Abstract. The occurrence of typhoon Herb in 1996 caused massive landslides in the Shenmu area of Taiwan. Many people died and stream and river beds were covered by meters of debris. Debris flows almost always take place in the Shenmu area during the flood season, especially in the catchment areas around Tsushui river and Aiyuzih river. Anthropogenic and natural factors that cause debris flow occurrences are complex and numerous. The precise conditions of initiation are difficult to be identified, but three factors are generally considered to be the most important ones, i.e. rainfall characteristics, geologic conditions and topography. This study proposes a simple and feasible process that combines remote sensing technology and multi-stage high-precision DTMs from aerial orthoimages and airborne LiDAR with field surveys to establish a connection between three major occurrence factors that trigger debris flows in the Shenmu area.

Key to predicting how a river will respond to a given flow rate is predicting the channel conveyance at the time of the flood (Sturm, 2010). In mountainous watersheds, channel conveyance is maintained by the capacity of the channel to move bedload. If bedload transport capacity falls below the supply rate, the channel fills with sediment and channel conveyance decreases until consummate increases in channel slope balance transport and supply rates (Lane, 1955).At gaged locations, an estimate of bedload transport capacity under historic hydrologic conditions can be determined from a survey of the reach geometry, measurement of bed surface grain size and the hydrograph (Wilcock et al., 2009). For ungauged locations or future climate scenarios, a hydrograph must be approximated. If the hydrograph is modeled from precipitation derived from observation networks like the US National Weather Service Cooperative Observer Network (NWS, 2020) and Livneh et al. (2013) data set, or a future climate meteorology data set such as the Salathé et al. (2014) data set, the precipitation may largely be recorded as daily average values.The frequency of precipitation observations necessary to accurately model floods was described decades ago (Bras, 1979;Eagleson & Shack, 1966;Singh, 1997) but time series of daily average precipitation are still commonly used to model hydrologic response to climate change (Dan et al., 2012;Shrestha et al., 2012), with precipitation assumed to fall at a constant rate over the day for hourly hydrologic simulations. As the understanding of hydrologic processes has improved, detailed physical models have confirmed the necessity of accurate temporal representation of precipitation for flood prediction at the basin outlet (Paschalis et al., 2014). Furthermore, over geologic timescales, precipitation variability has been shown to control erosion rates and channel

Channel wall scour may have debuttressed a deep-seated landslide located in a small, mountainous drainage in the Clearwater River Watershed in the western Olympic Mountains, Washington State. Debris flows and high peak flows, both of which can cause channel wall scour, may have been caused by upstream tree harvests. Through examination of stream flow, debris flow and precipitation records relative to the tree harvest and deep-seated landslide activity records, tree harvest effects on the deep-seated landslide are clarified. Following harvest on the landslide, evidence of deep seated landslide activity is undetectable until the trees upstream of the landslide are harvested. In total, three periods of landslide activity are observed. High stream flow (a 25-year event)coincides with one period of activity; however, the magnitude of flow events larger than a 1.1-year event are unaffected by tree harvests. High precipitation and snow melt (water input) events coincide with two of the landslide events but at the time of landslide activity, evapotranspiraiton rates of the plantation trees may have been nearly equal to that of the original forest. In contrast, debris flows, which coincide with all periods of deep-seated landslide activity, dramatically increase following harvest in the headwaters despite below average annual maximum 1-day precipitation and no change in annual maximum 30-day precipitation. Tree harvests in the headwaters of the watershed appear to have caused the increase in debris flow frequency which in turn triggered landslide activity.iii

No abstract