Spaceborne microwave sensors—such as scatterometers, Synthetic Aperture Radar (SAR), and Global Navigation Satellite System Reflectometry (GNSS-R)—are the backbone of global ocean weather monitoring. They allow scientists to track tropical cyclones, predict severe storms, and calculate ocean surface vorticity.
However, these sensors face a fundamental physical paradox: the extreme weather events they are most critical for monitoring are almost always accompanied by intense rainfall. Rain disrupts the precise relationship between microwave signals and the ocean surface, introducing significant measurement degradation and retrieval biases. Understanding why rain compromises these measurements, and how to detect it, is a major focus of modern remote sensing research.
The Dual Mechanism of Rain Interference
To understand why rain degrades satellite wind retrievals, we must look at how microwave remote sensing works. Active sensors emit a microwave signal and measure the "backscatter"—the energy reflected back from the ocean surface. Because wind creates capillary and ultra-gravity waves, a windier ocean is a rougher ocean, which yields a distinct radar backscatter profile.
Rain disrupts this physics through two distinct mechanisms:
- Atmospheric Attenuation and Scattering: As microwave signals travel through the atmosphere, liquid raindrops absorb and scatter the radar energy. This attenuates the signal on its way to the surface and on its way back to the satellite, altering the received power independently of the surface wind. This effect is highly frequency-dependent and is particularly severe for Ku-band scatterometers.
- Surface Roughness Modification: When raindrops physically strike the ocean surface, they alter its geometry. The impact creates ring waves and splashing (which increases roughness and artificially inflates backscatter), while simultaneously damping the wind-driven capillary waves (which decreases roughness).
The combination of these atmospheric and surface effects means the satellite receives a corrupted signal. As a result, wind retrieval algorithms interpret the rain-altered signal incorrectly, typically leading to an overestimation of wind speeds or severe errors in calculated wind directions.
How Rain Impacts Different Sensors
The severity and manifestation of rain-induced bias depend heavily on the sensor architecture and the retrieval algorithms being deployed.
Spaceborne Scatterometers
Scatterometers are highly sensitive to rain contamination. Research by Ye et al. evaluating the China-France Oceanography Satellite (CFOSAT) demonstrated that rain significantly degrades wind data accuracy, introducing higher biases and root-mean-square errors (RMSE) in both near-shore and offshore environments. This degradation is particularly severe for wind direction retrievals. When these corrupted wind fields are used to calculate higher-level meteorological features, the errors propagate; for instance, Wang et al. noted that rain-flagged data from the QuikSCAT satellite can introduce substantial noise into relative vorticity calculations, potentially obscuring the environmental signals that drive tropical cyclone genesis.
Synthetic Aperture Radar (SAR)
SAR provides high-resolution imaging of the ocean surface, making it valuable for capturing extreme wind events. However, Sun et al. showed that when Sentinel-1 SAR data is processed through Geophysical Model Functions (GMFs) to retrieve extreme wind speeds, intense precipitation consistently causes the models to overestimate wind speeds. Unraveling and eliminating the precise contribution of rainfall from the true wind signal under extreme cyclone conditions remains an ongoing challenge.
GNSS Reflectometry (GNSS-R)
Unlike scatterometers, GNSS-R is a passive technique that utilizes forward-scattered navigation signals (like GPS) reflecting off the ocean. Asgarimehr et al. investigated the impact of atmospheric rain attenuation on spaceborne GNSS-R wind retrievals. They found that while rain attenuation does cause an overestimation of wind speeds, the resulting biases generally fall within the acceptable uncertainty thresholds for missions like CYGNSS. This suggests that GNSS-R may offer a degree of resilience, keeping wind products usable for tropical storm monitoring even under pessimistic rain conditions.
The Confounding Role of Wind Speed
The relationship between rain interference and wind speed is highly non-linear, creating a dual-front problem for researchers.
Low Wind Speeds (< 10 m/s) High Wind Speeds (> 15 m/s)
-------------------------- ---------------------------
• Rain dominates the signal. • Wind waves dominate the signal.
• Splash effects mask minor waves. • Rain contribution is physically smaller.
• Easier to detect/flag rain. • Algorithms are hyper-sensitive to small perturbations.
• Harder to isolate and flag rain.
At low to moderate wind speeds, the physical impact of rain on radar signals is highly pronounced. As explained by Quan et al., when the ocean surface is relatively calm, the splashing and ring waves caused by rain dominate the surface geometry, easily masking the subtle wind-driven waves.
At high wind speeds (above 10–15 m/s), the massive wind-driven waves dominate the surface roughness, causing the relative physical effect of rain to diminish. While this makes the rain's physical footprint smaller, it introduces an algorithmic challenge: as Asgarimehr et al. pointed out, GNSS-R and scatterometer retrieval algorithms exhibit extreme sensitivity to minor changes in their primary observables at high wind speeds. Consequently, even a tiny, uncorrected perturbation from rain attenuation can explode into a massive bias in the final wind speed product.
Mitigation via Machine Learning Rain Flagging
Because physically modeling the simultaneous, coupled effects of wind and rain on microwave backscatter is incredibly complex, the prevailing strategy in the field is rain identification and flagging—detecting rain-contaminated pixels so they can be filtered out or treated with specialized region-specific adjustments.
Recent advancements have shifted away from relying on external, collocated meteorological data (which may suffer from timing and spatial mismatches) toward using machine learning to identify rain signatures directly from the satellite's own multi-parametric measurements.
- K-Nearest Neighbors (KNN): Peng et al. developed the HY2RRM rain identification model for the HY-2A Ku-band scatterometer. By constructing new rain-sensitive parameters from the scatterometer's backscattering coefficients, their KNN-based model successfully identified approximately 70% of rain-contaminated data with an overall accuracy of 88%, significantly improving the reliability of retrieved wind fields.
- Gradient Boosting (XGBoost): To handle the subtle variations in rain impact across different wind regimes, Quan et al. introduced a novel rain identification and intensity classification method for the CFOSAT scatterometer using an optimized XGBoost architecture (DBO-XGBoost). Their model capitalizes on the fact that rain signatures are distinct at lower wind speeds, allowing for highly accurate filtering of corrupted data without needing external validation data.
By deploying these intelligent flagging systems, remote sensing scientists can systematically cleanse satellite observations, ensuring that the wind and wave products delivered to meteorologists remain accurate—even when looking directly through the heart of a storm.
Key Foundational Work
Evaluating Impact of Rain Attenuation on Space-borne GNSS Reflectometry Wind Speeds (2019) — contributed 6 concept elements to this direction.
The relative vorticity of ocean surface winds from the QuikSCAT satellite and its effects on the geneses of tropical cyclones in the South China Sea (2007) — contributed 1 concept elements to this direction.