4 Discussion and Conclusion
Our study documents the first application of open-access standardized InSAR products from JPL ARIA to identify and monitor landslides across large regions. Although the coarse resolution of the ARIA InSAR product limits our ability to detect smaller landslides, it still provides valuable data that can be used to better understand landslide processes. Due to the large volume of open-access InSAR data that is currently available, and will continue to increase with time, especially with the upcoming launch of the NASA-ISRO SAR (NISAR) satellite, standardized InSAR products will become one of the primary ways to deliver InSAR data to the broader scientific community. The JPL Observational Products for End-Users from Remote Sensing Analysis (OPERA) project will soon be generating an operational high resolution displacement timeseries from Sentinel-1 and NISAR data over North America. With a spatial resolution of 30 m or better this product will be well-suited for identifying and monitoring landslides. Thus, it is important to continue to explore new approaches to analyze these InSAR products for scientific research, including use of automated or semi-automated detection and mapping techniques (e.g., Amatya et al., 2021; Milillo et al., 2021) .
Active slow-moving landslides across California occur in both wet and dry environments. Despite more than an order of magnitude difference in mean annual rainfall and factor of ~4 difference in estimated landslide thickness, these landslides exhibit similar first order behaviors including 1) seasonal and annual changes in displacement corresponding to local changes in rainfall, and 2) sensitivity to seasonal, annual, and multi-annual changes in rainfall. One explanation for climate- and size-independent sensitivity is that persistently active landslides maintain sufficiently high groundwater levels that keep them close to an acceleration threshold. Prior work has shown that active slow-moving landslides typically have high groundwater levels year round and become effectively saturated during the wet season (Finnegan et al., 2021; Iverson & Major, 1987; Malet et al., 2002; Schulz et al., 2009; 2018). The ability to maintain high groundwater levels may be a consequence of rock type-controlled critical zone (i.e., ground surface to unweathered bedrock) structure (Murphy et al., 2022). Hahm et al. (2019) showed that landslides in a wet region of northern California (mean annual rainfall ~1800 mm/yr) in the Franciscan mélange, the predominant bedrock of landslides in our inventory, have a critical zone characterized by a thin (< 3 m) seasonally unsaturated zone with low hydraulic conductivity that becomes effectively saturated after ~100 to 200 mm of seasonal rainfall. Similarly, Finnegan et al. (2021), showed that the Oak Ridge landslide, in an area of moderate rainfall in central California (mean annual rainfall ~640 mm/yr), also becomes effectively saturated after ~200 mm of seasonal rainfall. While not all landslides in our study occur in the Franciscan mélange, the other rock types are mostly marine and nonmarine sedimentary rocks that may bear similarity and thus may have a similar critical zone structure (Riebe et al., 2017). Additionally, the landslides themselves may create environments that allow water retention due to low permeability shear zones that inhibit water flow (Baum & Reid, 2000; Nereson et al., 2018). Therefore, landslides with thin seasonally unsaturated zones can often reach saturation in both the wetter and drier parts of the state. Once saturation occurs, excess precipitation should be shed as overland flow (e.g., Hahm et al., 2019), which may explain why landslides exhibit a muted response to large differences in annual rainfall across California (e.g., Murphy et al., 2022).
Another consequence of limited annual groundwater fluctuations is that slow-moving landslides display a relatively narrow range of velocities and rarely fail catastrophically. Finnegan et al. (2021) proposed that the narrow range of pore-water pressures (10-20 kPa) experienced by slow-moving landslides provides a natural limit to the stress changes that drive landslide acceleration. Yet, slow-moving landslides occasionally fail catastrophically. One example is the 2017 Mud Creek landslide in California (also in Franciscan mélange) that exhibited a minimum of 8 years of slow motion prior to rapid acceleration and collapse (Handwerger, Huang, et al., 2019). This transition from stable to unstable sliding may result from additional mechanisms that cause pore-water pressures to rise beyond the maximum from rainfall inputs. Both shear-induced contraction and longitudinal compression of landslide material can cause pore-water pressures to increase sharply (e.g., Iverson, 2005; Iverson et al., 2015), as can weakening mechanisms, such as slip localization (e.g., Handwerger, Huang, et al., 2019; Viesca and Rice, 2012). Although many mechanical-hydrological interactions contribute to the behavior of landslides, our results agree with Finnegan et al. (2021) and Murphy et al. (2022), which suggest that the volume of material that can accommodate water input exerts a primary control on the landslide sensitivity and response to precipitation.
Our study revealed that active slow-moving landslides moved seasonally during both dry and wet years and in dry and wet climates, indicating that even during dry periods and at dry landslides, there is often still sufficient water input to maintain downslope motion for many landslides. Climate models predict that rainfall in California is likely to become more seasonal (i.e., a higher proportion of rainfall delivered in December to March) (Swain, 2021) and dry to wet year extremes will become more common (Dong et al., 2019; Persad et al., 2020; Polade et al., 2017; Swain et al., 2018). Therefore, our study period may be representative of future precipitation and landslide behavioral patterns throughout California. While we currently cannot reliably predict landslide motion due to complex nonlinear relationships between precipitation, pore-water pressure, and velocity (e.g., Carey et al., 2019; Malet et al., 2002; Murphy et al., 2022), we may be able to predict relative changes in landslide velocity in response to relative changes in precipitation. Therefore, it is necessary to continue to document landslide behaviors during ‘normal’ years that may serve as baselines for comparison and prediction of future landslide behaviors.