GOES-13 Science Team Report SST Images and Analyses

Eileen Maturi, STAR/SOCD, Camp Springs, MDAndy Harris, CICS, University of Maryland, MDChris Merchant, University of Edinburgh, Edinburgh, ScotlandJon Mittaz, CICS, University of Maryland, MDWen Meng, Perot Systems, Camp Springs, MD

April 17, 2007

DATA CollectionGOES 13 Science Test began in December 7, 2006 and ended on January 5, 2007

Real-time GOES 13 GVAR data was available on GCR (140.90.195.68) remap server

The Imager channels for GOES-13 are:

GOES 13 GVAR data were collected for both north Hemispheric sector with image size 1827 by 3461 centered at latitude 14°19′53″, longitude 96°40′17″ and south Hemispheric sector with image size 613 by 3461 centered at latitude -31°55′10″, longitude 96°06′36″, every half hour from December 8, 2006 to January 5, 2007

Pre-processed VIS and IR imagery data were used to create multi-spectral imagery files as input of SST retrieval

ChannelsCentral Wavelength

(µm)Resolution

1 (visible) 0.65 1 km

2 (infrared) 3.9 4 km

3 (infrared) 6.5 4 km

4 (infrared) 10.7 4 km

6 (infrared) 13.3 4 km

GOES13 north sector channel 2 radiance image (upper left); GOES13 north sector channel 4 radiance image (upper right); GOES13 south sector channel 2 radiance image (lower left); GOES13 south sector channel 4 radiance image (lower right)

GOES-13 SST Coefficients• For 75 deg W:

3.9-11 coefficients (standard) Modeled RMS error =       0.48 K  -6.32    2.35  1.2027  0.0628 -0.1724 -0.0616  0.0000  0.0000

3.9-11-13 coefficients (experimental) Modeled RMS error =       0.25 K   3.13   13.58  1.0922  0.0669  0.1253  0.0174 -0.2327 -0.1369

For 105 deg W:

Modeled RMS error =       0.51367279   -2.54   -2.56  1.2122  0.0652 -0.1950 -0.0467  0.0000  0.0000 Modeled RMS error =       0.24545067    5.78   11.95  1.0932  0.0739  0.1173  0.0259 -0.2347 -0.1477

For 135 deg W:

Modeled RMS error =       0.41267461   -3.01   -0.52  1.1901  0.0750 -0.1711 -0.0639  0.0000  0.0000 Modeled RMS error =       0.24209366    3.88   10.18  1.1043  0.0693  0.0804  0.0100 -0.2006 -0.1181

NOTE: the coefficients were generated for all possible location of the GOES -13 satellite .

GOES-13 CRTM Coefficient Files

• Spectral Coefficients

• Tau Coefficients

GOES 13 SST Retrieval MethodologyRadiative-transfer-based SST retrieval algorithms are used to generate theGOES-13 SST retrievals

The form of the current GOES operational SST equation is:

where i is GOES-Imager channel number (2, 4, 6),S = sec (satellite zenith angle) – 1 Ti is channel brightness temperature in Kelvin.

SST retrievals were generated for dual and triple windowRTM Coefficients are shown:dual window (3.9 µm and 11 µm)

triple window (3.9 µm, 11 µm and 13 µm)

i

iii TSaaSaaSST ''00

ao a′o a2 a′2 a4 a′4 a6 a′6 modeled RMS

-2.54 -2.56 1.12122 0.0652 0.195 0.0467 0 0 0.51

ao a′o a2 a′2 a4 a′4 a6 a′6 modeled RMS

5.78 11.95 1.9032 0.0739 0.1173 0.0259 -0.2347

-0.1477 0.25

Bayesian Cloud Mask is applied to obtain clear-sky pixels:

– Bayes’ theorem applied to estimate the probability of a particular pixel being clear of cloud given the satellite-observed brightness temperatures, a measure of local texture and channel brightness temperatures calculated for the given location and view angle using NCEP GFS surface and upper air data and the CRTM fast radiative transfer model. The method is described in detail in a paper by Merchant et al. (2005).

Hourly SST is created by compositing three half hour SST McIDAS Area files with an applied threshold of ≥98% clear sky probability.

Satellite retrieval SST were matched with Buoy and NCEP GDAS data to create matchup dataset for validation.

GOES SST Retrieval Process

Hourly SST composite with applied 98% clear sky probability (left) and hourly composite clear sky probability

Validation – The comparison of GOES-13 SST with

operational GOES-12 SST

GOES-13 SST retrieval was compared with operational GOES-12 SST retrieval as following:

• Satellite retrieval SST vs. Buoy SST validation

• Selected hourly SST images

GOES-12 SST vs. Buoy SST validation (left for daytime, right for nighttime)

GOES-13 SST for dual window vs. Buoy SST validation (left for daytime, right for nighttime)

GOES-13 SST for triple-window vs. Buoy SST validation (left for daytime, right for nighttime). Note warm cluster of points where Buoy SST is ~24°C and Satellite SST is ~29°C

GOES-13 Day scatter plots of Satellite – Buoy SST vs. Satellite Zenith Angle for dual window (left) and triple window (right)

– Note angular dependence is reduced with triple-window algorithm, also change in Y-scale for RH plot

GOES-13 Day scatter plots of Satellite – Buoy SST vs. Satellite Zenith Angle for dual window (left) and triple window (right)

Since the 13.3 µm channel is a lower-tropospheric sounding channel, its use at angles above 65° – 70° is inevitably going to be compromised (see right panel). This increase in warm bias is responsible for the cluster of points noted previously

GOES 12 GOES 13 Dual window GOES 13 triple window

GOES 12 GOES 13 Dual window GOES 13 triple window

GOES 12 GOES 13 Dual window GOES 13 triple window

GOES 12 GOES 13 Dual window GOES 13 triple window

SUMMARY

The GOES-13 SST are somewhat noisier than the GOES-12 SSTs. This may be partially due to the unfavorable view angle conditions for the validation data.

The GOES-13 nighttime SST results suggest that 3.9 and 11 µm radiances are unbiased compared to the model used to generate the retrieval coefficients.

The daytime solar correction appears to be overcompensating the 3.9 channel effects, especially at high satellite angles (colder SSTs).

There may be a radiance bias issue in the 13 µm, which happens to compensate for the over-correction of solar contamination – the daytime triple probably looks good because of two opposing biases that occur at high satellite zenith angles.

Comparison of the dual window and triple night time plots also suggests that there is some residual cloud that affects the 13 µm channel severely but not the 3.9 or 11 µm – e.g., thin cirrus. This means that the night-time triple retrieval should not be introduced at night time until the 13 µm is also incorporated into the cloud screening information vector.

In light of the above results, the triple window should be applied in the day. Thin cirrus clouds are more likely to be caught when the visible channel is available.