Next Article in Journal
The Influence of Meteorological Parameters on PM10: A Statistical Analysis of an Urban and Rural Environment in Izmir/Türkiye
Previous Article in Journal
The Hemispheric Asymmetry of Gravity Wave Impact on the Polar Mesospheric Cloud, Based on the Aeronomy of Ice in the Mesosphere Satellite
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Atmosphere
Volume 14
Issue 3
10.3390/atmos14030420
atmosphere-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 thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Prediction of the SO2 Hourly Concentration for Sea Breeze and Land Breeze in an Urban Area of Split Using Multiple Linear Regression

by
Tanja Trošić Lesar
1,* and
Anita Filipčić
2
1
Croatian Meteorological and Hydrological Service, 10000 Zagreb, Croatia
2
Department of Geography, Faculty of Science, University of Zagreb, 10000 Zagreb, Croatia
*
Author to whom correspondence should be addressed.
Atmosphere 2023, 14(3), 420; https://doi.org/10.3390/atmos14030420
Submission received: 29 November 2022 / Revised: 13 February 2023 / Accepted: 15 February 2023 / Published: 21 February 2023
Browse Figures
Versions Notes

Abstract

:
The main goal of this paper is to study pollution during sea breeze days in the Split town center, which is placed near the industrial area with three cement plants and one asbestos cement plant, as well as a harbor with high traffic, and investigate the sources of pollution with SO2 and its relation to atmospheric parameters using stepwise multiple linear regression (MLR). The hourly temperature difference from the time of the sea breeze lull (dT) was considered in evaluating the influence of meteorological parameters on hourly pollutant concentrations. It was found that the wind direction index (WDI) is a significant predictor for the sea breeze, and wind speed, relative humidity, and dT are significant for the land breeze. A very high index of agreement of 0.9 was obtained by the MLR model for the land breeze, and 0.8 for the sea breeze. Low SO2 concentrations are observed at night, and increased values are found between 0800 and 1800 UTC. With WDI being the only predictor during sea breeze, local traffic is found to be the main anthropogenic source of SO2 pollution.

1. Introduction

A sea breeze occurs when the wind blows from sea to land during the day and from land to sea at night. In general, the sea breeze can affect the distribution of air pollution in coastal metropolitan areas due to the recirculation of the pollutants [1,2,3,4]. Due to acid deposition [5] and pollution problems such as acid rain, haze, and photochemical smog, as well as acidification of surface waters and damage to aquatic ecosystems, forests and vegetation, and corrosion of materials and structures, trace gas SO2 plays a significant role in the global climate [6,7,8,9]. Numerous research papers [10,11,12,13,14,15,16] investigated the effects of short-term exposure to high SO2 concentrations on human health, especially on mortality risk. Higher airborne SO2 concentrations have been associated with an increased risk of cardio- or cerebrovascular mortality [10,11,15]. It has also been shown to affect respiratory and total non-accidental mortality [12,13,14]. Asthmatic individuals exposed to high SO2 concentrations experience acute reactions characterized by bronchoconstriction (airway narrowing), and asthmatic children and adults are at much greater risk because acute reactions can occur at low SO2 concentrations [17]. Orellano et al. [16] have shown that SO2 has a negative effect on mortality even at low concentrations. The emission of plumes with high SO2 concentrations, resulting from processes such as coal and fuel combustions and sulfide oxide refining, is a characteristic of industrial areas [18,19], and the burning of biomass and volcanic eruptions are natural sources of SO2 (e.g., [20]). Anthropogenic sources of SO2 also include ship emissions (e.g., [21,22,23]). Bouchlagem et al. [24] concluded that a power plant is the main source of SO2 in Tunis, and that it is transported by the sea breeze with the highest concentrations. For the Gulf of Mississippi, Yerramilli et al. [25] reached similar conclusions. Recently, Bahtiyar et al. [26] used cluster analyses for PM10, SO2, NOX, and C at different monitoring stations and meteorological data (temperature, wind speed (WS), and wind direction (WD)) were assessed to determine potential source locations and to examine the seasonal variations.
Local orography has been shown to have a significant influence on sea breezes (e.g., [27,28,29,30,31,32]). This has also been confirmed for the Croatian part of the coast for sea breezes (e.g., [33,34,35]), as well as for the bora flow on the Adriatic coast (e.g., [36]), and this may have an impact on the transfer of pollutants. For the Croatian region, Telišman Prtenjak et al. [37] studied a severe SO2 episode during fog in the Northern Adriatic and concluded that high SO2 originated from an oil refinery and thermal power plant in Rijeka, where the local orography plays an important role in pollution transfer. Bralić et al. [38] conducted a study of monthly average pollution in the Split region, which included SO2. They concluded that SO2 and NO2 do not originate from the same source and that the highest SO2 concentrations occur in the summer months. A recent study by Trošić Lesar and Filipčić [39] showed that the sea breeze carries PM10 particles from the inland cement and asbestos cement industry across Split in the early afternoon. Pollution is also the highest during the morning sea breeze lull. Zhang et al. [40] showed that cement plants in China are associated with SO2 emissions leading to acid rain and smog problems in eastern and southern China. Ibrahim et al. [41] showed the same for cement plants in Lybia.
Multiple linear regression (MLR) has become a well-known and suitable method for modeling environmental systems, but it is not suitable for extreme or non-linear data (e.g., [42,43]). Lalas et al. [44] used the MLR model to determine the next-day SO2 concentration. He discovered that the cold season in Athens is controlled by WS, minimum temperature, and the amount of precipitation, while the warm season is controlled by WS, WD, and relative humidity (RH). Using spectral aspects of pollution, Tirabassi et al. [45] discovered that WS has a significant effect on SO2 concentration during sea breezes in Ravenna.
MLR models are often used for daily or monthly simulations, but rarely for hourly simulations or local winds. For instance, Shams et al. [46] used the MLR model and the Multilayer Perceptron (MLP) model from Artificial Neural Networks (ANN) to predict daily SO2 data in Tehran and found that MLP had an R2 of up to 0.9. Liu et al. [47], who used a stepwise MLP model for daily predictions, obtained the same result for SO2. Liu et al. [48] used the MLR model to study the relationship between pollution including daily SO2 concentrations and meteorological conditions. They concluded that the daily SO2 concentrations were negatively correlated with RH, temperature, precipitation, and surface pressure.
The hourly predictions of the MLR model are mainly for PM10 and PM2.5 particles (e.g., [49,50]). Hourly PM10 levels were obtained by Mukerjee et al. [49] and Kukkonen et al. [50] based on a correlation model between PMxx and NO2 with an IA between 0.45 and 0.65 and by Hrust et al. [51] using the MLP model for NO2, CO, O3, and PM10 between 0.6 and 0.9. There is less research on the hourly SO2 simulations. The MLR model was recently applied by Abdullah et al. [52] for Malaysia and SO2 for certain forecast hours with R2 = 0.5. Zhou et al. [53] recently analyzed the sea breeze cooling capacity (SBCC) using MLR simulation for the sea breeze. They showed that the variability of SBCC is explained by specific humidity and wind, and its spatial variation by the morning temperatures during the onset of the sea breeze. The hourly temperature change to the time of the sea breeze lull temperature is a significant predictor for the sea breeze, with an index of agreement (IA) of 0.8 during the sea breeze, according to Trošić Lesar and Filipčić [54], who used the MLR model for hourly simulations of PM2.5 particle concentrations in Split.
In this paper, the results of the simulation of SO2 hourly concentrations using the stepwise MLR model, which has not been used for SO2 concentrations and sea breeze, in an urban area of Split for selected sea breeze cases will be shown. The possible anthropogenic sources of pollution will also be discussed, as well as the SO2 concentration dependence on meteorological parameters (temperature, RH, WD, and WS). Measurements at the Kaštel-Sućurac station, where the industrial area is situated, will also be used in discussing the possible sources of pollution with SO2 in Split.

2. Methods

The MLR model is a method for modeling the linear relationship between a dependent variable (target) and one or more independent variables (predictors), and, in this case, the SO2 concentrations can be treated in response to the meteorological variables treated as predictors:
Xt = a0 + a1 G1,t + a2 G2,t + … + ai Gi,t + ɛt i = 0, …, k
where Xt is the pollutant concentration at time t, Gi,t stand for the meteorological variables at time t, ai are regression coefficients, and εt is the random error at time t. Multicollinearity between predictors leads to incorrect estimates of regression coefficients. To prevent multicollinearity between predictors, the VIF is calculated for each predictor in the model:
V I F = 1 1 − R 2 ,
where
R = M i − M i ¯ ⋅ S i − S i ¯ ¯ σ M ⋅ σ S ,
where Mi are hourly measurements and Si are simulations, and σ M and σ S are standard deviations for Mi and Si, respectively. M i ¯ and S i ¯ are average values of Mi and Si.
Measurements at the Split-1 station (see Figure 1) for the Split region are used for the MLR analysis for the period 2007–2009. These measurements include air temperature, WD and WS, RH, and SO2 concentrations.
Split is located on the Split (Marjan) peninsula in central Dalmatia (see Figure 1). The station AMS 3 Split-1, which monitors pollution levels from traffic as well as from industry, family homes, and smaller industrial facilities in the vicinity, is located in the center of the city, near a road with moderate traffic. The locations of inland cement plants, an asbestos cement plant, and the northern port of Split are also shown. The AMS 1 Kaštel-Sućurac station is situated 400 m inland from the cement plant in Kaštel-Sućurac, between two public roads, in the central part of the coastal area of the Bay of Kaštela. In the vicinity of this station, houses and industrial plants can be found. The main purpose of the station is to monitor the pollution that is a result of the local industry, but other sources of pollution cannot be excluded. SO2 is measured by a UV fluorescence method using an ML9850B instrument, which is reliable over a wide temperature range and has a measurement accuracy of 0.5 ppb.
Days with undisturbed coastal circulation at the Split-1 station were selected for the MLR model simulations, as in Trošić Lesar and Filipčić [54]. Days, where a visible change in WD from land to sea breeze and vice versa can be observed after and before the appearance of a sea breeze lull were selected. The objective method for selecting the time of the sea breeze lull was when the WS was below 1 m s−1 with a visible change in WD, before and after the sea breeze lull. The difference in air temperature between morning and evening sea breeze lulls was not allowed to exceed 1 °C when the coastal circulation was undisturbed [56]. The mean daily cloudiness was chosen to be below 4/10 in climatological terms [57]. The synoptic situations at 12 UTC (from the German Meteorological Service 2007–2009) show that there were no closed baric systems over southern Europe during the sea breeze days, and the mean sea level pressure was slightly elevated (1014–1020 hPa) with low gradients. The upper level map at 850 hPa shows weak geostrophic wind (up to 10 m s−1) over Croatia. One sea breeze case with an extreme hourly value of SO2 of 198.08 μg m−3 was excluded from further analysis.
The number of days with undisturbed coastal circulation is shown by month in Table 1, and a total of 40 days were selected at the Split-1 station. Table 2 shows the mean and maximum 24 h concentrations for selected days at the Split-1 station compared with the corresponding concentrations at the Kaštel-Sućurac station and limit values for human health. It is obvious that larger mean and maximum 24 h concentrations were measured at the Split-1 station and that the measured concentrations are far below the corresponding limit values from the EU directive (2008/EC/50).
Instead of WD, which is expressed in degrees, the wind direction index (WDI) is employed [58]:
WDI = 1 + sin(WD+ π/4).
Although the land breeze and the sea breeze are opposite winds and the WDI is not uniquely defined from 0 to 360°, this does not affect the MLR model calculations. For the sea breeze from the time of the morning sea breeze lull to the time of the evening sea breeze lull, as well as for the nighttime part of the coastal circulation, the air temperature difference (dT) is used as the temperature deviation from the temperature at the time of the morning sea breeze lull. C6, C7, and C8 are used as the SO2 concentrations at 6, 7, and 8 UTC, respectively, which are selected as predictors for the morning concentration initialization of the MLR model. A similar approach was made by Trošić Lesar and Filipčić [54] for hourly PM2.5 simulations.
The root mean square error (RMSE) between the output of the MLR model and the measured SO2 concentrations is calculated as follows:
R M S E = ∑ i N M i − S i 2 N ,
and the mean absolute error (MAE) is calculated with
M A E = ∑ i N M i − S i N ,
where N is the number of observations, Si are the hourly model-simulated predictions and Mi are the hourly measurements.
In addition, the index of agreement (IA) is calculated using the following equation:
I A = 1 − ∑ i N M i − S i 2 ∑ i N M i − S ¯ + S i − S ¯ 2 .

3. Results

Figure 2 shows the mean hourly change of air temperature and RH, and WS and vector mean WD, from April to June, and Figure 3 shows the same from July to September at the AMS 3 Split-1 station. The temperature difference between land and sea causes a thermal circulation known as the coastal circulation.
Figure 3 shows that the WS increases as the air temperature increases. In the morning, at the time of the sea breeze lull, the WD shifts from land breeze to sea breeze, and in the evening, the flow reverses.
Figure 4a,b show the box plots of SO2 change with meteorological predictors: WDI, WS, dT, and RH for sea breeze (a) and land breeze (b), respectively.
Figure 4 shows the relationship between SO2 concentrations and meteorological predictors at the Split-1 station. Larger air temperature differences to the temperatures at the time of the morning sea breeze lulls for the sea breeze and land breeze increase SO2 concentrations (see Figure 1). For a land breeze, the SO2 concentrations increase with increasing WS, and for a sea breeze, they change only slightly with WS. For the sea breeze, a larger WDI leads to a slight decrease in SO2 concentrations, while for the land breeze, there is essentially no change. SO2 concentrations decrease with an increase in RH for both sea breeze and land breeze.
Using Equation (2), the VIF is found to be between 1 and 2, and there is no multicollinearity between selected predictors using the MLR model (see Table 3), as noted by Zuur et al. [59].
The following equations were obtained using the MLR model simulations for SO2 for sea breeze (C_SO2_SB) and land breeze (C_SO2_LB), respectively:
C_SO2_SB = 7.0230 − 1.6209 × WDI + 0.5250 × C6 + 0.2914 × C8 − 0.2988 × time,
C_SO2_LB = 6.6896 − 1.7623 × WS + 0.8014 × dT − 0.0793 × RH + 0.5145 × C6.
These include WDI for the sea breeze, and dT, WS, and RH for the land breeze at the Split-1 station.
The measurements at the Split-1 station and of the MLR model show a large agreement, as shown in Table 4. Small MAE and RMSE and high IA values of 0.8 and 0.9 were calculated for the land breeze and sea breeze, respectively. This shows the same success for the sea breeze (IA = 0.8) and even better agreement for the nighttime component of the coastal circulation than for PM10 or PM2.5 by Trošić Lesar and Filipčić [54,60]. Moreover, MAE and RMSE are lower for the land breeze (2.2 and 3.2) than for the sea breeze (3.7 and 5.9), respectively. Even in months when severe values were measured, with the highest values occurring in April, the mean monthly hourly values of modeled and MLR-model-simulated SO2 accurately predicted the mean daily change in pollutant concentrations (see Figure 5).
Figure 6 shows the hourly MLR-model-simulated and measured SO2 concentrations for selected sea breeze cases at the Split-1 station in April 2007–2009. Due to rush hours and heavy traffic, the highest concentrations were measured during the day, mostly between 0800 and 1800 UTC. The SO2 concentrations were low at night, as can also be seen in Figure 5 for the average values. This is consistent with the results of Rogalski et al. [61], who studied mean daily SO2 concentrations and concluded that high daytime SO2 concentrations well coincide with rush hours and low SO2 concentrations are shown at night. However, high concentrations of PM10 particles, originating from cement plants in the Split hinterland (see Figure 1), were also observed at the Split-1 station during the night and in the morning hours [39,60], with dT proving to be significant predictor for the sea breeze. Similar conclusions for the morning hours’ maximum concentrations for industrial pollution sources were drawn by Bouchlagem et al. [24] and Yerramilli et al. [25], when SO2 pollution is transmitted with the sea breeze. Bahtiyar et al. [26] showed that a different WD is important for determining the source of pollution in Kagithane valley, with a southerly flow for SO2, pointing to a source from the industrial area, and a northerly flow for PM10 particles.
Additionally, hourly measurements at the Kaštel-Sućurac station (see Figure 6), where the industrial area inland of Split is situated, show lower SO2 values than those at the Split-1 station. The measurements of SO2 concentration from the Kaštel-Sućurac station peak only during the day, mostly in the morning and mid-day hours, which can be related to rush hours, and do not match up well with those at the Split-1 station.

4. Discussion

According to the results of the stepwise MLR model at the Split-1 station, the only meteorological predictor for a sea breeze is WDI, while the predictors for a land breeze are WS, RH, and dT. Lalas et al. [44] also found that WS, WD, and RH are significant predictors for SO2 concentrations in the summer season in Athens using the MLR model. However, the air temperature was not a significant predictor. WD was also found to be significant for determining the source of pollution by Bahtiyar et al. [26] for Kagithane valley. In the coastal city of Ravenna, Tirabassi et al. [45] discovered a strong correlation between WS, SO2, and particle concentrations for the sea breeze. Liu et al. [47] have shown a negative correlation for the daily SO2 concentrations with WS and RH, which well agrees with this study, as well as temperature for stations in China with significant parameters. On the other hand, Rogalski et al. [61] showed a positive correlation between the daily SO2 concentrations with RH in March and from May to August, and a negative correlation in other months.
When comparing the MLR model with hourly SO2 concentrations, high IA values of 0.8 for the sea breeze and even 0.9 for the land breeze are found. Comparison with IA for the hourly PM10 and PM2.5 particles from Trošić Lesar and Filipčić [54,60] show higher agreement for the land breeze and about the same agreement for the sea breeze at the Split-1 station. For example, Hrust et al. [51] used a unique technique based on families of hourly concentration-based univariate regression models to achieve an IA from 0.69 to 0.9. For hourly PM10 concentrations, Mukerjee et al. [49] and Kukkonen et al. [50] obtained IA values between 0.45 and 0.65. The MLR model and neural networks were used by Paschalidou et al. [42] to simulate hourly PM10 particle concentrations. The neural network model performed better, with a correlation coefficient R2 ranging from 0.65 to 0.76.
The highest SO2 concentrations at the Split-1 station were measured from 0800 to 1800 UTC, and the highest peaks were observed during the daytime, while the SO2 concentrations were low at night, similar to, e.g., Rogalski et al. [61]. Comparison with the measurements at the inland Kaštel-Sućurac station, which is the possible source of cement-plant industrial pollution, also excluded this source of pollution with SO2. One possible source of SO2 pollution is also the northern port of Split due to ship emissions (see Figure 1). However, since dT is not a significant predictor, this source can be excluded as the main source of pollution with SO2; however, occasional pollution from this source cannot be excluded. This is in contrast to the conclusions of Bouchlagem et al. [24] and Yerramilli et al. [25], where the highest SO2 concentrations were measured in the morning at the time of the sea breeze lulls, and when industrial pollution sources were present. High concentrations of PM10 particles, originating from a cement plant in Kaštela Bay, inland from Split, were observed at the Split-1 station during the night and morning hours, with dT proving to be a significant predictor during the sea breeze [39,60]. Certain cement plants are a large source of SO2 (e.g., [40,41]), but this is not the case in Split.
The MLR model, using the predictor dT in this case, performed well in simulating SO2 concentrations only for a land breeze with IA = 0.9, and WDI was a significant predictor only during sea breeze with IA = 0.8 with observations, indicating a different source of pollution than for PM10 and PM2.5 particles, when dT was found to be significant during sea breeze [54,60].

5. Conclusions

The MLR model, which used the predictor dT in this case, performed well with observations at the Split-1 station in simulating SO2 concentrations for land breezes at IA = 0.9, and for sea breezes at IA = 0.8. The meteorological predictor of SO2 for sea breezes in Split is WDI, indicating different pollution sources than for PM10 and PM2.5 particles when dT is determined to be significant predictor [54,60]. When PM10 particles are transported with the sea breeze in Split from nearby cement plants, with the Kaštel-Sućurac cement plant as the main source, dT, a predictor characteristic of sea breeze, is found to be significant [39,60]. In contrast to Bouchlagem et al. [24] and Yerramilli et al. [25], who found that SO2 is transmitted with the sea breeze from inland plants, the SO2 concentration peaks were recorded between 0800 UTC and 1800 UTC, which is consistent with Rogalski et al. [61], who concluded that pollution from local traffic and rush hours are the main sources of pollution with SO2.
The SO2 hourly measurements at Kaštel-Sućurac station show smaller values and the peak daily concentrations do not match up well with those at the Split-1 station. Trošić Lesar and Filipčić [62] for selected sea breeze cases in Kaštel-Sućurac also found smaller SO2 concentrations than at the Split-1 station. This all points to another source of SO2 pollution in Split, namely, local traffic and rush hours. Since dT is not a significant predictor, there should not be significant emissions from the northern port of Split as a source of SO2 pollution, but this source of pollution cannot be excluded. At night, dT as well as WS prove to be significant predictors, suggesting that SO2, which is in very low concentrations during the night, is transported with the land breeze.
Since SO2 can affect respiratory and total non-accidental mortality [13,14,15] and even a low concentration can endanger health and life, especially of asthmatic children and adults [17,18], finding the possible sources of pollution is of particular importance. The MLR model showed that WDI is a significant predictor during a sea breeze for SO2 when traffic is found to be the main source of pollution in Split, and this method can be used particularly for thermally induced winds.

Author Contributions

Conceptualization, T.T.L. and A.F.; methodology, T.T.L.; software, T.T.L.; validation, T.T.L.; formal analysis, T.T.L.; investigation, T.T.L.; resources, T.T.L.; data curation, T.T.L.; writing—original draft preparation, T.T.L.; writing—review and editing, T.T.L. and A.F.; visualization, T.T.L.; supervision, A.F. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The data are available on reasonable request and with approval of the Croatian Environment Agency.

Acknowledgments

The authors would like to thank the Croatian Environment Agency for pollutant measurements used in this paper and Cemex Croatia d.d. for the information on the measurements provided. The authors thank two anonymous reviewers, as well as the Academic Editor for valuable suggestions.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Lyons, W.A.; Olsson, L.E. Detailed Mesometeorological Studies of Air Pollution Dispersion in the Chicago Lake Breeze. Mon. Weather. Rev. 1973, 101, 387–403. [Google Scholar] [CrossRef]
  2. Keen, C.S.; Lyons, W.A. Land/lake breeze circulations on the western shore of Lake Michigan. J. Appl. Meteorol. 1987, 17, 1843–1855. [Google Scholar] [CrossRef]
  3. Simpson, J. Sea Breeze and Local Wind; Cambridge University Press: Cambridge, UK, 1994. [Google Scholar]
  4. Clappier, A.; Martilli, A.; Grossi, P.; Thunis, P.; Pasi, F.; Krueger, B.C.; Calpini, B.; Graziani, G.; Van den Bergh, H. Effect of sea breeze on air pollution in the Greater Athens Area. Part I: Numerical simulations and field observations. J. Appl. Meteorol. 2000, 39, 546–562. [Google Scholar] [CrossRef]
  5. IPCC. Climate Change 2022. In Impacts, Adaptation and Vulnerability: Working Group II Contribution to the Sixth Assessment Report of the IPCC; 2022 Intergovernmental Panel on Climate Change: Geneva, Switzerland, 2022. [Google Scholar]
  6. Jacobson, M.Z. Atmospheric Pollution: History, Science, and Regulation; Cambridge University Press: Cambridge, UK, 2002. [Google Scholar]
  7. He, Y.; Uno, I.; Wang, Z.; Ohara, T.; Sugimoto, N.; Shimizu, A.; Richter, A.; Burrows, J.P. Variations of the increasing trend of tropospheric NO2 over central east China during the past decade. Atmos. Environ. 2007, 41, 4865–4876. [Google Scholar] [CrossRef]
  8. Shon, Z.-H.; Kim, K.-H.; Song, S.K. Long-term trend in NO2 and NOx levels and their emission ratio in relation to road traffic activities in East Asia. Atmos. Environ. 2011, 45, 3120–3131. [Google Scholar] [CrossRef]
  9. Li, J.; Li, K.; Zhang, H.; Zhang, X.; Ji, Y.; Chu, W.; Kong, Y.; Chu, Y.; Ren, Y.; Zhang, Y.; et al. Effects of OH radical and SO2 concentrations on photochemical reactions and mixed antrophogenic organic gases. Atmos. Chem. Phys. 2022, 22, 10489–10504. [Google Scholar] [CrossRef]
  10. Shah, A.S.V.; Lee, K.K.; McAllister, D.A.; Hunter, A.; Nair, H.; Whiteley, W.; Langrish, J.P.; Newby, D.E.; Mills, N.L. Short term exposure to air pollution and stroke: Systematic review and meta-analysis. BMJ 2015, 350, h1295. [Google Scholar] [CrossRef] [Green Version]
  11. Yang, W.-S.; Wang, X.; Deng, Q.; Fan, W.-Y.; Wang, W.-Y. An evidence-based appraisal of global association between air pollution and risk of stroke. Int. J. Cardiol. 2014, 175, 307–313. [Google Scholar] [CrossRef]
  12. Çapraz, Ö.; Bahtiyar, E.; Ali, D. Study on the association between air pollution and mortality in İstanbul, 2007–2012. Atmos. Poll. Res. 2016, 7, 147–154. [Google Scholar] [CrossRef]
  13. Wang, L.; Liu, C.; Meng, X.; Niu, Y.; Lin, Z.; Liu, Y.; Liu, J.; Qi, J.; You, J.; Tse, L.A.; et al. Associations between short-term exposure to ambient sulfur dioxide and increased cause-specific mortality in 272 Chinese cities. Environ. Int. 2018, 117, 33–39. [Google Scholar] [CrossRef]
  14. Yorifuji, T.; Kashima, S.; Suryadhi, M.A.H.; Abudureyimu, K. Acute exposure to sulfur dioxide and mortality: Historical data from Yokkaichi, Japan. Arch. Environ. Occup. Health 2019, 74, 271–278. [Google Scholar] [CrossRef] [Green Version]
  15. Shen, S.; Li, X.; Yuan, C.; Huang, Q.; Liu, D.; Ma, S.; Hui, J.; Liu, R.; Wu, T.; Chen, Q. Association of short-term exposure to sulfur dioxide and hospitalization for ischemic and hemorrhagic stroke in Guangzhou, China. BMC Pub. Health 2020, 20, 263. [Google Scholar] [CrossRef] [Green Version]
  16. Orellano, P.; Reynoso, J.; Quaranta, N. Short-term exposure to sulphur dioxide (SO2) and all-cause and respiratory mortality: A systematic review and meta-analysis. Environ. Int. 2021, 150, 106434. [Google Scholar] [CrossRef] [PubMed]
  17. Sousa, S.I.V.; Pires, J.C.M.; Martins, E.M.; Fortes, J.D.N.; Alvim-Ferraz, M.C.M.; Martins, F.G. Short-term effects of air pollution on respiratory morbidity at Rio de Janeiro—Part II: Health assessment. Environ. Int. 2012, 43, 1–5. [Google Scholar] [CrossRef] [PubMed]
  18. Krotkov, N.A.; McLinden, C.A.; Li, C.; Lamsal, L.N.; Celarier, E.A.; Marchenko, S.V.; Swartz, W.H.; Bucsela, E.J.; Joiner, J.; Duncan, B.N.; et al. Aura OMI observations of regional SO2 and NO2 pollution changes from 2005 to 2015. Atmos. Chem. Phys. Discuss. 2016, 16, 4605–4629. [Google Scholar] [CrossRef] [Green Version]
  19. Zhang, L.; Lee, C.S.; Zhang, R.; Chen, L. Spatial and temporal evaluation of long term trend (2005–2014) of OMI retrieved NO2 and SO2 concentrations in Henan Province, China. Atmos. Environ. 2017, 154, 151–166. [Google Scholar] [CrossRef]
  20. Bani, P.; Oppenheimer, C.; Tsanev, V.; Scaillet, B.; Primulyana, S.; Boyson, S.; Alfianti, H.; Marlia, M. Modest volcanic SO2 emissions from the Indonesian archipelago. Nat. Commun. 2022, 13, 3366. [Google Scholar] [CrossRef]
  21. Eyring, V.; Isaksen, I.S.; Berntsen, T.; Collins, W.J.; Corbett, J.J.; Endresen, O.; Grainger, R.G.; Moldanova, J.; Schlager, H.; Stevenson, D.S. Transport impacts on atmosphere and climate: Shipping. Atmos. Environ. 2010, 44, 4735–4771. [Google Scholar] [CrossRef]
  22. Krause, K.; Wittrock, F.; Richter, A.; Schmitt, S.; Pöhler, D.; Weigelt, A.; Burrows, J.P. Estimation of ship emission rates at a major shipping lane by long-path DOAS measurements. Atmos. Meas. Tech. 2021, 14, 5791–5807. [Google Scholar] [CrossRef]
  23. Knežević, V.; Pavin, Z.; Čulin, J. Estimating Shipping Emissions—A Case Study for CargoPort of Zadar, Croatia. TransNAV Int. J. Mar. Navig. Saf. Sea Transp. 2021, 15, 615–620. [Google Scholar]
  24. Bouchlaghem, K.; Mansour, F.B.; Elouragini, S. Impact of a seabreeze event on air pollution at the Eastern Tunisian Coast. Atmos. Res. 2007, 86, 162–172. [Google Scholar] [CrossRef]
  25. Yerramilli, A.; Challa, V.S.; Indracanti, J.; Dasari, H.; Baham, J.; Patrick, C.; Young, J.; Hughes, R.; White, L.D.; Hardy, M.G.; et al. Some observational and modeling studies of the atmospheric boundary layer at Mississippi gulf coast for air pollution dispersion assessment. Int. J. Environ. Res. Pub. Health 2008, 5, 484–497. [Google Scholar] [CrossRef]
  26. Bahtiyar, E.; Yasar Burak, O.; Ali, D.; Alper, U. Analysis of air pollutants in Kagithane valley and Istanbul Metropolitan Area. Air Qual. Atmos. Health 2022, 15, 1–15. [Google Scholar]
  27. Physick, W.L. Numerical experimentson the inland penetration of the sea breeze. Q. J. R. Met Soc. 1980, 106, 735–746. [Google Scholar] [CrossRef]
  28. Abbs, D.J. Sea-breeze interactions along a concave coastline in southern Australia:observations and numerical modelling study. Mon Weather Rev. 1986, 114, 831–848. [Google Scholar] [CrossRef]
  29. Garratt, J.R. The Internal Boundary-Layer—A review. Bound.-Lay. Meteorol. 1990, 50, 171–203. [Google Scholar] [CrossRef]
  30. Kouchi, A.; Ohba, R.; Shao, Y. Gas diffusion in a convection layer near a coastal region. J. Wind. Eng. Ind. Aerodyn. 1999, 81, 171–180. [Google Scholar] [CrossRef]
  31. Augustin, P.; Delbarre, H.; Lohou, F.; Campistron, B.; Puygrenier, V.; Cachier, H.; Lombardo, T. Investigationof local meteorological events and their relationship with ozone and aerosols during an ESCOMPTE photochemical episode. Ann. Geophys. 2006, 24, 2809–2822. [Google Scholar] [CrossRef] [Green Version]
  32. Shang, F.; Chen, D.; Guo, X.; Lang, J.; Zhou, Y.; Li, Y.; Fu, X. Impact of Sea Breeze Circulation on the Transport of Ship Emissions in Tangshan Port, China. Atmosphere 2019, 10, 723. [Google Scholar] [CrossRef] [Green Version]
  33. Orlić, M.; Penzar, B.; Penzar, I. Adriatic Sea and Land breezes: Clockwise versus anti-clockwise Rotation. J. Appl. Met. Climat. 1998, 5, 675–679. [Google Scholar]
  34. Trošić, T.; Šinik, N.; Trošić, Ž. Available potential energy of the daily coastal circulation at Zadar (Croatia). Meteorol. Atmos. Phys. 2006, 93, 211–220. [Google Scholar] [CrossRef]
  35. Telišman Prtenjak, M.; Grisogono, B. Sea/landbreeze climatological characteristics along the north-ern Croatian Adriatic coast. Theor. Appl. Clim. 2007, 90, 201–215. [Google Scholar] [CrossRef]
  36. Trošić, T.; Trošić, Ž. Numerical Simulation and Observational Analysis of the Bora of Pag’s Ribs. Bound.-Layer Meteorol. 2010, 134, 353–366. [Google Scholar] [CrossRef]
  37. Telišman Prtenjak, M.; Jeričević, A.; Kraljević, L.; Bulić, I.H.; Nitis, T.; Klaić, Z.B. Exploring atmospheric boundary layer characteristics in a severe SO2 episode in the north-eastern Adriatic. Atmos. Chem. Phys. 2009, 9, 4467–4483. [Google Scholar] [CrossRef] [Green Version]
  38. Bralić, M.; Buljac, M.; Periš, N.; Buzuk, M.; Dabić, P.; Brinić, S. Monthly and Seasonal Variations of NO2, SO2 and Black-smoke Located Within the Sport District in Urban Area, City of Split, Croatia. Croat. Chem. Acta 2012, 85, 139–145. [Google Scholar] [CrossRef]
  39. Trošić Lesar, T.; Filipčić, A. Lagrangian particle dispersion (HYSPLIT) model analysis of the sea breeze case with extreme mean daily PM10 concentration in Split, Croatia. Environ. Sci. Pollut. Res. 2022, 29, 73071–73084. [Google Scholar] [CrossRef]
  40. Zhang, T.; Wu, C.; Li, B.; Wang, J.; Ravat, R.; Chen, X.; Wei, J.; Yu, Q. Linking the SO2 emission of cement plants to the sulfur characteristics of their limestones: A study of 80 NSP cement lines in China. J. Clean. Prod. 2019, 220, 200–211. [Google Scholar] [CrossRef] [Green Version]
  41. Ibrahim, H.G.; Okasha, A.Y.; Elatrash, M.S.; Al-Meshragi, M.A. Emissions of SO2, NOx and PMs from Cement Plant in Vicinity of Khoms City in Northwestern Libya. J. Environ. Sci. Eng. 2012, A1, 620–628. [Google Scholar]
  42. Paschalidou, A.K.; Karaloysios, S.; Kleanthous, S.; Kassomenos, P.A. Forecasting hourly PM10 concentration in Cyprus through artificial neural networks andmultiple regression models: Implications to local environmental management. Environ. Sci. Poll. Res. Intern. 2009, 18, 316–327. [Google Scholar] [CrossRef]
  43. Vlachogianni, A.; Kassomenos, P.; Karppinen, A.; Karakitsios, S.; Kukkonen, J. Evaluation of a multipleregression model for the forecasting of the concentrationsof NOx and PM10 in Athens and Helsinki. Sci. Total Environ. 2011, 409, 1559–1571. [Google Scholar] [CrossRef]
  44. Lalas, D.; Veirs, V.R.; Karras, G.; Kallos, G. An analysis of the SO2 concentration in Athens, Greece. Atmos. Environ. 1982, 16, 531–544. [Google Scholar] [CrossRef]
  45. Tirabassi, T.; Fortezza, F.; Vandini, W. Wind circulation and air pollutant concentration in the coastal city of Ravenna. Energy Build. 1991, 16, 699–704. [Google Scholar] [CrossRef]
  46. Shams, S.R.; Jahani, A.; Kalantary, S.; Moeinaddini, M.; Khorasani, N. The evaluation on artificial neural networks (ANN) and multiple linear regressions (MLR) models for predicting SO2 concentration. Urban Clim. 2021, 37, 100837. [Google Scholar] [CrossRef]
  47. Liu, Y.; Zhou, Y.; Lu, J. Exploring the relationship between air pollution and meteorological conditions in China under environmental governance. Sci. Rep. 2020, 10, 14518. [Google Scholar] [CrossRef]
  48. Liu, B.; Zhao, Q.; Jin, Y.; Shen, J.; Li, C. Application of combined model of stepwise regression analysis and artificial neural network in data calibration of miniature air quality detector. Sci. Rep. 2021, 11, 3247. [Google Scholar] [CrossRef]
  49. Mukerjee, S.; Shadwick, D.S.; Smith, L.A.; Somerville, M.C.; Dean, K.E.; Browser, J.J. Techniques toassess cross-border air pollution and application to a US-Mexico border region. Sci. Total Environ. 2001, 276, 205–224. [Google Scholar] [CrossRef]
  50. Kukkonen, J.; Harkonen, J.; Karppinen, A.; Pohjola, M.; Pietarila, H.; Koskentalo, T. A semi-empirical model for urban PM10 concentrations, and its evaluation againstdata from an urban measurement network. Atmos. Environ. 2001, 35, 4433–4442. [Google Scholar] [CrossRef]
  51. Hrust, L.; Klaić, Z.B.; Križan, J.; Antonić, O.; Hercog, P. Neural network forecasting of air pollutants hourly concentrations using optimised temporal averagesof meteorological variables and pollutant concentrations. Atmos. Environ. 2009, 43, 5588–5596. [Google Scholar] [CrossRef]
  52. Abdullah, S.; Napi, N.N.L.M.; Ahmed, A.N.; Mansor, W.N.W.; Mansor, A.A.; Ismail, M.; Abdullah, A.M.; Ramly, Z.T.A. Development of Multiple Linear Regression for Particulate Matter (PM10) Forecasting during Episodic Transboundary Haze Event in Malaysia. Atmosphere 2020, 11, 289. [Google Scholar] [CrossRef] [Green Version]
  53. Zhou, Y.; Guan, H.; Huang, C.; Fan, L.; Gharib, S.; Batelaan, O.; Simmons, C. Sea breeze cooling capacity and its influencing factors in a coastal city. Build. Environ. 2019, 166, 106408. [Google Scholar] [CrossRef]
  54. Trošić Lesar, T.; Filipčić, A. The hourly simulation of PM2.5 particle concentrations using the multiple linear regression (MLR) model for sea breeze in Split, Croatia. Water Air Soil Pollut. 2021, 232, 261. [Google Scholar] [CrossRef]
  55. Google Maps. 2022. Available online: https://www.google.com/maps/d/viewer?mid=1Fb1obUiEtNB8bJ-bu-WR7HA-Io&hl=en_US&ll=43.56611726780062%2C16.386492960821638&z=12 (accessed on 20 November 2022).
  56. Trošić, T. Climatological characteristics of the lower branch of coastal circulation along the Eastern Croatian coast. Croat. Meteorol. J. 2002, 37, 27–36. [Google Scholar]
  57. Telišman Prtenjak, M.; Grisogono, B. Idealised numerical simulations of diurnal sea breeze characteristics over a step change in roughness. Meteorol. Z. 2002, 11, 345–360. [Google Scholar] [CrossRef]
  58. Ordieres, J.B.; Vergara, E.; Capuz, R.; Salazar, R. Neural network prediction model for fine particulate matter (PM2.5) on the US–Mexico border in El Paso (Texas) and Ciudad Juarez Chihuahua. Environ. Modell. Soft. 2005, 20, 547–559. [Google Scholar] [CrossRef]
  59. Zuur, A.F.; Ieno, E.N.; Walker, N.J.; Saveliev, A.A.; Smith, G.M. Mixed Effects Models and Extensions in Ecology with R. Statistics for Biology and Health; Springer: New York, NY, USA, 2009; p. 574. [Google Scholar]
  60. Trošić Lesar, T.; Filipčić, A. Multiple Linear Regression (MLR) model simulation of hourly PM10 concentrations during sea breeze events in the Split area. Naše More 2017, 64, 77–85. [Google Scholar] [CrossRef]
  61. Rogalski, L.; Smoczynski, L.; Krzebietke, S.J.; Lenart, L.; Mackiewicz-Walec, E. Changes in sulphur dioxide concentrations in the atmospheric air assessed during short-term measurements in the vicinity of Olsztyn, Poland. J. Elem. 2014, 19, 735–748. [Google Scholar]
  62. Trošić Lesar, T.; Filipčić, A. The influence of sea breeze on the air pollution in the Split area. Geoadria 2018, 23, 135–151. [Google Scholar] [CrossRef]
Atmosphere 14 00420 g001 550
Figure 1. The position of the AMS 3 Split-1 measurement station (latitude: 43°30′39″ N, longitude: 16°26′56″ E, elevation: 46 m) in Split, Croatia, and the AMS 1 measurement station Kaštel-Sućurac (latitude: 43°32′53″ N, longitude: 16°26′5″ E) (the location of the station is marked with a square) [55]. The positions of cement plants in Kaštel-Sućurac (“K-S”), Solin (“S”), Klis (ˇK”), as well as the asbestos cement plant in Vranjic (“V”) (the location of the plant is marked with a triangle) and northern port (“NP”) of Split (the location is marked with a square).
Figure 1. The position of the AMS 3 Split-1 measurement station (latitude: 43°30′39″ N, longitude: 16°26′56″ E, elevation: 46 m) in Split, Croatia, and the AMS 1 measurement station Kaštel-Sućurac (latitude: 43°32′53″ N, longitude: 16°26′5″ E) (the location of the station is marked with a square) [55]. The positions of cement plants in Kaštel-Sućurac (“K-S”), Solin (“S”), Klis (ˇK”), as well as the asbestos cement plant in Vranjic (“V”) (the location of the plant is marked with a triangle) and northern port (“NP”) of Split (the location is marked with a square).
Atmosphere 14 00420 g001
Atmosphere 14 00420 g002 550
Figure 2. The mean hourly change in temperature mt (°C) and relative humidity mRH (%) (left), and mean wind speed mv (m s−1) and vek mdir (deg) showing the mean vector wind direction (right), for selected cases with undisturbed coastal circulation at the AMS 3 Split-1 station from April to June 2007–2009.
Figure 2. The mean hourly change in temperature mt (°C) and relative humidity mRH (%) (left), and mean wind speed mv (m s−1) and vek mdir (deg) showing the mean vector wind direction (right), for selected cases with undisturbed coastal circulation at the AMS 3 Split-1 station from April to June 2007–2009.
Atmosphere 14 00420 g002
Atmosphere 14 00420 g003 550
Figure 3. The same as in Figure 2, but from July to September 2007–2009.
Figure 3. The same as in Figure 2, but from July to September 2007–2009.
Atmosphere 14 00420 g003
Atmosphere 14 00420 g004a 550Atmosphere 14 00420 g004b 550
Figure 4. Box plots: SO2 with meteorological predictors at the AMS 3 Split-1 station: the hourly air temperature difference to the temperature of the sea breeze lull (dT (°C)), wind direction index (WDI), wind speed (m s−1), and relative humidity (RH (%)) for sea breeze (a) and land breeze (b). The x axis is divided into groups. The red line in the rectangle is the median, and the upper part of the rectangle corresponds to the 75% percentile and the lower to the 25% percentile. The lines outside the rectangle show the largest and smallest interval M ± 1.5 IR, where M is the median and IR the interquartile range. The crosses outside the rectangle are data out of the M ± 1.5 IR range.
Figure 4. Box plots: SO2 with meteorological predictors at the AMS 3 Split-1 station: the hourly air temperature difference to the temperature of the sea breeze lull (dT (°C)), wind direction index (WDI), wind speed (m s−1), and relative humidity (RH (%)) for sea breeze (a) and land breeze (b). The x axis is divided into groups. The red line in the rectangle is the median, and the upper part of the rectangle corresponds to the 75% percentile and the lower to the 25% percentile. The lines outside the rectangle show the largest and smallest interval M ± 1.5 IR, where M is the median and IR the interquartile range. The crosses outside the rectangle are data out of the M ± 1.5 IR range.
Atmosphere 14 00420 g004aAtmosphere 14 00420 g004b
Atmosphere 14 00420 g005 550
Figure 5. Mean monthly modeled and stepwise MLR-model-simulated SO2 hourly concentrations at the AMS 3 Split-1 station from April to September 2007–2009.
Figure 5. Mean monthly modeled and stepwise MLR-model-simulated SO2 hourly concentrations at the AMS 3 Split-1 station from April to September 2007–2009.
Atmosphere 14 00420 g005
Atmosphere 14 00420 g006 550
Figure 6. Hourly multiple linear regression (MLR)-model-simulated and measured SO2 concentrations for selected sea breeze cases in April 2007–2009 at the AMS 3 Split-1 station in comparison with the measurements at the AMS 1 Kaštel-Sućurac station.
Figure 6. Hourly multiple linear regression (MLR)-model-simulated and measured SO2 concentrations for selected sea breeze cases in April 2007–2009 at the AMS 3 Split-1 station in comparison with the measurements at the AMS 1 Kaštel-Sućurac station.
Atmosphere 14 00420 g006
Table
Table 1. Number of days with undisturbed coastal circulation for the period 2007–2009 at the Split-1 station.
Table 1. Number of days with undisturbed coastal circulation for the period 2007–2009 at the Split-1 station.
AprilMayJuneJulyAugustSeptemberTotal
Num. of days677134340
Table
Table 2. The measured 24 h average and maximum (max) values of SO2 concentration for the selected sea breeze days in Table 1 at AMS 3 Split-1 and AMS 1 Kaštel-Sućurac in comparison with the limit values for human health from the EU directive (2008/EC/50).
Table 2. The measured 24 h average and maximum (max) values of SO2 concentration for the selected sea breeze days in Table 1 at AMS 3 Split-1 and AMS 1 Kaštel-Sućurac in comparison with the limit values for human health from the EU directive (2008/EC/50).
Meas. Conc. at AMS 3 Split-1
(μg m−3)		Meas. Conc. at AMS 1 Kaštel-Sućurac
(μg m−3)		Limit Value for Human Health
(μg m−3)
24 h mean13.5911.84125
24 h max89.4273.79350
Table
Table 3. The selected meteorological predictors in the stepwise MLR model at the AMS 3 Split-1 station where WDI (wind direction index) is calculated from hourly wind direction, WS is the hourly wind speed (m s−1), dT (°C) is the hourly temperature deviation from the temperature at the time of the sea breeze lull, RH is the hourly relative humidity (%), C6, C7, and C8 (μg m−3) are SO2 concentrations at 6, 7, and 8 UTC, respectively, and the time is hours.
Table 3. The selected meteorological predictors in the stepwise MLR model at the AMS 3 Split-1 station where WDI (wind direction index) is calculated from hourly wind direction, WS is the hourly wind speed (m s−1), dT (°C) is the hourly temperature deviation from the temperature at the time of the sea breeze lull, RH is the hourly relative humidity (%), C6, C7, and C8 (μg m−3) are SO2 concentrations at 6, 7, and 8 UTC, respectively, and the time is hours.
WDIWSdTRHC6C7C8Time
Sea breezex x xx
Land breeze xxxx
Table
Table 4. Results of the stepwise multiple linear regression model are compared with the hourly measurements of SO2 concentration(μg m−3) and with the mean absolute error (MAE), root mean square error (RMSE), index of agreement (IA), and correlation coefficient (R) for the sea breeze and land breeze separately at the AMS 3 Split-1 station.
Table 4. Results of the stepwise multiple linear regression model are compared with the hourly measurements of SO2 concentration(μg m−3) and with the mean absolute error (MAE), root mean square error (RMSE), index of agreement (IA), and correlation coefficient (R) for the sea breeze and land breeze separately at the AMS 3 Split-1 station.
MAERMSEIAR
Sea breeze3.75.90.80.6
Land breeze2.23.20.90.7
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

Trošić Lesar, T.; Filipčić, A. Prediction of the SO2 Hourly Concentration for Sea Breeze and Land Breeze in an Urban Area of Split Using Multiple Linear Regression. Atmosphere 2023, 14, 420. https://doi.org/10.3390/atmos14030420

AMA Style

Trošić Lesar T, Filipčić A. Prediction of the SO2 Hourly Concentration for Sea Breeze and Land Breeze in an Urban Area of Split Using Multiple Linear Regression. Atmosphere. 2023; 14(3):420. https://doi.org/10.3390/atmos14030420

Chicago/Turabian Style

Trošić Lesar, Tanja, and Anita Filipčić. 2023. "Prediction of the SO2 Hourly Concentration for Sea Breeze and Land Breeze in an Urban Area of Split Using Multiple Linear Regression" Atmosphere 14, no. 3: 420. https://doi.org/10.3390/atmos14030420

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop