Seasonal river discharge forecast in alpine catchments using snow map time series and support vector regression approach

The prediction of monthly mean discharge is critical for water resources management. Statistical methods applied on discharge time series are traditionally used for predicting this kind of slow response hydrological events. With this paper we present a Support Vector Regression (SVR) system able to predict monthly mean discharge considering discharge and snow cover extent (250 meters resolution obtained by MODIS images) time series as input. Additional meteorological and climatic variables are also tested as inputs for the SVR. The prediction system has been evaluated on 14 catchments in South Tyrol (Northern Italy) showing improved prediction capacity compared with the expected value of the discharge, estimated as the average over the 10 years before the start of the simulation; the latter is an estimate used in common practice for water resources management in the study region. The percentage root mean square error (RMSE%) is reduced of 11% and 6% for a prediction lag of 1 and 3 months respectively.

CALLEGARI Mattia;  MAZZOLI Paolo;  DE GREGORIO Ludovica;  NOTARNICOLA Claudia;  PETITTA Marcello;  PASOLLI Luca;  SEPPI Roberto;  PISTOCCHI Alberto; 

2015-07-10

IEEE

JRC88276

http://ieeexplore.ieee.org/Xplore/defdeny.jsp?url=http%3A%2F%2Fieeexplore.ieee.org%2Fstamp%2Fstamp.jsp%3Ftp%3D%26arnumber%3D6946379&denyReason=-134&arnumber=6946379&productsMatched=null&userType=inst,    https://publications.jrc.ec.europa.eu/repository/handle/JRC88276,   

10.1109/IGARSS.2014.6946379,