This project calculates and analyzes the Zenith Wet Delay (ZWD) using 5 different methods and compares them with IGS tropospheric products as reference.
| # | Method | Type | Source |
|---|---|---|---|
| 1 | ERA5 | Reanalysis | ECMWF |
| 2 | VMF3 | Empirical | TU Wien |
| 3 | Saastamoinen | Empirical | Classical Model |
| 4 | Hopfield | Empirical | Classical Model |
| 5 | GPT3 | Empirical | Global Pressure/Temperature |
| Parameter | Value |
|---|---|
| Station Name | ALBH00CAN (Victoria, Canada) |
| Latitude | 48.39°N |
| Longitude | 123.68°W |
| Height | 31.8 meters |
| Type | Coastal (5 km from Pacific Ocean) |
| Climate | Temperate Oceanic |
| Study Period | September - November 2025 (91 days) |
| Method | RMSE (mm) | Bias (mm) | MAE (mm) | Correlation |
|---|---|---|---|---|
| VMF3 | 24.50 | +6.56 | 20.64 | 0.813 |
| ERA5 | 27.84 | +6.04 | 23.13 | 0.761 |
| Saastamoinen | 35.86 | +19.86 | 29.54 | 0.496 |
| GPT3 | 86.84 | -80.63 | 80.63 | 0.532 |
| Hopfield | 115.99 | -111.00 | 111.00 | 0.496 |
- VMF3 and ERA5 show the best performance
- GPT3 and Hopfield significantly underestimate ZWD
- VMF3 has the highest correlation (0.813) with IGS
- Coastal station shows strong correlation between ZWD and temperature (R = 0.646)
- Weather event detected on 2025-09-26 with ZWD increase of 69.46 mm
pip install -r requirements.txtImportant: Due to large file sizes, raw data is not included in this repository.
| Dataset | Source | Link |
|---|---|---|
| ERA5 | CDS | https://cds.climate.copernicus.eu/ |
| IGS Troposphere | NASA CDDIS | https://cddis.nasa.gov/ |
| VMF3 | TU Wien | https://vmf.geo.tuwien.ac.at/ |
| GPT3 | TU Wien | https://vmf.geo.tuwien.ac.at/codes/gpt3_1.grd |
| COSMIC-2 | UCAR | https://cdaac-www.cosmic.ucar.edu/ |
cd code
python main_project.pycd code
python occultation_analysis.py| # | Output | Status |
|---|---|---|
| 1 | Station location map | T |
| 2 | Vertical profiles (T, P, RH) | T |
| 3 | ZWD time series (5 methods + IGS) | T |
| 4 | Difference plots vs IGS | T |
| 5 | Scatter plots vs IGS | T |
| 6 | Correlation plots | T |
| 7 | Statistical table (Bias, RMSE, MAE, Corr) | T |
| 8 | Scientific analysis of results | T |
| 9 | Weather event analysis | T |
| Method | Mean PWV (mm) | Min (mm) | Max (mm) |
|---|---|---|---|
| ERA5 | 0.20 | 0.05 | 0.34 |
| VMF3 | 0.20 | 0.07 | 0.36 |
| Saastamoinen | 0.22 | 0.17 | 0.24 |
| GPT3 | 0.06 | 0.05 | 0.07 |
Correlation between PWV and ZWD: R > 0.997
A significant weather event was detected on 2025-09-26:
| Metric | Value |
|---|---|
| Date | 2025-09-26 |
| ZWD on event day | 142.5 mm |
| ZWD increase | 69.46 mm (from previous day) |
| Event type | Sudden moisture increase (rainfall system) |
Performance during event:
| Method | RMSE (mm) | Bias (mm) | Correlation |
|---|---|---|---|
| VMF3 | 23.80 | +3.84 | 0.735 |
| ERA5 | 31.30 | +6.67 | 0.646 |
| Saastamoinen | 34.34 | +14.72 | -0.313 |
- Extracted vertical profiles (37 pressure levels: 1000 to 1 hPa)
- Computed refractivity:
N = Nd + Nw - Integrated using cubic spline method
- ZWD = 10⁻⁶ × ∫ Nw dh
- Saastamoinen: ZHD = 0.002277×P / (1 - 0.00266×cos(2φ) - 0.00000028×H)
- Hopfield: ZWD = 10⁻⁶ × Nw_surface × Hw/5
- GPT3: Harmonic coefficients interpolation
- Direct extraction from TU Wien products
- Daily time series generation
- Radio occultation profiling
- Near real-time wet profiles
- Comparison with VMF3
PWV (Precipitable Water Vapor) calculated using:
PWV = Π × ZWD
Π = [10⁻⁶ × ρ × Rv × (k3/Tm + k'₂)]⁻¹
Where:
- ρ = 1000 kg/m³ (water density)
- Rv = 461.525 J/kg/K (gas constant)
- k'₂ = 24 K/hPa
- k₃ = 3.75×10⁵ K²/hPa
- Tm = 70.2 + 0.72×Ts (mean temperature)
This project is licensed under the MIT License - see the LICENSE file for details.
Mobin Ravan
Email: Mobinravan23@gmail.com
GitHub: @mobinravan
If you find this project useful, please give it a star!