Monitoring Industrial VOC Emissions and Geospatial Analysis

Abstract

Volatile organic compounds (VOCs) emissions from petroleum product storage pose not only a significant environmental concern but also a potential threat to occupational health. This study investigates geospatial analysis of VOCs on an industrial platform in Romania, utilizing a combination of portable field detectors and geostatistical modeling techniques. For more than 10 months, we conducted measurements at 41 georeferenced sampling points across three operational zones, using FID/PID instruments calibrated and validated in accordance with national standards. To evaluate dispersion conditions, meteorological data were simultaneously collected. VOC concentrations were measured under varying meteorological scenarios and analyzed using the Empirical Bayesian Kriging (EBK) method in ArcGIS Pro 3.1.0. Maximum concentrations reached up to 229.46 mg/m³ in central tank areas, with some point samples exceeding this level. Peripheral zones generally showed values below 65 mg/m³, although concentrations above 100 mg/m³ were still observed at 10% of the monitoring sites. The results indicate apparent spatial clustering of elevated VOC levels, particularly under low wind speed and high humidity. Our study highlights the relevance of continuous monitoring and site-specific mitigation strategies in high-risk industrial settings in Romania.

1. Introduction

Volatile organic compounds (VOCs) are key pollutants in industrial environments, particularly in petroleum product handling and storage facilities [1,2]. These emissions pose significant risks to air quality and occupational health. Previous research has highlighted the lack of continuous VOC monitoring in such industrial contexts across Romania [3]. Only a few studies have addressed this issue, including one carried out in Ploiești, a major Romanian center for petroleum refining and petrochemical production [4,5]. Sanda et al. [4] analyzed benzene levels and some associated health risks over a three-year monitoring period (2019–2021), emphasizing the temporal dynamics of benzene and its correlation with potential health outcomes. The air monitoring campaigns conducted in two cities in western Romania underscore the importance of determining BTEX (benzene, toluene, ethylbenzene, and xylenes) in areas with industrial activities, highlighting the need for further investigations to identify and control pollutant emission sources and improve urban air quality for support strategies [6,7].

Many approaches emphasize the importance of conducting detailed assessments directly within industrial areas, where multiple emission sources and meteorological factors significantly influence VOC dispersion. These findings provide significant support for the objectives of the present study, underscoring its importance and potential impact. The refineries are recognized as important sources of air pollutants, including highly concentrated volatile compounds and particulate matter, which significantly degrade local air quality [5]. The emissions from oil refineries primarily consist of hydrocarbons, in particular BTEX, which are the most significant because of their toxicity and persistence in the industrial environment [8]. A recent study that compared active and passive sampling methods for monitoring VOCs at industrial sites in northern Italy is particularly noteworthy. Borrelli et al. reported that active sampling methods deliver higher-resolution data but are affected by environmental humidity. In contrast, passive systems offer cost-effective and consistent coverage, particularly for organic compounds [9]. Monitoring VOCs in industrial environments presents specific challenges due to the wide range of concentrations, from very low levels in ambient air in neighboring areas or at the fenceline to very high levels near emission sources, such as gasoline storage tanks, petroleum product loading and unloading zones, or petroleum transport terminals. To address these challenges, FID- and PID-based detector systems are particularly suitable for industrial environments. The FID provides robust quantification of total hydrocarbons over a wide concentration range. It is less sensitive to humidity, while the PID offers high sensitivity to low-level VOCs and rapid response times. The combined use of FID–PID instruments fosters a comprehensive detection approach, reducing uncertainty caused by interfering gases and advancing data quality in heterogeneous industrial settings, such as petroleum storage facilities. Also, the presence of interfering gases (e.g., CO, NOx, O₃, or other non-hazardous compounds) requires sensors with high selectivity and sensitivity to reliably detect these pollutants under variable environmental conditions [10,11]. The need for accurate and timely VOC monitoring in industrial areas is critical, particularly in zones immediately adjacent to urbanized areas where strict environmental regulations apply [12]. Research studies increasingly incorporate spatial analyses using various modeling techniques, with GIS-based approaches frequently employed to examine the dispersion of air pollutants. However, most research has focused on urban environments rather than industrial areas. Sidharthan et al. [13] studied the spatial variation in air pollutants using GIS modeling. Their work demonstrates that GIS technology can be frequently utilized to investigate pollution concentrations both spatially and temporally, and to evaluate their impact on the air quality index. Additionally, Lee et al. [14] conducted fenceline monitoring at industrial sites, particularly petrochemical and refinery facilities, to measure VOC emissions in real time, demonstrate how pollution varies across the area, and create digital maps to help identify high-risk zones. Rashidi et al. [15] studied VOC emissions from several gas stations in a major city in western Iran. They used ArcGIS to visualize and analyze the spatial distribution of VOC concentrations, identifying hotspots and areas of higher exposure risk. The mean VOC concentrations were 3.58 mg/m³ in summer and 2.53 mg/m³ in winter, with the highest levels observed at one of the monitored stations. Spatial analysis showed that VOCs emitted from these stations contributed to elevated concentrations in the city’s central square. Choi and Chong [16] showed that geospatial integration methods significantly improve air quality assessment compared to earlier approaches, such as those by Kianisadr et al. [17]. Using GIS methodology, the authors mapped pollutant distributions across 45 monitoring stations during the summer and winter of 2016 from Khorramabad (Iran). The resulting maps enabled the identification of areas with varying human health risks and facilitated the calculation of air quality indices for selected locations. Another recent study employed GIS-based analysis to investigate VOCs in industrial and residential areas of Bucheon, Korea, using mobile laboratory measurements and mass spectrometry to assess three-dimensional spatial distributions [18,19]. Additionally, Choi et al. [20] created air pollution maps using GIS methods around an industrial complex and its surroundings, employed SIFT-MS for real-time VOC measurement, and identified the major substances contributing to pollution.

This study aims to assess VOC pollution at a major Romanian chemical platform, focusing on petroleum storage tanks, access roads, and loading and unloading ramps [21]. An integrated approach combining field measurements and geospatial methods was adopted to evaluate air quality and VOCs dispersion. The main stages of the study included field measurements at key points, calibration and verification of portable sensors, systematic data recording under varying environmental conditions, GIS-based spatial analysis, evaluation of containment structures, correlation analysis with tank features, and integration of field and spatial data to conduct a comprehensive air quality assessment. The study specifically focuses on: (1) analyzing spatial patterns around storage tanks; (2) assessing the effect of containment dikes on VOCs dispersion; and (3) investigating correlations between pollutant accumulation and tank operations, which have not been previously analyzed in research studies carried out in Romania.

Our study is a novel contribution by integrating in situ VOC monitoring with advanced spatial analysis at one of Romania’s petroleum refining platforms. This topic has received limited scientific attention to date. Unlike previous studies that rely primarily on modeled or generic urban/industrial emissions, our work provides reproducible measurement data from three monitoring zones over nine months. Additionally, this study establishes a transparent computational framework for converting ppmV sensor readings to mg/m³ under actual conditions and to g/Nm³ for BAT-AEL comparisons, while explicitly addressing uncertainties associated with the TVOC-to-NMVOC proxy conversion. These aspects enable a more accurate spatial representation of VOC distribution and risk areas within refinery facilities, contributing novel insights to GIS-based VOC mapping and management. Similar research articles have primarily focused on urban air quality [22,23] or indoor air of buildings [24,25]. This work directly investigates emission dynamics within industrial zones, offering new insights into pollutant dispersion patterns and potential accumulation hotspots. The findings are expected to support improved environmental management and decision-making processes for industrial operators and local authorities, thereby enhancing occupational safety and providing industrial operators with a sense of security and protection.

2. Experiments

2.1. Study Area

The experimental area is located in Romania, near the boundary of Navodari town, and approximately 200 m from the Black Sea coast (Figure 1). The entire industrial platform covers an area of 2 km², while the experimental zone, comprising the petroleum product storage tank park, spans approximately 0.5 km². The petrochemical complex is located on a coastal cordon characterized by generally low terrain. The region has a predominantly continental climate, moderated by the sea along a 10- to 15-km wide coastal strip. The local annual average temperature is 11.2 °C. Sea breezes shift between day and night: in summer, daytime winds blow from the sea toward the land and reverse at night, whereas in winter, cooler land temperatures generate winds from the land toward the sea. The temperature increases to 32–35 °C in summer, while winter averages around +1.1 °C.

Frequently, these temperature values generate upward air currents that facilitate the dispersion of pollutants into higher atmospheric layers. The annual mean humidity is 81%, influenced by continuous sea evaporation, which helps prevent excessive summer heat. Precipitation is generally low and stable, with yearly totals below 400 mm [26].

The sampling area lies between latitudes 44°20′12″ N and 44°20′30″ N and longitudes 28°38′56″ E and 28°39′6″ E, and is divided into three distinct zones, as shown in Figure 1. The research was conducted at an oil refinery located near both urban and ecological zones. The monitoring network consisted of 41 sampling points, distributed across three sites, each characterized by specific operational activities and storage tank arrangements. The criteria for selecting the three monitoring sites were based on the identification of: (a) areas comprising stationary sources of pollution (petroleum product storage tanks operating under steady-state conditions); (b) areas with lower levels of pollution (access roads within the industrial platform); (c) areas adjacent to sources with constant emissions over time.

Table 1. Key features of the work sites.

Site No.	Site Area (m2)	No. of Sampling Points	No. Tanks	Tank Capacity (m3)	Petroleum Product
1	8100	15	4	5000	gasoline
2	30,000	14	4	10,000
3	6700	12	2	3500

shows the key characteristics of the experimental area.

2.2. VOC Emission Sources

Petroleum storage tank farms are critical points in the fuel distribution chain and are prone to significant VOC emissions that can degrade air quality and affect nearby ecosystems. Atmospheric pollutants in crude oil refineries mainly originate from storage tank farms, petroleum loading and unloading areas, transport systems (road, rail, and marine), process installations, cooling towers, open-surface separator basins, dynamic equipment, pipelines, valves, flanges, and test connections. Storage tanks are constructed as vertical cylindrical tanks and are the most commonly used type for storing petroleum products. They operate at atmospheric pressure. These tanks are made from welded steel sheet plates. The tanks are equipped with a fixed steel or aluminum roof, an internal floating aluminum roof, or an external steel floating roof with double sealing. Vertical cylindrical tanks with floating roofs are designed so that the roof floats on the surface of the stored liquid, and the space between the roof and the tank shell contains a double seal that minimizes the release of volatile organic compounds into the atmosphere. The tanks include the following components: loading/unloading connections, manway holes, level indicators and sampling ports, roof hatches, vent valves, hydraulic safety valves, flame arresters, water-cooling systems, and foam fire-suppression systems. Each tank farm in a liquid hydrocarbon storage facility must be equipped with a containment dike, in which tanks are positioned at safe distances from the access road and can retain accidental product spills from the storage tanks. The capacity of the containment dike depends on the total capacity of the tank group. The design of a gasoline storage tank is carried out in accordance with the API 650 standards [27].

The study zones were defined based on the tank type, the specific stored petroleum product (e.g., gasoline), the product level in each tank, and the tank’s location relative to the access roads on the industrial platform. An origin point (e.g., the corner of the tank park) was chosen as the reference (Oxy) of each zone, from which distances to the VOC concentration measurement points were computed. The practical approach allowed the spatial distribution of measurements to be analyzed relative to a local reference. The distances between the monitoring points (Xi, Yi) and the reference point were calculated based on their coordinates using the Stereo 1970 projected coordinate system [28]. The ΔX and ΔY values represent the differences along the X and Y axes relative to the Oxy plane (Equations (1) and (2)). Simultaneously, the field distance corresponds to the Euclidean distance (ED) derived from these coordinate differences (Equation (3)).

$$\Delta \mathrm{X}={\mathrm{X}}_{\mathrm{i}}-{\mathrm{X}}_0;$$

(1)

where i $\in \left[1, 15\right]$ (zone 1); [1, 14] (zone 2), and [1, 12] (zone 3)

$$\Delta \mathrm{Y}={\mathrm{Y}}_{\mathrm{i}}-{\mathrm{Y}}_0;$$

(2)

where i $\in \left[1, 15\right]$ (zone 1); [1, 14] (zone 2), and [1, 12] (zone 3)

$$\mathrm{d}\left({\mathrm{P}}_{\mathrm{i}},\mathrm{O}\mathrm{x}\mathrm{y}\right)=\sqrt{{(∆\mathrm{X})}^2+({∆\mathrm{Y})}^2}$$

(3)

where d(${\mathrm{P}}_{\mathrm{i}}$,Oxy)—the Euclidean distance (m); ${\mathrm{P}}_{\mathrm{i}}$—the measured point of location (${\mathrm{X}}_{\mathrm{i}},{\mathrm{Y}}_{\mathrm{i}}$);

• $\Delta \mathrm{X}, \Delta \mathrm{Y}$—the relative horizontal and vertical displacement to the reference point (Oxy).

All spatial layers were referenced to the coordinate system Stereo 1970 (EPSG:31700), a planar Cartesian projection that ensures metric accuracy for distances and areas [28]. Figures S1–S3 from the Supplementary Materials illustrate these zones of interest.

2.3. VOCs Monitoring

The diversity and complexity of hydrocarbons emitted during the storage, transfer, and handling of fuels present significant challenges for accurately assessing their impact on air quality [5]. In practice, direct measurement techniques using FIDs (flame ionization detectors) or PIDs (photoionization detectors) are commonly employed to monitor VOCs. Continuous or periodic monitoring using both techniques is therefore a critical component of air quality management in petroleum storage and distribution operations. The U.S. EPA and European standardized procedures for instrument calibration and quality assurance facilitate the accurate quantification of VOC emissions, enable the identification of fugitive or leak sources, and support compliance with regulatory air quality standards [29,30]. European and Romanian legislation have established concentration limits for certain individual compounds (e.g., benzene, toluene, and styrene), primarily related to indoor air quality. In contrast, for outdoor workplace environments, volatile compounds are grouped into categories such as total volatile organic compounds (TCOV), without specific legislative limits [31,32]. Nevertheless, the sensors used must be capable of measuring not only the chemical compounds themselves but also their widely varying concentration ranges. These ranges extend from a few micrograms per cubic meter (µg/m³) in ambient air to several hundred µg/m³, and from single units to tens or even tens of thousands of milligrams per cubic meter (mg/m³) in industrial emissions.

2.4. Measurement Devices

In the Romanian oil refinery industry, dual-detector systems (FID and PID) are commonly employed. These systems allow rapid and efficient measurement of a wide range of organic and inorganic vapors, outperforming single-detector devices. For this study, two portable analyzers were used:

• Thermo Scientific TVA-1000B (Waltham, MA, USA)—a dual FID/PID device with measurement ranges of 0.5–2000 ppm (PID) and 0.5–5000 ppm (FID), 3.5 s response time, ±2.5 ppm accuracy, repeatability of ±1% (PID) and ±2% (FID), flow rate of 1000 mL/min, and eight hours operating time. Standard calibration was performed using methane (FID) and isobutylene (PID).
• Dräger Multi-PID 2 (Ion-Science, Fowlmere, UK)—featuring 1 ppb resolution, 0–20,000 ppm measurement range, <2 s response time, standard 100 ppm calibration with customizable isobutylene calibration, 220 mL/min flow rate, and a 10.6 eV PID lamp (additional 11.7 eV lamps available).

The TVA-1000B equipment (Thermo Fisher Scientific, Inc., Waltham, MA, USA) is primarily employed in points/zones with expected high VOC concentrations (e.g., tank roofs, vents, and loading/unloading areas). At the same time, the Dräger Multi-PID 2 (Ion Science Ltd., Fowlmere, UK) is used for lower concentration ranges. This complementary deployment ensured coverage of both high-intensity emission sources and background industrial levels, enhancing data robustness and reproducibility. The portable analyzers used for VOC monitoring comply with ATEX and IECEx standards [33,34], guaranteeing safe operation in potentially explosive atmospheres.

2.5. Data Collection

Measurements were conducted over one year (September 2023–August 2024) under varying meteorological conditions. Each sampling campaign was organized to capture a representative range of weather conditions (windy, calm, and humid days), enabling evaluation of the influence of meteorology on VOC dispersion. On each sampling day, hourly measurement sessions were conducted to primarily capture diurnal variations: in the morning (08:00–12:00) and, less frequently, in the afternoon (13:00–16:00). Each monitoring point was sampled 10–20 times per session. Measurements of VOC concentrations were performed at a height of 1.4 m above the ground, which is representative of potential human exposure near emission sources. Simultaneously with the VOC measurements, other atmospheric parameters, including air temperature, wind speed and direction, and humidity, were recorded using a local meteorological station. Although measurements were conducted under a range of meteorological conditions, spatial analysis using the EBK method was confined to datasets from stable storage phases and quasi-stationary atmospheric conditions (low, moderate, and constant wind speed; no precipitation; stable atmospheric pressure). These conditions are crucial for maintaining spatial coherence and ensuring the scientific validity of the interpolation. The measurement protocol ensured safety and data reliability, and consisted of general procedures and field campaign stages.

• General Procedures (applied at each monitoring session) include:

  (1)

  Preparation and Instrument Setup

  •

  Accessing the designated monitoring areas via safe routes and in compliance with ATEX/IECEx safety procedures;

  •

  Visual inspection and calibration of the portable analyzers (TVA-1000B and Dräger Multi-PID 2) before each measurement session (FID/PID measurements provide TVOC values, reflecting the total carbon content of volatile compounds, without distinguishing individual chemical species, and methane or isobutylene emissions are negligible in gasoline storage and loading operations).

  (2)

  Validate the Measurement Procedure

  •

  Activating the analyzers and confirming stable responses, allowing 10–15 s for the system to reach steady-state before measurements.

  •

  Performing a brief test run to ensure proper instrument function and procedural readiness.

  (3)

  Field Campaign Stages

The campaign was structured into two main stages, with measurement times selected based on meteorological conditions:

(a)

Preliminary Testing (September - November 2023) for VOC concentrations around storage tanks (four points) and on the tank roofs (e.g., five points, of which four were located around the central ventilation fittings, and one at the peripheral vent);

(b)

Main Testing (December 2023 – August 2024) for VOC concentrations at 41 sampling points (10–15 values per point).

The atmospheric parameters considered in the study were: wind direction, wind speed (m/s), air temperature (°C), relative humidity (%), and atmospheric pressure (mbar).

• The experimental data were grouped into four classes of atmospheric conditions:
  • Low humidity conditions (relative humidity = 48.72%; wind speed = 2.94 m/s; wind direction = NW; air temperature = 21.29 °C) on 2 November 2023, 13:45–14:20.
  • Calm atmosphere (wind speed = 0.5 m/s; wind direction = WSW; air temperature = 10.34 °C; relative humidity = 97%) on 14 December 2023, 10:00–11:00 a.m.
  • High humidity conditions (relative humidity = 99%; wind direction = WSW; wind speed = 0.9 m/s; air temperature = 5.27 °C) on 18 January 2024, 08:00–09:00 a.m.
  • Variable conditions in short intervals of 5–10 min (wind speed = 2.5 m/s; wind direction = SSE; air temperature = 6.4 °C; relative humidity = 80%), between 14:55 and 15:45 on 20 March 2024.

The calendar of the monitoring period is presented in

Table 2. VOCs monitoring schedule.

Date/Period	Time	Measurement Objectives
8 September 2023	01–02 p.m.	9 points at a single tank: 5 points on the tank’s roof; 4 points around the tank
2 November 2023	02–03 p.m.	16 points around four tanks
14 December 2023	10–11 a.m.	41 points around tanks at 1.4 m above ground level: zone 1–15 points (z1M01 to z1M15); zone 2–14 points (z2M01 to z2M14); zone 3–12 points (z3M01 to z3M12).
January 2024 (20 days)	08–12 a.m.
March 2024 (21 days)
May 2024 (20 days)	08–11 a.m.
June 2024 (19 days)
July 2024 (23 days)
August 2024 (21 days)

.

Hourly atmospheric data corresponding to the calendar dates and measurement intervals were retrieved from the on-site meteorological station (44°20′24.32″ N, 28°38′48.96″ E) located within the industrial platform.

2.6. Data Analysis

The VOC concentration data recorded in the three monitored zones (zone 1, zone 2, and zone 3) during the period December 2023–August 2024 yielded 125 data series, which were subsequently subjected to the following processing stages:

(1)

Conversion of the recorded values from ppmV to g/m³, to allow comparison with the thresholds established by EU Decision 2427/2022 concerning the implementation of Best Available Techniques (BAT) for industrial emission control [35]. This decision specifies the reference range for non-methane VOC (NMVOC) concentrations resulting from loading and unloading operations involving volatile petroleum fractions, namely between 0.15 and 10 g/Nm³ (hourly average).

VOC concentrations measured in ppmV were converted to g/m³ using an ideal gas approximation:

$$C\left({\displaystyle \frac{\mathrm{g}}{{\mathrm{m}}^3}}\right)=\mathrm{C}\left(\mathrm{p}\mathrm{p}\mathrm{m}\mathrm{V}\right){\displaystyle \frac{\overline{M_a}\left(\frac{\mathrm{g}}{\mathrm{m}\mathrm{o}\mathrm{l}}\right)}{V_m\left(\frac{{\mathrm{m}}^3}{\mathrm{m}\mathrm{o}\mathrm{l}}\right)}}\times {10}^{-3}$$

(4)

where C—VOC concentration; ($\overline{M_a}$)—average molar mass of the gasoline vapor mixture; Vm—molar volume at measurement conditions (20 °C and 1 atm).

A temperature of 20 °C was selected as the representative average of ambient conditions during the monitoring campaign. The values considered are: 1 ppmV = 1 × 10⁻⁶ (v/v); ($\overline{M_a}$) = 110 g/mol; Vm = 24.05 × 10⁻³ m³/mol.

The molar mass of the petroleum products during the measurement campaign ranged from 104 to 116 g/mol, as reported in the product technical datasheets. An average molar mass of gasoline vapors was adopted as a representative value for C₅–C₁₂ hydrocarbons. Concentration values were expressed in mg/m³ to facilitate the interpretation of results. Accordingly, the conversion factor of 4.57 used to approximate VOC concentrations in mg/m³ reflects the actual field conditions and ensures methodological consistency, whereas a conversion factor of 4.91 was applied for standard conditions corresponding to BAT-AEL values (0 °C, 1 atm) [36,37].

(2)

Comparison of VOC concentrations (expressed in mg/m³) recorded as atmospheric emissions with reference values applied in industrial practice. In the absence of an explicit legislative framework regulating maximum admissible VOC emissions at either the national or international level, the classification proposed by Tecamgroup Industrial was used as a reference [38]. This classification distinguishes the following emission levels: low (<0.3 mg/m³), acceptable (0.3–0.5 mg/m³), borderline (0.5–1 mg/m³), and high (1–3 mg/m³).

(3)

Descriptive statistical analysis of the data series, including the determination of minimum, maximum, and hourly mean VOC concentrations, aimed at providing an overall characterization of their distribution across the three monitored zones.

(4)

Selection of days with the highest VOC concentrations for EBK spatial analysis. Of the 125 monitoring days, nine were selected. These comprised four winter days (December and January), two spring days (March and May), and three summer days (June, July, and August) to analyze the spatial distribution of pollution under conditions that significantly affect industrial air quality. Criteria-based selection of nine days used for EBK analysis is presented in

Table 3. Criteria-based selection of days used for EBK analysis.

No.	Season	Day	Time (a.m.)	Windward Direction	Wind Speed (m/s)	Wind Variability	AT/AH * (°C/%)	Atmospheric Stability	Tank Operational State
D1.	winter	14 December 2023	10–11	WSW	0.5	low	10.4/97	very stable	idle mode
D2.	winter	3 January 2024	08–09	WSW	1.7	moderate	3.7/94	neutral	idle mode
D3.	winter	4 January 2024	08–09	W	2.6	moderate	6.4/95	slightly unstable	draining
D4.	winter	31 January 2024	09–10	NNE	0.2	low	−0.3/99	very stably	loading
D5.	spring	15 March 2024	08–09	NNE	0.8	low	5.6/93	weakly stable	revision
D6.	spring	31 May 2024	09–10	SSE	1.6	moderate	20.6/89	neutral	idle mode
D7.	summer	28 June 2024	10–11	NNE	2.0	moderate	26.6/61	slightly unstable	loading
D8.	summer	31 July 2024	09–10	NNE	1.1	moderate	24.9/56	neutral	draining
D9.	early autumn	29 August 2024	10–11	ENE	1.6	moderate	20.0/99	neutral	revision

* AT/AH—Air Temperature/Air Humidity.

. The selection of days used for EBK spatial analysis was based on a combination of meteorological representativeness, operational conditions, and data completeness. This approach aims to ensure the data reflects typical atmospheric conditions, which should instill confidence in the analysis for the audience.

(5)

Spatial interpolation of the data was performed using the EBK method in ArcGIS Pro (version 3.1.0, 2023 Esri Inc., Redlands, CA, USA), taking into consideration the methodological aspects and justifications presented in Section 3.

3. Spatial Analysis Method

Methodology

ArcGIS Pro software has introduced various extensions for geostatistical analysis. One of the most commonly used spatial interpolation methods in GIS is IDW (Inverse Distance Weighting). This approach estimates the concentration value at a known point using weighted contributions from surrounding points. An alternative technique is EBK (Empirical Bayesian Kriging), an advanced spatial interpolation method developed by ESRI and available in ArcGIS Pro [39]. It helps create digital maps of pollutant distributions, supporting environmental assessments and risk analyses in the industrial complex, and can also be applied to other fields and applications. VOC concentrations were spatially analyzed using EBK in ArcGIS Pro, version 3.1.0 (Esri). EBK is an advanced spatial interpolation method used to predict unknown values across a geographic area from a set of known data points (e.g., measured VOC concentrations within the monitoring area). EBK improves upon classical Kriging techniques by automatically incorporating the uncertainty of semivariogram model parameters into the interpolation process [40,41,42]. Conventional Kriging is a geostatistical spatial interpolation method that estimates unknown values at a given location based on measured values at nearby points, while accounting for the spatial structure of the data, i.e., the correlation of values as a function of distance [20]. Kriging provides not only an estimate but also a measure of uncertainty (standard error) for each predicted point.

Kriging interpolation estimated the value at a known location as a weighted linear combination of nearby known values, expressed by Equations (5) and (6):

$$\widehat{Z}\left(P_{(x,y)}\right)={\sum }_{i=1}^n{\lambda }_{i}Z\left(P_{i(x_i,y_i)}\right)$$

(5)

where

• $\widehat{Z}\left(P_{(x,y)}\right)$ = the estimated value at the prediction point, mg/m³;
• $Z\left(P_{i(x_i,y_i)}\right)$ = the measured value at point $P_{i(x_i,y_i)}$, mg/m³;
• (x,y) = the coordinates of the prediction location;
• ${\lambda }_i$= the Kriging weight assigned to each measured point;
• n = the number of measured points in the neighborhood of the interpolation point.

The Kriging weights (${\lambda }_i)$satisfy the constraint:

$${\sum }_{i=1}^n{\lambda }_i=1$$

(6)

The Kriging weight (${\lambda }_i)$ is determined from the semivariogram, which quantifies spatial correlation (7) or the equivalent form (8):

$$\gamma \left(h\right)={\displaystyle \frac{1}{2\mathrm{N}\left(\mathrm{h}\right)}} {\sum }_{\begin{array}{c}\left(i,j\right)\\ ‖P_i-P_j‖\end{array}}{\left[Z\left(P_{i(x_i,y_i)}\right)-Z\left(P_{j(x_j,y_j)}\right)\right]}^2$$

(7)

where

• $\gamma \left(h\right)$ = semivariogram function;
• h = lag distance (spatial separation between coordinate pairs of points), m;
• $\mathrm{N}\left(h\right)$ = number of pairs of points separated by distance h;
• i, j = the indices of the measured points forming pairs separated by the distance h;
• $‖P_i-P_j‖$ = the Euclidean distance between the two points, corresponding to the h, m;
• $Z\left(P_{i(x_i,y_i)}\right)$, $Z\left(P_{j(x_j,y_j)}\right)$ = the measured values at those locations, mg/m³.

$$\gamma \left(h\right)={\displaystyle \frac{1}{2\mathrm{N}\left(\mathrm{h}\right)}} {\sum }_{i=1}^{N\left(h\right)}{\left[Z\left(P_{i(x_i,y_i)}\right)-Z\left(P_i+h\right)\right]}^2$$

(8)

where

• $P_i+h$ = $(x_{i }+h_x, y_{i }+h_y).$

In the EBK method, semivariogram parameters are treated as random variables. Multiple realizations are generated using restricted maximum likelihood, producing a set of predictions. ${\widehat{Z}}_j\left(P_{(x,y)}\right)$. The EBK prediction is the mean of all realizations by (9):

$${\widehat{Z}}_{EBK}\left(P_{(x,y)}\right)=\left({\displaystyle \frac{1}{m}}\right){\sum }_{j=1}^m{\widehat{Z}}_j\left(P_{(x,y)}\right)$$

(9)

where

• m = the number of simulations.

The associated prediction variance includes both the Kriging variance of the simulation and the semivariogram uncertainty (10):

$${{\sigma }^2}_{EBK}\left(P_{(x,y)}\right)=\left({\displaystyle \frac{1}{m}}\right){\sum }_{j=1}^m{\sigma }_j^2\left(P_{(x,y)}\right)+Var\left[\widehat{Z_j}\left(P_{(x,y)}\right)\right]$$

(10)

where

• ${{\sigma }^2}_{EBK}\left(P_{(x,y)}\right)$ = the EBK variance for all simulations, (mg/m³)².

The EBK method generates different semivariogram models to describe spatial correlation. Semivariograms can take several theoretical forms depending on the spatial structure of the data. Common semivariograms include: (1) Linear: γ(h) = Nugget + b$\left|h\right|$; (2) Power: γ(h) = Nugget + b${\left|h\right|}^{\alpha }$, and (3) Thin Plate Spline: γ(h) = Nugget + b$\left|h^2\right|ln\left(\left|h\right|\right)$; where Nugget = the baseline variance at infinitesimal separation, b = the slope parameter, and α = the exponent (for the Power model).

A detailed description of the EBK method, its implementation steps, and its advantages is available in the online documentation of the ArcGIS Pro (Geostatistical Analyst) on Esri’s official website [40,41]. The resulting digital raster representations illustrate the estimated spatial distribution of VOC concentrations, allowing the identification of high-risk pollution zones and the depiction of emission gradients between the monitored points.

A GIS-based workflow (Figure S4) was implemented in ArcGIS Pro using ModelBuilder to ensure reproducibility of the spatial analysis. VOC point measurements were first filtered to select monitoring days corresponding to peak concentration events, and then EBK interpolation was used to generate continuous concentration surfaces. The resulting rasters were further analyzed using zonal statistics and visualized in 2D, highlighting the workflow’s role in supporting precise, reproducible spatial analysis.

4. Results

In this section, the data are presented in two complementary stages. First, the hourly-average VOC concentrations recorded on 14 December 2023 are analyzed, with values expressed in ppmV as measured in the field. Subsequently, the data are interpolated using the EBK method and converted to g/m³ to enable comparison with similar findings reported in other research studies or legislative requirements.

4.1. VOC Concentration Levels

The recorded data provide both quantitative and visual insights into the spatial distribution of VOCs across the three designated operational zones. Figure 2 illustrates the distributions for Zones 1–3, highlighting localized emission hotspots and concentration gradients. The tank covers represent different storage tanks within the monitored area.

This first complete dataset, obtained under real operating conditions and expressed in ppmV, reveals apparent heterogeneity in VOC dispersion, reflecting the combined influence of emission-source geometry, surface roughness, and prevailing meteorological conditions. Considering the meteorological characteristics on 14 December 2023 (10:00–11:00, WSW at 0.5 m/s, air temperature 10.4 °C, relative humidity 97%), the measured concentrations of VOCs in Zone 1 (Figure 2a) were generally low, ranging from 1 to 15 ppmV throughout the monitoring period. The highest concentrations were recorded at points z1M15 (x = 2 m, y = 21 m) and z1M08 (x = 60 m, y = 39 m), both of which reached approximately 15 ppmV. This peripheral area was not directly affected by petroleum product handling operations on that day, and favorable atmospheric conditions contributed to pollutant dilution and dispersion. Zone 2 (Figure 2b) exhibited intermediate concentrations, with a minimum of 10.20 ppmV at z2M12 (x = 117 m, y = 198 m) and a maximum of 30.10 ppmV at z2M14 (x = 117 m, y = 229 m), near the containment dike. The elevated levels in this area suggest localized accumulation due to reduced air circulation, while overall concentrations reflect moderate emission intensity influenced by site topography and micro-scale meteorological conditions. Zone 3 (Figure 2c) displayed the highest VOC concentrations, exceeding 50 ppmV at several central monitoring points near the tank area, with a maximum of 60.64 ppmV at z3M05 (x = 37 m, y = 66 m). Concentrations decreased with distance from the source, except at specific points z3M04 (x = 5 m, y = 66 m), indicating complex dispersion patterns shaped by wind speed, humidity, and temperature. The spatial distribution suggests a significant emission source in proximity to the storage tanks. Peripheral sectors consistently recorded lower values (<15 ppmV). The VOC concentrations in Zone 1 were significantly lower, while those in Zone 2 were intermediate compared to those in Zone 3. These findings, particularly in relation to the Romanian Law No. 264/2017 [32], underscore the importance and relevance of our study. It is noteworthy that in Zones 1 and 2, VOC concentrations were generally below the permissible limits for peripheral areas of the industrial platform (<30 ppmV), corresponding to the accepted levels for secondary storage areas and zones with limited petroleum product handling. These results are also significant in the context of European legislation for industrial emissions, including VOC management and treatment in the chemical sector [35]. The preliminary observations allow the identification of areas with elevated emissions and provide the foundation for subsequent spatial modeling using GIS applications.

4.2. Spatial Interpolation with EBK

Due to the heterogeneity of the data series recorded between December 2023 and August 2024 for the 41 measurement points in Zones 1, 2, and 3, a proper evaluation is complicated using simple statistical analysis. To interpret pollutant dispersion, data series were created by averaging VOC concentration values recorded during the hourly intervals on the monitoring days. Only days with complete datasets, defined as uninterrupted and quality-controlled VOC measurements across all monitoring locations and sensors, were retained for EBK spatial analysis. The results were analyzed using the geostatistical method (EBK), which allows estimation of pollutant concentrations at all positions between the monitoring points. The digital representations obtained through EBK interpolation are raster-type and depict the estimated spatial distribution of VOC concentrations. The results of EBK interpolation for the nine days mentioned in Section 2.4 are presented in Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8, Figure 9, Figure 10 and Figure 11, where the arrows show direction and wind speed. EBK interpolation allows the identification of areas with elevated pollution risk and highlights the emission gradient between the monitored points.

5. Discussion

5.1. Hourly Variation in VOC Concentrations

The EBK raster in Figure 3 indicates that Zone 3 exhibits the highest concentration (113–136 mg/m³), located in the northwestern part of the site. This area contains the most significant number of densely distributed measurement points (z3M05, z3M06, z3M09), suggesting an intense source of pollution, most likely due to loading/unloading activities or damage to elements of the internal floating roof on 14 December 2023, when the petroleum product level was 30% of the capacity of two storage tanks of 3500 m³ each. In Zone 2, lower concentrations (70–97 mg/m³) were recorded around two 10,000 m³ storage tanks, while in Zone 1, the lowest concentrations (52–62 mg/m³) were observed in the southern part of the site. Therefore, VOC pollution intensity is higher in the northern and northwestern areas of the site, primarily due to petroleum loading/unloading operations or the positioning of emission sources relative to the dominant wind direction during the 10–11 a.m. interval on 14 December 2023. The EBK raster in Figure 4 indicates a moderate to high VOC concentration in the vicinity of the 3500 m³ storage tanks located in the eastern part of Zone 3. The WSW wind direction favored pollutant dispersion towards the NE, which explains the expansion of the impact area in that direction. The relatively low average hourly temperature (3.7 °C) and high humidity (94%) contributed to the partial stagnation of pollutants in the lower air layer. Elevated values may be associated with reduced atmospheric ventilation, although the wind speed (1.7 m/s) was higher than on 14 December 2023. The EBK raster in Figure 5 shows a reduction in concentrations on 4 January 2024, compared to the previous day, particularly in the central part of Zone 3. However, there is a higher increase in the eastern part of Zone 1. The stronger wind speed (2.6 m/s) and westerly direction contributed to more efficient VOC dispersion, reducing local accumulations. The slight increase in temperature relative to 3 January 2024 favored faster evaporation of pollutants, an effect that was partially offset by good atmospheric circulation. A subtle tendency of accumulation is also observed in the western part of Zone 1, along the direction of wind movement (W). On 31 January 2024, a high accumulation of VOC emissions was observed in the eastern parts of Zones 1 and 2, near tanks with capacities of 5000 m³ (Zone 1) and 10,000 m³ (Zone 2), as shown in the EBK raster in Figure 6. The very low wind speed (0.2 m/s) and relatively high atmospheric pressure (1033 mbar) created thermal inversion and stagnation conditions, favoring the accumulation of pollutants at ground level. The 99% humidity and negative temperature (−0.23 °C) further accentuated the lack of dispersion, resulting in maximum concentration values in the immediate vicinity of the emission sources. The EBK raster in Figure 7 highlights a relatively moderate distribution of VOC concentrations across the three monitored zones, with higher values localized in Zone 3. Most concentrations in this zone range between 10 and 45 mg/m³, but a maximum VOC concentration was recorded at point z3M09 (125.09 mg/m³), located in the immediate vicinity of the 3500 m³ storage tank. In Figure 8, a significant increase in VOC concentrations is observed in the central part of Zone 3, across two areas extending from southwest to northeast, suggesting possible pollutant transport in this direction. A high-concentration zone is highlighted, possibly located between two storage tanks with fixed and floating roofs (capacity 3500 m³), with interpolated values exceeding 40 mg/m³. The EBK raster for 28 June 2024, shown in Figure 9, indicates a significant pollution peak in Zone 3, over relatively large areas, with VOC concentrations exceeding 50 mg/m³ in the western and northwestern parts of the site. Also, the EBK raster for 31 July 2024 (Figure 10) shows significant spatial variation in the northern part of Zone 3, with clearly defined areas of accumulation near the fixed- and floating-roof tanks. On 29 August 2024, under light wind conditions (1.6 m/s) from the ENE, lower temperature (20.08 °C) compared to 31 July 2024, and high atmospheric humidity (99%), VOC concentrations were higher in the eastern and northeastern parts of the site (Zones 2 and 3).

5.2. Variation in VOC Concentrations by Measurement Session

The distribution of VOCs in the industrial park suggests a significant risk to air quality and public health, especially under similar meteorological conditions. The results confirm that meteorological parameters significantly influence pollutant dispersion, particularly wind speed and direction, and air temperature.

Stable meteorological conditions, characterized by very low wind speeds, high atmospheric pressure, and elevated relative humidity, favored VOC accumulation near emission sources (e.g., tank ventilation systems). The most representative case of such conditions occurred on 31 January 2024 (Figure 6), when the wind speed dropped to 0.2 m/s, atmospheric pressure reached 1033 mbar, and relative humidity was 99%, accompanied by a negative air temperature (−0.23 °C). This combination created near-stagnant, thermally inverted conditions, resulting in the highest VOC concentrations across all monitored zones. Under these conditions, high VOC accumulation was observed in the eastern parts of Zone 1 and Zone 2, particularly near storage tanks with capacities of 5000 m³ and 10,000 m³, respectively. The EBK raster revealed sharply localized concentration peaks near the emission sources, indicating minimal atmospheric dispersion and strong source dominance. Similar accumulation tendencies were also observed on 14 December 2023 (Figure 3), when low wind speed (0.5 m/s) and stable conditions led to elevated VOC concentrations in Zone 3, with values ranging from 113 to 136 mg/m³. These concentrations were spatially clustered around densely distributed measurement points (z3M05, z3M06, z3M09), suggesting intense emissions associated with loading and unloading operations or potential issues related to floating roof elements.

Dynamic dispersion conditions, associated with higher wind speeds and improved atmospheric ventilation, resulted in a noticeable reduction in local VOC accumulation and a broader spatial spread of pollutants. This behavior is illustrated by the measurements conducted on 4 January 2024 (Figure 5), when wind speed increased to 2.6 m/s from the west. Under these conditions, VOC concentrations decreased in the central part of Zone 3, indicating more efficient dilution. At the same time, moderate increases were observed along the wind transport direction in the eastern part of Zone 1. A similar dispersion-driven pattern was observed on 3 January 2024 (Figure 4), when a wind speed of 1.7 m/s from the WSW favored pollutant transport toward the northeastern part of the site. Although the average temperature was relatively low (3.7 °C), the increased wind speed prevented substantial accumulation, resulting in moderate concentration levels distributed along the dominant airflow path. During the summer monitoring campaign, dynamic dispersion conditions were evident on 28 June 2024 (Figure 9), when wind speeds reached 2.0 m/s and temperatures exceeded 35 °C. Despite enhanced atmospheric mixing, a significant pollution peak was observed in Zone 3, particularly near storage tanks equipped with fixed and floating roofs. This suggests that under dynamic conditions, high VOC concentrations are more indicative of localized emission events or operational anomalies than of meteorological stagnation alone. A seasonal comparison of VOC spatial distributions reveals apparent differences in emission behavior and dispersion efficiency across winter, spring, and summer periods (

Table 4. Seasonal and zonal VOC concentrations based on recorded maxima.

Zone	Season	Day	Concentration (mg/m3)	Meteorological Condition	Dominant VOC Spatial Pattern	Key Factors
z1	winter	31 January 2024	>160	stable (thermal inversion)	accumulation near emission sources	wind speed—0.2 m/s air temperature < 0 °C
z1	summer	31 July 2024	<30	neutral	low and spatially uniform concentrations	reduced influence of sources
z2	winter	14 December 2023	70–97	stable	moderate accumulation around tanks	low wind speed and proximity to tanks
z2	late summer	29 August 2024	>70	neutral, high humidity	accumulation in East sector	reduced vertical dispersion
z3	winter	14 December 2023	113–136	stable	hotspot in the NW sector	loading activities
z3	early spring	15 March 2024	125	weakly stable	localized intensity core near tanks	local source dominance
z3	summer	28 June 2024	>50	dynamic	high-concentration areas in the NW sector	operational emissions outweigh dispersion

).

During the winter season, stable atmospheric conditions frequently led to pronounced VOC accumulation near emission sources, as observed on 14 December 2023 and 31 January 2024. In contrast, winter days with moderate winds (e.g., 3–4 January 2024) exhibited lower concentrations and more elongated dispersion patterns aligned with the prevailing wind direction. In spring, VOC distributions showed increased spatial variability. On 15 March 2024 (Figure 7), a localized accumulation core was identified in the NW part of Zone 3, while the surrounding areas exhibited lower, more uniform concentrations. On 31 May 2024 (Figure 8), the EBK raster indicated a broader dispersion pattern oriented along the SW–NE axis, likely influenced by SSE winds and operational activities related to fuel handling and transshipment. During the summer period, higher temperatures favored enhanced evaporation of volatile compounds, resulting in higher VOC concentrations. On 31 July 2024 (Figure 10), stable meteorological conditions (wind speed 1.1 m/s, NNE direction) led to localized accumulation near storage tanks in Zone 3, while Zone 1 remained less affected. On 29 August 2024, high relative humidity (99%) combined with moderate wind speed (1.6 m/s) resulted in elevated VOC concentrations in the E and NE parts of the site, particularly near the 3500 m³ storage tanks.

A comparative analysis of VOC pollution reported in various studies for industrial (oil refinery), urban, and rural areas due to significant differences in both emission levels and primary pollution sources. Maximum VOC concentrations were recorded in Izmir, Turkey (135.9 ppb in an industrial area during 2000–2001) [43], in Guangzhou, China (64.41 ppb in an urban area in 2004), and in Xinken, China (41.3 ppb in a rural area in 2004) [44]. In urban areas, emission levels are generally lower than those in oil-related industrial zones [45,46,47]. Regarding VOC levels specifically in industrial areas, the concentrations measured during our experimental campaign were higher than those reported in the similar studies: 60.64 ppm in Zone 3 (14 December 2023), 32.04 ppm in Zone 2 (15 November 2023), and 15.62 ppm in Zone 1 (20 February 2024), compared to 1.434 ppm recorded at Lynchburg Ferry/Houston Ship Channel in summer 2018 [45]. Prolonged exposure to VOCs in oil-related industrial areas can lead to serious health issues, including respiratory diseases and cancer. Temporal variations in VOC concentrations from petroleum industries are highly sensitive to seasonal changes. This observation is supported by higher values in summer, followed by sudden temperature increases of 10–15 °C, likely due to enhanced evaporation of volatile compounds at higher temperatures. Additionally, VOC concentrations are generally higher during the daytime, particularly at midday and in the afternoon, mainly when repeated pollutant accumulation occurs [48].

The results highlight the usefulness of spatial interpolation using the EBK method for identifying high-risk areas for atmospheric pollution in industrial environments, as well as the decisive role of meteorological conditions in controlling the dispersion of VOC concentrations. The EBK raster accurately reflects localized pollutant accumulations, demonstrating the robustness of the method under static meteorological conditions. The significant advantage of this approach is that the EBK method does not rely solely on a specific distribution of values, which makes it ideal for spatial modeling of environmental data characterized by high variability and nonlinear behavior. Compared to classical parametric methods, such as multiple linear regression, EBK is better able to capture local accumulation and dispersion phenomena of pollutants. EBK spatial models revealed distinct concentration gradients and pollutant hotspots, particularly under meteorological conditions limiting dispersion. Raster maps highlighted areas of elevated risk near tank vents and loading valves. The results demonstrate that the proposed EBK-based framework, with FID measurements under field conditions, provides a practical approach for high-resolution VOC mapping in complex industrial environments [49].

5.3. Compliance Evaluation

The comparison of measured VOC concentrations with legislative reference ranges should be interpreted with caution. Legislative gaps in emission limits for NMVOCs in Romanian industrial zones were noted [50]. Furthermore, discrepancies were observed between local practices and the requirements of EU Directive 2010/75/EU, indicating the need for harmonized monitoring frameworks [51]. The comparison serves as an order-of-magnitude assessment rather than a direct regulatory equivalence. Although TCOV emission concentrations measured at 1.4 m above gasoline storage tanks during the monitoring period did not exceed the BAT-associated emission levels (BAT-AELs), short-term concentration peaks remain relevant for assessing occupational and environmental risks. According to Directive 2010/75/EU and Commission Implementing Decision 2014/738/EU, BAT-AELs for VOC emissions from petroleum product storage and transfer activities generally fall within the regulatory limits specified in Section 2.6, while for high-risk substances such as benzene, the occupational exposure limit is set at one ppm or 3.54 × 10⁻³ g/Nm³ (8 h-Time-Weighted Average). Therefore, maintaining equipment tightness and implementing vapor recovery systems in accordance with Law No. 264/2017 are essential to protect both workers’ health and nearby populations, even when overall compliance with regulatory limits is achieved, as highlighted in guidance on refinery air emissions management and the cross-border impacts of oil refinery emissions [52,53]. The absence of real-time enforcement mechanisms also reduces the effectiveness of compliance control.

Our findings indicate that although the monitored zones generally comply with existing regulatory standards, the health and safety risks posed by VOC levels exceeding limits in some areas are significant. In contrast, several locations in adjacent areas of the industrial site have recorded VOC concentrations exceeding the recommended limits at various times [54]. The detected VOC levels pose significant risks to population health and safety, as well as to surrounding communities. To mitigate these risks, we recommend enhancing continuous VOC monitoring programs at the Romanian oil refinery and applying preventive measures to minimize pollutant exposure. Moreover, raising awareness and providing targeted training for workers and residents are essential components of effective risk reduction. This study highlights the critical role of robust risk management strategies in addressing VOC exposure in occupational and environmental contexts. Thus, our results emphasize the necessity of regularly updating and rigorously enforcing regulatory frameworks grounded in the latest scientific evidence to ensure adequate protection of worker health and environmental quality.

Furthermore, strengthening regulatory alignment and introducing mandatory continuous monitoring obligations could substantially improve compliance with international air quality standards. Currently, the absence of standardized methodologies for VOC dispersion assessment in Romania limits the comparability of local data with international datasets [55]. The development of integrated monitoring systems that combine satellite data with IoT-based sensors offers a promising approach to closing these compliance gaps and supporting more informed decision-making for industrial emission control.

6. Conclusions

This study demonstrates the feasibility of integrating sensor-based VOC detection with spatial modeling to evaluate emissions from petroleum-handling facilities. Results highlight the need for stricter monitoring protocols and adaptive management strategies to protect worker health and environmental quality. Our study advances air quality monitoring in industrial settings by applying EBK to spatially assess VOC concentrations. EBK integrates uncertainty and spatial variability, resulting in distribution maps that more accurately capture the complexity of pollutant dispersion.

A straightforward analysis of the recorded data indicates that VOC dispersion across the industrial platform is highly heterogeneous. It is strongly influenced by proximity to emission sources (tanks and other industrial equipment), terrain configuration, and local meteorological conditions, including wind speed, direction, and relatively elevated temperatures (around 10 °C on 14 December 2023). Among the monitored zones, Zone 3 was identified as critical, requiring continuous monitoring, emission reduction measures, and potentially remediation interventions. The analysis further shows that pollutant dispersion is primarily governed by local meteorology. Low wind speeds combined with high atmospheric pressure, as observed on 31 January 2024, led to pollutant accumulation near emission sources, whereas stronger winds and moderate pressure improved dispersion. During the warmer months of May through August 2024, VOC concentrations generally decreased compared to December and January, due to elevated temperatures and more dynamic atmospheric conditions that enhanced pollutant dispersion. Nonetheless, localized accumulations were still detected near storage tanks and operational areas, demonstrating that meteorological and operational factors can create high-risk zones even during warmer periods. This seasonal comparison underscores the combined influence of temperature, wind patterns, and industrial activities on VOC distribution throughout the year.

The EBK geostatistical method proved effective in mapping high-risk areas and identifying local concentration patterns, even when field data were unevenly distributed across the study area. VOC levels in the refinery were substantially higher than those reported in urban or rural areas, highlighting significant environmental and health risks. Seasonal and diurnal variations further modulate emission patterns. High VOC concentrations (>50 ppmV = 229 mg/m³) pose substantial risks to human health and industrial safety, particularly under conditions of accumulation or limited dispersion. While EBK demonstrated robustness in spatially representing VOC dispersion, its accuracy depends on field conditions and continuous monitoring.

The integration of real-time sensor networks could further enhance responsiveness to pollution events and improve industrial emission management. Scientifically, the study contributes to the development of a spatially explicit methodology for VOC assessment in complex industrial environments. From a socio-economic perspective, the outcomes may guide risk mitigation strategies and promote compliance with national and European environmental regulations.