Premium
Methods for the analysis of trends in streamflow response due to changes in catchment condition
Author(s) -
Letcher R. A.,
Schreider S. Yu.,
Jakeman A. J.,
Neal B. P.,
Nathan R. J.
Publication year - 2001
Publication title -
environmetrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.68
H-Index - 58
eISSN - 1099-095X
pISSN - 1180-4009
DOI - 10.1002/env.486
Subject(s) - streamflow , residual , calibration , environmental science , surface runoff , drainage basin , generalized additive model , hydrology (agriculture) , statistics , mathematics , geography , geology , ecology , cartography , geotechnical engineering , algorithm , biology
Two algorithms for analysing changes in streamflow response due to changes in land use and farm dam development, based on the Estimated Generalized Least Squares (EGLS) and the Generalized Additive Model (GAM) methods, were compared on three catchments in the Macquarie River Basin in NSW, Australia. In order to account for the influence of climatic conditions on streamflow response, the IHACRES conceptual rainfall‐runoff model was calibrated on a daily time step over two‐year periods then simulated over the entire period of concurrent rainfall, streamflow and temperature data. Residuals or differences between observed and simulated flows were calculated. The EGLS method was applied to a smoothing of the residual (daily) time series. Such residuals represent the difference between the simulated streamflow response to a fixed catchment condition (in the calibration period) and that due to the actual varying conditions throughout the record period. The GAM method was applied to quarterly aggregated residuals. The methods provided similar qualitative results for trends in residual streamflow response in each catchment for models with a good fitting performance on the calibration period in terms of a number of statistics, i.e. the coefficient of efficiency R 2 , bias and average relative parameter error (ARPE). It was found that the fit of the IHACRES model to the calibration period is critically important in determining trend values and significance. Models with well identified parameters and less correlation between rainfall and model residuals are likely to give the best results for trend analysis. Copyright © 2001 John Wiley & Sons, Ltd.