Next Article in Journal
Excessive Smoke from a Neighborhood Restaurant Highlights Gaps in Air Pollution Enforcement: Citizen Science Observational Study
Previous Article in Journal
Spatial-Temporal Assessment of Traffic-Related Pollutants Using Mobile and Stationary Monitoring in an Urban Environment
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Air
Volume 3
Issue 3
10.3390/air3030019
air-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Impact of Real-Time Boundary Conditions from the CAMS Database on CHIMERE Model Predictions

by
Anita Tóth
* and
Zita Ferenczi
HungaroMet, Kitaibel 1, H-1024 Budapest, Hungary
*
Author to whom correspondence should be addressed.
Air 2025, 3(3), 19; https://doi.org/10.3390/air3030019
Submission received: 5 June 2025 / Revised: 9 July 2025 / Accepted: 14 July 2025 / Published: 18 July 2025
Browse Figures
Versions Notes

Abstract

Air quality forecasts play a crucial role in informing the public about atmospheric pollutant levels that pose risks to human health and the environment. The accuracy of these forecasts strongly depends on the quality and resolution of the input data used in the modelling process. At HungaroMet, the Hungarian Meteorological Service, the CHIMERE chemical transport model is used to provide two-day air quality forecasts for the territory of Hungary. This study compares two configurations of the CHIMERE model: the current operational setup, which uses climatological averages from the LMDz-INCA database for boundary conditions, and a test configuration that incorporates real-time boundary conditions from the CAMS global forecast. The primary objective of this work was to assess how the use of real-time versus climatological boundary conditions affects modelled concentrations of key pollutants, including NO2, O3, PM10, and PM2.5. The model results were evaluated against observational data from the Hungarian Air Quality Monitoring Network using a range of statistical metrics. The results indicate that the use of real-time boundary conditions, particularly for aerosol-type pollutants, improves the accuracy of PM10 forecasts. This improvement is most significant under meteorological conditions that favour the long-range transport of particulate matter, such as during Saharan dust or wildfire episodes. These findings highlight the importance of incorporating dynamic, up-to-date boundary data, especially for particulate matter forecasting—given the increasing frequency of transboundary dust events.

1. Introduction

Air pollution remains one of the most significant environmental health risks globally, contributing to a wide range of adverse health effects even at low concentrations [1]. During specific pollution episodes—such as wildfires, dust storms, or elevated local anthropogenic emissions combined with stagnant meteorological conditions in the planetary boundary layer—pollutant levels can increase dramatically, resulting in higher mortality rates and hospital admissions [2]. Therefore, accurate air quality forecasts are important, especially to protect sensitive population groups by issuing health advisories and mitigation measures in advance.
Chemical transport models (CTMs) are essential tools for simulating the emission, transport, chemical transformation, and deposition of atmospheric pollutants [3]. These models are widely used for both air quality forecasting and regulatory assessments [4,5]. Depending on the spatial scale, different modelling systems are applied, i.e., global CTMs provide estimates of pollutant transport on intercontinental scales, while regional models focus on smaller domains with higher spatial resolution, allowing for a more detailed representation of atmospheric processes [6].
To operate effectively, regional air quality models require a range of input data, including meteorological fields, emission inventories, land use data, initial conditions, and boundary conditions. While initial conditions are necessary only for the first time step of a simulation, boundary conditions must be provided continuously along the edges of the modelling domain throughout the simulation period to ensure a consistent solution of the model’s three-dimensional equations [7].
In operational air quality forecasting, boundary conditions are usually derived from global model outputs. In contrast, for retrospective policy assessments or scenario modelling, boundary conditions may incorporate observationally constrained data to increase accuracy [8]. One of the most widely used global data sources is the Copernicus Atmosphere Monitoring Service (CAMS), part of the European Union’s Earth observation programme. The CAMS provides daily data on atmospheric composition using the Integrated Forecasting System (IFS), which is also used in ECMWF’s numerical weather prediction. The CAMS extends the IFS by including modules for aerosols, reactive gases, and greenhouse gases developed in the precursor projects GEMS and MACC [9,10].
The CAMS products—available both as forecasts and reanalyses—are widely used in regional air quality modelling, particularly as boundary conditions for European-scale and national models [11,12,13,14]. However, uncertainties remain, arising from both input data (e.g., emissions, meteorology, boundary conditions) and the model formulation itself (e.g., chemical mechanisms, parameterisations) [12]. Among these, boundary conditions play a critical role in influencing pollutant levels within regional domains, particularly for pollutants with strong long-range transport components.
Several studies have investigated the impact of boundary conditions on regional model performance. Im et al. [12] showed that long-range transport of ozone from global models significantly influences surface ozone levels in Europe and North America, whereas PM10 and PM2.5 are predominantly affected by local emissions. Their study also highlighted a greater dependency on boundary conditions during spring, when transboundary transport tends to be more active. Makar et al. [15] demonstrated improvements in ozone forecasts by applying a tropopause-height based dynamic adjustment to climatological boundary data. Similarly, Katragkou et al. [16] found that the impact of chemical boundary conditions on surface ozone, based on a series of sensitivity studies, is comparable to the changes caused by different meteorological forcing.
Jiménez et al. [7] examined the relative influence of initial and boundary conditions over time, concluding that the effects of initial ozone concentrations diminish after a 48 h spin-up period, whereas the influence of boundary conditions persists, especially near domain edges where short- to medium-range pollutant transport is significant. Akritidis et al. [17] tested three regional chemical transport model simulations using three different lateral boundary condition setups. In the first run, the boundary conditions were invariant in both space and time. In the second run, they were represented by monthly averages from a global model, allowing for seasonal variability. In the third run, the boundary conditions were also derived from a global model, but included both seasonal and inter-annual variability. The results showed that using boundary conditions with spatial and temporal variability improved the representation of ozone variability. However, incorporating inter-annual variability did not enhance the correlation between modelled and observed concentrations. Additionally, both the normalised standard deviation and the normalised mean bias were improved on a seasonal basis when time- and space-dependent boundary conditions were applied.
At HungaroMet, the Hungarian Meteorological Service, the CHIMERE chemical transport model is used operationally to generate daily air quality forecasts and to assess the previous year’s air quality over Hungary. In our earlier research, we explored the impact of meteorological uncertainties on simulated pollutant concentrations [18,19]. In the present study, we focus on another key source of uncertainty: the effect of boundary conditions on modelled concentrations.
Our approach involved a comparative analysis between the current operational configuration of CHIMERE, which uses climatological averages from the LMDz-INCA (Laboratoire de Météorologie Dynamique—INteraction with Chemistry and Aerosols) global database, and two test configurations incorporating the CAMS global forecast data. In the first test configuration, all boundary condition data—both gases and aerosols—were derived from the CAMS global forecast. In the second, only the aerosol-related species were replaced by the CAMS real-time data, while gas-phase pollutant boundary conditions remained based on climatological averages.
The primary aim of this study was to quantify the impact of using real-time boundary conditions, as opposed to climatological ones, on the predicted concentrations of key pollutants (NO2, O3, PM10, and PM2.5) over Hungary. To evaluate the model performance, we compared simulated results against measurements from the Hungarian Air Quality Monitoring Network using a set of statistical performance metrics. The findings contribute to a better understanding of how dynamic boundary data can enhance regional-scale air quality forecasting, particularly in the context of episodic events such as Saharan dust transport or wildfire plumes, which are expected to increase in frequency under changing climate conditions.
This study provides a systematic evaluation of the effects of real-time chemical boundary conditions on regional air quality modelling over Central Europe using the CHIMERE model. While previous studies have primarily focused on seasonal or idealised sensitivity experiments, our work presents an operationally relevant comparison based on the realistic, daily-varying CAMS forecast inputs versus climatological boundary conditions.

2. Materials and Methods

2.1. Models

CHIMERE is a Eulerian, offline, source-oriented chemistry-transport model. It calculates and provides the atmospheric concentrations of gas-phase and aerosol species over local to continental level [20]. The CHIMERE model (version CHIMERE2017) at HungaroMet runs every day at 00 UTC to make 48 h air quality forecasts for PM10, PM2.5, SO2, NO2, and O3 pollutants. The operational forecasts are generated with a 1 h time step and a horizontal resolution of 0.1° × 0.1° for the entire modelling domain, which covers the whole country (from latitudes 45° to 50° and longitudes 14° to 25°). Forecasts with finer horizontal resolution (0.02° × 0.015°) are available for three larger cities (Budapest, Miskolc, Pécs). The most recent CHIMERE forecasts are accessible on our website (https://legszennyezettseg.met.hu/ accessed on 27 March 2025), where the model outcomes can be viewed on maps and graphs.
CHIMERE relies on several data sources for its calculations. Figure 1 illustrates the various types of data that independently serve as an input for the model. As CHIMERE is an offline model, it does not calculate meteorological fields. Meteorological forecasts from the Hungarian Meteorological Service’s Application of Research to Operations at Mesoscale (AROME) [21] numerical model are used in the chemical transport model. The source of the anthropogenic emissions is the European Monitoring and Evaluation Programme (EMEP) [22] inventory for the year 2019. The biogenic emissions stem from the global Model of Emissions of Gases and Aerosols from Nature (MEGAN) [23]. CHIMERE uses the U.S. Geological Survey (USGS) data [24] to calculate land use categories. For the preparation of the chemistry within the model, the SAPRC chemical mechanism was chosen. CHIMERE is a limited-area model, and boundary and initial conditions are needed to obtain appropriate model results. In the operational setup, climatological averages are used as boundary conditions and the previous simulation produces the initial conditions for the next run. The boundary condition data in the case of gaseous species and non-dust aerosols originate from the LMDz-INCA [25] model. The specific version (LMDz4-INCA3) used in our model system includes 19 hybrid vertical levels extending up to 4 hPa. Its horizontal resolution is 1.9° in latitude and 3.75° in longitude. The global fields of the pollutants are monthly averages derived from a multi-year simulation. Further documentation of the model is described in Hauglustaine et al. [26]. Boundary condition information in the case of the dust aerosols is taken from the Goddard Chemistry Aerosol Radiation and Transport (GOCART) [27] model. The global distribution of dust aerosols is provided as monthly averages across 20 vertical levels, with a horizontal resolution of 2° in latitude and 2.5° in longitude.
CHIMERE enables the use of various boundary condition sources (simulations from different models), or even combinations of these sources. The model incorporates both top and lateral boundary conditions. Alongside the operational runs, we tested the CHIMERE model with two setups where we substituted the different types of boundary condition data with the CAMS global forecasts.
The CAMS uses ECMWF’s operational infrastructure and modelling tools to produce near-real-time global analyses and forecasts of atmospheric composition, alongside consistent long-term reanalyses [28]. It also provides four-day forecasts of pollutants for Europe based on the CAMS regional ensemble model and additional products designed to address the needs of users in policy support, industry, and the scientific community. The global atmospheric composition forecasts are produced twice a day. The forecasts cover over 50 chemical species (such as ozone, nitrogen dioxide, and carbon monoxide) and seven types of aerosols (including desert dust, sea salt, organic matter, black carbon, sulphate, nitrate, and ammonium aerosols) [29]. Additionally, a range of meteorological variables is provided.
To build the boundary condition input for CHIMERE, the parameters listed in Table 1 were required. We utilised the CAMS forecast data beginning at 00 UTC, downloading 48 h forecast data. The data retrieval process was automated through the Climate Data Store Application Program Interface (CDS API) system. The CAMS data were retrieved from a region defined by latitudes 43° and 53°, and longitudes 10° and 30°. The data have a horizontal resolution of 0.4° × 0.4° and a temporal resolution of 3 h. We used data from 10 vertical levels between 1013.25 hPa and 330 hPa.
After converting the CAMS data into a format compatible with the CHIMERE pre-processor, some minor adjustments to the pre-processor itself were necessary. This pre-processor interpolates the input data spatially and temporally, and performs the unit conversion from kg/kg to ppb using the temperature data. First, we had to define new builder files which are responsible for the matching of the components in the raw boundary condition database (the CAMS global forecast) with the CHIMERE species. Next, a new pressure formula had to be incorporated into the pre-processor code to calculate the pressure at the CAMS levels, enabling the recalculation of concentrations and temperature at the CHIMERE levels. Finally, small modifications were needed in the time dimension handling to properly interpolate the 3-hourly data into hourly data.
CHIMERE runs with the two different test configurations were carried out in parallel with the operational forecasts for the year 2024. The first run using the CAMS global boundary input will hereafter be referred to as Test1 and the second run using the mixed CAMS and climatological boundary input as Test2.

2.2. Measurements

The measured concentrations of NO2, O3, PM10, and PM2.5 were compared to the modelled values. Twelve stations were selected for the evaluation. These stations roughly cover the whole country. They can be grouped into four types: rural background, urban background, suburban background, and urban traffic. Figure 2 shows the location of the stations and the colour of the markers indicate their type. Unfortunately, not all stations measure all pollutants.
The modelled value of the grid cell closest to the site location was compared with the corresponding measurement to evaluate model performance. Based on these value pairs, several statistical indicators were calculated. The statistics used in the analysis are calculated with the formulas provided in Appendix A. In addition to the basic statistical performance indicators (R, BIAS, SD), the benchmarking methodology developed within the framework of the Forum for Air quality Modeling network (FAIRMODE) was also applied. The Modelling Quality Indicators (MQIs) were calculated, which compare modelled and measured concentrations while accounting for measurement uncertainty too [30]. Following the FAIRMODE recommendation, a model is fit for the purpose if the modelling quality objective is met, which means if at least 90% of the stations have MQI values lower than 1. The statistical indicators and MQI values were calculated based on hourly NO2, daily PM10, daily PM2.5, and daily maximum of 8 h mean O3 concentrations. The FAIRMODE recommendation was used to define the minimum data availability at the stations [30]. For the selected time period and time averaging interval, the requested percentage of the available data is 75%: if this is not fulfilled, the station is not to be taken into account for the evaluation of a given pollutant. The formula of MQI is the following:
M Q I = R M S E β R M S U ,
where β determines the stringency of the modelling quality objective, the numerator is the RMSE between observed and modelled values, and RMSU is a value representing the maximum allowed measurement uncertainty. β is equal to 2 in the case of all pollutants, which means that the difference between the measured and modelled concentrations can be at most twice the measurement uncertainty. The details of the calculation of the MQI can be found in the Guidance Document of FAIRMODE [30].

3. Results

3.1. Comparison of the Test Runs with the Operational CHIMERE Simulation

First, the modelled yearly concentrations simulated with the operational CHIMERE configuration were compared to those from the test runs. The concentrations were derived from the lowest model layer. Figure 3, Figure 4, Figure 5 and Figure 6 display the differences between Test1 and operational data in panel (a), and the differences between Test2 and operational data in panel (b), for the pollutants NO2, O3, PM10, and PM2.5. Test1 shows larger deviations from the operational run. In the case of NO2 (Figure 3), both test cases produced lower concentration values compared to the operational yearly average. The Test1 configuration shows larger concentration reduction, with a difference of about 20% closer to the domain boundaries.
During periods of unusually low pollutant concentrations (e.g., due to favourable weather conditions), real-time boundary conditions could result in lower NO2 concentrations compared to the average values. Weaker pollutant transport from neighbouring areas causes lower NO2 concentrations near the boundaries. If lower NO2 concentrations enter the model domain from surrounding regions, this may reduce the concentrations within city areas compared to those expected under constant climatological boundary conditions.
No significant change in annual O3 concentrations was observed for the replacement of only aerosol-type pollutants (Test2). In contrast, the Test1 configuration produced up to 20–40% higher O3 concentrations (Figure 4). The impact of real-time boundary conditions on surface ozone concentrations is larger within the inner parts of the domain. Global models can include long-range transport of ozone from distant regions, often from the upper troposphere or stratosphere. Therefore, the use of CAMS global forecasts as boundary conditions can lead to higher concentrations of ozone in the regional model. Figure 4 shows a more significant increase in O3 concentration in the areas around larger cities. Near cities, even if the decrease in NO2 is less significant, the increase in O3 can still be more pronounced. This may be because the local emissions of VOCs (from vehicles or the industry sector) may still be high, providing the necessary reactants for O3 formation while less NO is available to destroy it.
Regarding PM10, Figure 5 shows that both test configurations caused significant changes (even an increase of 40%) in the yearly averages in the southwestern part of the domain. The increase in the concentrations is more expanded in the case of Test2, although in most parts of the country the change is below 15%. A similar spatial distribution of concentration changes can be seen for PM2.5; however, their magnitude is lower. The activity of some natural sources of PM are timely integrated in the CAMS global system. Hence, sudden increases in PM10 concentrations at the domain boundaries due to regional dust storm or wildfire events could be captured with the two test configurations.
It can be concluded, that the usage of CAMS global forecasts as boundary conditions has a significant impact on the predictions made with the CHIMERE model. The extent of the deviation from the operational model varies depending on the area and the pollutant.

3.2. Model Performance Evaluation

The purpose of the evaluation is to compare the simulated concentrations with the measurements of the Hungarian Air Quality Monitoring Network and to see if the modelling system can accurately represent the surface concentrations of the pollutants. The analysis allows us to determine which configuration matches best the measured values. The assessment was conducted using hourly concentration data from 12 stations for a whole year. Due to the data availability criteria (75% for each time interval used for averaging, such as hourly values, daily averages, or daily maxima), the final dataset used for evaluation included 10 stations for NO2, 11 for O3, 11 for PM10, and 9 for PM2.5. In order to visualize the main aspects of model performance, we adapted the target diagram from FAIRMODE’s recommended benchmarking methodology. Figure 7 shows the target diagrams of the four pollutants. Each diagram offers a qualitative summary of the model’s performance, visually illustrating its accuracy in terms of BIAS and CRMSE (unbiased root mean square error) for each station. Each symbol represents the MQI value of that station, which is equal to the distance of the symbol from the origin. If a station falls within the green area, the MQI is less than 1. Additionally, the diagram also displays the 90th percentile of the MQI values for the stations in the top left corner and the parameters necessary for the derivation of the MQI values in the top right corner.
The target diagram of NO2 shows that the MQI values are below 1 for all stations and for all three model runs. The majority of the BIAS values is negative, only Budapest and Sarród lie in the upper half of the diagram. The performance of the three different configurations is similar at the stations; there are no outstanding differences in the position of the symbols representing the configuration types on the chart. Regarding O3, the Test1 configuration does not meet the requirement that 90% of the stations must have an MQI value lower than 1. All model runs have overestimated the measured concentrations. The difference between the performances of operational and Test2 configuration is small. While the symbols showing Test1 results are noticeably distant from those of the other configurations (shifted upwards), the relative positioning of the stations remains consistent. This indicates that the BIAS has increased significantly with the use of near-real-time boundary condition data and the stations were similarly affected. From the target diagram of PM10, we can see that none of the runs meets the model performance criteria, only in the case of the operational runs some stations fall inside the green area. The BIAS is overall negative; CHIMERE underestimates the measured concentrations. The only exception with positive BIAS is Nyírjes station. In the case of PM10, the Test1 and Test2 configurations produced similar performances; nevertheless, the stations were shifted together again. The CRMSE values increased; thus, the symbols indicating Test1 and Test2 are located to the left of the operative symbols. A higher CRMSE means that, although the average error remained almost the same, the spread of the errors around that average is greater. The PM2.5 target plot shows that the 90th percentile of the MQI values is over 1 for all the three model configurations. Similarly to PM10, with Test1 and Test2 configuration, we achieved similar model performances. Budapest and Nyírjes stations, with larger positive BIAS values, stand out from the rest. Again, we found that the CRMSE has changed more than the BIAS when compared to the operational model.
Figure 8 shows the correlation coefficient (R) values valid for the stations. The green bars visualize the R values of the operational, the red bars the Test1, and the orange bars the Test2 configuration. The correlation coefficients are generally higher for O3 than for the other pollutants. The test configurations did not change much the correlation between measured and modelled NO2 compared to the operational setup. The R values of PM10 are below 0.6, and the differences between two test configurations are small. The correlation increased at stations in the southwestern part of the country (Pécs, Szeged, Szombathely, and Veszprém) and at Nyírjes. In the case of PM2.5, the correlation increased only at Nyírjes when using near-real-time boundary conditions.
Figure 9 shows the normalised mean standard deviation (NMSD) values valid for the stations. The colouring of the three different model runs is the same as that in Figure 8. This metric provides insights into the variability of a model’s predictions relative to the observed data. For the majority of the cities, the modelled NO2 concentrations exhibit lower variability than observed. CHIMERE captures the observed variability the best at Budapest, K-puszta, and Szeged. Regarding O3, the NMSD values are positive for the Test1 case and negative for the other two cases at the stations, except for Miskolc, where all types of modelled standard deviation are larger than the observation. The NMSDs of PM10 changed a lot with the introduction of the test configurations. The operational model underestimates the observed PM10 values, but the usage of new boundary data increased the variability in the modelled concentrations. With the two test configurations the NMSDs reached closer to 1. This suggests that the model is able to capture episodic pollution events, but it seems to be overestimating the spread of the pollutant concentrations relative to the observed data. For PM2.5, it can be also seen that the test configurations improved the NMSD values. However, at Nyírjes and Budapest, the NMSD is above 1, which means that the model is too sensitive or overly influenced by small fluctuations in the input data at these locations.
Concerning the evaluation, there are some important findings. On the one hand, the statistical metrics that describe the accuracy of the NO2 predictions did not vary significantly between the different model configurations. One possible explanation is that the lifetime of NO2 is short, so even if the amount of pollutants that cross the borders is different from that of the operational model, there is very little change in the inner parts of the model domain. In the second place, the model performance of Test1 regarding O3 is different from the operational and Test2 results. Using real-time boundary conditions strengthened the overestimation of the O3 concentrations. Although the variability of the modelled concentrations has increased, the correlation has decreased. Using the global model output as boundary conditions can lead to higher ozone levels being imported into the regional model, as it does not account for smaller-scale local processes that might reduce ozone concentrations at the boundaries. Additionally, if the global model captures more stratosphere–troposphere exchange events, the amount of ozone in the lower troposphere can be higher than the climatological average. The Test1 configuration allows for more ozone entering the model domain, and, as the lifetime of ozone in the troposphere is long enough to allow it to be transported over long distances, ozone concentration is increasingly overestimated throughout the whole country. Regarding PM10, the overall underestimating behaviour of CHIMERE did not change with the introduction of real-time boundary data, but the degree of variability around this average error grew. Real-time boundary data may reflect short-term pollution events or spikes in PM10 concentrations that are not captured by climatological averages. As some of the episodic events, which can significantly raise PM10 levels (e.g., natural dust events), are presented in the CAMS global forecasts, the cross-border pollution could be reflected in our regional model results and this has improved the correlation and the matching between the modelled and measured standard deviation at some stations. In the case of PM2.5, all but one station showed a decrease in correlation with the Test1 and Test2 model configurations, and the NMSD values increased in all cases. When the model is exposed to more fluctuations in the boundary input data, the model captures the overall range of variability better, but the error variability increases too and this reduces the overall correlation strength.

3.3. Modelling Episode Situations: An Example of a Saharan Dust Event

From late March to early April 2024, warm air with a southerly flow brought significant amounts of Saharan dust towards Europe, resulting in high concentrations of aerosol particles across Hungary. On 1 April, the pollution level was very poor according to the Hungarian Air Quality Index and the daily average concentration exceeded 50 μg/m3 at all the PM10-measuring stations involved in this study. The elevated amount of desert dust could be modelled with the use of the CAMS forecasts as boundary conditions. The difference between the daily PM10 concentrations of the test configurations and the operational model is shown on Figure 10. We see that the differences are higher in the southern part of the model domain.
In Figure 11, the modelled and measured daily PM10 averages between 27 March and 2 April are displayed using bar charts. Two sites, Pécs and Szombathely, were selected to show the concentration-raising effect of the episode situation. Starting from these two cities, backward trajectories were drawn with the web version of the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model [31,32]. The two 72 h trajectories were plotted from a height of 1000 m above ground level to determine the dust transportation pathway. The air parcels originated from North Africa. The measured daily average PM10 concentrations (black bars) were above 140 μg/m3 at these two stations on 1 April. The operational model (green bars) could not follow the trend in the measured concentrations. The long-range transport of desert dust was observed in the two test configurations (red and orange bars). However, the maximum concentrations within the period were predicted by CHIMERE one day earlier. The maximum values of the test configurations’ daily averages were above 100 μg/m3, while the operational model’s maximum was significantly lower than 100 μg/m3.
The CAMS integrates satellite data and observation-based emission estimates to describe the state of the atmosphere accurately. By using the global CAMS forecasts as boundary conditions, the effect of long-range transport could be more realistically represented in our model system. CHIMERE can adjust immediately to extreme pollution events in the global model and provide more accurate forecasts. If there is a possibility of a Saharan dust intrusion, providing more accurate forecasts becomes especially important, not only to keep the public informed, but also to help experts better estimate solar power generation.

4. Conclusions

The impact of using the CAMS global forecasts as boundary conditions in the CHIMERE model run at HungaroMet has been investigated. The operational model, which uses climatological averages from the LMDz-INCA database for boundary conditions, was compared with two types of test configurations. In one configuration, the boundary conditions for all pollutant types were replaced with the data from the CAMS forecasts, while in the other, only the gaseous pollutants kept climatological averages. The year 2024 was selected for the analysis. Initially, 12 stations providing O3, NO2, PM10, and PM2.5 concentrations were chosen to evaluate the model. Unfortunately, not all station data could be included for every pollutant due to data gaps found in the measured time series. The model results were compared using maps, basic statistical metrics, and the FAIRMODE’s MQI values to determine whether the test configurations predicted the measured values more accurately.
The map comparison of the model results revealed that the test configuration led to lower NO2 concentrations and higher O3, PM10, and PM2.5 concentrations. No improvement in the MQI fulfilment was achieved with the test configurations, although some aspects of the model performance were improved. The effect of the choice of boundary conditions on NO2 is the smallest. The NO2 concentration is influenced mainly by the activity of local emission sources and its short lifetime limits the impact of transport from distant sources. However, Test2 has lower MQIs and slightly higher correlation coefficients compared to the other two model setups. Replacing the gaseous pollutants’ boundary conditions with the CAMS data (Test1) significantly degraded the model’s performance in predicting O3 concentrations. The overestimation of O3 increased significantly. This might be because the CAMS global model and our regional CHIMERE model have different spatial and temporal resolutions. The CAMS model, with its coarser resolution, may introduce errors that are then propagated during the interpolation process when downscaling to the finer resolution of the CHIMERE model. These propagated errors can affect the accuracy of the final forecast. Other shortcomings of using the CAMS global forecasts as boundary conditions in our system include differences in the representation of fluxes and the usage of different chemical schemes and reaction constants.
The BIAS of PM10 was reduced by the use of CAMS as the boundary data. However, the MQI values aggregated from station values were further from 1 in the test runs. Replacing the boundary conditions for aerosol-type pollutants with real-time predictions improved the correlation coefficients of PM10 at some stations. The benefits of using the CAMS data are significant when the meteorological situation favours the large-scale transport of aerosol particles from an area where Saharan dust disperses or a wildfire episode is occurring. During dust storm events, the concentration of the natural dust fraction increases within PM10, while during wildfires, the amount of carbonaceous aerosols rises. The increasing frequency of natural dust storms reaching Europe, along with the growing demand for a greater proportion of solar power in electricity generation, necessitates the development of more accurate forecasts of particulate matter (PM) concentrations. Since both test configurations offer advantages (through real-time consideration of the effects of dust, organic, and black carbon mixing ratios) over the operational model in terms of PM10 forecasts, and Test2 caused the least degradation in modelled NO2 and O3 values, the implementation of the Test2 configuration is recommended for operational usage.
Future efforts will focus on identifying the dominant sources of systematic error within the modelling system. While the CAMS-based boundary conditions improve preparedness for extreme pollution episodes, the evaluation revealed that modelled concentrations may still significantly overestimate observed values during such events. Therefore, the additional components of the modelling framework require investigation. Planned developments include an evaluation of how the choice of meteorological input from different numerical weather prediction (NWP) models influences the performance of the CHIMERE regional chemical transport model, as well as an assessment of revised planetary boundary layer (PBL) height calculations. A detailed analysis of vertical mixing processes may help explain the persistent overestimation of O3 and the model’s inability to reproduce observed PM10 peaks during specific episodes.

Author Contributions

Conceptualization, Z.F.; methodology, A.T.; writing—review and editing, A.T.; visualization, A.T. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Data Availability Statement

The CAMS global forecast data used in this study are available at the Copernicus Atmosphere Monitoring Service (CAMS) Atmosphere Data Store (ADS) (https://ads.atmosphere.copernicus.eu/datasets/cams-global-atmospheric-composition-forecasts?tab=overview, last access 7 May 2025); and the measurement data are published by HungaroMet (https://legszennyezettseg.met.hu/, last access 7 May 2025).

Use of Artificial Intelligence

AI or AI-assisted tools were not used in drafting any aspect of this manuscript.

Acknowledgments

The research was funded by the Sustainable Development and Technologies National Programme of the Hungarian Academy of Sciences (FFT NP FTA).

Conflicts of Interest

Authors Anita Tóth and Zita Ferenczi were employed by the company HungaroMet. They declare no conflicts of interest.

Appendix A

The statistical metrics employed in the study are defined below, where N represents the total number of values, M denotes the model value, and O corresponds to the observed value.
Table A1. The statistical metrics used in this study and their formulas.
Table A1. The statistical metrics used in this study and their formulas.
Statistical MetricFormula
Average   observed   values   ( O ¯ ) O ¯ = ∑ i = 1 N O i N
Average   modelled   values   ( M ¯ ) M ¯ = ∑ i = 1 N M i N
Standard   deviation   of   the   observed   values   ( σ O ) σ O = 1 N ∑ i = 1 N O i − O ¯ 2
Standard   deviation   of   the   modelled   values   ( σ M ) σ M = 1 N ∑ i = 1 N M i − M ¯ 2
Root mean square error (RMSE) R M S E = 1 N ∑ i = 1 N O i − M i 2
Centred root mean square error (CRMSE) C R M S E = 1 N ∑ i = 1 N O i − O ¯ − M i − M ¯ 2
Correlation coefficient
(R)		 R = ∑ i = 1 N M i − M ¯ O i − O ¯ ∑ i = 1 N M i − M ¯ 2 ∑ i = 1 N O i − O ¯ 2
BIAS B I A S = M ¯ − O ¯
Normalised Mean Bias (NMB) N M B = B I A S O ¯
Normalised Mean Standard Deviation (NMSD) N M S D = σ M − σ O σ O

References

  1. Strak, M.; Weinmayr, G.; Rodopoulou, S.; Chen, J.; De Hoogh, K.; Andersen, Z.J.; Atkinson, R.; Bauwelinck, M.; Bekkevold, T.; Bellander, T.; et al. Long term exposure to low level air pollution and mortality in eight European cohorts within the ELAPSE project: Pooled analysis. BMJ 2021, 374, 1904. [Google Scholar] [CrossRef]
  2. Kelly, F.J.; Fuller, G.W.; Walton, H.A.; Fussell, J.C. Monitoring air pollution: Use of early warning systems for public health. Respirology 2011, 17, 7–19. [Google Scholar] [CrossRef]
  3. Borge, R.; López, J.; Lumbreras, J.; Narros, A.; Rodríguez, E. Influence of boundary conditions on CMAQ simulations over the Iberian Peninsula. Atmos. Environ. 2010, 44, 2681–2695. [Google Scholar] [CrossRef]
  4. Stortini, M.; Arvani, B.; Deserti, M. Operational forecast and daily assessment of the air quality in Italy: A Copernicus-CAMS downstream service. Atmosphere 2020, 11, 447. [Google Scholar] [CrossRef]
  5. Monteiro, A.; Miranda, A.I.; Borrego, C.; Vautard, R. Air quality assessment for Portugal. Sci. Total Environ. 2007, 373, 22–31. [Google Scholar] [CrossRef]
  6. Takigawa, M.; Niwano, M.; Akimoto, H.; Takahashi, M. Development of a one-way nested global-regional air quality forecasting model. SOLA 2007, 3, 81–84. [Google Scholar] [CrossRef]
  7. Jiménez, P.; Parra, R.; Baldasano, J.M. Influence of initial and boundary conditions for ozone modeling in very complex terrains: A case study in the northeastern Iberian Peninsula. Environ. Model. Softw. 2006, 22, 1294–1306. [Google Scholar] [CrossRef]
  8. Simpson, D.; Benedictow, A.; Berge, H.; Bergström, R.; Emberson, L.D.; Fagerli, H.; Flechard, C.R.; Hayman, G.D.; Gauss, M.; Jonson, J.E.; et al. The EMEP MSC-W chemical transport model—Technical description. Atmos. Chem. Phys. 2012, 12, 7825–7865. [Google Scholar] [CrossRef]
  9. Flemming, J.; Huijnen, V.; Arteta, J.; Bechtold, P.; Beljaars, A.; Blechschmidt, A.; Diamantakis, M.; Engelen, R.J.; Gaudel, A.; Inness, A.; et al. Tropospheric chemistry in the Integrated Forecasting System of ECMWF. Geosci. Model Dev. 2015, 8, 975–1003. [Google Scholar] [CrossRef]
  10. Inness, A.; Ades, M.; Agustí-Panareda, A.; Barré, J.; Benedictow, A.; Blechschmidt, A.; Dominguez, J.J.; Engelen, R.; Eskes, H.; Flemming, J.; et al. The CAMS reanalysis of atmospheric composition. Atmos. Chem. Phys. 2019, 19, 3515–3556. [Google Scholar] [CrossRef]
  11. Akritidis, D.; Katragkou, E.; Zanis, P.; Pytharoulis, I.; Melas, D.; Flemming, J.; Inness, A.; Clark, H.; Plu, M.; Eskes, H. A deep stratosphere-to-troposphere ozone transport event over Europe simulated in CAMS global and regional forecast systems: Analysis and evaluation. Atmos. Chem. Phys. 2018, 18, 15515–15534. [Google Scholar] [CrossRef]
  12. Im, U.; Christensen, J.H.; Geels, C.; Hansen, K.M.; Brandt, J.; Solazzo, E.; Alyuz, U.; Balzarini, A.; Baro, R.; Bellasio, R.; et al. Influence of anthropogenic emissions and boundary conditions on multi-model simulations of major air pollutants over Europe and North America in the framework of AQMEII3. Atmos. Chem. Phys. 2018, 18, 8929–8952. [Google Scholar] [CrossRef]
  13. Thürkow, M.; Schaap, M.; Kranenburg, R.; Pfäfflin, F.; Neunhäuserer, L.; Wolke, R.; Heinold, B.; Stoll, J.; Lupaşcu, A.; Nordmann, S.; et al. Dynamic evaluation of modeled ozone concentrations in Germany with four chemistry transport models. Sci. Total Environ. 2024, 906, 167665. [Google Scholar] [CrossRef]
  14. Giordano, L.; Brunner, D.; Flemming, J.; Hogrefe, C.; Im, U.; Bianconi, R.; Badia, A.; Balzarini, A.; Baró, R.; Chemel, C.; et al. Assessment of the MACC reanalysis and its influence as chemical boundary conditions for regional air quality modeling in AQMEII-2. Atmos. Environ. 2015, 115, 371–388. [Google Scholar] [CrossRef]
  15. Makar, P.A.; Gong, W.; Mooney, C.; Zhang, J.; Davignon, D.; Samaali, M.; Moran, M.D.; He, H.; Tarasick, D.W.; Sills, D.; et al. Dynamic adjustment of climatological ozone boundary conditions for air-quality forecasts. Atmos. Chem. Phys. 2010, 10, 8997–9015. [Google Scholar] [CrossRef]
  16. Katragkou, E.; Zanis, P.; Tegoulias, I.; Melas, D.; Kioutsioukis, I.; Krüger, B.C.; Huszar, P.; Halenka, T.; Rauscher, S. Decadal regional air quality simulations over Europe in present climate: Near surface ozone sensitivity to external meteorological forcing. Atmos. Chem. Phys. 2010, 10, 11805–11821. [Google Scholar] [CrossRef]
  17. Akritidis, D.; Zanis, P.; Katragkou, E.; Schultz, M.; Tegoulias, I.; Poupkou, A.; Markakis, K.; Pytharoulis, I.; Karacostas, T. Evaluating the impact of chemical boundary conditions on near surface ozone in regional climate–air quality simulations over Europe. Atmos. Res. 2013, 134, 116–130. [Google Scholar] [CrossRef]
  18. Homolya, E. A Levegőminőség Várható Alakulásának Vizsgálata Újgenerációs Diszperziós Modellek Alkalmazásával. Ph.D. Thesis, Hungarian University of Agriculture and Life Sciences, Budapest, Hungary, 2021. [Google Scholar]
  19. Ferenczi, Z.; Homolya, E.; Lázár, K.; Tóth, A. Effect of the uncertainty in meteorology on air quality model predictions. Időjárás 2021, 125, 625–646. [Google Scholar] [CrossRef]
  20. Mailler, S.; Menut, L.; Khvorostyanov, D.; Valari, M.; Couvidat, F.; Siour, G.; Turquety, S.; Briant, R.; Tuccella, P.; Bessagnet, B.; et al. CHIMERE-2017: From urban to hemispheric chemistry-transport modeling. Geosci. Model Dev. 2017, 10, 2397–2423. [Google Scholar] [CrossRef]
  21. Szintai, B.; Szűcs, M.; Randriamampianina, R.; Kullmann, L. Application of the AROME non-hydrostatic model at the Hungarian Meteorological Service: Physical parameterizations and ensemble forecasting. Időjárás 2015, 119, 241–265. [Google Scholar]
  22. Vestreng, V.; Breivik, K.; Adams, M.; Wagner, A.; Goodwin, J.; Rozovskaya, O.; Oacyna, J. Inventory Review 2005—Emission Data Reported to CLRTAP and Under the NEC Directive—Initial Review for HMs and POPs; EMEP Status Report; Norwegian Meteorological Institute: Oslo, Norway, 2005. [Google Scholar]
  23. Guenther, A.; Karl, T.; Harley, P.; Wiedinmyer, C.; Palmer, P.I.; Geron, C. Estimates of global terrestrial isoprene emissions using MEGAN (Model of Emissions of Gases and Aerosols from Nature). Atmos. Chem. Phys. 2006, 6, 3181–3210. [Google Scholar] [CrossRef]
  24. Loveland, T.R.; Reed, B.C.; Brown, J.F.; Ohlen, D.O.; Zhu, Z.; Yang, L.; Merchant, J.W. Development of a global land cover characteristics database and IGBP DISCover from 1 km AVHRR data. Int. J. Remote Sens. 2000, 21, 1303–1330. [Google Scholar] [CrossRef]
  25. Hauglustaine, D.A.; Hourdin, F.; Jourdain, L.; Filiberti, M.A.; Walters, S.; Lamarque, J.F.; Holland, E.A. Interactive chemistry in the Laboratoire de Meteorologie Dynamique general circulation model: Description and background tropospheric chemistry evaluation. J. Geophys. Res. 2004, 109, D04314. [Google Scholar] [CrossRef]
  26. Hauglustaine, D.A.; Balkanski, Y.; Schulz, M. A global model simulation of present and future nitrate aerosols and their direct radiative forcing of climate. Atmos. Chem. Phys. 2014, 14, 11031–11063. [Google Scholar] [CrossRef]
  27. Ginoux, P.; Chin, M.; Tegen, I.; Prospero, J.M.; Holben, B.; Dubovik, O.; Lin, S.J. Sources and distributions of dust aerosols simulated with the GOCART model. J. Geophys. Res. 2001, 106, 20255–20273. [Google Scholar] [CrossRef]
  28. Peuch, V.; Engelen, R.; Rixen, M.; Dee, D.; Flemming, J.; Suttie, M.; Ades, M.; Agustí-Panareda, A.; Ananasso, C.; Andersson, E.; et al. The Copernicus Atmosphere Monitoring Service: From research to Operations. Bull. Am. Meteorol. Soc. 2022, 103, E2650–E2668. [Google Scholar] [CrossRef]
  29. Global Atmospheric Composition Analyses and Forecasts 2024. Copernicus Atmosphere Monitoring Service, European Centre for Medium-Range Weather Forecasts. Available online: https://atmosphere.copernicus.eu/ (accessed on 5 June 2024).
  30. Janssen, S.; Thunis, P. FAIRMODE Guidance Document on Modelling Quality Objectives and Benchmarking: Version 3.2; Publications Office of the European Union: Luxembourg, 2022. [Google Scholar] [CrossRef]
  31. Stein, A.F.; Draxler, R.R.; Rolph, G.D.; Stunder, B.J.B.; Cohen, M.D.; Ngan, F. NOAA’s HYSPLIT Atmospheric Transport and Dispersion Modeling system. Bull. Am. Meteorol. Soc. 2015, 96, 2059–2077. [Google Scholar] [CrossRef]
  32. Rolph, G.; Stein, A.; Stunder, B. Real-time Environmental Applications and Display SYSTEM: READY. Environ. Model. Softw. 2017, 95, 210–228. [Google Scholar] [CrossRef]
Air 03 00019 g001
Figure 1. The operational CHIMERE input structure. The model (centre box) receives input from various sources shown in the surrounding boxes (the specific data used are highlighted in green).
Figure 1. The operational CHIMERE input structure. The model (centre box) receives input from various sources shown in the surrounding boxes (the specific data used are highlighted in green).
Air 03 00019 g001
Air 03 00019 g002
Figure 2. The measurement stations used for the evaluation of the modelling results. Red: rural background; green: suburban background; brown: urban background; blue: urban traffic type of stations.
Figure 2. The measurement stations used for the evaluation of the modelling results. Red: rural background; green: suburban background; brown: urban background; blue: urban traffic type of stations.
Air 03 00019 g002
Air 03 00019 g003
Figure 3. Relative differences in the yearly NO2 concentrations between (a) Test1 and operational run and (b) Test2 and operational run.
Figure 3. Relative differences in the yearly NO2 concentrations between (a) Test1 and operational run and (b) Test2 and operational run.
Air 03 00019 g003
Air 03 00019 g004
Figure 4. Relative differences in the yearly O3 concentrations between (a) Test1 and operational run and (b) Test2 and operational run.
Figure 4. Relative differences in the yearly O3 concentrations between (a) Test1 and operational run and (b) Test2 and operational run.
Air 03 00019 g004
Air 03 00019 g005
Figure 5. Relative differences in the yearly PM10 concentrations between (a) Test1 and operational run and (b) Test2 and operational run.
Figure 5. Relative differences in the yearly PM10 concentrations between (a) Test1 and operational run and (b) Test2 and operational run.
Air 03 00019 g005
Air 03 00019 g006
Figure 6. Relative differences in the yearly PM2.5 concentrations between (a) Test1 and operational run and (b) Test2 and operational run.
Figure 6. Relative differences in the yearly PM2.5 concentrations between (a) Test1 and operational run and (b) Test2 and operational run.
Air 03 00019 g006
Air 03 00019 g007
Figure 7. Target diagrams for the four pollutants: (a) NO2, (b) O3, (c) PM10, and (d) PM2.5. Colours indicate different monitoring stations, while symbols represent the type of model runs: circles for operational, triangles for Test1, and rectangles for Test2.
Figure 7. Target diagrams for the four pollutants: (a) NO2, (b) O3, (c) PM10, and (d) PM2.5. Colours indicate different monitoring stations, while symbols represent the type of model runs: circles for operational, triangles for Test1, and rectangles for Test2.
Air 03 00019 g007
Air 03 00019 g008
Figure 8. Correlation (R) values at the different stations for (a) NO2, (b) O3, (c) PM10, and (d) PM2.5 pollutants. The colours of the bar charts indicate the type of the simulation: operational run as green, Test1 run as red, and Test2 run as orange.
Figure 8. Correlation (R) values at the different stations for (a) NO2, (b) O3, (c) PM10, and (d) PM2.5 pollutants. The colours of the bar charts indicate the type of the simulation: operational run as green, Test1 run as red, and Test2 run as orange.
Air 03 00019 g008
Air 03 00019 g009
Figure 9. The normalised mean standard deviation (NMSD) values at the different stations for (a) NO2, (b) O3, (c) PM10, and (d) PM2.5 pollutants. The colours of the bar charts indicate the type of the simulation: operational run as green, Test1 run as red, and Test2 run as orange.
Figure 9. The normalised mean standard deviation (NMSD) values at the different stations for (a) NO2, (b) O3, (c) PM10, and (d) PM2.5 pollutants. The colours of the bar charts indicate the type of the simulation: operational run as green, Test1 run as red, and Test2 run as orange.
Air 03 00019 g009
Air 03 00019 g010
Figure 10. The difference between the (a) Test1 and (b) Test2 configurations and the operational model’s daily PM10 concentrations on 1 April.
Figure 10. The difference between the (a) Test1 and (b) Test2 configurations and the operational model’s daily PM10 concentrations on 1 April.
Air 03 00019 g010
Air 03 00019 g011
Figure 11. The evaluation of a Saharan dust episode. (a) HYSPLIT 72 h backward trajectories from Szombathely (red) and Pécs (blue) launched at 1000 m above ground level on 1 April. (b) Daily mean PM10 concentrations at Pécs between 27 March and 2 April 2024. (c) Daily mean PM10 concentrations at Szombathely between 27 March and 2 April 2024. The colours of the bar charts indicate the origin of the concentration: measured marked as black, operational run as green, Test1 run as red, and Test2 run as orange.
Figure 11. The evaluation of a Saharan dust episode. (a) HYSPLIT 72 h backward trajectories from Szombathely (red) and Pécs (blue) launched at 1000 m above ground level on 1 April. (b) Daily mean PM10 concentrations at Pécs between 27 March and 2 April 2024. (c) Daily mean PM10 concentrations at Szombathely between 27 March and 2 April 2024. The colours of the bar charts indicate the origin of the concentration: measured marked as black, operational run as green, Test1 run as red, and Test2 run as orange.
Air 03 00019 g011
Table 1. Parameters retrieved from the CAMS global forecasts. Units are in brackets next to parameter names.
Table 1. Parameters retrieved from the CAMS global forecasts. Units are in brackets next to parameter names.
TypeParameter
Gascarbon monoxide (kg/kg), ethane (kg/kg), formaldehyde (kg/kg), hydrogen peroxide (kg/kg), isoprene (kg/kg), methane (kg/kg), nitric acid (kg/kg), nitrogen dioxide (kg/kg), ozone (kg/kg), peroxyacetyl nitrate (kg/kg), sulphur dioxide (kg/kg), glyoxal (kg/kg), ammonia (kg/kg)
Non-dust aerosolhydrophilic black carbon aerosol mixing ratio (kg/kg), hydrophobic black carbon aerosol mixing ratio (kg/kg), hydrophilic organic matter aerosol mixing ratio (kg/kg), hydrophobic organic matter aerosol mixing ratio (kg/kg), sulphate aerosol mixing ratio (kg/kg)
Dustdust aerosol mixing ratio (0.03–0.55 µm; 0.55–0.9 µm and 0.9–20 µm intervals) (kg/kg)
Meteorologicaltemperature (K)
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Tóth, A.; Ferenczi, Z. Impact of Real-Time Boundary Conditions from the CAMS Database on CHIMERE Model Predictions. Air 2025, 3, 19. https://doi.org/10.3390/air3030019

AMA Style

Tóth A, Ferenczi Z. Impact of Real-Time Boundary Conditions from the CAMS Database on CHIMERE Model Predictions. Air. 2025; 3(3):19. https://doi.org/10.3390/air3030019

Chicago/Turabian Style

Tóth, Anita, and Zita Ferenczi. 2025. "Impact of Real-Time Boundary Conditions from the CAMS Database on CHIMERE Model Predictions" Air 3, no. 3: 19. https://doi.org/10.3390/air3030019

APA Style

Tóth, A., & Ferenczi, Z. (2025). Impact of Real-Time Boundary Conditions from the CAMS Database on CHIMERE Model Predictions. Air, 3(3), 19. https://doi.org/10.3390/air3030019

Article Metrics

Back to TopTop