A GRU-Based Model Using GNSS-PWV and Meteorological Data for Forecasting Rainfalls
As one of the most important parameters for the atmospheric water vapor contents, precipitable water vapor derived from global navigation satellite systems (GNSS) has been a valuable source of information for forecasting rainfall events in recent years due to the distinctive advantages of high accur...
Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/11078771/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | As one of the most important parameters for the atmospheric water vapor contents, precipitable water vapor derived from global navigation satellite systems (GNSS) has been a valuable source of information for forecasting rainfall events in recent years due to the distinctive advantages of high accuracy, high temporal–spatial resolution, wide coverage, and low cost of GNSS data. In this study, a new gated recurrent unit (GRU) based model for forecasting rainfall events was developed using training dataset in the nine-year period of 2010−2018 at 54 GNSS stations located in the USA. Moreover, the length of input windows and the effectiveness of various combinations of seven meteorological parameters were also investigated for the determination of the optimal ones to be used in the new model. The performance of the new model was evaluated using the test dataset in the two-year period of 2019 and 2020, and its results were also compared with those of the threshold-based model developed in our previous study. It is showed that the optimal lengths of input windows at different stations were different; thus, they need to be specifically determined for each station; the model was improved by incorporating four meteorological parameters into the input data; and the mean values of probability of detection and false alarm rate resulting from the new model at all the above 54 stations were 93% and 45%, respectively, which were significantly improved over a threshold-based model. These results suggest that the GRU-based model can effectively forecast most rainfall events due to its utilization of more meteorological data in the input data. |
|---|---|
| ISSN: | 1939-1404 2151-1535 |