This study investigates the transboundary dust transport from the Sahara and Arabian Deserts to West Asia during a major event in spring 2015. The analysis integrates satellite observations, ground-based PM₁₀ measurements, and simulations from the WRF-Chem model configured at a 10 × 10 km grid resolution. The model was coupled with the Global Ozone Chemistry Aerosol Radiation and Transport (GOCART) dust emission scheme and employed the Four-Dimensional Data Assimilation (FDDA) technique. Model outputs were evaluated against in-situ PM₁₀ observations and satellite-derived aerosol data. Results indicate two distinct dust intrusion episodes: the first, originating from the Arabian Desert, affected northern Iraq and southeastern Türkiye on May 23–24; the second, associated with the Sahara Desert dust, arrived on May 27, merging with the earlier plume by May 28. This convergence led to elevated PM₁₀ concentrations, exceeding a daily mean of 500 µg/m³ across parts of West Asia. Aerosol observations confirmed substantial contributions from both desert sources, with Aerosol Optical Depth (AOD) values frequently reaching or exceeding 1.0 in several regions. These events were driven by strong Sahara-Arabian heating, which enhanced pressure gradients and generated Shamal winds and easterly surges, promoting dust uplift and long-range transport. Significant spatial variability was observed, from severe air pollution in some areas to minimal impacts elsewhere. During dust events, modeled–observed PM₁₀ correlations ranged 0.14–0.81 with systematic underestimation. These results highlight the need for improved dust emission parameterizations, early warning systems, and high-resolution regional models to better support air quality management and public health planning. This visual summary serves as a key entry point into the research, providing a concise overview of the study’s core findings and methodologies. The study investigates a major transboundary dust event in Spring 2015, during which dust from the Sahara and Arabian Deserts significantly affected air quality across West Asia. Using the WRF-Chem model, satellite data (MODIS, OMI, CALIOP), and ground-based PM₁₀ observations, two major dust intrusions were identified. Model results show Arabian dust reached northern Iraq and southeastern Türkiye by May 23, followed by Saharan dust arriving around May 27. These dust plumes merged, causing PM₁₀ levels to exceed a daily mean of 500 µg/m³ in some parts of the study area. The severity of the event varied across regions, influenced by topography, meteorological conditions, and urban exposure. Satellite analyses confirmed the long-range transport of dust from the Sahara and Arabian Deserts into West Asia. Elevated aerosol concentrations were recorded in several North African and West Asian countries, including Egypt, Iraq, Syria, Saudi Arabia, Türkiye, Iran, Armenia, and Azerbaijan. Daily observations revealed significant aerosol contributions from both desert regions, with Aerosol Optical Depth (AOD) values reaching 1.0 or higher in many areas. This widespread aerosol transport highlights the substantial impact of dust on air quality and atmospheric conditions. The findings underscore the urgent need for enhanced regional dust forecasting, early warning systems, and coordinated mitigation strategies such as sustainable land use practices and public health advisories to reduce the environmental and health risks of future dust events. The WRF-Chem model successfully simulated transboundary dust transport during a major 2015 event. Dust originating from the Sahara and Arabian Deserts elevated regional PM₁₀ concentrations to over daily mean 500 µg/m³. Model validation showed strong agreement with observations, achieving correlation coefficients (r) up to 0.81 across 16 monitoring stations. Results demonstrated significant spatial variability in dust impacts, underscoring the importance of location-specific mitigation strategies. The study emphasizes the critical need for cross-border forecasting systems and data-sharing frameworks across West Asia.

Air quality plays a vital role in determining the livelihood and well-being of a society. Among outdoor pollutants, Particulate Matter (PM) poses significant health risks. Based on the “State of Global Air” report by the Health Effects Institute, Türkiye experienced over 50 age-standardized deaths per 100,000 people annually in 2019 due to exposure to high levels of particulate matter, a figure that is double the European Union's average of 24.4 age-standardized deaths per 100,000 people. Transported dust from the Middle East and North Africa (MENA) region is the primary natural source of PM in this region. This study aims to investigate the aerosol pollution in Türkiye from 2000 to 2022 by utilizing satellite-based retrieval products such as Aerosol Optical Depth (AOD) and ground-level PM10 measurements from over 100 air quality stations. Machine Learning models and statistical methods such as Random Forest and Multiple Linear Regression (MLR) model were employed to model PM10 concentration in the Eastern and Southeastern Anatolia regions of Türkiye over a selected period from 2014 to 2016. This timeframe coincided with significant dust events affecting the Eastern and Southeastern Anatolia regions of Türkiye. Among all applied methods, Random Forest performed the best in terms of fitting the regression line, with R, R2 and RMSE values of 0.97, 0.93 and 17.90 μg/m3, respectively. Research examining also the link between drought and dust has revealed a direct correlation between the intensity of droughts and the levels of aerosols, suggesting that drought and desertification intensify air pollution by enhancing dust transportation.

We study regression-based data fusion under uncertainty, where multiple noisy and biased measurement sources are available but ground-truth labels are absent during training. This setting arises in sensor networks, simulation ensembles, and scientific monitoring systems where supervision is costly or infeasible. We propose the Neural Conjugate Aggregation Model (NCAM), a hierarchical Bayesian framework that combines neural networks with conjugate Gaussian inference for unsupervised multi-source fusion. NCAM learns source-specific bias and reliability conditioned on contextual covariates, yielding an analytically tractable posterior over a latent target variable with decomposed epistemic and aleatoric uncertainty. Structural non-identifiability is resolved through sensor anchoring and variance regularization, enabling stable and interpretable posterior aggregation. To complement Bayesian uncertainty with finite-sample guarantees, we integrate locally adaptive Monte Carlo conformal prediction, producing heteroscedastic prediction intervals with coverage guarantees under exchangeability assumptions. Experiments on synthetic and real-world air-quality datasets demonstrate improved predictive accuracy and well-calibrated uncertainty compared to unsupervised baselines, including mean aggregation, probabilistic PCA, and Kalman filtering.

Dust, composed of small solid particles, contributes to air pollution, especially in arid and semi-arid regions. Despite identifying substantial dust sources in many countries, effective measures to combat dust emissions in these regions remain limited. This study utilized ground-level PM10 data, alongside satellite data to investigate the impacts of transported dust from the Sahara Desert and Arabian Peninsula on the air quality of the Middle East and North Africa (MENA) countries in May 2015. WRF-Chem version 4.5 was utilized for dust modeling to assess the impacts of dust on particulate matter concentration in the study area. Based on the results of this study, dust has contributed significantly to the region's aerosol pollution with high AOD values close to or more than 1. The findings revealed elevated particulate matter concentrations originating from the MENA region, particularly in the Middle East countries with hourly PM10 values exceeding 1000 µg/m3 over the study period. Model performance of WRF-Chem through cross-validation of observed PM10 data against predicted values in six ground-observed air quality stations, yield a correlation coefficient values ranging between 0.35 and 0.77. The results of this study highlight the substantial impact of transported dust from the Sahara Desert and the Arabian Peninsula on air quality in the MENA region. They underscore the necessity of taking proactive measures to mitigate the environmental harm caused by dust.

This study evaluates the impact of climate change on tropospheric ozone (O3) concentrations in the Eastern Mediterranean using the Weather Research and Forecasting (WRF) and the Community Multiscale Air Quality (CMAQ) models. Simulations were conducted for a historical period (2012) and a future projection (2053) under SSP2-4.5 and SSP5-8.5 scenarios. Anthropogenic emissions were sourced from the EMEP/EEA inventory, while biogenic emissions were calculated using the MEGAN model. Model performance evaluations yielded R2 (RMSE) values of 0.71-0.85 (3.33-4.9) for WRF and 0.58 (8.35) for CMAQ, indicating reasonable predictive accuracy. Under SSP5-8.5 (SSP2-4.5), the WRF model projects an average summertime temperature increase of 1.6 degrees C (1.2 degrees C) and a significant decline in precipitation across the Eastern Mediterranean. Air quality simulations show a regional increase in summertime O3 concentrations by 3.5 ppb (3.0 ppb) under SSP5-8.5 (SSP2-4.5), with the most pronounced increases occurring in the southeast. Conversely, a significant reduction in O3 concentrations is observed over the Marmara Sea and parts of Istanbul in both scenarios. This reduction is attributed to climate- induced processes, including accelerated O3 photolysis in moist conditions and enhanced O3 consumption by NOx in the NOx-saturated regime of the Marmara Sea region. Additionally, analyses reveal a significant increase in O3 levels in Istanbul under SSP5-8.5, while Bursa shows notable increases under both scenarios. These findings underscore the need for targeted emission control measures to mitigate future O3 pollution in the region.

The spatial allocation of emissions in air quality models introduces uncertainties that significantly impact pollution exposure assessments. This study quantified the effects of emission allocation uncertainty on atmospheric concentrations and exposure levels using the CMAQ modeling system. The research focused on the Af & scedil;in-Elbistan Power Plant (AP), with substantial emissions of SO2 (similar to 300,000 t/y) and PM2.5 (similar to 6000 t/y), evaluating the variability in concentrations from emission allocation in gridded inventories. 13 model simulations were conducted, including a base case (c0) where emissions were spatially allocated based on intersection ratios and 12 scenario cases (c1-c12) where emissions were assigned to different grids for 2018. Results showed significant variability in pollution levels and population exposures across scenario cases. In the Maximum Impact Zone (MIZ), annual mean PM2.5 concentrations ranged from 5.0 to 41.3 mu g/m(3), with differences up to 24.9 mu g/m(3) from the base case. SO2 exhibited even greater variability, with maximum differences reaching 338.2 mu g/m(3). The 95 % probability range of uncertainty for PM2.5 was estimated at -45 % to +96 %, while for SO2, it reached -84 % to +240 %. Grids A-F represent six selected regions with high population density, used to evaluate differences in concentration and exposure across scenarios. In Grid A-F, meteorology influenced these patterns, with low wind speeds causing pollutant build-up in Grid A, while pollutant transport affected Grids D-F in summer. Annual population exposure in Grid C ranged from 1.0 to 2.1 kg/y for PM2.5 and from 3.9 to 16.7 kg/y for SO2. This paper highlights the importance of not only absolute emission inventories but also spatial emission allocation in air quality models to enhance regulatory effectiveness and protect public health.

Several groups in the United States, including communities of color and low-income communities, are frequently disproportionately exposed to ambient (i.e., outdoor) air pollution, reflecting unjust placement of emission sources, systemic bias, and historic race-based land use planning. Eliminating these inequities is critical for advancing environmental justice. This review synthesizes methodological innovations for characterizing and mitigating ambient air pollution inequities, focusing on the past 10 years, mostly in the United States. Advances in exposure assessment (e.g., empirical models, satellite remote sensing, mobile monitoring, sensor networks) provide new tools for characterizing disparities. Advances in techniques for attributing pollution to specific sources (e.g., reduced-complexity models) reveal how emission-reduction approaches may or may not eliminate disparities. Spatially targeted emission reductions are critical for eliminating relative disparities; conventional approaches (e.g., sectoral emission reductions, national concentration standards) are unlikely to eliminate those disparities. This article provides insights for effective interventions to promote equity in ambient air pollution exposure.

Fine particulate matter (PM2.5) posing significant risks due to its ability to penetrate deep into the respiratory system. This study introduces the Regional PM2.5 Predictor (RPP), a machine learning-based framework designed to estimate PM2.5 concentrations across Turkiye, especially in regions with limited PM2.5 monitoring infrastructure. Leveraging satellite-derived Aerosol Optical Thickness (AOT) data, meteorological variables from ERA5, and ground-based air quality measurements, the model integrates diverse datasets spanning 2018 to 2023, the RPP employs XGBoost algorithms to address spatial monitoring gaps. The model demonstrates strong predictive performance across multiple evaluation scenarios: the seasonal analysis yielded RMSE values of 4.39-10.01 mu g/ m3 and R2 values of 0.66-0.84; temporal evaluations achieved an average RMSE of 8.28 mu g/m3 and R2 of 0.76; spatial (station-blinded) cross-validation maintained reliable predictions with average RMSE of 9.21 mu g/m3 and R2 of 0.71; while random sampling achieved RMSE of 6.82 mu g/m3 and R2 of 0.85 with an 80-20 % split. The framework successfully captured Turkiye's air quality trend, with PM2.5 levels decreasing from 25.52 mu g/m3 (2018) to 18.88 mu g/m3 (2023), while identifying performance variations across diverse topographical regions. The model demonstrated remarkable stability during the COVID-19 pandemic period, achieving its best performance in 2020 (RMSE: 7.54 mu g/m3, R2: 0.80). This approach demonstrates how machine learning can complement traditional monitoring networks, providing cost-effective air quality assessments for public health interventions and environmental policy evaluation.

Levels of fine particulate matter (PM2.5) air pollution in the United States have declined substantially in recent decades, yielding substantial benefits to public health. This study evaluates emission reductions across five key economic sectors-electricity, industrial, transportation, agriculture, and residential-and their impact on air quality and health. We employ a recently developed sector-specific inventory that provides emissions and their spatial disaggregation across time in a self-consistent framework. Using a national source-receptor matrix, we estimate annual PM2.5-attributable mortality and its variability spatiotemporally and by sector. We find that annual PM2.5-attributable mortality decreased 51% between 2002 (197,000 deaths) and 2019 (96,000 deaths). The largest reductions were from electricity and transportation, especially secondary PM2.5 from NOx, SOx, and VOC emissions. Emissions reductions from industrial and residential sectors were more modest. In contrast, agricultural emissions, especially NH3, increased over time; the importance of agriculture among the five sectors increased from second-smallest (2002) to the largest (2019). While the reductions in PM2.5-attributable mortality have been large (approximately a factor of 2), future progress may need to focus greater attention on agricultural emissions, in addition to traditionally dominant sources such as transportation and industry.

The global increase in population has led to higher emissions from livestock and synthetic fertilizers. This study investigates the impact of agricultural ammonia emissions on NH3 concentrations and provides insights into PM2.5 levels and their components in agriculturally intensified areas. We developed a bottom-up emission inventory focused on fertilizer application over croplands and livestock, instead of relying on the EMEP database. This approach utilized an improved spatial and temporal distribution of these emissions. We compared annual total NH3 emissions from livestock and fertilizer, estimated at 598.5 kt and 187.2 kt in the EMEP inventory (Base case), and 245.2 kt and 536 kt in the bottom-up inventory (Scenario case). Using the CMAQ modelling framework, we estimated atmospheric concentrations for both cases and evaluated the model results by comparing them with IASI-NH3 satellite retrievals. This comparison revealed significant differences in column concentrations between the Base and Scenario cases, with the Scenario case showing substantial improvement. Over a period of seven months, which contributed 80 % of the annual agricultural emissions for the Scenario case, the domain averages of NH3 were 3.02 × 1015, 4.15 × 1015, and 4.17 × 1015 molecules/cm2 for the Base and Scenario cases and IASI-NH3, respectively. The Scenario case closely matched IASI measurements, indicating a more accurate representation of NH3 emissions and concentrations. This enhanced reliability underscores the effectiveness of the bottom-up inventory approach. Additionally, using the CMAQ model, we found that in the IASI hotspots, the averages were 1.67 μg/m3 for sulfate, 0.57 μg/m3 for nitrate, and 0.62 μg/m3 for ammonium, with a total PM2.5 mean of 10.45 μg/m3.

Seasonal forecasting is a crucial task when it comes to detecting the extreme heat and colds that occur due to climate change. Confidence in the predictions should be reliable since a small increase in the temperatures in a year has a big impact on the world. Calibration of the neural networks provides a way to ensure our confidence in the predictions. However, calibrating regression models is an under-researched topic, especially in forecasters. We calibrate a UNet++ based architecture, which was shown to outperform physics-based models in temperature anomalies. We show that with a slight trade-off between prediction error and calibration error, it is possible to get more reliable and sharper forecasts. We believe that calibration should be an important part of safety-critical machine learning applications such as weather forecasters.

High concentrations of Particulate Matter (PM) have become a major problem in Turkey because of its economic development over the past decades, as well as its geographical proximity to natural dust source areas. In this study, PM10 data (for the period of 2010-2020) of 36 ground-based stations in 12 metropolitan cities over different regions in Turkey was used and Kolmogorov-Zurbenko (KZ) filter was implemented to decompose the data into its temporal components. A stepwise Multiple Linear Regression (MLR) model was developed to establish relationships between PM10 concentrations and a set of meteorological variables for each city to quantify the long-term meteorological as well as emission impacts on changes and trends in air quality. In this study analysis has revealed that out of twelve major cities in Turkey, only three of them has PM10 levels below or around 45 mu gmi 3,which is 24 h Air Quality Guideline (AQG) level identified by WHO (WHO, 2021). Overall, over the selected period, long-term change in PM10 concentration is negative for 10 out of 12 cities, ranging between i 39.6 mu gmi 3 and i 1.2 mu gmi 3. Only Adana and Kayseri have a positive long-term change ranging between 1.0 mu gmi 3 and 3.1 mu gmi 3. Long-term change in meteorology related component (chi LTmet) is relatively constant, hence, long-term change in PM10 concentration (chi LT) is heavily influenced by emission related component (chi LT emis). As emission impact reduces over time, there is a decrease in PM10 levels in most of the cities. This finding is further supported by reductions in national emissions data.

In this study, NH3 and NMVOC emissions from agricultural activities were updated and compared with EMEP inventory for a region known for agricultural activities in Turkey. An advanced air quality model, CMAQ, was used to determine the spatio-temporal distribution of NH3 and NMVOC, which have seasonal patterns. Model was performed for base (EMEP 2018 emissions) and scenario (updated 2019 emissions) cases and 2018 summer months with high-resolution grid size. Moreover, secondary particle formation over the domain was identified. This study is significant due to the agriculture emissions to be updated and re-gridded using high-resolution CORINE land use data and examined of pollutants which are climate dependent. EMEP agriculture emissions were 955.92 kt/annual and 486.60 kt/annual (Turkey totals), and 284.35 kt/annual and 227.54 kt/annual (Turkey emissions in the domain) for NH3 and NMVOCs, respectively; updated agriculture emissions were found as 579.58 kt/annual and 289.97 kt/annual (Turkey totals), and 195.01 kt/annual and 98.66 kt/annual (Turkey emissions in the domain) for NH3 and NMVOC. Not only the emissions were different in overall quantities, but also there were spatial differences in concentrations between cases. For NH3, the largest source of which is agriculture, monthly averaged concentration differences were found up to 60% in some areas of the domain due to the distribution of total emissions over agricultural areas unlike the EMEP spatial distribution methodology. VOC emissions were overestimated in EMEP inventory; thus, a difference had also become in concentrations between two cases. Highest concentrations were found in July for PM2.5, PM2.5(OC), PM2.5(SO4), and PM2.5(NH4).

The energy demand is increasing day by day, although the installation and operation of coal-fired power plants slowly fade out throughout the world, it is still increasing in Turkey. Air pollutant emissions from public power sector, mainly composed of power plants contribute significantly to Turkey national totals (70.4% SO2, 38.9% NOx, 9.8% PM2.5). Along with the emission estimation, the temporal and spatial distribution of these emissions are also crucial for accurate simulations with low uncertainty in air quality modeling. Sector-specific profiles are widely used as temporal profiles to input hourly emissions for modeling. In this study, a new temporal profile was established using hourly electricity generation of a large-capacity coal-fired power plant: Afsin Elbistan Power Plant (AEPP). The WRF/CMAQ model was run for two cases; one with sector-specific temporal profile (base case) and another with facility-specific temporal profile (scenario case). EMEP 2018 emissions were used and model was run for 2018 summer months (JJA) with high-resolution 4 × 4 km2 grid size. Temporal profile comparisons for the summer period indicated maximum differences between 00.00 and 07.00 h with higher ratios for scenario case. Day of the week comparison showed consistently high differences on the weekends, especially on Sundays (average of +40.4% for JJA) with base case always allocating lower emissions on Saturdays and Sundays than scenario case. The largest daily differences were observed between 14 and 22 July where the highest was on 22 July (+91%) which were again weekend days. Hourly concentrations over AEPP showed the highest differences in July with scenario case being usually higher. The peak concentrations and concentration differences usually occur around midnight (00.00). The days with the highest hourly differences were on 18 July (1614 ppbV for SO2, 26 ppbV for NO2, and 9 ppbV for VOC) for Layer 7, which is around midnight.

Understanding seasonal climatic conditions is critical for better management of resources such as water, energy, and agriculture. Recently, there has been a great interest in utilizing the power of Artificial Intelligence (AI) methods in climate studies. This paper presents cutting-edge deep-learning models (UNet++, ResNet, PSPNet, and DeepLabv3) trained by state-of-the-art global CMIP6 models to forecast global temperatures a month ahead using the ERA5 reanalysis dataset. ERA5 dataset was also used for fine-tuning as well performance analysis in the validation dataset. Ten different setups (with CMIP6 and CMIP6 + ERA5 fine-tuning) including six meteorological parameters (i.e., 2m temperature, 10meastward component of wind, 10mnorthward component of wind, geopotential height at 500 hPa, mean sea-level pressure, and precipitation flux) and elevation were used with both four different algorithms. For each model 14 different sequential and nonsequential temporal settings were used. The mean absolute error (MAE) analysis revealed that UNet++ with CMIP6 with 2mtemperature + elevation and ERA5 fine-tuning model with "Year 3 Month 2" temporal case provided the best outcome with an MAE of 0.7. Regression analysis over the validation dataset between the ERA5 data values and the corresponding AI model predictions revealed slope and R2 values close to 1 suggesting a very good agreement. The AI model predicts significantly better than the mean CMIP6 ensemble between 2016 and 2021. Both models predict the summer months more accurately than the winter months.

The ability to accurately predict seasonal and sub-seasonal weather patterns is of great significance in climate modeling, as extreme weather events and climate change become more prevalent. Furthermore, current climate models require extensive computational resources to generate monthly forecasts. Recent advancements in machine learning, accessibility of the vast amount of data, and efficiency of the ML models motivate the use of machine learning based approaches for seasonal forecasting.In this study, we present a sub-seasonal weather forecast model using a UNet based deep learning architecture that enables the learning of long-range spatiotemporal information. Our model is trained with surface air temperature data obtained from 10 different state-of-the-art CMIP6 model output, and finetuned using ERA5 atmospheric reanalysis dataset to increase the generalization in forecasting with real weather modeling data. We focus on the monthly temperature forecast of continental Europe and compare our results with physics-based climate models and ML methods. We evaluate our models using Rooted Mean Square Error (RMSE). Moreover, we analyze the effect of incorporating ancillary data such as topography maps into our model. Our method outperforms linear regression techniques by a high margin. We have found that using the temperature information from the preceding 3 to 4 years of data can improve the performance.We design different experimental settings to arrange monthly historical temperature information given to the deep learning model. In addition to the use of historical data consequently, we investigate the performance of periodical arrangements.

Considering an integrated approach to assess all of the measured pollutants in a diurnal, monthly, seasonal, and annual time scales and understanding the mechanisms hidden under low air quality conditions are essential for tackling potential air pollution issues. Konya, located in central Anatolia, is the largest province of Turkey with a surface area of 40,838 km2 and has different industrial activities. The lack of recent detailed studies limits our information on the underlying air pollution levels in Konya and obscuring policymakers to develop applicable mitigation measures. In this study, we used hourly monitored air quality data of CO, NO2, NOx, PM10, PM2.5, and SO2 from five stations in Konya and investigated the temporal and spatial variabilities for the 2008–2018 period via statistical analysis. Upon analysis, particulate matter was found to be the dominant pollutant deteriorating the air quality of Konya. The highest 2008–2018 periodic mean value of PM10 was found in Karatay Belediye as 70.5 µg/m3, followed by 67.4 µg/m3 in Meram, 58.7 µg/m3 in Selçuklu, and 43.7 µg/m3 in Selçuklu Belediye. The 24-h limit value of PM10 given as 50 µg/m3 in the legislation was violated in all of the stations, mainly during winter and autumn. High positive correlations were found among the stations, and the highest correlation was obtained between Selçuklu Belediye and Karatay Belediye with a Pearson correlation coefficient of 0.77. Long-term data showed a decreasing trend in PM10 concentrations. Diurnal variability is found to be more pronounced than weekly variability. For almost all of the pollutants, except for photochemical pollutants like O3, a prominent result was the nighttime and morning rush hours high-pollutant levels. A case study done for the January 29, 2018 to February 05, 2018 episode showed the importance of meteorology and topography on the high levels of pollution. Limitation of the pollutant transport and dilution by meteorological conditions and the location of Konya on a plain surrounded by high hills are believed to be the main reasons for having low air quality in the region.

Approximately 15.4 million people are continuously exposed to various air pollutants in the megacity of Istanbul. Anthropogenic activities in Istanbul generate emissions from mobile sources (i.e., vehicles, aircraft, ships, etc.) and stationary sources such as industrial and residential heating emissions. Biogenic emissions and long-range transport such as desert dust from Sahara and pollutants from Balkan countries also contribute to the local air pollution in Istanbul. In this work, fine (Dp < 2.5 μm) and coarse (Dp > 2.5 μm) particles were collected with a high-volume sampler during four seasons of the period Jan 2017 - Jan 2018. A total of 15 PAHs and 28 n-alkanes were identified and quantified with a newly developed thermal desorption – gas chromatography with mass spectrometry method (TD-GC–MS). Source analysis was performed with PAH and n-alkane diagnostic ratios, and source apportionment was performed with principal component analysis. The yearly averages of PAHs and n-alkanes in the fine fraction were 21.6 ng m−3 and 103.8 ng m−3, with daily averages of 7.1–80.8 ng m−3 and 55.3–204.2 ng m−3, respectively. Approximately 90% of the PAHs and n-alkanes were found in the fine PM fraction in this traffic site. The BaP carcinogenic (BaP-TEQ) and mutagenic (BaP-MEQ) equivalents were on average 5.47 ± 0.64 and 4.72 ± 0.8 ng m−3 in the fine fraction, respectively, and were approximately 6–7 times lower in the coarse fraction. This has important implications due to the respirable nature of fine aerosols and their longer lifetimes in the atmosphere. Multivariate analysis coupled with principal component analysis led to an important result that the organic aerosol mainly originates from two local sources: road traffic (50.4%) and shipping emissions (26.6%). The results found in this work indicate the urgent need for the application of mitigation measures to control road traffic and minimize the emissions from ships passing through the İstanbul Bosphorus.