Monte Carlo simulation to characterize runoff uncertainty in a changing climate

Gregory S. Karlovits

Climate change has the potential to intensify precipitation, affecting design storms that are based on historical, stationary data. This decreases the ability to accurately predict the magnitude of runoff due to extreme precipitation events, so a method for assessing the range of possibilities becomes necessary. This paper presents a framework for predicting runoff due to climate change and understanding uncertainty in the prediction. Historical and future precipitation were modeled with the Generalized Extreme Value distribution fit to the annual maximum 24-hour precipitation event for gridded data at 1/2 degree resolution over the Pacific Northwest (PNW) using the method of L-moments. The rainfall intensities for the 2, 25, 50 and 100-year storms were determined for 1915-2006 and for a number of future climate scenarios for the 2040s, projected by two emissions scenarios and ten global climate models (GCMs). To determine the range in runoff depths projected due to climate change, Monte Carlo simulation was coupled with the Variable Infiltration Capacity (VIC) hydrology model. For the Monte Carlo simulation, each GCM was weighted by its ability to re-produce 20th century precipitation and temperature over the PNW. Snowpack and soil moisture conditions were simulated for each future climate scenario and fit to a normal distribution. For each return interval, 5000 randomly-selected runoff scenarios varying emissions scenario, GCM, soil moisture and snowpack were simulated with VIC. The results of the Monte Carlo simulation show increases in runoff for the future with large uncertainty in the forecast of runoff depths. The largest source of uncertainty is from selecting emissions scenarios, which affects all other parts of the projection. The range of runoff was most sensitive to GCM selection and antecedent soil moisture. Scenarios that are warmer and wetter produced the highest runoff forecasts. The most at-risk locations in the PNW, the Puget Sound region and the Olympic Peninsula, were also subject to the largest uncertainty in projecting future runoff depths. We conclude that a probabilistic approach is favorable for assessing the large amount of uncertainty and risk involved in forecasting hydrologic fluxes and states in a changing climate.

Monte Carlo simulation to characterize runoff uncertainty in a changing climate

Files and links (1)

Abstract

Metrics

Details