshore wind farm located in the Belgian North Sea is considered to assess modeling capabilities for local wind behavior at farm level. For precipitation, results are shown to be very sensitive to model setup, but no clear trends can be observed. For wind-related variables on the other hand, results show a definite improvement in accuracy when both scale-aware cumulus and PBL parameterizations are used in combination with 1-hourly temporal resolution reanalysis data and extended domain sizes.

A micro-genetic algorithm (GA v1.7.1a) for combinatorial optimization of physics parameterizations in the Weather Research and Forecasting model (v4.0.3) for quantitative precipitation forecast in Korea

One of the biggest uncertainties in numerical weather predictions (NWPs) comes from treating the subgrid-scale physical processes. For more accurate regional weather and climate prediction by improving physics parameterizations, it is important to optimize a combination of physics schemes and unknown parameters in NWP models. We have developed an interface system between a micro-genetic algorithm (µ-GA) and the WRF model for the combinatorial optimization of cumulus (CU), microphysics (MP), and planetary boundary layer (PBL) schemes in terms of quantitative precipitation forecast for heavy rainfall events in Korea. The µ-GA successfully improved simulated precipitation despite the nonlinear relationship among the physics schemes. During the evolution process, MP schemes control grid-resolving-scale precipitation, while CU and PBL schemes determine subgrid-scale precipitation. This study demonstrates that the combinatorial optimization of physics schemes in the WRF model is one possible solution to enhance the forecast skill of precipitation.

A micro-genetic algorithm for combinatorial optimization of physics parameterizations in Weather Research and Forecasting model for quantitative precipitation forecast in Korea

One of biggest uncertainties in Numerical Weather Predictions (NWPs) comes from treating the subgrid-scale physical processes. For the more accurate regional weather/climate prediction by improving physics parameterizations, it is important to optimize a combination of physics schemes as well as unknown parameters in NWP models. We have developed an interface system between micro-Genetic Algorithm (μ-GA) and the WRF model for the combinatorial optimization of CUmulus (CU), MicroPhysics (MP), and Planetary Boundary Layer (PBL) schemes in terms of quantitative precipitation forecast for heavy rainfall events in Korea. The μ-GA successfully improved simulated precipitation despite the non-linear relationship among the physics schemes. During the evolution process, MP schemes control grid-resolving scale precipitation while CU and PBL schemes determine subgrid-scale precipitation. This study has demonstrated the combinatorial optimization of physics schemes in the WRF model is one of possible solutions to enhance the forecast skill of precipitation.

Comprehensive Methodology for the Evaluation of High-Resolution WRF Multiphysics Precipitation Simulations for Small, Topographically Complex Domains

ABSTRACTA stepwise evaluation method and a comprehensive scoring approach are proposed and implemented to select a model setup and physics parameterizations of the Weather Research and Forecasting (WRF) Model for high-resolution precipitation simulations. The ERA5 reanalysis data were dynamically downscaled to 1-km resolution for the topographically complex domain of the eastern Mediterranean island of Cyprus. The performance of the simulations was examined for three domain configurations, two model initialization approaches and 18 combinations of atmospheric physics parameterizations. Two continuous and two categorical scores were used for the evaluation. A new extreme event score, which combines hits and frequency bias, was introduced as a complementary evaluator of extremes. A composite scaled score was used to identify the overall best performing parameterizations. The least errors in mean daily and monthly precipitation amounts and daily extremes were found for the domain configuration with the largest extent and three nested domains. A 5-day initialization frequency did not improve precipitation, relative to 30-day continuous simulations. The parameterization type with the largest impact on precipitation was microphysics. The cumulus parameterization was also found to have an impact on the 1-km nested domain, despite that it was only activated in the coarser “parent” domains. Comparison of simulations with 12-, 4-, and 1-km resolution revealed the better skill of the model at 1 km. The impact of the various model configurations in the small-sized domain was different from the impact in larger model domains; this could be further explored for other atmospheric variables.

The Inclusion of chemistry modules into the NOAA UFS Weather Model with the Common Community Physics Package (CCPP)

<p>Online atmosphere-chemistry coupled models have been rapidly developed in recent years. In online models, the atmospheric model can impact air quality and atmospheric composition, while the aerosol feedbacks also impact the atmosphere through direct, semi-direct and indirect effects. At NOAA GSL, in collaboration with scientists from the Chemical Science Laboratory (CSL) and Air Resource Laboratory (ARL), we developed an atmospheric composition suite (based on WRF-Chem) and coupled it online with FV3GFS through the National Unified Operational Prediction Capability (NUOPC)-based NOAA Environmental Modeling System (NEMS) software. This modeling system has been operational since September 24<sup>th</sup>, 2020 as an ensemble member of the Global Ensemble Forecast System (named as GEFS-aerosols) for global aerosol predictions. When using the NUOPC coupler, there are two independent components for atmosphere and chemistry that communicate via the NUOPC coupler every time-step. Because of the interactive and strongly couple nature of chemistry and physics, it is natural to allow for some of the atmospheric composition modules to be called directly from inside the physics suite. This can be accomplished through the use of the Common Community Physics Package (CCPP). CCPP, designed to facilitate a host-model agnostic implementation of physics parameterizations, is a community development and will be used by many different organizations. All the physics parameterizations in the NOAA Unified Forecast System (UFS) Weather Model are CCPP-compliant. Here we broke up the chemistry suite used in GEFS-aerosols, and all the chemical modules were embedded into UFS Weather Model using CCPP as subroutines of physics. This newly developed model with CCPP has been running in real-time starting in the middle of November, 2020. Because of this development we were able to include the CCPP-compliant modules of sea salt, dust, and wild-fire emissions into the NWP model to provide input for the double moment Thompson microphysics parameterization. The inclusion of smoke and aerosol emission modules into the Rapid Refresh Forecast System (RRFS) with CCPP is also ongoing. We will show results from real-time experiments for medium range weather forecasting and compare results with runs that do not include aerosol impacts.</p>

Comprehensive Methodology for the Evaluation of High-Resolution WRF Multi-Physics Precipitation Simulations for Small, Topographically Complex Domains

<p> </p><p>A stepwise evaluation method and a comprehensive scoring approach are proposed and applied to select a model setup and physics parameterizations of the Weather Research and Forecasting (WRF) model for high-resolution precipitation simulations. The ERA5 reanalysis data were dynamically downscaled to 1-km resolution for the topographically complex domain of the eastern Mediterranean island of Cyprus. The performance of the simulations was examined for three domain configurations, two model initialization frequencies and 18 combinations of atmospheric physics parameterizations (members). Two continuous scores, i.e., Bias and Mean Absolute Error (MAE) and two categorical scores, i.e., the Pierce Skill Score (PSS) and a new Extreme Event Score (EES) were used for the evaluation. The EES combines hits and frequency bias and it was compared with other commonly used verification scores. A composite scaled score (CSS) was used to identify the five best performing members.</p><p>The EES was shown to be a complete evaluator of the simulation of extremes. The least errors in mean daily and monthly precipitation amounts and daily extremes were found for the domain configuration with the largest extent and three nested domains. A 5-day initialization frequency did not improve precipitation, relative to 30-day continuous simulations. The use of multiple and comprehensive evaluation measures for the assessment of WRF performance allowed a more complete evaluation of the different properties of simulated precipitation, such as daily and monthly volumes and daily extremes, for different dynamical downscaling options and model configurations. The scores obtained for the selected five members for a three-month simulation period ranged for BIAS from zero to -25%, for MAE around 2 mm, for PSS from 0.25 to 0.52 and for EES from 0.19 to 0.26. The CSS ranged from 0.56 to 0.83 for the same members. The proposed stepwise approach can be applied to select an efficient set of WRF multi-physics configurations that accounts for these properties of precipitation and that can be used as input for hydrologic applications.</p>

Impact of Physics Parameterizations on High-Resolution Air Quality Simulations over the Paris Region

The accurate simulation of meteorological conditions, especially within the planetary boundary layer (PBL), is of major importance for air quality modeling. In the present work, we have used the Weather Research and Forecast (WRF) model coupled with the chemistry transport model (CTM) CHIMERE to understand the impact of physics parameterizations on air quality simulation during a short-term pollution episode on the Paris region. A lower first model layer with a 4 m surface layer could better reproduce the transport and diffusion of pollutants in a real urban environment. Three canopy models could better reproduce a 2 m temperature (T2) in the daytime but present a positive bias from 1 to 5 °C during the nighttime; the multi-urban canopy scheme “building effect parameterization” (BEP) underestimates the 10 m windspeed (W10) around 1.2 m s−1 for the whole episode, indicating the city cluster plays an important role in the diffusion rate in urban areas. For the simulation of pollutant concentrations, large differences were found between three canopy schemes, but with an overall overestimation during the pollution episode, especially for NO2 simulation, the average mean biases of NO2 prediction during the pollution episode were 40.9, 62.2, and 29.7 µg m−3 for the Bulk, urban canopy model (UCM), and BEP schemes, respectively. Meanwhile, the vertical profile of the diffusion coefficients and pollutants indicated an important impact of the canopy model on the vertical diffusion. The PBL scheme sensitivity tests displayed an underestimation of the height of the PBL when compared with observations issued from the Lidar. The YonSei University scheme YSU and Boulac PBL schemes improved the PBL prediction compared with the Mellor–Yamada–Janjic (MYJ) scheme. All the sensitivity tests, except the Boulac–BEP, could not fairly reproduce the PBL height during the pollution episode. The Boulac–BEP scheme had significantly better performances than the other schemes for the simulation of both the PBL height and pollutants, especially for the NO2 and PM2.5 (particulate matter 2.5 micrometers or less in diameter) simulations. The mean bias of the NO2, PM2.5, and PM10 (particulate matter 10 micrometers or less in diameter) prediction were −5.1, 1.2, and −8.6 µg m−3, respectively, indicating that both the canopy schemes and PBL schemes have a critical effect on air quality prediction in the urban region.

Assessment of the Forecast Skill of Multiphysics and Multistochastic Methods within the GRAPES Regional Ensemble Prediction System in the East Asian Monsoon Region

To more comprehensively and accurately address model uncertainties in the East Asia monsoon region, a single-physics suite, where each ensemble member uses the same set of physics parameterizations as the control member in combination with multiple stochastic schemes, is developed to investigate if the multistochastic schemes that combine different stochastic schemes together can be an alternative to a multiphysics suite, where each ensemble member uses a different set of physics parameterizations (e.g., cumulus convection, boundary layer, surface layer, microphysics, and shortwave and longwave radiation). For this purpose, two experiments are performed for a summer monsoon month over China: one with a multiphysics suite and the other with a single-physics suite combined with multistochastic schemes. Three stochastic schemes are applied: the stochastically perturbed parameterizations (SPP) scheme, consisting of temporally and spatially varying perturbations of 18 parameters in the microphysics, convection, boundary layer, and surface layer parameterization schemes; the stochastically perturbed parameterization tendencies (SPPT) scheme; and the stochastic kinetic energy backscatter (SKEB) scheme. The combination of the three stochastic schemes is compared with the multiphysics suite in the Global and Regional Assimilation and Prediction Enhanced System–Regional Ensemble Prediction System with a horizontal grid spacing of 15 km. Verification results show that, overall, a single-physics suite that combines SPP, SPPT, and SKEB outperforms the multiphysics suite in precipitation verification and verification for upper-air weather variables, 10-m zonal wind, and 2-m temperature in the East Asian monsoon region. The indication is that a single-physics suite combining SPP, SPPT, and SKEB may be an appropriate alternative to a multiphysics suite. This finding lays a foundation for the development and design of future regional and global ensembles.