The BETR-GLOBAL model was used to simulate the global transmission of PCBs in 288 multimedia regions ( Figure 1 ). The seven representative regions (regions 37, 57, 60, 63, 86, 94 and 133) selected in this study are mainly based on the following points: they cover different geographical regions around the world and are spatially representative; the pollution concentration levels vary greatly, covering high, medium and low pollution areas; the intensity of human activities in the region is different, which is conducive to analyzing the relationship between pollution and human activities.

In this research, the BETR-GLOBAL model was used to simulate the distribution of PCBs concentrations. The model divides the global environment into 288 regions. Simulations were conducted for seven representative regions worldwide during the period from 2001 to 2010, and the spatiotemporal evolution of PCBs concentrations in these regions was obtained. The random forest training meteorological function is used to predict the seven representative regions of the world from 2031 to 2040.

Specifically, the PCBs concentration data from 2001 to 2010, simulated by the BETR-GLOBAL model, were used as training samples. For the period of 2031-2040, meteorological functions were artificially defined, as shown in Equations (1)-(6), assuming that temperature (T), humidity (H), wind speed (W), and solar radiation (S) influence PCBs concentrations. The variation functions T(t), H(t), W(t), and S(t) were used to modify future meteorological scenarios according to Equations (1)-(4). In Equation (5), k₁, k₂, k₃ and k₄ represent the influence coefficients of temperature, humidity, wind speed, and solar radiation, respectively. F(t) denotes the combined influence factor. The predicted PCBs concentration D(t) is assumed to be linearly related to F(t), as expressed in Equation (6), where a and b are coefficients representing the sensitivity of PCBs concentration to the combined factor and the baseline concentration. A Random Forest regression model was trained using the adjusted meteorological functions, and the future trend of PCBs concentrations was predicted accordingly. Since the meteorological conditions for 2031-2040 were artificially defined, the predicted concentrations reflect possible trends under specific scenarios.

To explore the correlation of PCBs pollution between different regions, Spearman’s rank correlation coefficient was used, as shown in Equation (7). In this equation, d_i is the rank difference of the i^th regional variable, and n is the number of data points. The correlation coefficient ranges from −1 to 1, with values near 1 indicating a strong positive correlation, values near −1 indicating a strong negative correlation, and values near 0 indicating little or no correlation.

${\rm T}\left(t\right)=20+10\sin \left(\frac{2\pi t}{365}\right)$ (1)

$H\left(t\right)=50+20\cos \left(\frac{2\pi t}{365}\right)$ (2)

$W\left(t\right)=2+\sin \left(\frac{2\text{π}t}{365}\right)$ (3)

$S\left(t\right)=8+2\cos \left(\frac{2\text{π}t}{365}\right)$ (4)

$F\left(t\right)=k_1\cdot T\left(t\right)+k_2\cdot H\left(t\right)+k_3\cdot W\left(t\right)+k_4\cdot S\left(t\right)$ (5)

$D\left(t\right)=a\cdot F\left(t\right)+b$ (6)

$\rho =1-\frac{6{\displaystyle {\sum }_{i=1}^n{\text{d}}_i^2}}{n\left(n^2-1\right)}$ (7)

The data were further processed to map the spatial distribution of PCBs through BETR-GLOBAL (see Figure 2 ), which shows the distribution of PCBs concentration in each region over time from 2001 to 2010. It could be seen from Figure 2 that PCBs concentrations in 2006 and 2007 were relatively high. In 2006, the PCBs concentration reached 0.01 ng/m³ in several regions, which may be attributed to large-scale industrial production and the absence of emission limits during those periods.

The pollutant concentration data predicted for 2031-2040 using the Random Forest model are presented in Table 1 , and the predicted concentration distribution is shown in Figure 3 . It could be observed that the PCBs concentration ranges in Regions 37 and 57 exhibit relatively smooth changes, while Regions 94 and 133 show a sharp increase in PCBs concentrations in 2035 and 2036, followed by a decline beginning in 2037 and reaching a lower level by 2040.

The Spearman rank correlation coefficient matrix of the above 7 regions predicted by the BETR-GLOBAL model is shown in Table 2 . Correlation information of PCBs concentrations among the regions was provided the correlation coefficient matrix.

As could be viewed from Figure 4 and be analyzed from Table 2 , and the correlation coefficient between District 94 and 133 reaches 0.98, indicating a strong positive correlation exists between them; the correlation coefficient between district 37 and 133 is 0.63; the correlation coefficient between district 37 and 94 is 0.60, indicating a strong positive correlation exists between them.