Advanced Grid Model of Weighted Mean Temperature Based on Feedforward Neural Network Over China
Abstract Weighted mean temperature (Tm) is a key parameter in Global Navigation Satellite System meteorology. In this study, European Centre for Medium‐Range Weather Forecasts Re‐Analysis product with a spatial resolution of 0.5° × 0.5° from 1999 to 2018 was used to study the spatiotemporal behavior...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
American Geophysical Union (AGU)
2021-08-01
|
| Series: | Earth and Space Science |
| Subjects: | |
| Online Access: | https://doi.org/10.1029/2020EA001458 |
| _version_ | 1819134358647209984 |
|---|---|
| author | Mingchen Zhu Wusheng Hu Wei Sun |
| author_facet | Mingchen Zhu Wusheng Hu Wei Sun |
| author_sort | Mingchen Zhu |
| collection | DOAJ |
| description | Abstract Weighted mean temperature (Tm) is a key parameter in Global Navigation Satellite System meteorology. In this study, European Centre for Medium‐Range Weather Forecasts Re‐Analysis product with a spatial resolution of 0.5° × 0.5° from 1999 to 2018 was used to study the spatiotemporal behaviors of Tm in China. Decomposed by Fast Fourier Transformation, Tm and lapse rate (β) variations are highly latitude‐dependent and exhibit periodicities on annual, semi‐annual, and diurnal scales. Meanwhile, Tm keeps increasing at a rate of 0.25 K per decade across China. Based on these discoveries, this study build a new grid Tm model based on feedforward neural network (FNN) with a spatial resolution of 0.5° × 0.5°, known as Grid‐FNN model. FNN is applied to each grid point to compensate the residual error of the corresponding periodic functions. And the fitting accuracy at each grid is improved by the FNN algorithm. ERA‐Interim product with a spatial resolution of 0.4° × 0.4° and Radiosonde data in 2018 are used to validate the new model, and the accuracy of Grid‐FNN model is proved 8.6% and 10.9% better than GPT2w‐Tm model, respectively. The Grid‐FNN model also shows better performance than IGPT2w, GTm‐III, and GTrop model in autumn and winter and in high‐altitude regions. |
| first_indexed | 2024-12-22T10:01:55Z |
| format | Article |
| id | doaj.art-1bceb98f9d2b4d7398f08e0dd7efd5ae |
| institution | Directory Open Access Journal |
| issn | 2333-5084 |
| language | English |
| last_indexed | 2024-12-22T10:01:55Z |
| publishDate | 2021-08-01 |
| publisher | American Geophysical Union (AGU) |
| record_format | Article |
| series | Earth and Space Science |
| spelling | doaj.art-1bceb98f9d2b4d7398f08e0dd7efd5ae2022-12-21T18:30:05ZengAmerican Geophysical Union (AGU)Earth and Space Science2333-50842021-08-0188n/an/a10.1029/2020EA001458Advanced Grid Model of Weighted Mean Temperature Based on Feedforward Neural Network Over ChinaMingchen Zhu0Wusheng Hu1Wei Sun2School of Transportation Southeast University Nanjing ChinaSchool of Transportation Southeast University Nanjing ChinaSchool of Civil Engineering & Architecture Tongling University Tongling ChinaAbstract Weighted mean temperature (Tm) is a key parameter in Global Navigation Satellite System meteorology. In this study, European Centre for Medium‐Range Weather Forecasts Re‐Analysis product with a spatial resolution of 0.5° × 0.5° from 1999 to 2018 was used to study the spatiotemporal behaviors of Tm in China. Decomposed by Fast Fourier Transformation, Tm and lapse rate (β) variations are highly latitude‐dependent and exhibit periodicities on annual, semi‐annual, and diurnal scales. Meanwhile, Tm keeps increasing at a rate of 0.25 K per decade across China. Based on these discoveries, this study build a new grid Tm model based on feedforward neural network (FNN) with a spatial resolution of 0.5° × 0.5°, known as Grid‐FNN model. FNN is applied to each grid point to compensate the residual error of the corresponding periodic functions. And the fitting accuracy at each grid is improved by the FNN algorithm. ERA‐Interim product with a spatial resolution of 0.4° × 0.4° and Radiosonde data in 2018 are used to validate the new model, and the accuracy of Grid‐FNN model is proved 8.6% and 10.9% better than GPT2w‐Tm model, respectively. The Grid‐FNN model also shows better performance than IGPT2w, GTm‐III, and GTrop model in autumn and winter and in high‐altitude regions.https://doi.org/10.1029/2020EA001458weighted mean temperatureERA‐Interimradiosondefeedforward neural networkFast Fourier Transformationprecipitable water vapor |
| spellingShingle | Mingchen Zhu Wusheng Hu Wei Sun Advanced Grid Model of Weighted Mean Temperature Based on Feedforward Neural Network Over China Earth and Space Science weighted mean temperature ERA‐Interim radiosonde feedforward neural network Fast Fourier Transformation precipitable water vapor |
| title | Advanced Grid Model of Weighted Mean Temperature Based on Feedforward Neural Network Over China |
| title_full | Advanced Grid Model of Weighted Mean Temperature Based on Feedforward Neural Network Over China |
| title_fullStr | Advanced Grid Model of Weighted Mean Temperature Based on Feedforward Neural Network Over China |
| title_full_unstemmed | Advanced Grid Model of Weighted Mean Temperature Based on Feedforward Neural Network Over China |
| title_short | Advanced Grid Model of Weighted Mean Temperature Based on Feedforward Neural Network Over China |
| title_sort | advanced grid model of weighted mean temperature based on feedforward neural network over china |
| topic | weighted mean temperature ERA‐Interim radiosonde feedforward neural network Fast Fourier Transformation precipitable water vapor |
| url | https://doi.org/10.1029/2020EA001458 |
| work_keys_str_mv | AT mingchenzhu advancedgridmodelofweightedmeantemperaturebasedonfeedforwardneuralnetworkoverchina AT wushenghu advancedgridmodelofweightedmeantemperaturebasedonfeedforwardneuralnetworkoverchina AT weisun advancedgridmodelofweightedmeantemperaturebasedonfeedforwardneuralnetworkoverchina |