A comparative study of mesoscale flow-field modelling in an Eastern Alpine region using WRF and GRAMM-SCI

doi:10.1016/j.atmosres.2020.105288

Atmospheric Research

Volume 249, February 2021, 105288

https://doi.org/10.1016/j.atmosres.2020.105288 Get rights and content

Highlights

•
Successful coupling of GRAMM-SCI with ERA5 reanalysis data
•
Reasonable model simulations of mountain wind systems in the Alps
•
Comparable modelling results between GRAMM-SCI and WRF demonstrated
•
Research needs for modelling radiation fog illustrated
•
New modelling quality objective for wind direction in complex terrain suggested

Abstract

Recently, the mesoscale model GRAMM-SCI has been further developed to make use of the freely available global ERA5 reanalysis dataset issued by the ECMWF. In this study, first results are discussed for the federal state of Styria, which is situated in the eastern Alps of Austria. Additional simulations were made with the mesoscale model WRF, which serve as a benchmark in this work. The model runs covered one week in summer and another one in winter dominated by fair weather conditions. These were characterized by the development of complex mountain wind systems in the Alps, which play an important role for the dispersion of pollutants. Regarding the bias and the root mean square error both models perform very well in comparison with existing studies for Alpine areas and are able to capture the main features of observed surface flows such as valley-wind systems or katabatic flows at slopes. In addition, observed calm wind conditions at many stations during the winter period were reproduced by the models. However, the correct simulation of wind directions in these conditions was found to be extremely challenging. Existing model quality criteria for wind direction seem to be too strict for low-wind-speed conditions. Therefore, based on theoretical and empirical considerations, a new model evaluation benchmark for wind direction is proposed, which takes into account the random nature of horizontally meandering flows in stagnant weather situations.

Introduction

The main motivation for this study has its origin in the need for reliable wind fields for routine air quality assessments in the eastern Alps (Federal State of Styria, Austria). Typically, several hundreds of air pollution and odour assessments are carried out at the Air Quality Department of Styria each year. Due to the highly complex terrain, the prognostic, nonhydrostatic mesoscale model GRAMM has been in use since more than a decade now. Legislation in Austria requires air pollution concentrations being permanently kept below national thresholds. Therefore, air quality assessments are carried out for a reference year in the past with unfavourable dispersion conditions. This led to the development of so-called wind-field libraries established for a single calendar year in many regions in Austria (e.g. Oettl, 2014), whereby the horizontal resolution is 200 m in the current library for the reference year 2015. It should be noted that the establishment of wind-field libraries for practical purposes seems to become increasingly popular. For instance, the US-EPA makes available a nationwide dataset for dispersion modelling applications, where WRF was run at 12-km horizontal resolution (Misenis et al., 2019). Berchet et al., 2017a, Berchet et al., 2017b established wind-field libraries for the cities of Lausanne and Zurich, Switzerland, with a horizontal resolution of 100 m using GRAMM. Another example, though in a different context than dispersion modelling, is outlined in Bernhardt et al. (2008), who used MM5 with a horizontal resolution of 200 m for generating a library consisting of 220 wind fields used to drive a snow transport model in an Alpine area.

It is remarkable that the number of mesoscale meteorological modelling studies in highly complex terrain like the Alps with horizontal resolutions in the sub-kilometre range, which are necessary for local air quality assessments, is still rather low (most of the relevant studies will be discussed later in this work). This might be an indication of the difficulties involved in obtaining reasonable and numerically stable results in such regions. For instance, Cantelli et al. (2017) reported about numerical problems at steep slopes when applying WRF with a horizontal resolution of 370 m in an alpine region. Moreover, there is certainly a lack in knowledge regarding turbulent exchange processes in highly complex terrain, which introduces uncertainty in subsequent simulations of dispersion processes when turbulence quantities are based on the output of numerical weather models. In such applications, tailored modifications of the turbulence parametrization are needed to accurately estimate, for instance, the standard deviation of wind velocity components and the Lagrangian time scales (Tomasi et al., 2019). In addition, the understanding of the surface energy balance and the prediction of surface wind fields become even more challenging when regions with complex topography like the Alps, are characterized by the presence of urbanized areas that may cause a nocturnal urban heat island. Giovannini et al. (2014) demonstrated that the WRF model applied in the Alpine Adige Valley at 500 m resolution, combined with suitable urban parametrization scheme could improve the understanding of local valley winds when urbanized areas are present in complex inhomogeneous terrain. Another attempt to reduce numerical model deficiencies in reproducing atmospheric turbulence phenomena are long-term observations of near-surface exchange processes carried out in the Inn valley in Tyrol, Austria, by the University of Innsbruck (e.g. Rotach et al., 2017) for example.

The main goal of this study was to evaluate the performance of the mesoscale model GRAMM, which has been further developed to make use of the ERA5 reanalysis data (Copernicus Climate Change Service, 2017) issued by the European Centre for Medium-Range Weather Forecasts (ECMWF). ERA5 is a global climate reanalysis dataset, currently covering the period from 1979 to present on a spatial grid of 0.25 deg. Data is generally provided either at 37 pressure levels or at 137 model levels, the latter are used in GRAMM currently.

For comparison purposes, simulations have been performed with the WRF model, too. These results are used as benchmark for GRAMM-SCI, because WRF undoubtedly is one of the most advanced mesoscale models available to the public and most modelling studies in Alpine regions have been using WRF in recent times. The focus of the model evaluations are the simulated surface winds, as this is the most important input to dispersion models for air quality assessments for regulatory purposes. In addition, 2 m-temperatures were compared with observations as it allows an indirect insight in the capability of the models to calculate near-surface exchange processes (e.g. heat fluxes, radiation). Large deviations between observed and modelled temperatures would quite certainly indicate deficiencies with regard to this exchange processes.

Section snippets

GRAMM-SCI

The Graz Mesoscale Model - Scientific (GRAMM-SCI) is a new branch of the public and open source version GRAMM 20.01 (Oettl, 2020). It has been further developed and is capable of using ERA5 reanalysis data for initialization and for prescribing transient boundary conditions. The origin of GRAMM dates back to the early 1990s and was then carried out at the Graz University of Technology (e.g. Almbauer et al., 1995) until 2006. Since then GRAMM has been mainly further developed at the Air Quality

Study area

Fig. 2 illustrates the approximate location of the innermost WRF and GRAMM-SCI modelling domains. It covers the eastern Alps in Austria, where the highest ridges exceed 3.000 m. To the southeast the alpine foreland stretches out, which is characterized by hilly terrain with heights varying between 100 and 300 m above the valley floors. Typically, diurnal mountain winds develop during fair weather conditions. For instance, valley wind systems, mountain-plain wind systems, or slope winds as

Results

The following model quality objectives (MQO) have been used for evaluating model results:

Pearson correlation coefficient (r): $r = \frac{\sum_{i = 1}^{n} (M_{i} - \bar{M}) (O_{i} - \bar{O})}{\sqrt{\sum_{i = 1}^{n} {(M_{i} - \bar{M})}^{2}} \sqrt{\sum_{i = 1}^{n} {(O_{i} - \bar{O})}^{2}}}$

Mean error: $BIAS = \frac{1}{n} \sum_{i = 1}^{n} (M_{i} - O_{i})$

Root mean square error: $RMSE = \sqrt{\frac{1}{n} \sum_{i = 1}^{n} {(O_{i} - M_{i})}^{2}}$

Mean wind direction error or gross error: ${BIAS}_{dir} = \frac{\sum_{i = 1}^{n} min [|M_{i -} O_{i}|, |M_{i -} O_{i} + 360|, |M_{i -} O_{i} - 360|]}{n}$

If not stated otherwise the model runs taken for comparison purposes were WRF1000 and GRAMM1000.

Conclusions

By taking WRF simulation results as benchmark, it can be stated safely that GRAMM-SCI has been successfully coupled with ERA5 reanalysis data. Both models perform quite similar based on the model quality objectives used in this work. As the selected episodes were dominated most of the time by fair weather conditions and weak synoptic-scale forcing, the majority of monitoring stations were influenced by mountain wind systems. Under such conditions nesting techniques become less important

Author statement

The contribution of the authors were as follows:

Dietmar Oettl: Conceptualization, Methodology, Software, Validation, Visualisation, Writing

Giogio Veratti: Methodology, Validation, Visualisation, Wirting.

Declaration of Competing Interest

None.

References (64)

R.A. Almbauer et al.
Simulation of the air quality during a field study for the city of Graz
Atmos. Environ.
(2000)
A. Berchet et al.
A cost-effective method for simulation city-wide air flow and pollutant dispersion at building resolving scale
Atmos. Environ.
(2017)
A. Gsella et al.
Evaluation of MM5, WRF and TAMPER meteorology over the complex terrain of the Po Valley
Italy. Atmos. Environ.
(2014)
R. Lazar et al.
An urban climate analysis of Graz and its significance for urban planning in the tributary valleys east of Graz (Austria)
Atmos. Environ.
(1999)
Y. Liu et al.
Simultaneous nested modeling from the synoptic scale to the LES scale for wind energy applications
Journal of Wind Engineering and Industrial Aerodynamics
(2011)
D. Oettl et al.
A new hypothesis on meandering atmospheric flows in low wind speed conditions
Atmos. Environ.
(2005)
D. Oettl et al.
Lagrangian particle model simulation to assess air quality along the Brenner transit corridor through the Alps
Develop. Environ. Sci
(2007)
P. Thunis et al.
MESOCOM an inter-comparison exercise of mesoscale flow models applied to an ideal case simulation
Atmos. Environ.
(2003)
E. Tomasi et al.
Turbulence parameterizations for dispersion in sub-kilometer horizontally non-homogeneous flows
Atmos. Res.
(2019)
G. Veratti et al.
Towards the coupling of a chemical transport model with a micro-scale Lagrangian modelling system for evaluation of urban levels in a European hotspot
Atmos. Environ.
(2020)

R.A. Almbauer

A new finite volume discretisation for solving the Navier-Stokes-equations

Numerical Methods in Laminar and Turbulent Flow

(1995)

R.A. Almbauer et al.

Estimation of the pollution concentration in the vicinity of a cellulose plant in an alpine valley using the non-hydrostatic mesoscale model GRAMM. Air Pollution III

(1995)

R.A. Almbauer et al.

Analysis of the daily variations of wintertime air pollution concentrations in the city of Graz-Austria

Environ. Monit. Assess.

(2000)

D. Belušić et al.

Can mesoscale models reproduce meandering motions?

Q.J.R. Meteorol. Soc.

(2010)

A. Berchet et al.

Evaluation of high-resolution GRAMM–GRAL (v15.12/v14.8) NOx simulations over the city of Zürich, Switzerland

Geosci. Model Dev

(2017)

M. Bernhardt et al.

Using wind fields from a high-resolution atmospheric model for simulating snow dynamics in mountainous terrain

Hydrol. Process.

(2008)

A. Cantelli et al.

Numerical simulations of mountain winds in an alpine valley

Wind Struct.

(2017)

F. Chen et al.

Coupling an advanced land-surface/hydrology model with the Penn State/NCAR MM5 modeling system, part I: model description and implementation

Mon. Weather Rev.

(2001)

Copernicus Climate Change Service (C3S)

ERA5: Fifth generation of ECMWF atmospheric reanalyses of the global climate

A. De Meij et al.

The impact of MM5 and WRF meteorology over complex terrain on CHIMERE model calculations

Atmos. Chem. Phys.

(2009)

Doms, G., And M. Baldauf (2018) A description of the nonhydrostatic regional COSMO-Model. Part I: Dynamics and...

J.A. Dutton et al.

Approximate equations of motion for gases and liquids

J. Atmos. Sci.

(1969)

EEA

The application of models under the European Union’s Air Quality Directive: A technical reference guide

Rep. Nr. 10/2011

(2011)

J.H. Ferziger et al.

Computational Methods for Fluid Dynamics

(2002)

L. Giovannini et al.

Numerical simulations of boundary-layer processes and urban-induced alterations in an Alpine valley

Int. J. Climatol.

(2014)

B. Goger et al.

The Impact of Three-Dimensional Effects on the simulation of Turbulence Kinetic Energy in a Major Alpine Valley

Boundary-Layer Meterol.

(2018)

J.J. Gómez-Navarro et al.

Sensitivity of the WRF model to PBL parametrisations and nesting techniques: evaluation of wind storms over complex terrain

Geosci. Model Dev.

(2015)

G.A. Grell et al.

A scale and aerosol aware stochastic convective parameterization for weather and air quality modeling

Atmos. Chem. Phys.

(2014)

W. Greuell et al.

Elevational changes in meteorological variables along a mid-latidtude glacier during summer

J. Geophys. Res.

(1997)

S. Hanna

Diurnal variation of horizontal wind direction fluctuations in complex terrain at Geysers

California. Bound.-Layer Meteor.

(1981)

S.-Y. Hong

A new stable boundary-layer mixing scheme and its impact on the simulated East Asia summer monsoon Q. J. R

Meteorol. Soc.

(2010)

G. Kirkil et al.

Implementation and Evaluation of Dynamic Subfilter-Scale stress Models for Large-Eddy simulation using WRF Mon

Wea. Rev.

(2012)

Cited by (12)

Evaluation of the GRAMM/GRAL model for high-resolution wind fields in Heidelberg, Germany
2024, Atmospheric Research
Independent verification of mitigation efforts for climate and air quality action in cities relies on inferring emissions from atmospheric concentration measurements. As emissions are dispersed in the atmosphere before they reach an instrument, the quantitative estimation of emissions requires an understanding of the atmospheric transport and associated uncertainties. In this study, we analyse the catalogue of steady-state flow fields generated by the Graz Mesoscale Model (GRAMM) coupled to the Graz Lagrangian Model (GRAL) for an entire year in Heidelberg, Germany. We use a loss function for the wind field selection, which assigns a best-matching catalogue entry to any given hour by exploiting observation data. We introduce a new loss function which finds an optimal balance between differences in wind speed and wind direction. We evaluate the performance of the model based on 15 meteorological measurement sites, of which 14 are in the inner high-resolution and building-resolving GRAL domain (12.5 km × 12.5 km, 10 m resolution). Performance metrics include mean bias (MB) and root mean square errors (RMSEs) of simulated and observed wind speed and wind direction for all individual stations. On average, we find a mean underestimation of wind speed of 0.14 ms⁻¹ corresponding to about 7 % of the mean wind speed and a mean RMSE of 1.03 ms⁻¹. For wind direction, a mean overall bias smaller than 1° is achieved, but individual stations show larger biases (mean absolute bias: 37°), especially at stations where wind speeds are low on average. Evaluation benchmarks for mean biases of wind direction and wind speed of mesoscale models provided by the European Environmental Agency (EEA) are met at 11 and 14 out of 15 stations at low measurement heights, respectively. Recently suggested extended benchmarks for complex terrain are met at almost all stations. Additionally, for the first time, we analyse the model's ability to simulate the vertical wind profile and we analyse the benefit of implementing a wind profile measurement into the process. We find that the model does not fully capture the vertical profile in our setting. We further study the required measurement network size and find that a high number (> 6) of meteorological stations improves the selection of flow fields over the entire GRAL domain substantially. The conducted comprehensive analysis of the wind fields in the GRAL domain are the basis for detailed quantitative analysis of greenhouse gas and air pollutant emissions using the GRAMM/GRAL modelling framework.
High-frequency fluctuation of air temperature during a heatwave event in urban environment and the physical mechanism behind
2023, Building and Environment
Heatwaves threaten human health and power systems. Urban climate is non-stationary and wide-spectrum, with high-frequency temperature and wind-speed variations that could overload power grids and expose people to extreme heat. In this study, Hilbert-Huang transform (HHT) was unprecedentedly used to decompose the urban-scale temperature (IMFθ1 to IMFθ6) and wind-speed (IMFW1 to IMFW6) signals during a 5-day heatwave event into 6 intrinsic mode functions (IMFs). The spatio-temporal characteristics, physical mechanism, and effective ranges of high-frequency components (IMF1 to IMF4) were unveiled. Temperature (wind speed) IMFθ1 to IMFθ4 (IMFW1 to IMFW4) had a temporal scale of 2.63 h (2.53 h), 5.88 h (5.78 h), 13.16 h (9.84 h), and 22.72 h (19.05 h); as well as a spatial scale of 2.31 km (0.99 km), 4.29 km (1.65 km), 5.94 km (2.64 km), and 6.6 km (2.97 km), respectively. The physical mechanisms of IMF1 to IMF4 were composed of turbulence and heat storage/release; disturbance induced by mountainous terrain and slope flows; land/sea breeze, together with anthropogenic heat. Besides, the peaked amplitudes of IMFθ1 were most risky in compact/open high-rise urban (1.4 °C–1.6 °C) rather than rural (0.6 °C–1.0 °C) areas. The foothill areas within 8-km coverage were susceptible to IMFθ2 (1 °C–2.1 °C). IMFθ3 (0.6 °C–3.6 °C) was effective in urban areas within 10 km from coastline. IMFθ4 (2.5 °C–3.5 °C) exhibited the most intense fluctuation in urban/suburban areas. The outcome provides references for policy makers to mitigate heat-related risks.
Environmental efficiency of the fossil fuels to electricity transition in Eastern Siberia cities
2023, Atmospheric Pollution Research
In many regions of the world, the primary source of domestic heating and cooking is still coal or wood, which severely threatens people's health. Low-pipe stove emissions negatively impact about 2.4 million urbanized residents in Eastern Siberia. Through Lagrangian dispersion modelling, this study estimated the effectiveness of coal and firewood to electricity transition in the context of reducing hazardous concentration zones of Benzo [a]pyrene (BaP) and Particulate matter (PM10) in six cities. The calculation experiments showed that a transition of half of the households to electricity could be practically reduced to zero zones of the negative impact of pollutants in the ground layer. The GRAMM/Gral model enabled it to implement pollutant particle tracking and calculate hourly changes in their concentrations at 2 and 10 m. The study results are oriented to support practical solutions to improve the quality of life in cities by reducing the pollution of their atmosphere.
Urban population exposure forecast system to predict NO<inf>2</inf> impact by a building-resolving multi-scale model approach
2021, Atmospheric Environment
Citation Excerpt :
Observed wind speed and direction at the same three urban stations have been used to validate modelled wind fields by both WRF-Chem and Micro-SWIFT. The temporal correlation of simulated wind speed with observations is between 0.50 and 0.54 for both the models, which is comparable to the values obtained in simulations carried out in other European cities (Gsella et al., 2014; Kuik et al., 2016; Oettl and Veratti, 2021). Wind speed is overestimated at all the three sites with MB between +0.18 and + 0.99 m s−1 (corresponding to +7% and +44% of NMB), respectively at OSS and POL stations for WRF-Chem and PMSS.
Operational forecasting systems based on chemical transport models (CTMs) nowadays generally produce concentration maps with a resolution in the order of 2–5 km, very rarely exceeding the sub-kilometre scale. The main reason for this restriction is the prohibitive computing cost that a simulation covering an entire country would have if set-up with a resolution in the order of meters. In this paper a hybrid forecast system, relying on the WRF-Chem model coupled with the PMSS Lagrangian modelling suite, has been developed and applied for each day of February 2019, to predict hourly NO₂ and NO_x concentrations with a spatial resolution of 4 m, for the urban area of Modena (a city located in the central Po Valley). Simulated meteorological fields (temperature, wind speed and direction) were assessed at three urban stations, compliant with WMO standards, and modelled concentrations were compared with measurements at two urban air quality stations located at background and traffic sites.
Results show that meteorological variables are well captured by the hybrid system and statistical performances are in line with the benchmark values suggested by the European Environmental Agency and with similar case studies focusing on the same area. Modelled NO₂ and NO_x concentrations, notwithstanding a slight underestimation mainly evident at urban traffic stations for NO_x, present a large agreement with related observations. The NO₂ Model Quality Objective, as defined by Fairmode guidelines, was met for both the urban stations and the other statistical indexes considered in the evaluation fulfilled the acceptance criteria for dispersion modelling in urban environment, for both NO₂ and NO_x concentrations.
In the second section of the study, the population exposure to forecasted NO₂ concentrations has been evaluated adopting a generic model of dynamic population activity. The population was distributed at hourly time steps in specific urban micro-environments at the same resolution of the concentration maps (4 m) and the short-term exposure has been computed as the product between the population density in each model cell and related surface NO₂ concentrations. An infiltration factor was also applied to estimate indoor concentrations. The hybrid system was shown to be particularly suited for assessing short-term peak exposure in areas influenced by traffic emissions. On the other hand, due to the limited time spent by the population within traffic related environments, the long-term population exposure calculated by the hybrid system tends to be similar to the WRF-Chem stand-alone estimate.
The Performance of GRAMM-SCI and WRF in Simulating the Surface-Energy Budget and Thermally Driven Winds in an Alpine Valley
2023, Boundary-Layer Meteorology
Recent developments in high-resolution wind field modeling in complex terrain for dispersion simulations using GRAMM-SCI
2023, Air Quality, Atmosphere and Health

View all citing articles on Scopus

View full text

A comparative study of mesoscale flow-field modelling in an Eastern Alpine region using WRF and GRAMM-SCI

Highlights

Abstract

Introduction

Section snippets

GRAMM-SCI

Study area

Results

Conclusions

Author statement

Declaration of Competing Interest

Atmos. Environ.

Atmos. Environ.

Italy. Atmos. Environ.

Atmos. Environ.

Journal of Wind Engineering and Industrial Aerodynamics

Atmos. Environ.

Develop. Environ. Sci

Atmos. Environ.

Atmos. Res.

Atmos. Environ.

A new finite volume discretisation for solving the Navier-Stokes-equations

Numerical Methods in Laminar and Turbulent Flow

Estimation of the pollution concentration in the vicinity of a cellulose plant in an alpine valley using the non-hydrostatic mesoscale model GRAMM. Air Pollution III

Analysis of the daily variations of wintertime air pollution concentrations in the city of Graz-Austria

Environ. Monit. Assess.

Can mesoscale models reproduce meandering motions?

Q.J.R. Meteorol. Soc.

Evaluation of high-resolution GRAMM–GRAL (v15.12/v14.8) NOx simulations over the city of Zürich, Switzerland

Geosci. Model Dev

Using wind fields from a high-resolution atmospheric model for simulating snow dynamics in mountainous terrain

Hydrol. Process.

Numerical simulations of mountain winds in an alpine valley

Wind Struct.

Coupling an advanced land-surface/hydrology model with the Penn State/NCAR MM5 modeling system, part I: model description and implementation

Mon. Weather Rev.

ERA5: Fifth generation of ECMWF atmospheric reanalyses of the global climate

The impact of MM5 and WRF meteorology over complex terrain on CHIMERE model calculations

Atmos. Chem. Phys.

Approximate equations of motion for gases and liquids

J. Atmos. Sci.

The application of models under the European Union’s Air Quality Directive: A technical reference guide

Rep. Nr. 10/2011

Computational Methods for Fluid Dynamics

Numerical simulations of boundary-layer processes and urban-induced alterations in an Alpine valley

Int. J. Climatol.

The Impact of Three-Dimensional Effects on the simulation of Turbulence Kinetic Energy in a Major Alpine Valley

Boundary-Layer Meterol.

Sensitivity of the WRF model to PBL parametrisations and nesting techniques: evaluation of wind storms over complex terrain

Geosci. Model Dev.

A scale and aerosol aware stochastic convective parameterization for weather and air quality modeling

Atmos. Chem. Phys.

Elevational changes in meteorological variables along a mid-latidtude glacier during summer

J. Geophys. Res.

Diurnal variation of horizontal wind direction fluctuations in complex terrain at Geysers

California. Bound.-Layer Meteor.

A new stable boundary-layer mixing scheme and its impact on the simulated East Asia summer monsoon Q. J. R

Meteorol. Soc.

Implementation and Evaluation of Dynamic Subfilter-Scale stress Models for Large-Eddy simulation using WRF Mon

Wea. Rev.