![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
Medium-term predictions of cumulative runoff in a Mediterranean mountain river. |
VerfasserIn |
Zacarias Gulliver, Javier Herrero, María José Polo |
Konferenz |
EGU General Assembly 2016
|
Medientyp |
Artikel
|
Sprache |
en
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 18 (2016) |
Datensatznummer |
250133620
|
Publikation (Nr.) |
EGU/EGU2016-14250.pdf |
|
|
|
Zusammenfassung |
It is important to find patterns and hidden connections between data to assess the
development of decision-making tools for water management. The climate variability
of the Mediterranean environments makes it necessary the establishment of
methodological/hydrological frameworks that allow us to limit the uncertainty on the decision
for further periods within the year, and thus achieve better resource utilization. For that, a
study of different machine learning methods has been applied in a Mediterranean
mountainous basin in South Spain, by means of an ensemble classification and
regression approach to predict the river flow volumes for further periods on a quarterly
scale. The predictions are made within the same hydrological year and under two
different time schemes, after three (A-scheme) and six months (B-scheme), testing
the further periods. The study was carried out with the longest streamflow time
series registered in the basin (43 years), collected at a high mountain gauge station
(Narila, 975 metres above sea level) in the Guadalfeo River. This station is located
in the upstream part of the river (with an associated 67 km2 contributing area),
where there are not significant human alterations of the natural hydrological cycle
(withdrawals or discharges) and with a strong influence of the snow regime. The set of
selected predictors for the river water volumes includes cumulated runoff, cumulated
rainfall and the average of different Climate indexes. The results show that the
nature of future periods can be classified accurately in our study case by the methods
proposed, classifying correctly more than 90 % of the values during the testing period. |
|
|
|
|
|