Details for 'Hydro-meteorological thresholds for landslide initiation and forest operations shutdowns in the Kalum and North Coast Forest Districts.'

Category: Geomorphology and Airphoto Interpretation

Submitted by: David Melville

Date Submitted: May 29 2006

Author(s): Author : Jakob, M.

Publisher: Prepared for BC Ministry of Forest. Forest Science Program. Project No. 0275-003-03.

Description: Abstract

The objective of this research is to determine hydrometeorologic thresholds for the initiation of shallow landslides that can be used as guidelines for forest operation shutdowns. Three principal sources of data were used for this research: The first is a dataset of landslide dates and locations, which was compiled by Ministry of Forest staff and tree farm licensees. The second source is climatological and hydrological data collected for antecedent rainfall, rainfall intensity and stream discharge from Environment Canada and the Ministry of Transportation. Finally, meteorological data were gathered for 51 rainstorms that were known to have triggered landslides. The initial research design aimed at a comparison of hydro-meteorological data of landslide initiating and non-landslide initiating rainstorms to determine those variables that explain the greatest variance and combine the most significant variables in a discriminant function. However, the spatial coverage and reliability of the landslide database in combination with the vastness of the two forest districts did not allow the unambiguous designation of a storm to the non-landslide triggering group. It was not possible to demonstrate that no landslides occurred somewhere in the two forest districts during a storm where forest workers recorded the absence of landslides. This realization necessitated a change in research design, which focused on integrating rainfall and meteorological data into a four-class storm classification system Antecedent rainfall amounts prior to a landslide-triggering storm were plotted against rainfall durations up to four-weeks. Thresholds obtained at the Terrace A and Prince Rupert A rain gauges based on the 10th, 30th and 40th percentile values were low and would be exceeded too many times in any given year to be practical as forest operation shutdown guidelines. The same analysis was repeated with the 24-hr rainfall intensity and yielded similar results. Even in combination, the four-week antecedent and 24-hr rainfall intensity thresholds at the 40th percentile are exceeded 18 and 9 times in Terrace A and Prince Rupert A, respectively. This result is unsatisfactory because it implies that, despite the relatively high number of shutdowns, 40% of all observed landslides would have occurred without a shutdown being announced. Rainfall intensities for landslide triggering storms were plotted on intensity duration frequency (IDF) graphs to determine their return periods. This analysis demonstrated that most landslides occur at rainfall intensities that are exceeded several times a year at the Terrace or Prince Rupert rain gauges and rarely reach or exceed low (~5 year) return periods. In contrast to previous work in the study region, which cite 2-year storm return periods for landsliding, this study showed that landslides occur during storms with return periods of half a year or less. This analysis suggests that either rainfall intensity alone is a poor variable in explaining landslide occurrence, and/or those conditions favourable for landsliding are exceeded many times per year. Using a return period of 0.5 years for Terrace A and Prince Rupert A, 24-hr rainfall thresholds of approximately 50 mm and 70 mm respectively are obtained. It is reasonable to assume that 24-hr rainfall significantly exceeds 50 mm and 70 mm in many mountainous locations in both forest districts when such values are recorded in Prince Rupert or Terrace. This implies that rainfall will have to be measured near or at locations where forestry operations are ongoing. A synthesis of the analysis to this point suggests that given the poor spatial coverage of longterm quality automated rain gauges, a combination of antecedent rainfall and rainfall intensity to produce a threshold for forestry operation shutdowns is of insufficient quality and reliability to be practical if derived from the two rain gauges analysed. The third approach used in this study integrates meteorological indices of storms that are known to have triggered landslides. This analysis deciphered common storm characteristics that could be used to pre-determine storm impact with regard to landslide occurrence. The dominant variables used in the analysis were the 850 millibar (mb) wind speed and direction, the existence of subtropical moisture flow and a strong west to southwest jet stream at 250 mb; and the existence of a warm layer marked by high thickness values of the 500 mb to 1000 mb pressure levels. These variables were ranked subjectively and resulted in a threefold classification of storms with regard to landslide susceptibility. The scale ranges from low to high landslide hazard. Based on the three-class system a decision tree was created to facilitate the process of recommending forestry operation shutdowns and work resumption. This decision tree contains three decision levels that include forecasted rainfall, the meteorologically-derived storm classes, the four-week antecedent rainfall and the on-site 24-hr rainfall. This system allows for a flexible decision making and yields three different levels of landslide hazard advisories ranging from low to high. While local operators will have the responsibility to decide shutdowns based on this system for low and moderate advisory levels, the high landslide advisory ought to result in mandatory shutdowns. The proposed hydro-meteorological landslide warning system should be updated following significant storms that trigger landslides, and thus should improve in reliability over time.
,

PDF File: