Figure 2. Landslide and precipitation time series for 38 selected landslides in California. (a) Cumulative displacement time series projected onto the downslope direction and separated by water year (WY). The time series for each landslide are smoothed using a moving median temporal filter. Solid lines correspond to landslides occurring within the Franciscan mélange rock unit. Colors correspond to 30-year mean water year precipitation (WY1990-WY2019) for each landslide. Inset shows the location of the selected landslides on the 30-year mean precipitation map. (b) Cumulative precipitation time series for each landslide separated by WY and colored by 30-year mean water year precipitation.
Despite the large variability in rock type throughout California (Figure S4), we found that 228 of 247 landslides occurred in host rocks broadly defined as marine and nonmarine sedimentary or metasedimentary rock units (Figure S4 and Table S3). Of these landslides, 176 (71% of total) landslides are hosted within the Franciscan complex mélange (Figure 1b), which indicates a strong lithologic control on the distribution of slow-moving landslides. Numerous recent studies have made similar findings indicating that rock type exerts a primary control on slow-moving landslides in California (Bennett, Miller, et al., 2016; Handwerger, Fielding, et al., 2019; Scheingross et al., 2013; Xu et al., 2021) and throughout the world (see refs. in Lacroix, Handwerger, et al., 2020).
Estimates of precipitation from PRISM data showed that active landslides are occurring in both wet and dry environments (Figure 1). We found more than an order of magnitude precipitation variation between the wettest active landslide in northern California (30-year mean = 2180 mm/yr) and driest landslide in southern California (30-year mean = 216 mm/yr) (Figure 1a).
3.2 Seasonal and Annual Landslide Behavior
To assess the behavior of landslides occurring in wet and dry climates, we selected a subset of 38 landslides spanning more than an order of magnitude in 30-year mean precipitation (Figure 2 and Table S4). The subset of landslides consisted of different landslide types, rock types, and occurred in different environments including coastal and inland regions, as well as developed and undeveloped areas.
Annual downslope velocities (± bootstrap uncertainty) averaged over the full study period ranged from 0.85 ± 0.56 to 9.7 ± 1.2 cm/yr (Table S4). Landslides occurring in both wet and dry regions of California exhibited seasonal kinematic changes in response to seasonal precipitation each year (Figure 2). Each landslide accelerated in response to infiltrating rainfall during the wet season before decelerating back to lower rates, or completely stopping, during the dry season.
The landslides also responded to changes in seasonal rainfall each year. We found large changes in seasonal precipitation caused large changes in displacement (Figures 2 and 3). While there is no clear relationship between velocity, precipitation, and landslide size (Figures 3 and S5), the landslides moved faster than average during the wetter WY2017 and WY2019, and slower than average during the drier WY2016 and WY2018 (Figure 3). Interestingly, we observed the largest displacement at a single landslide during the drier than average WY2018 (Figure 2). This landslide is located on the coast and is likely subject to other driving forces such as debuttressing from wave erosion at its toe (see location of “landslide 22” in Table S4). However, we cannot constrain additional driving forces with our dataset. While we did detect several other coastal landslides, most of their motion appears to be driven by rainfall (Figure 2).
Landslide sensitivity to precipitation also appeared to change during the study period. For instance, WY2016, which was the final year of the historic California drought, had both wetter and drier than average conditions in certain places, however all of the landslides were moving slower than average (velocity ratio < 1). WY2018 was drier than WY2016 across California, but a few landslides exhibited velocity ratios > 1. Our findings suggest that antecedent rainfall from the previous wet seasons, particularly the lingering impact of long-term droughts, likely play an important control on landslide behavior and sensitivity to rainfall.