TL;DR
This paper presents an optimized framework for mapping precipitable water vapor (PWV) using Sentinel-1 Interferometric Synthetic Aperture Radar (InSAR) and Global Navigation Satellite System (GNSS) data. By carefully processing InSAR data and calibrating with GNSS, the study achieves high-resolution, spatially continuous PWV maps, demonstrating robust consistency with GNSS measurements.
What Problem Does This Paper Address?
Accurate monitoring of precipitable water vapor (PWV) is crucial for weather forecasting and understanding climate change. While InSAR offers advantages like all-weather capability and high spatial resolution for PWV mapping, two key challenges exist: accurately separating the tropospheric wet delay from other atmospheric effects, and converting the differential PWV (ΔPWV) derived from InSAR into absolute, non-differential PWV. Traditional methods often struggle with spatial continuity and require synchronous atmospheric data, which is not always available. This paper aims to address these limitations by developing an optimized framework that leverages both InSAR and GNSS data for improved PWV retrieval.
Methods
The study employs a multi-stage approach:
- Small Baseline Subset (SBAS) InSAR: This technique is used to process Sentinel-1 SAR images, generating interferograms with optimized coherence by selecting pairs with small spatial and temporal baselines. This helps mitigate decorrelation issues.
- Tropospheric Delay Separation: The atmospheric phase delay in the interferograms is decomposed into hydrostatic and wet delay components.
- Hydrostatic Delay: This is simulated using ERA5 reanalysis data, and its accuracy is validated by comparing it with hydrostatic delays calculated using the Saastamoinen model and GNSS observations. A threshold is set to exclude datasets with significant deviations.
- Wet Delay: This is obtained by subtracting the hydrostatic delay from the total atmospheric delay.
- Water Vapor Conversion Factor (Π): To convert zenith wet delay (ZWD) into PWV, a precise water vapor conversion factor is needed. Instead of assuming a constant value, this study calculates a spatially continuous factor using ERA5 data (temperature, relative humidity, and geopotential) to derive the weighted average temperature (Tm). This approach is more accurate than empirical models like the Bevis model, especially for regional applications.
- GNSS Calibration: Since InSAR inherently provides differential measurements (ΔPWV), GNSS PWV data from ground-based stations is used to calibrate and convert the InSAR ΔPWV into absolute PWV. This step addresses the challenge of obtaining non-differential PWV and leverages the temporal and spatial coverage of GNSS.
Datasets
- Sentinel-1A SAR Data: A total of 29 Sentinel-1A images acquired in 2020 over Jinan, Shandong Province, China, were used for InSAR processing.
- ERA5 Reanalysis Data: This dataset provided temperature, relative humidity, and geopotential data, crucial for calculating the hydrostatic delay and the water vapor conversion factor.
- GNSS Data: Data from 10 Continuous Operational Reference System (CORS) stations in the Jinan region, provided by the Shandong Provincial Institute of Land and Resources Surveying and Mapping, were used for calibration and validation.
- SRTM-DEM: The Shuttle Radar Topography Mission Digital Elevation Model was used to remove topographic phase contributions from the interferograms.
- POD Precise Orbit Ephemerides: These were used to correct satellite orbit errors.
Key Insights and Ideas
- Importance of Hydrostatic Delay: The study highlights that hydrostatic delay cannot be disregarded in InSAR-based PWV retrieval, especially for larger InSAR swaths, as variations can lead to significant errors.
- Spatially Continuous Water Vapor Conversion Factor: Moving beyond constant or sparsely sampled conversion factors is critical for accurate, pixel-wise PWV mapping. Using ERA5 data to derive a spatially continuous factor based on weighted average temperature is a key innovation.
- GNSS as a Calibrator: GNSS data, despite its discrete nature, proves invaluable for calibrating the differential InSAR measurements, enabling the retrieval of absolute PWV and addressing the lack of synchronous atmospheric observations.
- Synergy of InSAR and GNSS: The combination of InSAR's spatial continuity and high resolution with GNSS's accuracy and temporal sampling offers a powerful approach for comprehensive atmospheric water vapor monitoring.
Key Results
- The proposed framework demonstrated a robust consistency between InSAR-derived PWV and GNSS PWV, achieving a correlation coefficient of 0.96.
- The Root Mean Square Error (RMSE) between InSAR PWV and GNSS PWV was 1.62 mm.
- The deviation between InSAR PWV and BDS PWV ranged from 0.74 mm to 2.63 mm, with an overall average deviation of less than 2.7 mm.
- The study successfully mapped the spatio-temporal distribution of PWV in Jinan, reflecting seasonal variations consistent with the region's temperate monsoon climate.
- The accuracy of the weighted average temperature (Tm) calculated using the proposed method was significantly better than the Bevis empirical model, with an absolute average deviation of 3.9 K compared to radiosonde data, versus 15.7 K for the Bevis model.
Limitations and Future Work
- Data Availability: The study acknowledges that the availability of GNSS data is crucial for calibration. While BDS data was used, the spatial distribution of CORS stations can still influence the accuracy of interpolated GNSS PWV.
- Liquid Cloud Water: The influence of liquid cloud water on InSAR was ignored in this study. Future work could investigate its impact and potentially incorporate it into the retrieval process.
- Resolution Mismatch: A difference in resolution between the high-resolution InSAR PWV and the coarser resolution ERA5 data is noted as a potential source of deviation.
- CORS Station Density: The study suggests that increasing the number of CORS stations would further reduce errors associated with the interpolation process.
- Generalizability: While validated for Jinan, further studies could assess the framework's performance in different geographical and climatic regions.
Summary updated