Related Experiment Videos
Application of an ensemble technique based on singular spectrum analysis to daily rainfall forecasting
Daniela Baratta1, Giovambattista Cicioni, Francesco Masulli
1Istituto Nazionale per la Fisica della Materia, Via Dodecaneso 35, I-16146, Genova, Italy.
Summary
This study introduces a novel temporal data learning method for rainfall forecasting. The approach achieved an average Root Mean Square (RMS) error of less than 3mm for individual rainfall intensities.
Area of Science:
- Hydrology
- Data Science
- Signal Processing
Background:
- Temporal data learning is crucial for environmental forecasting.
- Previous work established a constructive methodology using embedding theorems and singular spectrum analysis (SSA).
- SSA helps mitigate signal discontinuity and enables efficient ensemble methods.
Purpose of the Study:
- To present new results applying the established temporal data learning methodology.
- To forecast individual rainfall intensities within the Tiber basin.
- To evaluate the forecasting accuracy using Root Mean Square (RMS) error.
Main Methods:
- Application of a constructive methodology for temporal data learning.
- Utilizing embedding theorems and singular spectrum analysis (SSA).
- Forecasting individual rainfall intensity series from 135 monitoring stations.
Main Results:
- Successful application of the temporal data learning approach to rainfall forecasting.
- Forecasting of individual rainfall intensity series across the Tiber basin.
- Achieved an average Root Mean Square (RMS) error below 3mm for rainfall predictions.
Conclusions:
- The proposed temporal data learning methodology is effective for rainfall intensity forecasting.
- The approach demonstrates high accuracy in predicting hydrological events.
- This method offers a valuable tool for water resource management and flood prediction.