Ensemble Forecasting of Long-Term Erosion at a Hazardous Waste Site

Abstract:

Progressive erosion at sites with long-lived hazardous waste can threaten infrastructure and potentially release contaminants into the environment. Management of such sites therefore requires forecasts of potential erosion over time periods of thousands of years. Because of the considerable uncertainty in such long-term forecasts, quantitative estimates of the uncertainty are critical. We discuss erosion projections on a 10,000-year time frame for a small watershed in New York, USA. We use an Analysis-of-Variance approach to assess and partition uncertainty arising from different sources. Uncertainty sources include: model structure, selection, and calibration; future incision along downstream rivers; future climate; and near-term anthropogenic modification of the topographic surface. Our results suggest that about 1/6 of the watershed will experience erosion exceeding 5 meters in the next 10,000 years. Projection uncertainty grows with time. Uncertainty in each source manifests in a distinct spatial pattern. Model structure uncertainty is relatively low, which reflects our ability to constrain parameter values and reduce the model set through calibration to the recent geologic past. Uncertainty arising from potential near-term topographic modification translates into uncertainty in the position of individual gullies, and is therefore important source of uncertainty in future erosion along the edges of plateau surfaces. Our results represent the first application of a comprehensive multi-model uncertainty analysis for long-term erosion forecasting.