Premium
A Statistical Method for Flow Prediction, River Murray Example
Author(s) -
Veitch L. G.,
Shepherd K. J.
Publication year - 1971
Publication title -
water resources research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.863
H-Index - 217
eISSN - 1944-7973
pISSN - 0043-1397
DOI - 10.1029/wr007i006p01469
Subject(s) - principal component analysis , uncorrelated , regression , statistics , mathematics , stability (learning theory) , linear regression , flow (mathematics) , mean squared prediction error , interpolation (computer graphics) , principal component regression , regression analysis , econometrics , hydrology (agriculture) , computer science , geology , artificial intelligence , geotechnical engineering , machine learning , motion (physics) , geometry
Use of multiple linear regression enabled prediction equations to be obtained for the flow of water leaving the River Murray in the Edward River 1‐12 days forward. The predictors, being flow readings at appropriate stations taken over the previous 5 days, are highly correlated and subject to error, so a direct application of regression may give unreliable results. Therefore the correlations between the predictors were investigated by principal component analyses to overcome these difficulties. A reduced number of new variates, approximating the principal components, were used that retained the relevant information, were nearly uncorrelated, and gave interpolation in time. The regression equations were tested for both heterogeneity and stability. The results indicate a fairly reliable forecasting scheme, but suggest that it should be updated at periodic intervals.