Test Results and Analysis

Slopes treated only with microbially induced carbonate precipitation

Rainfall erosion observation

After treatment with MICP, the surface erosion pattern of model slopes differed significantly. The results from the rainfall erosion observation under simulated rainfall at 5, 10, and 20 min are shown inFig. 3 . For untreated slopes (A2, B2, and C2), the surface erosion pattern was similar and surface erosion could be clearly observed at 5 min. At the beginning, the loess sample showed a collapsing phenomenon. After 5 min, a great amount of loess soils at the top of the slope was washed into the collection vessel, as shown in Fig. 3(a) . As shown in Fig. 3(b) , at 10 min, more loess was washed out from slopes. After rainfall for 30 min, the slope was destroyed and only a small amount of loess remained in the container, as shown in Fig. 3(c) .
After MICP treatment, the stability of slopes improved and there was no loess collapse. In contrast to untreated slopes, soil at the bottom of the slope (A1), instead of the top of slope, was washed out at 5 min, as shown in Fig. 3(d) . The reason of this loess loss might not be surface erosion, but slope collapse. Fig. 3(e) shows that at 10 min, the amount of soil loss on the upper slope was still small, but damage began to spread from the bottom of the slope. After 30 min, the slope remained significantly damaged because of the smaller volume of the mixed solution (Fig. 3(f) ).
With regard to slope B1, the damage of B1 was smaller and had better integrity at 5 min, as shown in Fig. 3(g) . After 10 min, damage spread similarly from the bottom of the slope, but the amount of soil loss on the upper slope was smaller (Fig. 3(h) ). In contrast to slope A1, at 30 min, there was still a whole part on the upper of B1, as shown in Fig. 3(i) .
Compared with slopes A1and B1, slope C1 (which was treated by 6 L/m2 mixed solution) achieved better erosion mitigation. Fig. 3(j) shows that a portion of the slope surface at the bottom of the slope was washed away within 5 min. However, after that, less loess soil was washed out from the bottom of the slope. This was because a large amount of the mixed solution remained at the bottom of the slope because of gravity, which cemented the loess particles together and moved the erosion position to the upper part of the sample, as shown in Fig. 3(k) . After 30 min of exposure to rainfall, soil loss was much less significant than in the other two MICP cases (Figs. 3(l) ).

Soil loss weight and surface strength

The accumulative soil loss weights were measured, as shown inFig. 4 . For untreated slopes A2, B2, and C2, spraying more water would not decrease the amount of soil washed out, indicating that water did not aid erosion mitigation. The percentage of accumulative soil loss weight exceeded 50% during the initial 10 min. After erosion for 50 min, the value even reached 90%. With MICP treatment, the percentage of eroded soil decreased significantly. The rate of the percentage of accumulative soil loss weight decreased with time, eventually leading to less eroded loess. The total percentages of eroded loess in A1, B1, and C1 after 50 min of exposure to rainfall were about 80%, 65%, and 47%, respectively. The curing effect was worse than that reported by Jiang et al., (2019) . The reason was that the total volume of the bacterial solution and the cementation solution used was less and the curing time was shorter. The accumulative loess loss weight in C1 was lower than in A1 and B1. The results were consistent with the rainfall simulation test. A better curing effect was contributed to a higher spraying dosage.
Several researchers used unconfined compressive strength (UCS) to evaluate the cementation effect (Qabany et al., 2012; Mortensen et al. 2011; Collins and Sitar 2009 ). However, the depths of slopes used in this study were insufficient to obtain samples for UCS. This was why surface strengths were reported instead of UCS. Ulusay and Erguler, (2012) also used surface strength as a comparative indicator.Fig. 5 shows that in response to spraying more water, the surface strength increased from 156 kPa to 327 kPa. This was because the loess sample contained a portion of clay soil, which would shrink and thus increase strength. After treatment with MICP, the surface strength further increased because of the produced calcium carbonate. The loess-slope C1 had a larger surface strength (442 kPa), which resulted from more calcium carbonate precipitation. According to previous studies, the cementing properties