Landslides

The irrigation experiment was situated in the Rhone Valley near Leuk (Valais, Switzerland, 46°18′ N, 7°37′ E, 615 m a.s.l.) in a Scots pine (P. sylvestris) forest with occasional interspersed pubescent oak (Q. pubescens). Permission for the field experiment was issued by the forest service of the canton Wallis (CH) (Kantonaler Forstdienst, Kreis Oberwallis, Kantonsstrasse 275, 3902 Brig-Glis). Additionally, the permission for use of the forest for research purpose was approved by the owner of the forest, the Burgerschaft Leuk (http://www.burgerschaft-leuk.ch). The geological properties are dominated by gravel input from the Rhone river and from the Illgraben alluvial cone. A more pristine pedogenic event was the landslide from Siders. The mean annual precipitation measured in Sion (20 km) was 518 mm and the mean annual temperature 10.7°C from 2003 to 2012 [48] . The irrigation experiment had 8 plots (25 × 40 m) of which four were randomly chosen for irrigation, whereas the remaining four served as control. The plots were separated by a buffer zone of 5 m (Fig. 1). From 2003 to 2012, the irrigation system was activated in rainless nights during the vegetation period (May-October), doubling the annual rainfall amount. Water from the Rhone-channel situated along the experiment site (Fig. 1) was used for irrigation. Nutrient input through irrigation was minor: phosphate was below the detection threshold (PO₄ <0.15 kg ha⁻¹ yr⁻¹) and the input of nitrogen (2.4–3.3 kg ha⁻¹ yr⁻¹) was less than the amount that could be expected to be deposited by a doubling of rainfall (N ≤ 3.5 kg ha⁻¹ yr⁻¹) [49] (link), [50] . Three identical trees per plot with the lowest crown transparency value, which refers to trees with the highest foliation, were chosen for our study [23] (link). In the first two plots, the volumetric soil water content was monitored hourly at a soil depth of 10 cm at four different locations using time domain reflectometry (Tektronix 1502B cable tester, Beaverton, OR).

In this study, Sensitivity (SST), Specificity (SPF) and Accuracy (ACC) are popular statistical indexes used for validation of model performance. Out of these, the SST and SPF are the proportion of the landslide and non-landslide instances which are correctly predicted as landslide and non-landslide, respectively [37 (link),63 (link)]. Values of these indexes are calculated using the values extracted from confusion matrix as below: $\mathrm{SST}=\frac{\mathrm{TP}}{\mathrm{TP}+\mathrm{FN}}$
$\mathrm{SPF}=\frac{\mathrm{TN}}{\mathrm{TN}+\mathrm{FP}}$
$\mathrm{ACC}=\frac{\mathrm{TP}+\mathrm{TN}}{\mathrm{TP}+\mathrm{NT}+\mathrm{FP}+\mathrm{FN}}$
$\mathrm{Kappa}\mathrm{index}(\mathrm{K})=\frac{{\mathrm{P}}_{\mathrm{C}}-{\mathrm{P}}_{\exp }}{1-{\mathrm{P}}_{\exp }}$
${\mathrm{P}}_{\mathrm{C}}=(\mathrm{TP}+\mathrm{TN})/(\mathrm{TP}+\mathrm{TN}+\mathrm{FN}+\mathrm{FP})$
$\mathrm{Pexp}=\left(\left(\mathrm{TP}+\mathrm{FN}\right)\left(\mathrm{TP}+\mathrm{FP}\right)+\left(\mathrm{FP}+\mathrm{TN}\right)\left(\mathrm{FN}+\mathrm{TN}\right)/\sqrt{(\mathrm{TP}+\mathrm{TN}+\mathrm{FN}+\mathrm{FP})}\right)$
$\mathrm{RMSE}=\sqrt{\frac{1}{\mathrm{n}}}{\displaystyle {\sum }_{\mathrm{i}=1}^{\mathrm{n}}({\mathrm{X}}_{\mathrm{pred}.}}-{\mathrm{X}}_{\mathrm{act}.}{)}^2$
where TP (true positive) and TN (true negative) are the number of instances predicted correctly, whereas FP (false positive) and FN (false negative) refer the numbers of instances predicted erroneously. P_c is the proportion of number of pixels that have been classified correctly as landslide or non-landslide pixels. P_exp means the expected agreements. ${\mathrm{X}}_{\mathrm{pred}.}$ is the predicted values in the training dataset or the validation dataset. ${\mathrm{X}}_{\mathrm{act}.}$ is the actual (output) values from the landslide susceptibility models [20 (link)].

This study investigated the plant communities along elevational gradients representing different flooding strengths within a freshwater hydro-fluctuation belt of the TGRR in China (a subtropical mega reservoir; Figure 1). It includes 16 counties. The climate is mainly influenced by the subtropical monsoon, with average annual temperature (MAT) and precipitation (MAP) being 18.22°C and 1110 mm (Zheng et al., 2021a (link)), of which the primary annual precipitation (about 80%) happens in the rainy season, with daily temperatures ranging from 28°C to 30°C (Yi et al., 2020 (link)). Since the first full impoundment of the TGRR in 2010, the periodic inundation and drainage of the TGRR drive large hydraulic fluctuations at elevations of 145-175 m (Ren et al., 2018 (link)). The newly formed hydro-fluctuation belt occupies 344.22 km² and covers 639.38 km along the main waterway (Zheng et al., 2021a (link)). The annual human-induced inundation, including rising and falling water levels, lasts for more than eight months, resulting in lower elevations experiencing inundation for longer periods of time (Chen et al., 2020 (link)). Under extreme inundation situations, flood-tolerant perennials (e.g., Cynodon dactylon, Hemarthria compressa) and annual herbs (e.g., Xanthium strumarium, Echinochloa crusgalli) are dominant in the TGRR (Hu et al., 2022 (link)). These plants can still rapidly colonize and establish distinct community types when the inundation recedes, even if their growth is also limited by local environmental factors (e.g., soil).
Under hydrological variations of up to 30 m in the TGRR, lower elevations (145 - 160 m) are subject to longer periods of inundation per year (Figure S1 and Table S1), as well as these areas are usually disturbed by natural flooding in summer (Chen et al., 2020 (link)). Considering this scenario, the field surveys were conducted in 2019 and 2020 during the period (June-August) when TGRR’s water level reaches its minimum. We collected a total of 327 transects from 36 linked rivers in the riparian zones of the TGRR, encompassing 16 counties within a 58,000 km² landscape (Figure 1). Because the plants in our study area have a consistent growing season and are subject to similar inundation disturbances, under these conditions we can use data from two adjacent years for supplementary investigations (Moore et al., 2011 (link); Zheng et al., 2021b (link)). All plant communities from the tail to the dam area were classified according to the differences in water level and assigned to four elevation intervals at 170-175 (zone I), 165-170 (zone II), 160-165 (zone III) and 145-160 m (zone IV), representing 68, 112, 152, and 204 days of inundation continuity, respectively (Table S1; Wang et al., 2021 (link)). Within each elevation interval, we defined a transect (100 m long) parallel to the river and set up three 2 × 2 m quadrats at an interval of 50 meters (Zheng et al., 2021b (link)). Species presence/absence data were obtained for each elevation interval within 16 counties. In some counties, a sample from four elevation intervals was prevented from being collected simultaneously due to limitations in high water levels (e.g., reservoir tail areas), steep slopes, and landslides (Zheng et al., 2021b (link)). Finally, 981 quadrants were collected in this investigation, including 186, 192, 294, and 309 quadrats in zone I, zone II, zone III, and zone IV, respectively (Table S1).