Satellite radar systems for climate parameters
The prediction of rainfall using Precipitable Water Vapour (PWV) derived from GPS signal delays has gained popularity in recent years. Rainfall, however, is dependent upon a wide range of atmospheric variables. This project proposes a deep-learning neural network called U-Net to predict rainfall for...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2024
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/176539 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | The prediction of rainfall using Precipitable Water Vapour (PWV) derived from GPS signal delays has gained popularity in recent years. Rainfall, however, is dependent upon a wide range of atmospheric variables. This project proposes a deep-learning neural network called U-Net to predict rainfall for the next 6 hours and analyse various weather parameters affecting rainfall. Different datasets of input weather parameters, including Precipitable water vapour (PWV), Relative humidity (RH), Temperature, Total Electron Content, Convergence of gradient, and its direction, are identified, and a detailed correlation study is presented. While all features are essential for classifying rainfall, only PWV, relative humidity, convergence, and its direction are noteworthy for predicting rainfall. With the use of normalised parameters for training, the employment of regularisation, and the optimal adjustment of the batch size to the UNET model, the proposed algorithm achieved approximately 80% accuracy compared to the previous model configurations. |
|---|