Hydrologic monitoring to support landslide forecasting in Sitka, Alaska
Abstract
Landslide forecasting and landslide warning have the potential to reduce loss of life and infrastructure. Recent advances in landslide forecasting at Seattle, Washington, and other regions worldwide demonstrate the value of antecedent soil moisture for improving landslide prediction. However, questions remain about how to select in-situ monitoring sites, the ideal number of sites, and the best combinations of hydrologic and atmospheric predictor variables. Our experimental landslide warning system in Sitka, Alaska, uses a logistic regression model to estimate landslide probability based on 3-hour rainfall intensity. We found that the addition of variables for longer timescales of rainfall (several hours to weeks) did not substantially improve model fit and performance. We hypothesize that rapid hydrologic response in the highly permeable volcanic-ash derived colluvial soils near Sitka limit the predictive value of antecedent rainfall conditions. To investigate this hypothesis, we present an analysis of recent soil moisture and groundwater pressure data. We compare hydrologic response at three sites installed near Sitka to evaluate the value of multiple monitoring locations and the utility of soil moisture versus water pressure data for improving landslide forecasting in this region. At all three sites we observed limited responses in the shallow soil moisture sensors and rapid hydrologic responses in piezometers and shallow wells, with peak water pressures at depths of 25-80 cm within 0-3 hours of peak rainfall, but qualitatively responses varied between sites with different sediment types. Water pressure response and recession behaviors were more sudden at a site with cobble colluvium, whereas the recession of peak pressures was slower at sites underlain by shallow glacial till deposits. These results can be used to inform the selection of optimum locations and number of hydrologic monitoring stations for future landslide warning systems.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFMNH15B..01P