Kernel distributed residual function in a revised multiple order autoregressive model and its applications in hydrology. Issue 12 (10th September 2018)

• Record Type: Journal Article
• Authors: Ilich, Nesa; Gharib, Amr; Davies, Evan G. R.
• Is Part Of: Hydrological sciences journal. Volume 63:Issue 12(2018)
• Page Start: 1745
• Page End: 1758
• Publication Date: 2018-09-10
• Subjects: regression -- autoregressive models -- residual function -- missing data -- hydrologic time series; Hydrology -- Periodicals; 551.4805
• DOI: 10.1080/02626667.2018.1541090
• Languages: English
• ISSNs: 0262-6667

Abstract

ABSTRACT: This paper describes a new approach to fill missing data in hydrologic series. Based on a multiple-order autoregressive model, our algorithm represents the random term with an empirical distribution function that includes different parameters for the low, medium and high ranges of the modelled hydrologic variable. The algorithm involves a corrective mechanism that preserves the original statistical distribution of the series that are filled, while also eliminating the possibility of obtaining negative values for low flows. The algorithm requires multiple correlated hydrologic time series with sufficient data to permit accurate calculation of their statistical properties. It ensures that both the original statistical dependence among the data series and the statistical distribution functions will be preserved after the missing data had been filled. The model has been tested using 15 streamflow series in the Upper Bow River watershed in Alberta, Canada.