AMSR2 TC-winds V8.2 released
AMSR2 TC-Winds V8.2 data release:
A new version (V8.2) of the AMSR2 winds for Tropical Cyclones (TC-winds) has been released. This new version uses a new wind algorithm developed to reduce the impact of rain contamination. The version V8.1 has been discontinued.
Analyses of the previous version AMSR-2 TC-winds V8.1 (Meissner et al, 2021) have revealed that occasionally (<10%) the winds in TCs showed significant positive bias due to residual rain/ice contamination in the wind retrievals. An examination of many such biased cases revealed that most biased cases are associated with extremely cold brightness temperatures (TB) in the 89 GHz channels (lower than 200 K) indicating areas of very cold convective towers/presence of ice, often in the early days/developing phase of the storm. Note that the 89 GHz channels were not used in the V8.1 TC-winds algorithm. These biases suggest that the original V8.1 TC-wind algorithm was not sufficiently trained for these environmental conditions.
The new V8.2 TC-wind algorithm is based on the regression of all AMSR-2 TB channels with colocated HWRF surface winds. It was trained with over 100,000 AMSR-2/HWRF colocations over 38 Tropical Cyclones in the period 2018-2023. The retrievals from this new algorithm display a significant improvement in the wind speed, both in TC core and over the outer rain bands, and in the consistency with the SMAP wind retrievals.
Additionally, some sources of Radio Frequency Interference (RFI) at 6-7 GHz were detected in the Bay of Bengal and Arabian Sea. A new RFI flag has been implemented, resulting in missing data where RFI is detected.
Notice that the new algorithm has been trained only in Tropical Cyclone conditions, therefore valid retrievals are allowed only for wind speed > 5 m/s, and SST > 20 o C.
More details about the algorithm can be found here:
Ricciardulli, L., T., Meissner, A. Manaster, 2024:“Remote Sensing of Extreme Winds with Microwave
Radiometers”, presented at EUMETSAT Meteorological Satellite Conference, Sep 30-Oct 4, 2024,
Wurzburg. Germany
Figure 1: Wind field for Typhoon Haishen in the Western Pacific on 2 September, 2020 at 4:03 UTC. Top, from left to right: AMSR2 TC-winds V8.1, HWRF resampled and interpolated at the AMSR2 time, Bias between AMSR and HWRF, and AMSR2 TB observed with the 89 GHz V-pol channel. Notice the location of the strong wind bias corresponds to unusually cold 89V TBs, which are below 200K. Bottom row: similar to top row, but with AMSR2 TC-winds V8.2, which uses a new regression algorithm including all frequency channels between 6-89 GHz. The bias in the rain-band is significantly reduced.
Figure 2: Timeseries of the intensity (maximum 10-minute sustained wind) from the SMAP (red circles, Meissner et al, 2017) and AMSR-2 (blue squares) V8.1 TC-winds during the recent typhoon Kong-rey. Also displayed are the intensities from the Best-Track data (orange, https://www.metoc.navy.mil/jtwc/jtwc.html?best-tracks) and from the satellite consensus product SatCon (black, https://tropic.ssec.wisc.edu/real-time/satcon/), which includes an uncertainty range (dashed black lines), both scaled from 1-minute to 10-minute sustained winds for comparison with the satellite intensities. Notice that SMAP is always consistent with the validation data, AMSR2 TC-winds are consistent in most instances, with the exception of some outliers due to residual rain/ice contamination in the wind retrievals, during the developing phase of the storm. The wind retrievals (purple diamonds) from the new AMSR2 TC-winds V8.2 do not display large biases, and are more consistent with the validation data and SMAP.