4.1 Trend
At all gauges combined by altitude groups, median values of SSLtrends increase with the height (Figure 6 ). This is a piece of evidence that higher gauging stations are less exposed to a considerable reduction in suspended sediment load. On the other hand, there is a lack of suspended sediment measurements in the 2000s and 2010s in this database. Our last findings (Tsyplenkov et al. , 2021) from the Djankuat river station near Elbrus mountain (ca. 2700 m a.m.s.l.) suggested a decline in 20-37% in suspended sediment load during the 2015-2019 period. Therefore, a tendency of declining absolute trend values with the altitude should be discussed with caution.
We found that the mean annual SSD change pattern has considerably altered after 1988-1994 depending on the gauging station from the performed CUSUM analysis. Therefore, although it is hard to determine an exact transition year for the whole Caucasus region, we can assume with the high confidence level (cf. Table 3 ) that it was in the 1988-1994 period. This finding is in line with other studies on the water runoff for the Caucasus region (Rets et al. , 2018, 2020), whereas it was suggested 1986 as a transition year.
To a significant extent, calculated SSL trends correspond to water discharge trends (Rets et al. , 2020) and suspended sediment load changes (Gusarov et al. , 2021) calculated previously for the Greater Caucasus (Table 6 ). Therefore, we performed a correlation analysis using a nonparametric Spearman correlation coefficient to compare our findings with previous studies. As a result, we observed an insignificant negative correlation between the trendSSL values from our research and a change in suspended sediment load from Gusarov et al. (2021) (Spearman r = −0.8, p = 0.1).
Also, a significant negative correlation between change in mean July’s water discharge and SSL trends was observed (Spearman r = −0.7, p = 0.06). The correlation with June’s (Spearman r = −0.45, p = 0.26) and August’s (Spearman r = −0.42, p = 0.3) mean monthly discharges were lower. Since the significance level is weak, this result should be interpreted with caution. Nevertheless, it is likely that monthly runoff in July to some extent more reflects the climatic sensitivity of catchments in terms of sediment export than monthly runoff in June and August. However, a comparison of peak water discharge trends with SSL suggests that not always a drop in SSL trend results from water discharge reduction. For example, the observed positive trend in peak annual water discharge at Kam-Ol (+0.27% per year) corresponds to a decline in annual SSL at −1% per year.
### FIGURE 6 ###
### TABLE 6 ###
### FIGURE 7 ###
### FIGURE 8 ###
We did observe a negative correlation between the SSL trend values and the catchment area (Spearman r = −0.25, p = 0.17). However, further exploration of this scale dependency indicated significant differences, depending on the altitude zone. Hence, when only the low to a middle mountain (500-1000 m) catchments were considered, a highly significant positive correlation between SSL trend and catchment area was observed (Spearman r = 0.86, p < 0.0001). The catchment area explains 71% of the variation in SSL trend values of low to middle mountain (500-1000 m) catchments (Fig. 7a ). Since the observed total correlation is relatively weak and cannot be straightforwardly explained by a transparent mechanism, this result should be interpreted cautiously.
Nevertheless, likely, suspended sediment load trends are indeed scale-dependent in Northern Caucasus. The gauging station altitude most likely explains this scale-dependency since smaller catchments tend to locate higher in the mountains and be more glacierized and less cultivated (Fig. 7b-c ).
Similar results were obtained for 1000 generated sets of SSLtrend values (Section 2.6), confirming that these correlations are not attributable to calculating or measuring errors (Fig. 8 ). We found that SSL trend values of a low mountain (< 500 m) gauging stations are significantly positively correlated to altitude and glacier fraction and significantly negatively to cropland area. However, in another altitude zone (500-1000 m), the catchment area became the most valuable positive factor, while the correlation with altitude and glacier should be considered insignificant. A correlation with controlling factors in high mountain catchments was insignificant.