Trends in Mean and Variability of Hydrologic Series Using Regression
Abstract
Concern for design and prediction under nonstationarity has led to research into trend detection and development of nonstationary probabilistic models. This work introduces a method for using least squares regression to test for trends in the mean and variance, which can be an appropriate tool for water managers and decision makers. Regression has the advantages of (1) ease of application, (2) application for both nonlinear and linear trends, (3) graphic visualization of trends, (4) analytically estimating the power of the trend, and (5) analytically estimating the prediction intervals related to trend extrapolation. Though this general method can be applied for a variety of hydrologic variables, we present a case based on annual maximum flows from the Mekong basin. We outline a generalized method for hypothesis testing and modeling trends for a log normal variable. We also document development of a nonstationary model to assess the impact of trends in both the mean and variance on the future magnitude and frequency of floods in the Mekong basin.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2013
- Bibcode:
- 2013AGUFM.H24B..01R
- Keywords:
-
- 1873 HYDROLOGY Uncertainty assessment;
- 1821 HYDROLOGY Floods;
- 1817 HYDROLOGY Extreme events;
- 4318 NATURAL HAZARDS Statistical analysis