Keywords

Flood event, classification algorithms, time-delay embedding, variable selection

Start Date

26-6-2018 5:00 PM

End Date

26-6-2018 7:00 PM

Abstract

Application of supervised classification to short-term predictions of hydrological events rely on data routinely collected by Conservation Authorities with high frequencies on streams and their watersheds. It implies that the quality of predictions depends on location of monitoring stations and representativeness of data sets used in the analysis. Given that application of classification algorithms requires data transformation, an attempt was made to improve the performance of these algorithms by extending the set of variables of black box models which are supplied by stream and rain gauges and include their derivatives which may carry very important information as well. The original computational scheme was based on time-delay embedding applied to data from all observation sites of a watershed. The variable selection was implemented using both hydrological knowledge and computational procedures. The computational experiments were conducted on data of various granularity and years with different hydrological characteristics. The results of the study are presented in the paper.

Stream and Session

Stream B: (Big) Data Solutions for Planning, Management, and Operation and Environmental Systems

Session B2: Hybrid modelling and innovative data analysis for integrated environmental decision support

Download
COinS
 
Jun 26th, 5:00 PM Jun 26th, 7:00 PM

Variable selection for improving predictions of hydrological events

Application of supervised classification to short-term predictions of hydrological events rely on data routinely collected by Conservation Authorities with high frequencies on streams and their watersheds. It implies that the quality of predictions depends on location of monitoring stations and representativeness of data sets used in the analysis. Given that application of classification algorithms requires data transformation, an attempt was made to improve the performance of these algorithms by extending the set of variables of black box models which are supplied by stream and rain gauges and include their derivatives which may carry very important information as well. The original computational scheme was based on time-delay embedding applied to data from all observation sites of a watershed. The variable selection was implemented using both hydrological knowledge and computational procedures. The computational experiments were conducted on data of various granularity and years with different hydrological characteristics. The results of the study are presented in the paper.