THE 1997 FLOOD: A TOOL FOR REFINING PREDICTIONS OF LANDSLIDE SEDIMENT IN THE KLAMATH MOUNTAINS, NORTHERN CALIFORNIA
The effect of human activities on landslide rates is a key issue in the management of western forests. In the Klamath Mountains, sedimentation has been identified as a key issue relative to anadromous fish habitat, and landslides contribute most of the sediment to the stream system (about 90% in the Salmon River sub-basin). On the Klamath National Forest, a simple empirical model is currently used for estimating future sediment delivery. It utilizes GIS coverages (bedrock, geomorphology, roads, etc.), and a concept of geomorphic terranes, lands expressing similar slope processes and similar landslide rates. A landslide rate is assigned to each terrane for undisturbed, harvested or burned, and roaded conditions. These rates are based on air photo/field inventories of landslides which occurred in the Salmon River basin from 1965-1975. The model is used to predict future sediment delivery by geomorphic terrane under various roading and vegetation management scenarios, assuming a 1965-1975 type climatic episode. The 1997 flood was used to test the predictive ability of this model. It was found that undisturbed land produced 34% of the volume in 1997, compared to a model prediction of 39%. For road corridors, it was 40% versus the model prediction of 39%, and for harvested or burned lands, it was 27%, versus 21%. Similarly, the increases in sedimentation rate for harvested and roaded areas over undisturbed rates measured after the 1997 flood were very similar to those predicted by the model. For example, data from 1997 revealed an increase of 4.5 times for harvested or burned land over undisturbed rates, while the model predicted 3.6 times. For road corridors, the increase was 62.7 times, compared to 58.3 predicted by the model. Landslide sediment delivery rates (cubic yards per acre per year) measured from 1997 data were 0.09 for undisturbed land, 0.39 for harvested or burned land, and 5.39 for road corridors. This compares to 0.28, 1.01, and 16.33 respectively, for modeled values. Thus, model values were 2.6 to 3.2 times higher than the rates measured for the 1997 flood.