Towards a Correction of ASCAT Ocean Measurements for Rain Effects

W. Lin1, M. Portabella1, A. Stoffelen2, A. Verhoef2, D. Weissman3, T. Johnson3, and J. Wolf3

(1)Institut de Ciències del Mar (ICM-CSIC), Barcelona, Spain (wenminglin@icm.csic.es),

(2) Royal Netherlands Meteorological Institute (KNMI), De Bilt, The Netherlands,

(3) School of Engineering and Applied Science, Hofstra University, Hempstead, New York, U.S.

2. Rain/Wind model The rain-splasing-induced bias is estimated from one year (2008) of ASCAT 25-km data collocated

with ECMWF winds and TRMM/TMI rain data. The TMI data have been obtained from the Remote

Sensing Systems Web site (http://www.ssmi.com). The collocation criteria for TMI rain data are less

than 30-min time and 0.25° spatial distance from the ASCAT measurement.

Figure 1 illustrates the mean ASCAT backscatter as a function of wind speed for WVC number

11. The bias is estimated for each WVC number in log space.

Fig. 1 The mean ASCAT σ0 as a function of ECMWF wind speed for WVC number 11, (a): fore-

beam; (b) mid-beam; (c) aft-beam.

(b) (a) (c)

3. Experimental results The proposed model is applied to 6 months (January 2013- June 2013) of collocated ASCAT 25-km measurements and TMI/ECMWF data. Due to the lack of heavy rain

data, the correction is only applied to the categories with TMI-RR [0.1 1) mm/hr, [1 3) mm/hr, and [3 6) mm/hr.

Figure 2 presents the ASCAT wind speed quality w.r.t. ECMWF under the three mentioned rain conditions. The upper row shows the statistics without correction, while

the lower row shows those with rain correction. Figure 3 presents the same statistics as Fig. 2 but for wind direction.

(b-1) (a-1) (c-1)

(b-2) (a-2) (c-2)

(b-1) (a-1) (c-1)

(b-2) (a-2) (c-2)

Fig. 2 Two dimensional histograms of ASCAT wind speed versus ECMWF wind

speed for the three RR categories (a): [0.1 1) mm/hr, (b): [1 3) mm/hr, and (c): [3

6) mm/hr. The upper row shows the statistics without rain correction, while the

lower row shows the ones with rain correction.

Fig. 3 Two dimensional histograms of ASCAT wind direction versus ECMWF

wind direction for the three RR categories (a): [0.1 1) mm/hr, (b): [1 3) mm/hr,

and (c): [3 6) mm/hr. The upper row shows the statistics without rain correction,

while the lower row shows the ones with rain correction.

4. Conclusions and Outlook The results indicate that the proposed rain correction method increases the consistency between the corrected backscatter triplets and the

forward model or Geophysical Model Function (see Fig. 4), and thus, reduces the VRMS difference between ASCAT and ECMWF winds.

It improves the retrieval of wind speed significantly. However, it does not have much impact on the retrieval of wind direction. Meanwhile,

the VRMS scores of the rain free and heavy rain conditions categories slightly improve although no correction is performed over such

WVCs. This indicates that the rain correction applied in neighboring WVCs (contaminated by RR in the range [0.1,6] mm/hr) has a positive

impact on the ambiguity removal of rain-free and heavy-rain-contaminated WVCs.

Since ECMWF wind is not well resolved under rainy conditions, it is not optimal for the study. An independent wind source, such as

buoy wind information, is required to improve the model. Future model improvements should also account for the backscatter impact of the

wind variability increase with increasing rain rate, i.e., heavy rain conditions. Further investigation using high resolution ASCAT and rain

data will be carried out. We are currently using high resolution rain measurements from the Next Generation Weather Radar data (NEXRAD)

to improve the model. Rain-induced wind variability will be also considered in the future model improvement.

Rain free [0.1 1)

mm/hr

[1 3)

mm/hr

[3 6)

mm/hr

>=6

mm/hr

Non-

correction

1.5958 2.52 3.24 3.97 5.30

Correction 1.5949 2.30 2.74 2.97 5.24

Table 1. The mean Vector-Root-Mean-Square (VRMS, m/s)

difference between ASCAT and ECMWF winds for the

studies with and without rain correction.

Reference [1] C. Nie and D.G. Long, “A C-Band Wind/Rain

Backscatter Model”, IEEE Trans. Geosci. Remote

Sens., 2007, 45(3), 621-631.

[2] M. Portabella, , A. Stoffelen, W. Lin, A. Turiel,

A. Verhoef, J. Verspeek and J. Ballabrera-Poy,

“Rain effects on ASCAT-retrieved winds: toward

an improved quality control”, IEEE Trans. Geosci.

And Remote Sens., 2012, 50(7), 2495-2506.

1. Introduction Rain is known to both attenuate and scatter the microwave signal.

In addition, the roughness of the sea surface is increased because of

splashing due to raindrops. This increases the radar backscatter (σ0)

measured, which, in turn, will affect the quality of scatterometer

wind retrievals. For C-band radar systems such as the AMI

scatterometer and ASCAT, the rain-induced backscatter is

dominated by rain perturbation on the sea surface [1] [2]. Thus, a

simplified model for the ASCAT backscatter under rainy conditions

is presented as a summation of the wind-induced backscatter and

the rain-splashing-induced backscatter (namely bias) in this paper.

The backscatter bias varies with incidence angle (or wind vector

cell), rain rate and wind speed. Then, provided radiometer rain data,

the rain-contributed bias is subtracted from the total backscatter

measurements to improve the scatterometer wind retrieval under

rainy conditions.

The performance of this simple model is evaluated by running the

ASCAT Wind Data Processor (AWDP) for six months of ASCAT

with and without the rain correction model, and using Tropical

Rainfall Measuring Mission’s (TRMM) Microwave Imager (TMI)

and European Centre for Medium-range Weather Forecasts

(ECMWF) collocations for validation purposes.

Fig. 4 Histograms of ASCAT first-rank MLE for the three RR categories (left): [0.1 1) mm/hr, (mid): [1 3) mm/hr, and

(right): [3 6) mm/hr. Solid line-- without rain correction; Dash line-- with rain correction.