Drifters in Hurricane Rita, September 2005

Drifters

On September 21, 2005, 20 Scripps drifting buoys were deployed by the 53rd WRS squadron ahead of Hurricane Rita in the Gulf of Mexico. This deployment is the most successful to date - all twenty drifters survived the air drop and the passage of the hurricane, and are sending data! There are 12 Minimets and 8 ADOS, reporting air pressure, SST, wind direction and wind speed. The 8 ADOS drifters are equipped with 100m thermistor chains. All the following figures are based on preliminary data - no quality control applied.

The figures were last updated with data up to 9/26/2005 00:00UTC.

For the following figures, click on the images for enlarged versions of plots.

1. Drifter Tracks during hurricane Rita

Solid dots indicate deployment locations. The numbers represent the last two digits of the WMO drifter ID numbers. The 3-hourly track data of Rita is shown as a solid line. The deployment locations were based on the forecast track of Rita, as of 9/20 11pm EDT. This forecast is indicated with a dashed line. Forecast and actual hurricane centers at 00:00UTC at consecutive days are indicated. Colors of drifter tracks in the detailed plot (fig. 1.1.b) are the same as in the next following figures of drifter sensor data.

Fig. 1.1 Drifter tracks and track of hurricane Rita, (a) larger, and (b) detailed map.

2. Drifter SST during hurricane Rita

The time series of SST observations are presented below for all drifters. The SST minimum from each drifter is indicated on the figure, as well as the observed decrease in SST. The two drifters with the largest drop in SST are Minimet 41945 (SST change of -3.3° C) and ADOS 41612 (-2.1° C). These changes occured over the first 42 hours after deployment.

Fig. 2.1 Drifter observed SST, during the first (a) 4 days and (b) 7 days.

3. Drifter Air Pressure during hurricane Rita

The time series of of air pressure observations are presented below for all drifters. The pressure minimum from each drifter is indicated on the figure. The two drifters with the lowest pressures in their time series are Minimet 41945 (926mb) and ADOS 41612 (938mb). Their deployment locations are to the right of the hurricane path.

Also shown are three days for drifters 41945, 41612, and 41671. These drifters recorded the largest pressure drops. Drifter 41945 displays a few data measurement cycles with, as it seems, anomalous pressure fluctuations after the encounter with the center of the hurricane at a minimum distance of 12km (during days 266.4 - 266.8).

Fig. 3.1 Drifter observed sealevel pressure: (a) all drifters during the first 4 days, and (b) 3 drifters during 3 days.

In the next figure, drifter pressure observations are plotted with respect to distance to center of Rita. An exponential fit to the data is computed:

pressure = f(distance) = p0 + (pe-p0) · e-(a/distb) ,

where p0 is specified as 920mb, and the other coefficients are found to be: pe=1018.62mb, a=22.67, and b=0.84.

Fig. 3.2 Drifter sealevel pressure vs. distance to hurricane center: (a) all drifters, and (b) with exponential fit (Stdev. indicated with dotted lines).

4. Drifter Wind Speed during hurricane Rita

Drifters measured surface wind speed by measuring acoustic sound pressure under water. The following plots compare drifter wind speeds with collocated QSCAT data. Drifter and QSCAT observations were collocated when both measurements were made within 1 hour and 100km. On average, the collocations were made within 8min and only 11km apart. The scatterometer signal has been flagged for rain.
Fig. 4.1 Drifter and QSCAT wind speeds, during first 6 days, (a) drifters labeled, and (b) rain-flagged QSCAT data.

The next two figures present data from the first 2.5 days after deployment, and then from the following 2.5 days.

Fig. 4.2 Drifter and QSCAT wind speeds, (a) during first 2.5 days, and (b) next 2.5 days.

For the next comparison, drifter hourly averages of wind speed were computed, by averaging only the two lowest speed observations in each hour. The scatter does not decrease very much. The same drifters that appear as outliers in fig. 4.2.b also exhibit relatively large wind speeds in the hourly data set.

Fig. 4.3 Drifter (hourly averages) and QSCAT wind speeds, first 6 days.

5. Drifter and QSCAT Wind Speed and Direction (hourly averages)

In the following twenty figures, all drifter observed wind speeds and directions (hourly averaged) are compared with collocated QSCAT data. The top panel shows distance to the center of hurricane Rita (in km). The second panel shows wind speed (collocated QSCAT in red, rain-flagged data as open squares). The third panel shows wind speed differences (drifter-QSCAT); average differences and rms are printed on the plot. The fourth panel shows wind directions. And the last panel shows direction differences. In the direction difference panel, the average difference and rms for all data is printed, as well as for a sub-sample of differences less than 120° ("SUB"). When data in the sub-sample are adjusted for the computed offset, the resulting rms of the adjusted wind directions is also indicated.

Fig. 5.1 Drifter (a) 41923 and (b) 41930, wind speed and directions.

Fig. 5.2 Drifter (a) 41933 and (b) 41936, wind speed and directions.

Fig. 5.3 Drifter (a) 41937 and (b) 41938, wind speed and directions.

Fig. 5.4 Drifter (a) 41940 and (b) 41941, wind speed and directions.

Fig. 5.5 Drifter (a) 41942 and (b) 41943, wind speed and directions.

Fig. 5.6 Drifter (a) 41944 and (b) 41945, wind speed and directions.

Fig. 5.7 Drifter (a) 41593 and (b) 41612, wind speed and directions.

Fig. 5.8 Drifter (a) 41615 and (b) 41619, wind speed and directions.

Fig. 5.9 Drifter (a) 41670 and (b) 41671, wind speed and directions.

Fig. 5.10 Drifter (a) 41852 and (b) 41919, wind speed and directions.

Wind speed and direction differences are summarized in table 5.1. The sub-samples of direction differences, refer to data when the differences are less than 120° and drifter data have been adjusted by the indicated offsets.

Table 5.1: Wind speed and direction differences (drifter-QSCAT)
Number of Data Speed Direction Direction sub-sample
Drifter days N av Diff rms comment av Diff rms Nsub offset rms
41923 40 69 -3.4 4.4 bad -21 47 66 -15 28
(a) 41930 40 93 -0.9 2.7 - 8 45 89 - 9 30
41933 40 73 -0.9 2.6 -15 27 72 -14 19
41936 40 84 0.0 2.1 good 1 51 78 - 6 31
41937 40 58 1.1 2.5 4 46 55 - 5 26
41938 40 71 24.1 28.2 bad - 2 27 70 - 4 19
41940 40 78 5.2 6.4 trend - 3 41 73 - 5 21
41941 40 85 -0.5 2.0 4 35 84 2 30
41942 40 73 -0.6 1.9 -17 37 72 -20 24
(b) 41943 40 79 -0.9 2.0 - 5 32 77 - 9 21
(c) 41944 40 81 -0.7 2.1 short -12 34 79 -12 21
(d) 41945 40 71 -1.7 3.6 -13 37 68 -11 19
41593 33 70 0.8 3.3 - 2 41 66 - 6 20
41612 40 79 -2.1 3.0 -11 40 77 - 7 31
41615 41 81 -1.7 2.4 - 4 29 79 - 4 18
(e) 41619 40 57 -1.9 2.9 short 2 38 56 - 2 31
41670 40 49 8.3 9.5 bad - 8 34 48 - 5 22
41671 4 short
41852 40 49 0.2 3.3 trend 4 55 44 - 4 26
41919 40 73 1.1 4.4 trend -14 37 71 -14 21

For speed: (a) exclude days 275-278.3, (b) only first 21 days, (c) first 10 days, (d) first 21 days, (e) first 6 days;
bad = erratic speed, good = very good speed, trend = trend in speed, short = very short speed record.


6. Drifter Winds and QSCAT swaths

QSCAT revs of hurricane Rita in the Gulf of Mexico are shown in the next several figures. Rain-flagged satellite data are plotted in red. Drifter obs within 1hr of the QSCAT data are shown in blue (last 2 digits of drifter ID labeled on plot). Also indicated are the NOAA "best track" hurricane center (green cross) and max wind speed (in subtitle). The (blue) subtitle for drifters indicates: 2 last digits of drifter ID, time difference from QSCAT data (in minutes), and measured wind speed (in m/s).

Fig. 6.1 Drifter winds and QSCAT, during (a) rev 32601 (9/22 12:20) and (b) rev 32608 (9/23 00:42).

Fig. 6.2 Drifter winds and QSCAT, during (a) rev 32615 (9/23 11:54) and (b) rev 32622 (9/24 00:16).

Fig. 6.3 Drifter winds and QSCAT, during (a) rev 32629 (9/24 11:28) and (b) rev 32636 (9/24 23:50).



7. Drifter Wind Speed vs. distance to hurricane center

The following figures show drifter measured wind speed (hourly averages) versus distance to hurricane center. In these figures, three drifters with bad speed data have been excluded (41923, 41938, and 41670). And for two drifters only partial data are shown: drifter 41970 up to day 270, and drifter 41945 after day 267.

Fig. 7.1 Drifter wind speed vs. hurricane distance (17 drifters).

The difference between the two types of drifters is emphasized below: Minimet data are plotted in blue, and ADOS drifters in black. Also shown are collocated QSCAT data (in red).

Fig. 7.2 Drifter wind speed vs. hurricane distance: (a) Minimet (blue), ADOS (black), and (b) QSCAT (red).

In the following figure, the Gradient Wind solution is shown (computed from pressure function fitted to observed pressure, see Fig. 3.2b), as well as the gradient wind data from Rick Lumpkin's figure (R.L.), and his H*Wind data, for the left side of Rita.

Fig. 7.3 Drifter wind speed vs. hurricane distance, gradient wind solutions and H*Wind.

In the following figures, scatterometer wind speeds are plotted vs. hurricane distance. Standard QSCAT is shown (25 x 25km resolution), as well as David Long's high-res data (2.5 x 2.5km resolution). In all figures, the gradient wind solution from fig. 7.3 is plotted for comparison. Also shown are the 50km bin averages and standard deviation (standard QSCAT in red, and high-res QSCAT in blue).

The high-res data have a bigger spread, but the standard deviations of bin averages are very similar. Standard QSCAT bin averages are about 1m/s higher than high-res averages.

Fig. 7.4 QSCAT wind speed vs. hurricane distance, rev 32608, for (a) standard QSCAT and (b) high-res QSCAT. 50km bin averages and standard deviations are shown.

Fig. 7.5 QSCAT wind speed vs. hurricane distance, rev 32615, for (a) standard QSCAT and (b) high-res QSCAT. 50km bin averages and standard deviations are shown.

Fig. 7.6 QSCAT wind speed vs. hurricane distance, rev 32622, for (a) standard QSCAT and (b) high-res QSCAT. 50km bin averages and standard deviations are shown.

During rev 32615 the drifters are located mostly in the southwest quadrant of the hurricane (see fig. 6.2a). The scatterometer data in the SW quadrant is emphasized in the next figure (7.7.a). In this quadrant the wind speed tends to be lower than in other quadrants.
All of the above figures are based on QSCAT-R1, the originally processed scatterometer data. JPL has released a revised product, which has been completed in February 2007. The new data, "R-2", includes more conservative, i.e. less often, rain-flagging, and wind speeds that are about 10% higher at speeds greater than 15m/s. QSCAt-R2 rev 32615 is shown below (fig. 7.7.b).

Fig. 7.7 QSCAT wind speed vs. hurricane distance, rev 32615 for standard (a) QSCAT-R1 (SW quadrant data is in red), and (b) QSCAT-R2.



8. Drifter Acoustic Data and derived Wind Speed

The drifter wind speed data presented above (incl. figs. 4.1-4.3 and 5.1-5.10, and in table 5.1) are based on the mid frequency band, using the following conversion formula:

wind speed (m/s) = { 10 (spl/20) + 104.5 } / 53.91 (Equation I)


Here spl is the sound pressure level. Acoustic data are available from the low (0-2kHz) , mid (6-10kHz) , and high (14-20kHz) frequency bands. Timeseries of all acoustic data are shown in the next plots, including wind speed derived from the mid frequency, using Equ. I. There were deployed 12 Minimet and 8 ADOS drifters. Included are collocated QSCAT wind speeds (red dots), and the distance to the center of hurricane Rita. The high band has been shifted by -10 in the plot, in order to see each band more clearly.

Fig. 8.1 Minimet acoustic data, 15min measurements, for drifters (a) 41923, and (b) 41930.

Fig. 8.2 Minimet acoustic data, 15min measurements, for drifters (a) 41933, and (b) 41936.

Fig. 8.3 Minimet acoustic data, 15min measurements, for drifters (a) 41937, and (b) 41938.

Fig. 8.4 Minimet acoustic data, 15min measurements, for drifters (a) 41940, and (b) 41941.

Fig. 8.5 Minimet acoustic data, 15min measurements, for drifters (a) 41942, and (b) 41943.

Fig. 8.6 Minimet acoustic data, 15min measurements, for drifters (a) 41944, and (b) 41945.

Fig. 8.7 ADOS acoustic data, 15min measurements, for drifters (a) 41593, and (b) 41612.

Fig. 8.8 ADOS acoustic data, 15min measurements, for drifters (a) 41615, and (b) 41619.

Fig. 8.9 ADOS acoustic data, 15min measurements, for drifters (a) 41670, and (b) 41671.

Fig. 8.10 ADOS acoustic data, 15min measurements, for drifters (a) 41852, and (b) 41919.

There are several drifters with bad acoustics data: Minimets 41923 (long gaps, and very spiky), 41938 (bad mid band), 41940 (trend); and ADOS 41670 (bad in all bands).

There seems to be stronger high-frequency variablity in the Minimet than in the ADOS data, in all frequency bands. For Minimets, the mid range has stronger high-frequency variablity than the low bands. For ADOS, both mid and low have about the same variability. The high band has a smaller range than the mid band, but in some drifters the high band agrees very well with the mid band (e.g. Minimet 41937).

The acoustic variability differences between ADOS and Minimet drifters are due to the fact that ADOS drifters had long thermistor chains attached, and Minimets didn't. As a result, the Minimet buoys tend to bounce up and down more than the ADOS drifters.

Hourly estimates have been created for all acoustics data, by averaging all 15min per hour (usually 4), and computing the average of the two lowest values. The results are shown below. For better comparison, the mid range has been shifted to overlay the low and high frequency data. The shifts used in the plots are indicated in the bottom right of the figures. Drifters with bad data are not shown.

Fig. 8.11 Minimet acoustic data, hourly averages , for drifters (a) 41930, and (b) 41933.

Fig. 8.12 Minimet acoustic data, hourly averages , for drifters (a) 41936, and (b) 41937.

Fig. 8.13 Minimet acoustic data, hourly averages , for drifters (a) 41941, and (b) 41942.

Fig. 8.13 Minimet acoustic data, hourly averages , for drifters (a) 41943, and (b) 41944.

Fig. 8.14 Minimet acoustic data, hourly averages , for drifter 41945.

Fig. 8.15 ADOS acoustic data, hourly averages , for drifters (a) 41593, and (b) 41612.

Fig. 8.16 ADOS acoustic data, hourly averages , for drifters (a) 41615, and (b) 41619.

Fig. 8.17 ADOS acoustic data, hourly averages , for drifters (a) 41852, and (b) 41919.

Mid and low bands tend to agree very well, but usually there are some periods when they differ significantly (e.g. Minimet 41937). Like in the case of the 15-min data, there is stronger high-frequency variability in the houlry Minimet than in the hourly ADOS data.
The average difference of low minus mid bands (using all hourly data) varies for Minimets between 24 and 27 (drifter 41937 is an outlier with an average difference of 19), and for ADOS between 15 and 21.

After subtracting the average difference between low and mid (using hourly averages) for each drifter, scatter plots of low vs. mid frequencies have been made. They are shown below: for all 42 days, only the first 6 days during passage of Rita, and for just the ADOS drifters.
Also shown is the scatter of mid vs. low frequencies.

Fig. 8.18 Low vs. mid scatter plots, all drifters, for (a) 42 days, and (b) 6 days.

Fig. 8.19 Low vs. mid scatter plots, only ADOS drifters, for (a) 42 days, and (b) 6 days
In figure (b) all data within 100km of the hurricane center are marked with "+".

Fig. 8.20 Mid vs. low scatter plots, only ADOS drifters, for (a) 42 days, and (b) 6 days
In figure (b) all data within 100km of the hurricane center are marked with "+".

In the following figures, the log of low and mid frequencies is plotted as a function of distance from the hurricane center. 15min and hourly data are shown, for 17 drifters with good acoustic data: excluded are 41923, 41938, and 41670.

Fig. 8.21 15min acoustic data vs. distance to hurricane center, for (a) low, and (b) mid frequencies.

Fig. 8.22 Hourly-averaged acoustic data vs. distance to hurricane center, for (a) low, and (b) mid frequencies.

Fig. 8.23 15min acoustic data, ADOS only, for (a) low, and (b) mid frequencies.

Fig. 8.24 Hourly-averaged acoustic data, ADOS only, for (a) low, and (b) mid frequencies.

In the following section the conversion algorithm between acoustic noise measurements and wind speed is investigated. Usually, it is only the mid-frequency values that are used to derive the wind speed, based on the following conversion function:

wind speed (m/s) = { 10 (spl/20) + 104.5 } / 53.91 (Equation I)

For each drifter acoustics measurement the distance from the hurricane center is calculated, and then the gradient wind solution from the observed drifter pressure measurements is used for the "truth" wind speed. The gradient wind solution vs. hurricane distance is shown in fig. 7.3 above. The resulting scatter plots, (gradient wind) speed vs. sound pressure level, are presented below, for the hourly ADOS data. Only data within 1,000km of the hurricane center are used. The red line is the best fit of the data to a function as in Equ. I, and the resulting coefficients are listed below the figure. In fig. 8.25.b, the black line represents the conversion function with the coefficients as in Equ. I (a=20., b=104.5, and c=53.91).

Fig. 8.25 Gradient wind speed vs. hourly-averaged acoustic data, ADOS only, for (a) low, and (b) mid frequencies.
Red line is fit to data, and black line are coefficients as in Equ. I.

last modified on September 10, 2008
Go Back