3.1. Dissolved Heavy Metal Concentration Removal Effects
The concentration removal efficiency of Cu, Zn, and Cd in different media is shown in
Figure 2. The experimental results showed that 1# (local soil) had no outflow. However, considerable overflows were observed in 9 tests, and it may be related with the infiltration efficiency of the fine and dense soil. The water scour led to the formation layer with low infiltration efficiency at the bottom. So the concentration removal efficiency of Cu, Zn, and Cd in 1# (local soil) was not considered. The concentration removal efficiency of Cu and Zn fluctuated greatly. The average removal efficiencies of 2# (local soil+sand) and 3# (BSM) were 61.31% and 74.67%. The 4# (BSM+WTR), 5# (BSM+Gz), 6# (BSM+Ms), and 7# (BSM+fly ash) had higher removal efficiency of Cu than 3# (BSM). The average removal efficiencies were 76.38%, 75.29%, and 77.04%. The removal efficiencies of 8# (BSM+vermiculite), 9# (BSM+Ts), and 11# (BSM+Ms+Ts) for Cu were lower than that of 3# (BSM) but higher than that of 2# (local soil+sand), and the average removal efficiencies were 65.65%, 73.89%, and 71.62%. The removal efficiencies of 10# (BSM+coconut bran) and 12# (BSM+Gs+Ts) for Cu were relatively low. The removal efficiencies of 4# (BSM+WTR), 5# (BSM+Gz), 6# (BSM+Ms), 7# (BSM+fly ash) and 8# (BSM+vermiculite) for Zn were higher than those of 2# (local soil+sand) and 3# (BSM), and the average removal efficiencies were 74.06%, 74.93%, 79.61%, 82.27%, and 75.03%. The removal efficiencies of 9# (BSM+Ts), 10# (BSM+coconut bran), 11# (BSM+Ms+Ts), and 12# (BSM+Gs+Ts) for Zn were relatively low. The removal efficiency of Cd fluctuated slightly, which may be related to low inflow concentration. The 7# (BSM+fly ash) and 9# (BSM+Ts) had high Cd removal efficiency, and their average removal efficiencies were 70.64% and 69.45%, respectively. The removal efficiencies of 10# (BSM+coconut bran) and 12# (BSM+Gs+Ts) for Cd were low, and the rest were relatively stable.
The difference in inflow, outflow, and overflow water volumes leads to different load reduction efficiency and accumulation of heavy metals in the media. The volume reduction efficiency for the 9 tests is shown in
Figure 3. The load reduction efficiency in different media combinations is shown in
Figure 4.
The first test volume reduction efficiency of the 12 columns was generally high. This condition is due to the fact that the media inside the columns were dry before test 1. The volume reduction efficiencies in the remaining 8 tests of 1#–12# columns fluctuated within the range of 24%–46%, 8%–24%, 10%–19%, 15%–36%, 14%–42%, 18%–22%, 22%–42%, 24%–50%, 13%–40%, 24%–43%, 19%–40%, and 16%–49%, respectively. The 1# (local soil) had overflow and no outflow. The 2# (local soil+sand), 3# (BSM), and 7# (BSM+fly ash) also overflowed at different degrees when the inflow volume was large. The average volume reduction efficiencies of 4#–12# were higher than those of 2# (local soil+sand) and 3# (BSM). The 10# (BSM+coconut bran) had the highest average volume reduction efficiency of 39.25%. Low volume reduction efficiency and high overflow concentration decreased load reduction efficiency. The load reduction efficiencies of 3#–12# on Cu, Zn, and Cd were all more than 70%. The load reduction efficiencies of 4# (BSM+WTR), 5# (BSM+Gz), and 6# (BSM+Ms) on Cu can reach more than 80%. The load reduction efficiencies of 4# (BSM+WTR), 5# (BSM+Gz), 6# (BSM+Ms), 7# (BSM+fly ash) and 8# (BSM+vermiculite) on Zn reached more than 80%. The load reduction efficiencies of 4#–11# were higher than those of 2# (local soil+sand) and 3# (BSM), and the highest was that of 8# (BSM+vermiculite), which reached 77.83%.
The modified media had a better treatment effect on dissolved Cu, Zn, and Cd. The treatment effect of 4# (BSM+WTR), 5# (BSM+Gz), 6# (BSM+Ms), 9# (BSM+Ts), and 11# (BSM+Ms+Ts) in dissolved Cu, Zn, and Cd was better than the traditional BSM, which can replace the traditional BSM as the new modified media.
3.2. Relationship between Treatment Effect and Test Factors
Nine orthogonal tests were conducted to change the inflow concentration, discharge ratio, and precipitation, and the other conditions were kept constant. The influence of the three factors on the treatment effect of biological retention facilities was analyzed. Combining the load reduction efficiency of the nine tests, the best test results were screened for each of the three factors, and the corresponding factor levels were the optimal test condition. The range of load reduction efficiencies under the three factors of each column was calculated, and the best configuration level was determined. The results are shown in
Table 5 and
Figure 5.
The greater the range is, the easier the treatment effect by this factor. The treatment effects of 2#–9# and 11# (BSM+Ms+Ts) were most affected by inflow concentration, followed by discharge ratio and recurrence interval. The ranges of the three conditions for 10# (BSM+coconut bran) were close. The optimal test conditions of 1#–12# were obtained by comparing the load reduction efficiency of different factors and different levels. The optimal test conditions for 1# (local soil), 3# (BSM), 7# (BSM+fly ash), and 12# (BSM+Gs+Ts) were A1B1C1, A1B2C3, A1B1C2, and A1B2C1, respectively, and for the other 8 devices are A1B2C2. For the optimal test conditions of 12 devices, A1, B2, and C2 appeared 12, 10, and 9 times, respectively. The optimal setting levels that correspond to the three factors were 1 mg·L−1 for Cu, 1.5 mg·L−1 for Zn, 0.5 mg·L−1 for Cd, 15:1, and 2a.
3.3. Adsorption and Accumulation of Heavy Metals and Difference of Enzyme Activity in Different Media
Removal of heavy metals from stormwater runoff in bioretention is mainly caused by filtration, sorption, and plant/microorganism uptake. Most heavy metals accumulate inside the media. The study showed that 88%–97% of Cu, Zn, and Cd in stormwater are trapped in the media and 0.5%–3.3% are absorbed by plants [
28]. Cu, Zn, and Cd contents in the upper and middle layers were determined. Meanwhile, the contents of urease and protease in 1# (local soil), 4# (BSM+WTR), 5# (BSM+Gz), 7# (BSM+fly ash), and 10# (BSM+coconut bran) were analyzed by colorimetry. The content of heavy metals in media before and after treatment is shown in
Figure 6. The content change rates of heavy metal/enzyme with depth are shown in
Figure 7.
The inflow concentrations of Zn, Cu, and Cd in this test were sequentially decreased. The average load difference of 4#–12# was Zn>Cu>Cd and the average load difference of 1# (Local soil), 2# (local soil+sand), and 3# (BSM) was Cu>Zn>Cd, as determined by comparing the average load difference of three depths in
Figure 6 (4#). The accumulations of Cu, Zn, and Cd in 6# (BSM+Ms), 7# (BSM+fly ash), 8# (BSM+vermiculite), and 9# (BSM+Ts) were relatively low. This conclusion is inconsistent with the high load reduction efficiency of 6# (BSM+Ms), 7# (BSM+fly ash), and 9# (BSM+Ts), which is related to the absorbed heavy metals by plants or microorganisms. Future studies on the uptake of heavy metals by plants or microorganisms should be conducted. For vertical comparison of media in each column, the content change rate of Cd was volatile, which is related to small influent concentration. The content of Cd in a large number of columns is less than that before the test, which is due to the growth and absorption of plants and microorganisms. For Cd and Zn, the curve slightly fluctuates but does not show the law of depth. For Cu, the content change rates of the upper layer in 1# (Local soil), 5# (BSM+Gz), 6# (BSM+Ms), 7# (BSM+fly ash), 8# (BSM+vermiculite), 10# (BSM+coconut bran), and 11# (BSM+Ms+Ts)are more than that of the middle and lower layers. However, the difference between the three layers in 1# (Local soil), 2# (local soil+sand), 7# (BSM+fly ash), 8# (BSM+vermiculite), 10# (BSM+coconut bran), and 12# (BSM+Gs+Ts) is not obvious. For Zn, the content change of the middle layer in 1# (Local soil), 2# (local soil+sand), 3# (BSM), 5# (BSM+Gz), 7# (BSM+fly ash), 9# (BSM+Ts), and 11# (BSM+Ms+Ts) are the highest. However, the difference between the three layers in 1# (Local soil), 2# (local soil+sand), 3# (BSM), 4# (BSM+WTR), and 12# (BSM+Gs+Ts) is not obvious. This result is due to the fact that dissolved heavy metals can only be adsorbed by media or absorbed by plants and microbes.
Pearson correlation analysis was conducted on heavy metal content and enzyme activity in media (
Table 6). Heavy metals and enzymes were correlated. The contents of the three metals were positively correlated with urease and negatively correlated with protease. The correlation between each metal content and enzyme was low, whereas the correlation between the summation of the three heavy metals and enzyme was relatively high. The Pearson correlation coefficient of summation and urease was 0.52, and that of the summation and protease was −0.45.
Currently, no systematic study is reported on the existing forms and transformation rules of heavy metals in bioretention. Granular heavy metals are mainly accumulated in the upper media of bioretention and gradually decrease by depth [
29]. However, no obvious rule is established for the accumulation of dissolved heavy metals. This condition is because dissolved heavy metals can only be absorbed by fillers and absorbed by microbes and plants and cannot be intercepted by fillers. Metals adsorbed to organic matter in the bioretention media are not permanently immobilized. Any processes that result in leaching of the media organic matter, such as dissolution or biotransformation, can result in mobilization of the organic matter-associated metals. The release of Cu from bioretention media has been linked to the release of dissolved organic matter [
30,
31].