Following a large continental earthquake, information on the spatial distribution of triggered landslides is required as quickly as possible for use in emergency response coordination. Synthetic Aperture Radar (SAR) methods have the potential to overcome variability in weather conditions, which often causes delays of days or weeks when mapping landslides using optical satellite imagery. Here we test landslide classifiers based on SAR coherence, which is estimated from the similarity in phase change in time between small ensembles of pixels. We test two existing SAR-coherence-based landslide classifiers against an independent inventory of landslides triggered following the Mw 7.8 Gorkha, Nepal earthquake, and present and test a new method, which uses a classifier based on coherence calculated from ensembles of neighbouring pixels and coherence calculated from a more dispersed ensemble of ‘sibling’ pixels. Using Receiver Operating Characteristic analysis, we show that none of these three SAR-coherence-based landslide classification methods are suitable for mapping individual landslides on a pixel-by-pixel basis. However, they show potential in generating lower-resolution density maps, which are used by emergency responders following an earthquake to coordinate large-scale operations and identify priority areas. The new method we present outperforms existing methods when tested at these lower resolutions, suggesting that it may be able to provide useful and rapid information on landslide distributions following major continental earthquakes.
27 November 2020 in Natural Hazards and Earth System Sciences
Emergency responders require information on the distribution of triggered landslides within 2 weeks of an earthquake or storm. Useable satellite radar imagery is acquired within days of any such event worldwide. Recently, several landslide detection methods that use these data have been developed, but testing of these methods has been limited in each case to a single event and satellite sensor. Here we systematically test five methods using ALOS-2 and Sentinel-1 data across four triggering earthquakes. The best-performing method was dependent on the satellite sensor. For three of our four case study events, an initial ALOS-2 image was acquired within 2 weeks, and with these data, co-event coherence loss (CECL) is the best-performing method. Using a single post-event Sentinel-1 image, the best-performing method was the boxcar–sibling (Bx–S) method. We also present three new methods which incorporate a second post-event image. While the waiting time for this second post-event image is disadvantageous for emergency response, these methods perform more consistently and on average 10 % better across event and sensor type than the boxcar–sibling and CECL methods. Thus, our results demonstrate that useful landslide density information can be generated on the timescale of emergency response and allow us to make recommendations on the best method based on the availability and latency of post-event radar data.
Monsoon rainfall triggers hundreds of landslides across Nepal every year, causing significant hazard and mass wasting. Annual inventories of these landslides have been mapped using multi‐spectral satellite images, but these images are obscured by cloud cover during the monsoon, making it impossible to use them to constrain landslide timing. We employ recently developed techniques to derive individual timings from Sentinel‐1 for 579 landslides in 2015, 2017, 2018, and 2019 in Nepal. We use this new timing information alongside satellite rainfall data to identify spatio‐temporal clusters of landslides and associate these with periods of particularly intense rainfall. We also observed that during the 2015 monsoon, many landslides failed earlier and in dryer conditions than in 2017–2019. We use physical models to demonstrate how this requires a temporary loss of hillslope strength following the Mw 7.8 Gorkha Earthquake sequence and suggest a modeled cohesion loss in the range 1–3 kPa.
17 August 2022 in Natural Hazards and Earth System Sciences
Heavy-rainfall events in mountainous areas trigger destructive landslides, which pose a risk to people and infrastructure and significantly affect the landscape. Landslide locations are commonly mapped using optical satellite imagery, but in some regions their timings are often poorly constrained due to persistent cloud cover. Physical and empirical models that provide insights into the processes behind the triggered landsliding require information on both the spatial extent and the timing of landslides. Here we demonstrate that Sentinel-1 synthetic aperture radar amplitude time series can be used to constrain landslide timing to within a few days and present four techniques to accomplish this based on time series of (i) the difference in amplitude between the landslide and its surroundings, (ii) the spatial variability in amplitude between pixels within the landslide, and (iii) geometric shadows and (iv) geometric bright spots cast within the landslide. We test these techniques on three inventories of landslides of known timing, covering various settings and triggers, and demonstrate that a method combining them allows 20 %–30 % of landslides to be timed with an accuracy of 80 %. Application of this method could provide an insight into landslide timings throughout events such as the Indian summer monsoon, which triggers large numbers of landslides every year and has until now been limited to annual-scale analysis.
7 October 2021 in Natural Hazards and Earth System Sciences
Information on the spatial distribution of triggered landslides following an earthquake is invaluable to emergency responders. Manual mapping using optical satellite imagery, which is currently the most common method of generating this landslide information, is extremely time consuming and can be disrupted by cloud cover. Empirical models of landslide probability and landslide detection with satellite radar data are two alternative methods of generating information on triggered landslides that overcome these limitations. Here we assess the potential of a combined approach, in which we generate an empirical model of the landslides using data available immediately following the earthquake using the random forest technique and then progressively add landslide indicators derived from Sentinel-1 and ALOS-2 satellite radar data to this model in the order they were acquired following the earthquake. We use three large case study earthquakes and test two model types: first, a model that is trained on a small part of the study area and used to predict the remainder of the landslides and, second, a preliminary global model that is trained on the landslide data from two earthquakes and used to predict the third. We assess model performance using receiver operating characteristic analysis and r2, and we find that the addition of the radar data can considerably improve model performance and robustness within 2 weeks of the earthquake. In particular, we observed a large improvement in model performance when the first ALOS-2 image was added and recommend that these data or similar data from other L-band radar satellites be routinely incorporated in future empirical models.
Earthquake-triggered landslides can be mapped using optical satellite images, but assessing how they evolve during earthquake sequences is difficult due to cloud cover in these data. This information is crucial for understanding their triggering conditions. Here we use Sentinel-1 amplitude and a new, coherence-based method to characterise the evolution of rapid landslides during an earthquake sequence that occurred over a 23 d period in 2018 in Lombok, Indonesia. While most new landslides were triggered during the largest earthquake in the sequence on 5 August, we also identified landslide activity associated with other, lower magnitude earthquakes on 28 July, 9 and 19 August, with around half of the landslides studied active in more than one earthquake. In particular, many landslides triggered by the 5 August earthquake were then reactivated later in the sequence. These reactivations were triggered by accelerations as weak as 0.1 g, while new failures generally did not occur below 0.15 g, suggesting a post-seismic weakening effect driven by the landslides themselves rather than general landscape weakening. We also identified an example where possible precursory motion detected during the first earthquake in the sequence was later followed by larger scale failure. Overall, we demonstrate that, although they are not sensitive to all landslides and are more likely to detect larger events, Sentinel-1 amplitude and coherence are valuable tools to study how landslide hazard and mass wasting evolve during sequences of triggers.