Comparative Analysis of Multiple Ensemble Forecasting Methods of Areal Rainfall in Lower Reaches of Jinsha River
- Article
- Figures
- Metrics
- Preview PDF
- Reference
- Related
- Cited by
- Materials
Abstract:
The ensembled method is beneficial in improving the accuracy and predictability of precipitation element forecasts. This paper is based on grid data and smart grid forecasts, southwestern regional numerical forecasts, ECMWF model forecasts and GRAPES model forecast data, with area rainfall as the research object, using the multiple regression method, BP neural network method, scoring weight method, weighted ensembled forecasting method and the arithmetic average method to obtain the ensembled areal rainfall forecast, and then the average absolute error, fuzzy score, correct rate, TS score, deviation analysis and other methods are used to compare and analyze the forecast effect of the lower reaches of the Jinsha River from April to October 2020. The results show that the prediction effect of the multiple regression method and the BP neural network method are generally better than those of the other ensemble methods. When considering the forecast magnitude of the area rainfall in the basin, the model and ensemble method with smaller forecast magnitude can be employed downstream. After ensemble, the deviation percentages are reduced, and the multiple regression method and the BP neural network method have a corrective effect on the models with smaller forecast magnitudes. In the forecast of whether there is precipitation, light rain and moderate rain, the multiple regression method has a better ensemble effect. In the heavy rainfall forecast, the BP neural network method has a better ensemble effect. These conclusions can provide references for future surface rainfall forecasting in the river valley.
Keywords:
Project Supported:
Clc Number:
You are thevisitors
Supported by:Beijing E-Tiller Technology Development Co., Ltd.
Supported by:Beijing E-Tiller Technology Development Co., Ltd.