Using HF radar coastal currents to correct satellite altimetry

Carolyn Roesler, William J. Emery and Waqas Qazi

CCAR

Aerospace Eng. Sci. Dept.

University of Colorado at Boulder

IGARSS 2011 , 24-29 July, Vancouver, Canada

Outline

• CODAR data set

• Coastal temporal and spatial scales

• Retrieval of synthetic CODAR heights

• Comparison with J-2 and PISTACH retrackers

• J-2 SSH errors: Waveform behavior and sea state

• Conclusions

Goal: to determine how the Jason waveforms might be retracked to better fit the CODAR synthetic heights

CALIFORNIAN CODAR DATA SET

At least 2 radar overlapping fields are needed to derive the currents

Using 2008 & 2009 hourly CODAR measured ocean surface currents on a resampled and post processed 6 km and 2 km grid. Courtesy of Sung Yon Kim, Scripps Institute of Oceanography.

Long range

Short range

Frequencies

4.66 MHz 13.54 MHz

Resolution

6 km 2 km

Up to 150 km 50 km offshore

Bathymetry and geography in the coastal transition zone

Data set geography Around Monterey Bay

Red Lines: bathymetry, 100 m and 250 m

Blue Lines: bathymetry, every 1000 m from 1000 to 4000 m every 1000 m from 500 to 3500 m

Green Line: boundary of 6 km CODAR set

~ 150 km from shore

In order to retrieve the geostrophic currents from the total velocity surface derived from HF radar, an analysis of the time and spatial scales of the coastal oceanic features is done. These features are expected to vary with the distance to the coastline and or bathymetry, and thus may vary regionally

CODAR TIME and SPATIAL SCALES

Time scale 10 days in the open ocean 3 days near the coast

From temporal covariance of velocities at zero spatial lag averaged over 2008

3 day averaging as a first approximation to the geostrophic flow

binned according to spatial lag X and Y normalized and averaged over 2008.

Looking at the shape of the velocity covariances we will

assume that locally, the homogeneous isotropic turbulent model is adequate.

Spatial Scale

Covariances of CODAR velocity Cuu and Cvv for 10 km width regions

Cuu

u cross shelf velocity

Cvvv along shelf velocity

Spatial scales

For homogeneous isotropic turbulence, velocity covariances are related to the stream-function covariance

We use the Waldstat (1991) streamfunction spatial covariance :

And find parameters :

For zone 50-60 km : a = 50 km, b = 70 km For zone 20-30 km: a = 35 km, b = 50 km

Cψψ= ( 1 - (r⁄b)2 ) exp( -(r⁄a)2 )

Fitted Observed

Optimal Interpolation (OI) [Bretherton et al. (1976)] to derive the estimated streamfunction

CODAR Geostrophic heights field

Region of P221 going through Monteray bay. The complex bathymetry generates the formation of Eddies. And this region is prone to have large SLA variations.3 day-averaged CODAR total currents minus a Mean current derived from the 0.02° gridded DNSC08 MSS

This mean was chosen due to the short span of the CODAR time series. A Mean dynamic topography should be removed, but would have more errors in the coastal region.

In the OI, the spatial scales L vary locally as a function of the grid point distance to coast, and only the surrounding observations at a distance less than L are considered.

The noise error for the CODAR was chosen to be constant = 15 cm/s although in reality it varies depending on the radar geometry, and weather conditions, and in our case to how close the geostrophic assumption is valid.CODAR ocean

currents u,v

3 day- averaged -Mean derived from DNSC08MSS

OI varying spatial scales

Synthetic height fields

CODAR 2 km Geostrophic field

Right: Effect of using a single spatial scale chosen at the 50 km zone (top), and the improved resolution using a varying spatial scale (bottom)

Left: A close up of the improved map underlying the input CODAR vectors cm

6 & 2 km CODAR SSH Along track P221

At the time of J1 passage

More variations in the 2 km set. But the 6 km set enables us to get information further offshore.

Later we combine both resolution products.

CODAR & weekly Merged MSLA

Aviso product: Weekly merged MSLA on a 1/3 °x1/3° grid, for 2008

CODAR SSH computed directly on the MSLA grid.

Interpolated at the J along track P221

JASONCorrected SSH Along track P221Distances more than 20 km off coast

Smoothed 25 km

Smoothed 50 km

Mean

of

ti

me

Series removed

CODAR ocean currents u,v

3 day- averaged -Mean current derived from DNSC08MSS

OI varying spatial scales

Synthetic height fields

Comparison

6-km-Codar and J2 SSH on P221 for 2009

J2 PISTACH coastal product resampled to 1 Hz using the MLE4 Ocean Brown model retracked range : Sea level relative to the DNSC08 MSS, all corrections applied.

We generate two smoothed SSH sets: cut-off frequency of 25 km cut-off frequency of 50 km

Codar 6 km Along track SSH amplified by 2

Mean of all time series removed

6-km-Codar and J2 SSH on P221 for 2009

OCE

6-km-Codar and J2 SSH on P221 for 2009 ( cont.)

2x{Codar}SSH filt 25 kmSSH filt 50 km

OCE

Correlation

Red points : statistically non significant

Mean( blue points) = 0.65

For the sets with a high correlation the

slope of the regression coefficient is around 2

Correlation Codar 6 km and 50-km-smoothed J2 between Along track distance 50 – 150 km

J2 PISTACH Retrackers

Data : Aviso 20 Hz-J2 PISTACH coastal product

Oce: ku Band range: Deep ocean Brown model: MLE4 (range, amplitude, significant wave height & mispointing angle)

Red3: MLE3

Ice3: 30 % threshold

Time period: 2009

Extracted range from 3 retrackers: ( from PISTACH handbook)

Principe of the Ice3 algorithm ( from PISTACH handbook)

Done on a smaller window selected

around the leading edge

J2 retrackers & 6-km-CODAR for P221

2x{Codar}SSH filt 25 kmSSH filt 50 km

If we assume Codar SSH to be the best estimate of the geostrophic field, then we can evaluate the retracking techniques. At this point, the shapes are more important than the exact values.

cm

C006 C007

Along track distance ( km)

For these first 4 cycles it seems that Ice3 fits better to the Codar SSH. Although they are quite similar except for the third case on C021, Feb -03

OCE

OCE OCE

OCE

OCE

J2 retrackers & 6-km-CODAR for P221

Here we only keep the best visually retracker fit either Ice3 or OCE, for the first 16 cycles of 2009

2x{Codar}SSH filt 25 kmSSH filt 50 km

The averaged correlation is now 0.82

Closer to shore use of 20 hz Pistach & Codar 2 km

To filter the noisy 20 hz data: iterative median and low Pass filter with a 3-sigma data selection [ Dufau Claire,2011] With wave length L = 21 points ( 7 km ) L = 61 points ( 21 km )

Sea level relative to a MSS, all corrections applied: use of decontaminated water vapor corr, GIMP ionospheric corr, SSB included ( may not be reliable over continental shelf)

3x{ Codar 2 km}

Mean of time series removed

Ice3 retracked range

Good fit Phase shift

wiggles

Jason-2 Ku 20 Hz Waveform time series for P221median tracker algorithm, Diode acquisition

mode

In some places the leading edge of the WF deviates from the predetermine tracker gate 31. They correspond with large offsets between Codar and altimetry SSH.

The median tracker algorithm uses the WF to update the tracker range. So large deviations could give us a clue to possible changes in WF shape and errors in retracking ranges.

Gate number

Alo

ng

tra

ck d

ista

nce

fro

m c

oast

sig0 and unretracked range relative to MSS

MLE4 Ku-band sig0 (dB)Unretracked range relative to MSSOce retracked – unretracked range

Strong correlation ( > 0.7) when sig0 > 15 dBIf not lower ( 0.6 ) due to signal to noise level

Sig0 = radar return backscattering cross section

C019, Jan 15 C031, May 14

C030, May 04

sig0 and unretracked range relative to MSS

J2, Jan 15

J1, Jan 15

Median tracker algorithm

Split gate tracker algorithm

J2 and J1 in flying phase formation, same orbit , 54 s delay

Both tracker algo behave similarly with a slight difference in the sensitivity of the echo shape

However the (sig0, unretracked range) relationship disappear when J2 is in the Diode/DEM coupling mode, because the tracking operations do not depend on echoes analysis anymore.

Diode/DEM mode

J2, Jun12

Sigma0 blooms

Altimetry data degraded by high sig0 [ Tournadre 2006, Thibaut 2007]

Occur in presence of weak winds ( cm scale waves absent) surface slick

WF may be corrupted due to the non uniform sig0 in the altimeter footprint with localized highly reflecting patches. various possible WF distortions

Consecutive 20 Hz WF, 30 km offshore in presence of a blooming event.

Trailing edge increasing

Peakiness increased

V shape orRound pattern similar to rain events

J2 C021 Feb03

Variations in sea states : Sig0 and SWH MLE4 retracked

Sig0 (dB)2x{SWH} (m)

Occurrence, size and strength of blooming events variable

Dots: regions where Ice3 and Red3 ranges diverge

1Blooms 30 km and 140 km

4Blooms before 40 km and after 60 kmSWH variable

5Sig0 > 15 dBSWH high and variable

2Bloom 30 km SWH high

3Sig0 < 14 dBNo blooms

6Sig0 ~ 15 dBA lit bloom at 140 km

C019, Jan 15 C021, Feb 03

C030, May 04

C034, Jun 12C031, May 14

C026, Mar 25

Ice 3 retracker Red 3 retracker

C019 Jan 15 C019 Jan 15

No blooming events Ice 3 performs well and better. Sig0 stays below 15 dBSWH < 2.5 m

Ice3 behaves well closer than 50 km , then Red3 until a little blooming event around 140 km Sig0 is about 15 dBSWH < 1.5 m

Ice3 fits well closer than 50 km when a strong blooming event starts, then Red3, until the next blooming event around 140 kmSWH < 2 m

Behavior of Ice3 and Red3 range related to the sea state

3

6

1

Along track distance from coast Along track distance from coast

Ice 3 retracker

Red 3 retracker

After the close to shore blooming event ( < 30 km), Red3 performs better, though it seems noisy.High SWH

Sig0 > 15 dB and high ; High variations of SWH

Ice3 and Red3 behave in a similar way and do not perform well.

A different retracking method should be adapted to this situation.

Behavior of Ice3 and Red3 range related to the sea state

2

5

Along track distance from coast

Conclusion and future work

• We have processed the CODAR current derived velocities into synthetic heights using a varying spatial scale OI technique, assuming a locally isotropic and homogeneous field.

• We have shown that there is a good relationship between JASON & CODAR SSH, further than 20 km off the coast.

• It seems promising as a tool to evaluate the several retracking techniques. They seem to perform differently depending on the sea state.

• In the presence of blooming events the altimetry data is corrupted. These could happen more frequently over the Californian coastal areas due to upwelling.

FUTURE WORK

• To correct for the blooming events using the Codar SSH as a reference.

• Redefine the retracking strategy: using the Codar SSH, find the corresponding retracking gate and deduce what type of retracking technique would fit.

• This should allow to make improvements on the coastal altimetry SSH accuracies.

Questions?

Thank you

Carolyn Roesler, William J. Emery and Waqas Qazi

CCARAerospace Eng. Sci. Dept.

University of Colorado at Boulder