OTT rate deployment GPOD experiment SMOS vs ARGO comparison QWG10 – ESRIN 4-6 February 2013

SMOS-BEC – Barcelona (Spain)

OTT rate deployment GPOD experiment

SMOS vs ARGO comparison

QWG10 – ESRIN 4-6 February 2013

Justino Martínez & Carolina Gabarróand BEC team

SMOS Barcelona Expert CentrePg. Marítim de la Barceloneta 37-49, Barcelona SPAINE-mail: smos-bec@icm.csic.esURL: www.smos-bec.icm.csic.es

SMOS-BEC

Motivation of GPOD experimentAfter last reprocessing campaign important differences in SSS L3 maps were detected.

DPGS operational L2 products distributed by DPGS in real-time

High variability in SSS bias These differences are correlated between ascending and descending

ReprocessedLast reprocessing campaign from 2010/01/12 to 2011/12/22 More stable behavior

Reason for the differences? Difficult to conclude, different processing conditions were applied:

L2 data set OTT L2 configuration file version L1/L2 processors versions TEC

DPGS operational

Monthly OTTApplied two weeks

after NIR events

2 different versionsfrom 2011/10/20 to 2012/01/22

version 021_8Tg_TEC_gradient=20

Tg_num_outliers_max=50

version 022_8 from 2012/01/23

Year 2011: Wide variety of L1/L2 versions

Year 2012:Last versions of L1/L2

processors

predicted

reprocessed15-day OTT

Validity centered on NIR calibration events

Version 022_8Tg_TEC_gradient=5000

Tg_num_outliers_max=20Last versions of processors consolidated

2/20different different same only 1.5 months different

SMOS-BEC

Tested zones

Zone Lat Lon Description

Global 60S : 60N 180W : 180E Tropics and mid-latitudes

Tropic 30S : 30N 180W : 180E Tropics

122 30S : 0N 150W : 120W A region of the South Eastern Pacific

124 24S : 10S 165E : 165W A region of the South Western Tropical Pacific

126 10S : 10N 180W : 180E Equatorial Oceans

127 15N : 37N 50W : 0E North Atlantic (Tarfaya model region)

131 60S : 40S 180W : 180E Southern Ocean

132 5N : 15N 110W : 180W Intertropical Pacific

133 45N : 60N 170E : 140W A region of the North Pacific

3/20

SMOS-BEC

BINNING PROCESS

HIGH WIND FILTER

POOR GEOPHYSICAL

POOR RETRIEVAL

VALID L1C RATE

ACROSS TRACK FILTER

L2 - REPR - DPGS total L1c measures <= 90

L2 filtering rules

valid L1c measures less than

total L1c measures / 3

L2PP CNF - fileMax iterations reached > 20Poor fit qualityHigh retrieval sigma > 5Values out of range (0 .. 42) …

L2PP CNF - fileSuspect ICE > 50%Rain > 2 l/hMany Outliers > 20%Galactic noise > 10%#measures < 30 …

Wind speed > 12 m/s

1x1 grid size10 days

average overretrieved SSS

-every 3 days

+/- 400 km

4/20

SMOS-BEC

Argo values computation

Robust interpolation applied to 7.5 m depth

1x1 grid size10 days

average overArgo SSS

-every 3 days

BINNING PROCESS

3 interpolation methods on T and SSS

SSS SSS Argo 100

5SSS

SSSSSS

SSS31SSS

i

3

1ii

Differences between the mean and each interpolated T and SSS below the 5%

compare cells to study thetemporal evolution of

SMOS-ARGO

5/20

SMOS-BEC

Differences DPGS vs reprocessed

ASCENDING

MEAN(SMOS-ARGO)

DPGS shows high temporal variations

Oscillations not clear correlated with NIR

Pattern independent of the zone. Linked to the OTT generation zone?.

122South Eastern Pacific - OTT

124South Western Tropical Pacific

127Tarfaya

GLOBAL

131Southern Ocean

6/20

SMOS-BEC

Differences DPGS vs reprocessed

DESCENDING

MEAN(SMOS-ARGO)

Tropics

Correlation: Oscillations take place with the same period in ascending and descending. Maxima at

February/March May/June October/November

.

122South Eastern Pacific - OTT

126Equatorial Oceans

131Southern Ocean

GLOBAL

7/20

SMOS-BEC

Differences DPGS vs reprocessed

Descending case:SUN affects OTT generation in Feb/Mar and Oct/Nov

Ascending case:SUN effects OTT generation in May/June

A fast response in OTT deployment is critical in a rapidly changing environment as when Sun is crossing FOV (in DPGS the OTT is computed using L1C from 1-2 weeks earlier and applied during a month)

ASCENDING DESCENDING

SUNSUNSUN

As reported by ARGANS: Sun

in OTT

8/20

GLOBAL

SMOS-BEC

G-POD study

Comparison study focused in the OTT deployment rate.

Will increasing the OTT generation rate solve the problem?

G-POD Recently reprocessed data using G-POD system from 2011/12/23 to 2012/02/29

L2 data set OTT VERSIONS IN THIS PERIOD TEC

DPGS operationalMonthly OTT

Applied up to two weeks after NIR events

L2 configuration file 022_8Tg_TEC_gradient=5000

Tg_num_outliers_max=20

L1 processor: 5.04L2 processor: 5.50

predicted

G-POD15-days OTT

Validity start time matches the NIR calibration events

L2 configuration file 022_8L1 processor: 5.04L2 processor: 5.50 predicted

9/20

different same same

SMOS-BEC

G-POD study

Ascending

GLOBAL60S:60N

MEAN(SMOS-ARGO)Descending

Purple line: Reprocessing campaign. It can be compared with G-POD experiment results (red line)

Red line: G-POD experimentVery different from DPGS (blue line)

At a first glance it seems to improve results but it is necessary to carry out a more accurate study10/20

SMOS-BEC

Analysis procedureWe depart from 10-day SSS maps computed every 3 days

Our target is to compare temporal stability of G-POD and DPGS

Problem: low number of points in our time series Proposed solution: how well fit our series to a straight line?

Procedure to quantify the degree of improvement

1. Compute drift (linear fitting) f(t)=at +b of each time series y(t) (G-POD and DPGS)We use the nonlinear last-squares Levenberg-Marquardt method algorithm to find linear fitting (a and b) and its error (Da and Db)

2. Substract dritf to time seriese(t) = y(t) - f(t); De(t) = Dy(t) + Df(t)

3. Compute a measure about how far is each point of the value e=0

This measure d can be computed for each point (L3 maps) and compare histograms. The lower d, the best fit

11/20

SMOS-BEC

Results by zone

GLOBAL

ASCENDING

In this case 15-days OTT introduces a clear Improvement

1

2

3

12/20

Each point indicates the central date of 10-days L3 mapsHorizontal errorbars include each 10-days period. Vertical errorbars indicate the error of the mean.

SMOS-BEC

Tropics

ASCENDINGResults by zone

122South Eastern Pacific - OTT

13/20

SMOS-BEC

124South Western Tropical Pacific

ASCENDINGResults by zone

126Equatorial Oceans

14/20

SMOS-BEC

127Tarfaya

ASCENDINGResults by zone

131Southern Ocean

15/20

SMOS-BEC

Tropics

DESCENDINGResults by zone

122South Eastern Pacific - OTT

16/20

SMOS-BEC

124South Western Tropical Pacific

DESCENDINGResults by zone

126Equatorial Oceans

17/20

SMOS-BEC

127Tarfaya

DESCENDINGResults by zone

131Southern Ocean

18/20

SMOS-BEC

Conclusions

- A fast response in OTT generation and deployment is critical in a rapidly changing environment as when Sun is crossing the FOV

- Two strategies:

Center L1C orbits used for OTT generation in the OTT validity period

Generate OTT more frequently (15 days) To implement this option it is necessary an automatic method to generate OTT with the same quality as the current method (manual)

L2 production should be in delayed mode

19/20

SMOS Barcelona Expert Centre (SMOS-BEC)Pg. Marítim de la Barceloneta 37-49, E-08003 Barcelona, SPAINTel. (+34) 93 230 95 00; Fax. (+34) 93 230 95 55URL: www.smos-bec.icm.csic.es

SMOS-BEC

Differences DPGS vs reprocessedASCENDING

21/20

SMOS-BEC 22/20

Differences DPGS vs reprocessedASCENDING

SMOS-BEC

Differences DPGS vs reprocessedDESCENDING

23/20

SMOS-BEC 24/20

Differences DPGS vs reprocessedDESCENDING

SMOS-BEC 25/20

G-POD results by zoneASCENDING

SMOS-BEC 26/20

G-POD results by zoneASCENDING

SMOS-BEC 27/20

G-POD results by zoneDESCENDING

SMOS-BEC 28/20

G-POD results by zoneDESCENDING

SMOS-BEC 29/20

G-POD results by zoneASCENDING

SMOS-BEC 30/20

G-POD results by zoneASCENDING

SMOS-BEC 31/20

G-POD results by zoneDESCENDING

SMOS-BEC 32/20

G-POD results by zoneDESCENDING

SMOS-BEC 33/20

G-POD results by zoneASCENDING

SMOS-BEC 34/20

G-POD results by zoneASCENDING

SMOS-BEC 35/20

G-POD results by zoneDESCENDING

SMOS-BEC 36/20

G-POD results by zoneDESCENDING