The incidence of tuberculosis in various regions of Xinjiang (One in 100,000) is sourced from the Xinjiang Uygur Autonomous Region Center for Disease Control and Prevention. Rigorous auditing is conducted on all data reports to guarantee data integrity and authenticity. Due to Medical and health conditions, only the incidence data from 2011 to 2022 are available.

Daily statistical data on air pollutants from 2012 to 2023 were obtained from the Environmental Monitoring Station of the Xinjiang Environmental Protection Bureau and the Center for Disease Control and Prevention of the Xinjiang Uygur Autonomous Region. Some data were collected from 153 stations of the China National Environmental Monitoring Center.¹ The daily monitoring data for SO₂ (μg/m³), NO₂ (μg/m³), PM_2.5 (μg/m³), PM₁₀ (μg/m³), CO (μg/m³), and O₃ (24 h average concentration (μg/m³)) were acquired based on historical monitoring data from air quality monitoring stations across various regions. The air quality index (AQI) is determined by the highest value of the individual air quality index (IAQI) corresponding to each pollutant. The classification of each pollutant is based on the air quality index reference guide (29).

The data used in the inverse trajectory model for the second quarter (April to June, the peak incidence period of tuberculosis in the Xinjiang region) from 2015 to 2022 are derived from the Global Data Assimilation System (GDAS) of the National Centers for Environmental Prediction (NCEP). Meteorological information is obtained from the National Meteorological Science Data Sharing Service Platform.² These datasets were initially recorded starting from 2015. This study aligns with the truncation year of tuberculosis statistics and extends until 2022.

The Kriging method is a scientific method based on the theory of variogram and structural analysis, which is used for unbiased optimal estimation of regionalized variables in a limited area (30). It is one of the important components of geostatistics and is often used in epidemiological studies to construct and predict disease transmission models. The advantage of this method is that it can take into account the correlation of geographical space, that is, there may be more similar disease transmission patterns between close locations, thus improving the accuracy of prediction. In our study, we used this method to describe the distribution of the incidence of tuberculosis in 2011–2022 and the distribution characteristics of the median size of AQI in 2012–2023. Among them, the ordinary Kriging method is the most used interpolation method in the Kriging method, and its expression is:

In Equation 1, Zrepresents the estimated value of the interpolation point, while Zi denotes the measured value of the sample point i. The weight coefficient λi corresponds to the sample point i, and m indicates the number of measured sample points used in the calculation. The weight coefficients for Kriging interpolation are determined based on the semi-variance function, which is expressed as follows:

In Equation 2, Nh is the number of sampling points with distance h. In the application, it is necessary to select the appropriate semi-variance function model according to the characteristics of the test semi-variance. The selection of weights should ensure that the estimated value Z is unbiased and the estimated variance is smaller than the variance generated by other linear combinations of the observed value. The minimum variance expression of Z is:

When the following formula is satisfied, the minimum variance of Z can be obtained.

In this Equations 3, 4, m is the number of measured sample points involved in the calculation, ϕ is the Lagrange multiplier, γμiμj is the semi-variance function between the i sample point and the j sample point, and γμiμ is the semi-variance function between the sample point and the trajectory point.

Potential Source Contribution Function (PSCF) (31) is a widely used method in recent years for identifying sources of high-concentration pollutants, which helps in understanding the origins, transport pathways, and spatial distribution of atmospheric pollutants (32). In this study, this method was used to indirectly identify the potential source regions of particulate matter and the possible Mycobacterium tuberculosis it may carry. This method is based on HYSPLIT regional pollution source identification method. The PSCF at the ij grid cell is calculated, and its expression is:

In Equation 5, mij is the total number of trajectory endpoints in the same grid unit where the measured pollutant concentration exceeds the specified threshold of the pollutant, and nij is the total number of all trajectory endpoints passing through the ij grid unit. Regions with high PSCF values indicate a greater contribution to pollutants and may represent high-risk areas for tuberculosis transmission (33). In Equation 6, the PSCF value is multiplied by any weight function Wnij to eliminate the uncertainty in the pixels with smaller nij, so as to better reflect the uncertainty of the median value of these pixels. Based on this, a model was constructed to assign weight to the common spatial distribution of tuberculosis and particulate matter. The weight function is defined as follows:

Weight Potential Source Contribution Function (WPSCF) is often used to analyze the proportion of pollution trajectories in grid cells, but for areas with the same WPSCF value, it is impossible to determine the impact of pollution on the target area accurately. The concentration-weighted trajectory analysis (CWT) method can break through the above limitations and calculate the relative contribution of different source areas, which can better explore the spatial and temporal characteristics of air pollutants (20). By assuming that the particles may carry Mycobacterium tuberculosis, the potential source area of epidemic tuberculosis is proposed. The CWT calculation is as follows:

In Equation 7, Cij is the weighted average concentration of the ij grid unit, i is the trajectory index, m is the total number of trajectories, Cl is the pollutant concentration corresponding to the trajectory l passing through the network unit ij, and τijl is the time that the trajectory l stays in the grid unit ij. Because the CWT method also has uncertainty, this paper uses the same weight factor Wij as WPSCF to reduce its uncertainty, which called Weight concentration-weighted trajectory analysis (WCWT). The WCWT model was used to better understand how different meteorological conditions and elevation factors, as well as desert environments, affect the distribution of pollutants and the size of potential source areas for TB transmission.

Based on the hour-by-hour data of PM_2.5 and PM₁₀ from 2015 to 2022 in different regions of Xinjiang, the potential source areas of aerosol transmission routes and the potential source areas of tuberculosis transmission were analyzed by WPSCF and WCWT models. The potential source contribution factor and concentration weight trajectory of PM_2.5 and PM₁₀ are calculated by multiplying the potential contribution factor and concentration weight trajectory by the weight factor, respectively. Based on the calculation results of the WPSCF model, we divided the potential source areas into mild (0–0.3), moderate (0.3–0.7), and severe (0.7–1.0) pollution source areas. Combined with the spatial distribution of aerosol and tuberculosis incidence, and based on the WCWT model, we defined areas with particulate matter concentrations below 50 μg/m³ as low contribution source areas, 50–150 μg/m³ as moderate contribution source areas, and above 150 μg/m³ as high contribution source areas. This was done to explore the impact of particulate matter source areas on tuberculosis transmission (34).

Kriging interpolation and time series observations were used to analyze spatiotemporal trends and identify hot and cold spots of tuberculosis (TB) incidence from 2011 to 2023 and air quality in Xinjiang from 2010 to 2022. A predictive model for TB incidence and AQI was constructed using quadratic fitting, and model performance was evaluated using the R² metric and p-value. The significance level was set at α = 0.05.

Spatial autocorrelation was analyzed for TB incidence and air quality in Xinjiang using Moran’s I index, where positive values near 1 indicate clustering and negative values near −1 suggest dispersion. A z-score of 7.49886, greater than the critical value of 1.65, confirmed significant clustering. Hotspot and cold-spot areas were identified based on 90, 95, and 99% confidence intervals.

Local spatial autocorrelation was assessed using the Gi* statistic, where values near 1 indicate strong clustering and near 0 suggest no significant clustering. Combined analysis of Moran’s I and General G index pinpointed Southern Xinjiang as a hotspot for high-value clustering, revealing spatial heterogeneity in TB incidence and air quality. All analyses were performed using ArcGIS 10.8.2 software (ESRI Inc., United States).

To assess regional differences in TB incidence and air quality, non-parametric statistical tests were conducted. Specifically, the Kruskal-Wallis test was applied to examine the significance of differences between groups. Multiple comparisons were used to further explore and highlight significant regional disparities between different locations within Xinjiang.

Meteorological clustering and trajectory analysis were carried out to identify the pathways of pollutants and potential source areas contributing to air pollution in the region. Using GADS meteorological data and hourly PM_2.5 and PM₁₀ concentration data from 2015 to mid-2022, backward trajectory simulations were conducted to model the movement of atmospheric pollutants over a 72-h period. The simulations were performed at a height of 500 meters and at four daily intervals (00:00, 06:00, 12:00, and 18:00 UTC).

The TrajStat 1.4.9 plug-in in Meteoinfo software was used for clustering and analyzing air mass trajectories across different seasons. The air mass trajectories were categorized into five types based on their movement patterns. Additionally, PM_2.5 and PM₁₀ pollution trajectories were classified according to the secondary standard limits outlined in the “Ambient Air Quality Standard (GB3095-2012) (35)”, which sets thresholds of 75 μg/m³ for PM_2.5 and 150 μg/m³ for PM₁₀. The number of pollution trajectories for each pollutant was calculated to assess the transport pathways of these pollutants in relation to TB transmission.

Finally, based on the identified pollution pathways and source areas (WPSCF and WCWT), hypotheses were proposed regarding the potential transmission routes of TB, predicting the source areas and the contribution levels of tuberculosis in this region.

Figure 2 shows the geographical, administrative, and environmental distribution, along with site locations in northeast and southwest Xinjiang. Located in northwest China, Xinjiang is highlighted by a purple dot marking Kashgar, the focus of our analysis. Our Kriging interpolation sites include meteorological observation sites (red circles) and tuberculosis monitoring sites (black triangles). The region is divided into southern, northern, and eastern Xinjiang based on local customs.

Kashgar, in southwest Xinjiang, is surrounded by the Tianshan Mountains, Kunlun Mountains, and Pamir Plateau, and bordered by the Taklimakan Desert. Xinjiang has three main desert zones: Taklimakan, Gurbantünggüt, and Kumtag, classified as migratory, semi-fixed, and fixed deserts. Migratory Deserts: The Taklimakan Desert, in the Tarim Basin (Southern Xinjiang), is characterized by shifting sand dunes and harsh, arid conditions. It extends from Hotan to near Aksu and Korla, with highly mobile dunes. Semi-fixed Deserts: The edges of the Taklimakan near Kashgar, Aksu, and Hotan have semi-fixed dunes with some stabilizing vegetation. Parts of the Kumtag Desert, extending east toward Lop Nur (Eastern Xinjiang), are also semi-fixed. Fixed Deserts: The Gurbantünggüt Desert, in Northern Xinjiang’s Junggar Basin, features stabilized sand dunes due to sparse vegetation. The northern Taklimakan and parts of the Kumtag Desert also show characteristics of fixed deserts. Overall, Xinjiang’s deserts vary in dune mobility, with the Taklimakan dominant in the south and the Gurbantünggüt more stable in the north.

The descriptive statistics for both tuberculosis incidence and air quality index (AQI) across four regions of Xinjiang (Eastern, Northern, Southern, and overall Xinjiang) are presented in Table 1. The incidence rate of tuberculosis is “n/100,000.”

The analysis of data for the Air Quality Index (AQI) and pulmonary tuberculosis (TB) incidence did not reveal any significant correlation or regression between the two variables. Both correlation analysis (p > 0.05) (Supplementary Table S1) and regression analysis (p > 0.05) (Supplementary Table S2) indicated a weak association, suggesting that changes in AQI have a limited impact on TB incidence.

As shown in Figures 3, 4, both the Air Quality Index (AQI) and tuberculosis (TB) incidence in Xinjiang exhibited significant trends from 2012 to 2023. Both AQI and TB incidence declined from 2012 to 2018, followed by an upward trend in AQI and a sharp decline in TB incidence after 2018, mainly due to active interventions by the government and health authorities. Following the TB outbreak in 2018, disease control departments learned from the incident and implemented proactive measures in collaboration with the government and healthcare sectors. The relationship between the pandemic’s lockdown measures and the decline in tuberculosis cases lies in the strict control of movement and reduced public interaction, which limited the spread of infectious diseases, including TB. The lockdown and isolation measures likely contributed to a decrease in TB transmission, as they reduced the opportunities for individuals to be exposed to infected people, especially in crowded settings.

There were significant regional fluctuations in air quality, particularly in Southern Xinjiang, which saw high AQI levels in 2012, 2015, 2018, and 2022, indicating severe air quality deterioration. TB incidence also showed notable regional differences, with Southern Xinjiang experiencing a sharp increase from 250 cases per 100,000 in 2015 to over 450 cases per 100,000 in 2018, followed by a steady decline, returning to lower levels by 2022. Meanwhile, Eastern and Northern Xinjiang consistently maintained the lowest TB incidence rates during the study period.

The fitted curve in the figure indicates that the AQI has shown a long-term upward trend, which is associated with the industrialization and urbanization of the region. Among the models, the predictions for Southern and Northern Xinjiang are relatively accurate, with R² values of 0.45 and 0.58, respectively. The prediction model for the incidence rate of tuberculosis shows even higher accuracy, particularly in Northern Xinjiang, where R² reaches 0.90, followed by Eastern Xinjiang with an R² of 0.81. The fitting results for all regions, except Eastern Xinjiang, are statistically significant (p < 0.05).

Figure 5 illustrates that from 2010 to 2022, areas with high incidence of pulmonary tuberculosis in Xinjiang include Kashgar, Hotan, Akesu, and Kezilesu (Southern Xinjiang), while Kelamayi (Northern Xinjiang) has a low incidence. The correlation between AQI distribution (2013–2023, Figure 6) and tuberculosis incidence is evident in Southern Xinjiang, where higher pollution aligns with higher tuberculosis rates. In contrast, the relationship between AQI and low-incidence areas, like Northern Xinjiang, is less clear. In 2018, the tuberculosis incidence from high to low was Kashgar, Hotan, Akesu, and Kezilesu, mirroring the AQI pattern. From 2010 to 2022, the average tuberculosis incidence in Xinjiang was 166.66 per 100,000 people, with Kashgar reaching 377.43. The top counties-Yingjisha, Zepu, and Maigaiti (Southern Xinjiang)-had the highest rates. Severe air pollution in Kashgar, particularly from 2013 to 2023, correlates with the high incidence of tuberculosis cases. Referring to the tuberculosis trend in Kashgar from 2010 to 2022 (Figure 7), Tashkurgan (Southern Xinjiang) had the lowest prevalence, likely due to its sparse population in a desert area, while other parts of Kashgar remain highly affected.

Figure 8 shows the spatial distribution of pulmonary tuberculosis incidence in the Xinjiang Uygur Autonomous Region from 2011 to 2022. The southern Tarim Basin, surrounded by the Tianshan Mountains (including Kashgar, Hotan area, Aksu, and other areas of southern Xinjiang), has higher incidence rates, while the northern Zhungeer Basin (including kelamayi, Urumqi, Changjihuizu Autonomous Prefecture, and other areas of northern Xinjiang) shows relatively lower rates. This may be related to factors such as higher population density, relatively limited medical resources, and environmental and climatic conditions in southern Xinjiang, leading to significant regional differences in incidence rates between southern and northern Xinjiang.

Due to non-normal distribution and unequal variances in tuberculosis incidence (Levene statistic = 3.144, p = 0.03355), a Kruskal-Wallis test was performed, revealing significant regional differences (p < 0.0001). Dunn-Sidak post-hoc comparisons showed significantly higher tuberculosis incidence in Southern Xinjiang compared to Northern Xinjiang (Table 2).

Meanwhile, Figure 9 illustrates the spatial distribution of AQI, which largely aligns with the spatial distribution of tuberculosis incidence. Southern Xinjiang, including Kashgar, Hotan area, and Aksu, experiences heavier pollution, while northern Xinjiang, such as Urumqi and Hami, sees relatively lighter pollution.

Notably, Urumqi, located in the central region of Xinjiang and surrounded by the Tianshan Mountains, shows higher tuberculosis incidence at specific sites according to Kriging interpolation, although the overall spatial analysis classifies the area as blue, indicating lower incidence. This contrasts with the AQI distribution, where Urumqi is marked red due to high pollution. This discrepancy is likely due to the uneven distribution of monitoring stations. Urumqi, a valley city near the Gurbantunggut Desert (semi-fixed), located in the northern foot of the middle section of the Tianshan Mountains and the southern margin of the Junggar Basin, is hindered by Mountain range that limit pollutant dispersion, resulting in consistently poor air quality. Despite its high pollution, Urumqi and Hami, unlike Kashgar and Hotan, are not directly impacted by desert dust. The overall pattern indicates that high tuberculosis incidence in southern Xinjiang, particularly in deeper areas of the Kunlun and Tianshan mountain regions, may be influenced by windblown desert particulate matter, as suggested by the hotspot areas in Figure 2. This supports the trend of southern Xinjiang being a high-risk area for tuberculosis, while northern Xinjiang remains relatively low-risk, as reflected in both Figures 8, 9.

This phenomenon may be influenced by various factors. It is important to note that these analyses are based solely on the available data and represent hypothetical interpretations. The actual situation may be affected by multiple factors, including but not limited to local environmental policies, industrial development, population density, increased industrial activities, or insufficient public health measures. The Kruskal-Wallis test revealed no statistically significant differences in AQI across regions (χ² = 4.722, p = 0.1933), suggesting that regional variations in AQI are not significant.

The Kruskal-Wallis test revealed significant differences in tuberculosis incidence, but AQI did not show a corresponding significant difference, despite following similar regional patterns. This may be due to the uneven distribution of monitoring stations, with more stations in rural and suburban areas compared to urban centers, leading to reduced statistical power in urban regions. Tuberculosis monitoring stations are concentrated in hospital districts, while AQI monitoring stations are mostly located in suburban areas, and the scarcity of urban stations may affect AQI’s statistical results. By including all available stations rather than random sampling, we enhanced the reliability of the spatial analysis. Southern Xinjiang, particularly regions like Kashgar, Hotan, and Aksu, showed higher tuberculosis incidence and AQI values, suggesting a potential correlation between air quality and health outcomes. These areas, surrounded by mountains and the Taklamakan Desert, may experience higher levels of particulate matter, contributing to both elevated AQI and tuberculosis rates. In contrast, Northern and Eastern Xinjiang exhibit better air quality and lower tuberculosis incidence, likely due to fewer environmental stressors such as desert dust. These spatial patterns are shown in Figure 10.

Figures 11, 12 clearly show the differences between southern and northern Xinjiang through hot and cold spot analyses. The spatial hotspots for both AQI and tuberculosis incidence are located in southern Xinjiang, particularly in Kashgar, Hotan, and Aksu, which are surrounded by mountains on three sides and face the Tarim Basin and the Taklamakan Desert (migratory). The cold spots are mostly found in northern Xinjiang or other sparsely populated areas. Notably, in Figure 10, the AQI hotspots include the Urumqi area (located in the Tianshan mountain pass, in the central part of the Xinjiang map), whereas Figure 9 shows no corresponding tuberculosis hotspots in this region. This could be due to the significant spatial variability in tuberculosis distribution, where smaller Individual cases fail to influence the overall pattern of hot and cold spots.

Trajectory ① (blue line) in Figure 13 traverses from the northern junction of the West Tianshan Mountains to the southern junction of the West Kunlun Mountains, extending westward and eastward through Tajikistan to Kashgar. Trajectory ② (Green Line) extends from east to west, passing through the Aksu region, Bayinguolengmenggu Mongol Autonomous Prefecture, and Hotan area region to Kashgar in the west. Trajectory ③ (yellow line) extends from northwest to southeast, passing through Kyrgyzstan and the Kezilesukeerkezi Kyrgyz Autonomous Prefecture to Kashgar. Trajectory ④ (red line) extends from southwest to northeast, passing through Pakistan, including Islamabad, to Kashgar. Trajectory ⑤ (purple-red line) extends from northeast to southwest, passing through Kyrgyzstan and the Aksu region to Kashgar.

The Kashgar region experiences increased susceptibility to dust during the second quarter of each year. Trajectories originating from the Taklimakan Desert contribute 30.88 and 6.73% to Trajectories ② and ⑤, respectively, transporting a significant amount of dust aerosols to the Kashgar region. The proportion of PM₁₀ pollution trajectories reaches 59.73 and 65.79%, with average trajectory concentrations of 500.72 and 471.47 ug/m³, respectively, higher than those in other directions, indicating substantial influence of dust on PM₁₀ pollution levels in the second quarter. Trajectory 4 exhibits the shortest airflow transmission trajectory, primarily influenced by the southern Kunlun Mountains, leading to limited pollutant diffusion and accumulation of atmospheric pollutants, with higher average concentrations of PM_2.5 and PM₁₀. Trajectories ① (49.09%) and ③ (9.96%) cover longer transmission distances, originating mainly from Tajikistan and Pakistan, with lower pollution concentrations. Further details are reported in Table 3.

Notably, the main wind direction and the direction of topographic barriers are consistent with the previous hot and cold spot results, as well as the spatial correlation results. The meteorological cluster extending deeper into the desert has the largest proportion.

The Weighted Potential Source Contribution Function (WPSCF) and Concentration Weighted Trajectory (WCWT) analysis provide insights into the influence of potential aerosol source areas on the Kashgar region. Figures 14, 15 present the PSCF and CWT analyses for PM_2.5 and PM₁₀ concentrations during the second quarter (April–June, which coincides with the peak incidence period for tuberculosis in this region.) from 2015 to 2022.

For PM_2.5, the predominant pollution sources are located in the northeast, with severe pollution clusters concentrated in Aksu, Alar, Korla, and Urumqi, primarily originating from the Taklimakan Desert. The airflow patterns disperse moderate and mild pollution clusters toward the southwest and west. This is critical for epidemiology, as aerosols, including PM_2.5, can facilitate the suspension of pathogens such as Mycobacterium tuberculosis, allowing them to remain airborne longer due to their attachment to fine particles. The combination of dust, air pollution, and the presence of resistant pathogens increases the risk of tuberculosis (TB) transmission, especially given Kashgar’s high population density and crowded living conditions.

The trajectory analysis for PM_2.5 shows that WCWT values exceed 150 μg/m³ in regions such as Aksu, Alar, and Hotan, with peak values reaching 300 μg/m³ in some areas. This indicates a significant influence of aerosols in southwestern China on the region’s air quality and their potential role in facilitating the spread of TB.

In contrast, PM₁₀ pollution is widely dispersed but severe in specific areas. Notably, the southern region of Kashgar, particularly areas affected by the Kunlun Mountains and located on the leeward slope, experiences significant pollution accumulation due to the blockage of air masses. WCWT values for PM₁₀ exceed 400 μg/m³ in the northwest Taklimakan Desert and reach over 250 μg/m³ in the southern region of Kashgar, where the Kunlun Mountains contribute to the buildup of pollution. This geographic and meteorological pattern correlates with the high incidence of tuberculosis in Kashgar, demonstrating distinct regional characteristics.

These findings highlight the critical interaction between air pollution (specifically PM_2.5 and PM₁₀), the geographic environment, and tuberculosis incidence in Kashgar. The unique topographic and environmental factors exacerbate pollution levels, particularly in regions adjacent to major deserts and mountainous barriers, further intensifying the potential for airborne transmission of Mycobacterium tuberculosis.