H33D-1652
Categorical forecasts for dam inflows based on flexible generalised linear models, Huites Mexico
Wednesday, 16 December 2015
Poster Hall (Moscone South)
Floris Frederik van Ogtrop1, Martin Montero2, Waldo Ojeda2, Sergio Santana2, Ivan Rivas2 and Rutger W Vervoort1, (1)University of Sydney, Sydney, Australia, (2)IMTA Mexican Institute of Water Technology, Jiutepec, Mexico
Abstract:
Developing forecasts for dam inflows was identified as a key area for development by Mexican and Australian delegates at recent bilateral talks. A simple categorical forecast using a generalised linear models was developed based on suspected teleconnections between sea surface temperatures (SSTs) and climate in Northern Mexico, and dam inflows in Huites, North West Mexico. Using a split validation, the final results show reasonable skill in predicting the inflow terciles (0-25, 25-75, and 75 – 100 percentile) for key periods identified by dam managers. A Heidke skill score of 0.35 (0 no skill and 1 perfect skill) was found for the forecast for total June-September inflow based on February-April sea surface temperatures and 0.36 for total September-December inflows based on May-July SSTs. While promising, the model had a tendency to over predict low inflows for both periods.
