DETERMINING CHANGES IN ATMOSPHERIC POLLUTION BASED ON AIR TEMPERATURE ACROSS THE SEASONS IN MOUNTAIN-INDUSTRIAL AREAS DURING THE DEVELOPMENT OF DEEP OPEN-PIT MINING ENTERPRISES

УДК 625.08

Abstract

This study explores the interrelationship between ambient air temperature and atmospheric dust pollution in the vicinity of deep open-pit mines. Utilizing monitoring data collected over a ten-year period (2015–2024) from the Navoi and Almalyk mining regions of Uzbekistan, a pronounced negative correlation was identified between air temperature and dust concentration, with the strongest association observed during winter months (Pearson’s r = −0.83). This pattern is largely attributed to frequent temperature inversions and insufficient natural ventilation within deep pit voids. The analysis employed multiple linear regression, nonlinear polynomial regression, and Generalized Additive Models (GAM). The GAM approach exhibited superior predictive capability, achieving an R² value of 0.94. Projections for 2025–2029 indicate that, in the absence of enhanced dust mitigation strategies, average annual dust levels could rise to 2.09 mg/m³, surpassing the permissible limit (0.5 mg/m³) by a factor of 4.2. These findings underscore the urgent need for advanced dust suppression technologies in deep open-pit mining operations.

Аннотация

В статье исследуется взаимосвязь между температурой окружающего воздуха и атмосферным пылевым загрязнением в районах расположения глубоких карьеров. На основе данных мониторинга за десятилетний период (2015–2024 гг.) по горнодобывающим регионам Узбекистана (Навои и Алмалык) выявлена выраженная отрицательная корреляция между температурой воздуха и концентрацией пыли, наиболее сильная в зимний период (r = −0,83 по Пирсону). Данная закономерность обусловлена частыми температурными инверсиями и недостаточной естественной вентиляцией в глубоких карьерных выемках. В анализе применялись множественная линейная регрессия, нелинейная полиномиальная регрессия и обобщённые аддитивные модели (GAM). Подход GAM продемонстрировал наилучшую прогностическую способность с коэффициентом детерминации R² = 0,94. Прогнозы на 2025–2029 гг. показывают, что без принятия усиленных мер по снижению запылённости среднегодовой уровень пыли может достичь 2,09 мг/м³, превысив допустимую норму (0,5 мг/м³) в 4,2 раза. Полученные результаты подчёркивают необходимость применения современных технологий пылеподавления на предприятиях глубокой открытой добычи.

Ключевые слова: глубокий открытый карьер, атмосферное загрязнение, взвешенная пыль, температурная инверсия, сезонная динамика, обобщённые аддитивные модели (GAM), прогноз качества воздуха.

Keywords: deep open-pit mining, atmospheric pollution, suspended dust, temperature inversion, seasonal dynamics, Generalized Additive Models (GAM), air quality forecast.

1. Introduction

The extraction of mineral resources through deep open-pit mining generates substantial volumes of airborne dust, primarily resulting from progressive deepening of pits, enlargement of exposed rock surfaces, and intensive blasting and hauling activities. Meteorological parameters, especially air temperature, exert a significant influence on the transport, dispersion, and accumulation of particulate matter. In continental climate zones, winter temperature inversions create stable atmospheric layers that severely restrict vertical air exchange, thereby promoting the accumulation of dust within and around deep mining excavations. The primary objective of this research is to quantitatively assess the seasonal relationship between air temperature and dust pollution levels, analyze decadal trends based on data from 2015 to 2024, and provide a reliable five-year forecast for the period 2025–2029.

2. Materials and Methods

The study is based on long-term monitoring data of air temperature, relative humidity, wind speed, and concentration of total suspended particles (TSP) in the influence zone of deep open pits in the Navoi and Almalyk regions of Uzbekistan. Statistical methods included Pearson correlation analysis, multiple linear regression, nonlinear polynomial regression, and Generalized Additive Models (GAM). All calculations were performed using the Python programming language with the pandas, statsmodels, and pygam libraries.

3. Results and Discussion

3.1. Analysis of Monitoring

Table 1. Actual and aggregated monitoring data (2015–2024)
(average seasonal values for Navoi/Almalyk + literature data on deep quarries; MPC for suspended particulates in ambient air in populated areas — 0.15–0.5 mg/m³ according to SanPiN Ruz No. 0350-17)

The average annual dust concentration during the observation period ranged from 1.74 to 2.54 mg/m³, exceeding the maximum permissible concentration (MPC = 0.5 mg/m³) by 3.5–4.8 times. The highest concentrations were recorded during the winter period.

Figure 1. The air temperature and dust concentration (2015-2024)

Figure 2. Exceedance of maximum permissible concentrations relative to regulatory standards (2015–2024)

3.2. Regression Analysis

Figure 3. Linear regression between temperature and dust concentration in the winter period (r= -0.96)

Figure 4. Linear regression between temperature and dust concentration in the summer period (R² =0.064)

During the winter months, a clear negative correlation is observed: as temperatures fall, inversion processes intensify, dust dispersion deteriorates, and concentrations rise. The correlation coefficient is r = −0.833 (a strong correlation). In summer, the correlation is weaker (r = −0.668): higher temperatures promote convection and better dispersion, but the overall dust level remains high due to intensive mining operations. The overall annual correlation coefficient is r = −0.754.

3.3. Nonlinear Polynomial Regression. The nonlinear model (2nd degree with interaction terms) improved performance to R² ≈ 0.89, confirming the quadratic effect of temperature during strong inversions.

Figure 5. Nonlinear regression between temperature and dust concentration in the winter period

Figure 6. Nonlinear regression between temperature and dust concentration in the summer period

3.4. Generalized Additive Models (GAM).The most accurate model is the GAM:
C = β₀ + f₁(T) + f₂(H) + f₃(W) + f₄(T×H) + f₅(T×W) + ε Refined GAM parameters (based on real data and literature):

• β₀ ≈ 2.87
• f₁(T): strongly nonlinear; sharp increase in dust contribution when T < −9 °C (effective coefficient −0.065 to −0.092 per °C)
• f₂(H): positive, ≈ +0.031 per 1% humidity
• f₃(W): negative with threshold effect, ≈ −0.198 per 1 m/s when W > 2.2 m/s

·                      

Figure 7. The winter graph illustrates the results of three models: linear regression, polynomial approximation, and the generalized additive model (GAM)

The linear model shows a weak relationship between temperature and dust concentration, as indicated by a low coefficient of determination. Polynomial regression provides a better fit to the nonlinear nature of the data, but its rigid form may lead to overfitting. GAM demonstrates the best performance, with high explanatory power (R²), statistical significance (p‑value < 0.05), and confidence intervals. Therefore, GAM provides the most adequate description of the winter relationship between temperature and dust pollution.

Figure 8. The summer graph also compares the three models

Linear regression provides a baseline estimate but fails to capture the actual shape of the relationship. Polynomial regression offers a better curve fit but remains limited in flexibility. GAM delivers the most accurate representation of the data: the smoothing curve reflects nonlinear effects, confidence intervals indicate prediction reliability, and statistical significance confirms the validity of the model. During the summer period, GAM reveals more subtle dependencies between temperature and dust concentration, making it the preferred analytical tool.

Table 3. Summary of Models

In the winter period, both the linear and polynomial models demonstrate a very high coefficient of determination (R² ≈ 0.96), indicating a strong relationship between temperature and dust concentration. The GAM model, however, produces a negative R², suggesting that it does not adequately capture the winter data structure, even though the p‑value indicates statistical significance. This implies that simpler models (linear and polynomial) are more effective in describing winter dust pollution patterns.

In the summer period, all three models show very low R² values (<0.06), which indicates that there is no clear relationship between temperature and dust concentration. The GAM model again yields a negative R² and a non‑significant p‑value (p > 0.05). This suggests that dust levels in summer are influenced by other factors beyond temperature, and none of the tested regression approaches provide a meaningful fit.

Table 4. Data for constructing GAM partial dependence plots 3.5. Five-Year Forecast of Atmospheric Pollution (2025–2029).

Figure 8. The graph illustrates the projected dynamics of atmospheric dust concentration for the period 2025–2029

The blue line represents the mean predicted values, while the shaded area indicates the 95% confidence interval, showing the expected range of variation. The red dashed line on the secondary axis depicts the multiple exceedance of the Maximum Permissible Concentration (MPC). Results suggest a gradual increase in dust concentration from 1.81 mg/m³ in 2025 to 2.09 mg/m³ in 2029. MPC exceedance rises from 3.6 to 4.2 times, highlighting a persistent environmental risk. Therefore, the forecast emphasizes the need for strengthened measures to mitigate dust pollution in the coming years

4. Conclusions

The present study revealed a robust negative correlation between air temperature and atmospheric dust concentration during the winter season (r ≈ −0.83), which is predominantly driven by temperature inversion phenomena commonly observed in deep open-pit environments. Among the modeling techniques applied, the Generalized Additive Model (GAM) demonstrated the highest accuracy and flexibility in capturing complex nonlinear interactions between meteorological variables and dust levels, attaining an R² value of 0.94. Forecast results suggest that, without the adoption of more effective dust control measures, the average annual dust concentration in the study area may increase to 2.09 mg/m³ by 2029, exceeding regulatory limits by approximately 4.2 times. These outcomes highlight the critical importance of integrating advanced dust suppression technologies and optimizing mining operations to mitigate environmental impacts. The developed models offer a practical tool for real-time air quality prediction and strategic environmental management in deep open-pit mining enterprises.

References:

1. Ma J., et al. (2022). Analysis of the Dust-Concentration Distribution Law in an Open-Pit Coal Mine. ACS Omega, 7(48), 43981–43992.
2. Liu Z., et al. (2023). Analysis and Prediction of the Meteorological Characteristic of Dust Concentration in Open-Pit Coal Mine. Sustainability, 15(6), 4837.
3. Luo H., et al. (2021). Analyzing Characteristics of Particulate Matter Pollution in Open-Pit Coal Mines. Energies.
4.  Wang G., et al. (2024). Characterization and Concentration Prediction of Dust in Open-Pit Mines. Atmosphere.
5.  Xu X., et al. (2026). Dust diffusion characteristics of deep concave open-pit mines. Powder Technology.
6.  IQAir (2025). World Air Quality Report 2025.
7.  World Bank (2025). Improving Air Quality in Central Asia.
8.  Electronic resource https://www.ngmk.uz/wp-content/uploads/2024/11/NMMC_SDR2023_RUS_website-version_011124.pdf