Prescriptive Use

Back Home Next

 

Title Page
Preface
Background
Conceptual Framework
Theory
Application
Testing
Prescriptive Use
Conclusion
References
Table3
Figure Captions
Figure1
Figure2a
Figure2b
Figure3
Figure4
Figure5
Figure6
Figure7
Figure8a
Figure8b
Figure9
Figure10
Figure11
Figure12a
Figure12b
Figure13
Figure14
Figure15
Figure16
Figure17a
Figure17b
Figure18
Figure19
Figure20
Figure21

PRESCRIPTIVE USE

There are at least four prescriptive uses to be made of the SHALSTAB model: 1) hazard mapping for public safety, 2) guiding forest practices to minimize potential for shallow landsliding and debris flows, 3) redesign of road network to reduce road failures, and 4) coarse screen ranking of watersheds to prioritize them for watershed analysis.

In all four cases, there are many practical considerations at play. A decision must be made as to how to define the boundaries between high, medium and low potential slope instability. The best approach would do the following: 1) obtain the best possible topographic data base; 2) use field observations and aerial photographs to create a map of landslide scars and locate accurately these scars on the data base; 3) use output from SHALSTAB to determine a log(q/T) value for each scar; and 4) use the number of landslides associated with different log(q/T) values to guide in the decision as to what threshold values to assign. Additionally, it can be useful to compare the results from mapped scars with that produced by a random model to insure that SHALSTAB is significantly better than random (see above comments about the random model). Plots of the cumulative percent of landslides falling below each log(q/T) value can then be constructed to aid in threshold selection.

Figure 19 shows an example from Northern California in which the cumulative percent of the drainage area falling under the different stability categories is also plotted. Here 54% of all landslide scars were found with log (q/T) values of less than -3.1; 68% had -2.8 or less, and 90% had less than -2.5. These q/T values correspond to 5% , 10% and 20% of the total watershed area, respectively.

In our experience (and as noted earlier), we find that the threshold value for which the majority (about 60 to 80%) of shallow landslide scars occur depends on the quality of the base map, but in general, a value of log(q/T) -2.5 will capture the vast majority of the scars (up to 100%). In the three small test sites reported by Montgomery and Dietrich (1994) between 83 and 100% of all scars fell below the -2.5 threshold. A study in the upper Chehalis watershed in Washington in which 629 landslides were mapped (including 470 they were road-related) found 86% of all the scars in values below -2.5 (K. Sullivan, pers. com. 1994). In the Berkeley Hills, just south of the UC campus, 84% of the 78 scars were found below -2.5. In a validation study in six Northern California forest lands, in all but one watershed between 66 and 95 % of all scars fell below the -2.5 value (Dietrich et al., in prep). In the one exception, only 45% were accounted for at this threshold value, but this amounted to just 6% of the land area of the relatively gentle topography of this landscape.

Based on our experience in order to capture greater than 60% of the landslides, for 30 m grid data, a threshold of -2.5 appears to be needed, for 10 m data (from digitized 7.5' quadrangles) a threshold of -2.8 may be adequate, and for still higher resolution data this threshold may be pushed to -3.1. Specific studies may find otherwise. For example, Pack and Tarboton (1997) accounted for 91% of their mapped landslides using a threshold log (q/T) of -3.3 for a 20 m grid.

Shallow landsliding is a natural process, hence some fraction of the mapped scars may not be related to current or past landuse practices. In addition, it may be possible to refine the definition of the threshold by noting which sites produce the largest landslide events (including the debris flow runout). Furthermore, the threshold between high and medium slope instability potential may vary with perceived importance of that risk to other values (public safety or aquatic habitat, for example). To some degree, then, the delineation of hazard thresholds for log (q/T) also becomes a policy decision, and we think the output of the SHALSTAB model provides at least an objective and rational way to discuss this policy in a way that can not be done with more intuitive, and operator-dependent landslides maps.

 

Hazard mapping

SHALSTAB may be useful in identifying high hazard areas as part of a public safety program. We have not added a runout model to the basic version of SHALSTAB described here, but for public safety this may be very important, as usually structures and their occupants are hit by debris flows that have traveled some distance along a valley. The simplest and crudest thing to do is to delineate all channel slopes less than some threshold value, say 5% and to assume that channels upslope of the 5% percent reach can convey debris flows. Then SHALSTAB can be used to identify all areas that feed into channels steeper than this threshold channel gradient. Figure 20 illustrates this approach and shows that while identifying some of the runout hazard, it is not sufficient. This is because unchanneled valleys down which debris flows can rapidly travel are not identified by this approach. All steep valleys, channeled or not, which are fed by sufficiently steep hillslopes should be considered corridors of debris flow runout.

 

 

Forest Practices

A second use is to assign forest practices according to the SHALSTAB relative potential for slope instability. If reduction of shallow landsliding is a significant concern, then the log(q/T) values can be grouped as representing high, medium and low risk for potential instability and appropriate forest practices could be assigned to each group. The high risk sites may be assigned a prescription of no timber cutting, effectively creating "leave" areas. Medium categories may also receive some limits on forest practices (avoid road construction, etc.).

Practical application of SHALSTAB in forest management prescriptions is evolving and there are some important issues to resolve, hence some flexibility is required. Field work must be done with a heighten care given to location of activity. High hazard areas while well defined on topographic maps may be difficult to delineate on the ground when the local topography differs significantly from the topographic map portrayal of it. There are practical matters, as well, of designing logging operations around high hazard areas, such as 1) dealing with the artifact that the hazards are delineated as squares when, of course, the real hazard area is never square and 2) the model will produce isolated cells of predicted local instability that are difficult to work around and may not really provide much protection as no-cut sites. These observations and the fact that the topography is often locally different that the mapped topography requires that the hazard rating from STALSTAB always be one that can be re-evaluated based on field evidence from a qualified professional.

It is our opinion that any assignment of hazard and consequent landuse prescription based on SHALSTAB have as a condition that qualified professionals can use field observations and related calculations to reassign the relative hazard rating. Sites mapped as high hazard, for example, may be not as steep or convergent as shown on the map, or may be all bedrock (no soil to fail). If a site, however has topography and a soil mantle similar to other places that have failed and is rated as a high hazard, then extensive field work should be necessary to make a compelling case that the site should have a lower hazard, and, say cutting of trees be permitted. It would not be sufficient to attempt a back analysis based on revised soil mechanics parameters (friction angle, bulk density and root strength) unless the hydrologic processes were also well documented. Based on our experience in our study site in Coos Bay, this is not a trivial matter. We welcome efforts in this area, but predict that it will be difficult to make a convincing case that a high hazard, strongly convergent, steep soil mantled site can be clear cut, for example, without significantly elevating its risk of failure. It should be noted, too, that the professional review in each site should also be able to increase the hazard rating, for example in areas where inner gorges, or small steep unchanneled valleys are not portrayed on the topographic map.

Hazard thresholds relative to forest practices in which public safety is an issue must meet an even more stringent test. At present, it is difficult to imagine the rationale for permitting forest practices on any land upslope of a dwelling or heavily used public road in which a landslide can occur and reach the feature as a debris flow.

 

Road design

A SHALSTAB map of log(q/T) values or values classified into high, medium and low hazard can be overlain with existing or anticipated roads and skid trails to delineate the possible risk of failure associated with this disturbance. Roads crossing high risk sites might be selected for elimination if that is part of the management goal. Future road layouts can be designed to avoid high risk areas. Such considerations was used as part of Louisiana-Pacific's proposed sustained yield plan in Northern California in recent years.

 

Coarse screen of watersheds

Because every grid cell in a landscape is assigned a stability category in SHALSTAB, it is convenient to compare percent of area in each category between landscapes to get a first cut estimate of the landscape-scale relative potential for shallow landsliding. Table 3 compares SHALSTAB results for six 7.5' quadrangles in the Oregon Coast Range (based on 30 m data). Five of the quadrangles have very similar values, but the sixth, Cedar Butte in the northern end of the Coast Range is steeper and has a much greater percentage of its area in the low q/T values. Figure 21 compares cumulative percentage values for 7 watersheds (ranging in size from 143 km2 for Noyo to 4.8 km2 for Caspar/Spitler) in Northern California (using 10 m data derived from digitized 7.5' quadrangles). Large differences exist, with Rockport and James having a large proportion of their watersheds with low values of q/T and Maple being predicted to be very stable in regard to shallow landsliding. These expectations match observed landslide frequencies in these watersheds: annotating each curve is the total number of landslides per km2 mapped from 1996 aerial photographs in each watershed by John Coyle, ranging from 11 for Rockport to 0.9 at Maple.

If the road network is digitized and overlain on the SHALSTAB results, then watersheds can also be rated according to the proportion of the network crossing different hazard categories.

Finally the in-unit failure potential and the road hazard rating can be combined with some rating of the habitat potential of the channel (say based on percent of channel slope less than some threshold values) or other attributes of the watershed to develop an overall rating of the system. This approach has been used by personnel at Stillwater Sciences (Berkeley) to develop what they call a Watershed Relative Risk rating system which is used to prescribe forest practices (including road management) and to prioritize watersheds for watershed analysis.

 

Copyright 1998, William Dietrich and David Montgomery
For problems or questions regarding this web contact bill@geomorph.berkeley.edu.
Last updated: November 29, 1998.