Journal cover Journal topic
Hydrology and Earth System Sciences An interactive open-access journal of the European Geosciences Union
https://doi.org/10.5194/hess-2017-75
© Author(s) 2017. This work is distributed under
the Creative Commons Attribution 3.0 License.
Research article
16 Mar 2017
Review status
A revision of this discussion paper is under review for the journal Hydrology and Earth System Sciences (HESS).
Residual uncertainty estimation using instance-based learning with applications to hydrologic forecasting
Omar Wani1,2,a, Joost V. L. Beckers2, Albrecht H. Weerts2,3, and Dimitri P. Solomatine1,4 1UNESCO-IHE Institute for Water Education, Delft, the Netherlands
2Deltares, Delft, the Netherlands
3Hydrology and Quantitative Water Management Group, Department of Environmental Sciences, Wageningen University, Wageningen, the Netherlands
4Water Resource Section, Delft University of Technology, Delft, the Netherlands
acurrently at: Institute of Environmental Engineering, Swiss Federal Institute of Technology (ETH), Zurich and Swiss Federal Institute of Aquatic Science and Technology (Eawag), Dübendorf, Switzerland
Abstract. A non-parametric method is applied to quantify residual uncertainty in hydrologic streamflow forecasting. This method acts as a post-processor on deterministic model forecasts and generates a residual uncertainty distribution. Based on instance-based learning, it uses a k-nearest neighbour search for similar historical hydrometeorological conditions to determine uncertainty intervals from a set of historical errors, i.e. discrepancies between past forecast and observation. The performance of this method is assessed using test cases of hydrologic forecasting in two UK rivers: Severn and Brue. Forecasts in retrospect were made and their uncertainties were estimated using kNN resampling and two alternative uncertainty estimators: Quantile Regression (QR) and Uncertainty Estimation based on local Errors and Clustering (UNEEC). Results show that kNN uncertainty estimation produces accurate and narrow uncertainty intervals with good probability coverage. The accuracy and reliability of these uncertainty intervals are at least comparable to those produced by QR and UNEEC. It is concluded that kNN uncertainty estimation is an interesting alternative to QR and UNEEC for estimating forecast uncertainty. An advantage of this method is that its concept is simple and well understood, it is relatively easy to implement and it requires little tuning.

Citation: Wani, O., Beckers, J. V. L., Weerts, A. H., and Solomatine, D. P.: Residual uncertainty estimation using instance-based learning with applications to hydrologic forecasting, Hydrol. Earth Syst. Sci. Discuss., https://doi.org/10.5194/hess-2017-75, in review, 2017.
Omar Wani et al.
Omar Wani et al.
Omar Wani et al.

Viewed

Total article views: 622 (including HTML, PDF, and XML)

HTML PDF XML Total BibTeX EndNote
494 105 23 622 5 26

Views and downloads (calculated since 16 Mar 2017)

Cumulative views and downloads (calculated since 16 Mar 2017)

Viewed (geographical distribution)

Total article views: 622 (including HTML, PDF, and XML)

Thereof 615 with geography defined and 7 with unknown origin.

Country # Views %
  • 1

Saved

Discussed

Latest update: 28 Jun 2017
Publications Copernicus
Download
Short summary
We generate uncertainty intervals for hydrologic model predictions using a simple instance-based learning scheme. Errors made by the model in some specific hydrometeorological conditions in the past are used to predict the probability distribution of its errors during forecasting. We test it for two different case studies in England. We find that this technique, even though conceptually simple and easy to implement, performs as well as some other sophisticated uncertainty estimation methods.
We generate uncertainty intervals for hydrologic model predictions using a simple instance-based...
Share