Loading...
Please wait, while we are loading the content...
Similar Documents
Modélisation hydrologique probabiliste par réseaux de neurones : calibration de la distribution prédictive
| Content Provider | Semantic Scholar |
|---|---|
| Author | Boucher, Marie |
| Copyright Year | 2006 |
| Abstract | Since the last few years, there has been an increasing interest for probabilistic forecasting in the meteorological and hydrological community. This enthusiasm arises in a large extent from the possibility of uncertainty assessment that probabilistic forecasting has brought. Hydrological ensemble forecasting may be constructed using neural networks, which is a very useful tool regarding its simplicity and exécution speed. Since the mathematical description the neural model remains unknown by the user, it is not possible to evaluate the forecast's uncertainty in a strict mathematical way. Some methods like the bootstrap can be used to overcome this difficulty. The bootstrap générâtes an ensemble of forecasts at each time step. Then, the ensemble can be used to fit a law of probability in order to obtain a prédictive distribution. However, to the extent of our knowledge, the reliability of this probabilistic forecast has never been investigated. In addition, the calibration of this distribution as well as methods to correct it does not appear to hâve been investigated up to day in a hydrological context. This document présents graphical and numerical methods used to assess the quality of probabilistic hydrological forecasts obtained from neural networks. The methods employed hère apply to any type of probabilistic forecasts of continuous variable and are not restricted to neural networks forecasting. The calibration of the prédictive distribution will be evaluated and corrected, and the impact of the bootstrap will be investigated. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://corpus.ulaval.ca/jspui/bitstream/20.500.11794/18609/1/24021.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
