240 likes | 391 Views
LO calibration frequency impact Part II. C. Gabarró , J. Martínez, V. González, A. Turiel & BEC team SMOS Barcelona Expert Centre Pg. Marítim de la Barceloneta 37-49, Barcelona SPAIN E-mail: smos-bec@icm.csic.es URL: www.smos-bec.icm.csic.es. LO CAL FREQ ANALYSIS. INDEX
E N D
LO calibrationfrequencyimpact Part II C. Gabarró, J. Martínez, V. González, A. Turiel & BEC team SMOS Barcelona Expert Centre Pg. Marítim de la Barceloneta 37-49, Barcelona SPAIN E-mail: smos-bec@icm.csic.es URL: www.smos-bec.icm.csic.es
LO CAL FREQ ANALYSIS • INDEX • BRIEF SUMMARY OF QWG-4 (March-2011) & • QWG- 5 (June) RESULTS & CONCLUSIONS : • - ICM STUDY • - UPC STUDY • NEW ANALYSIS WITH DECIMATED DATA FROM MARCH 2010 (processors v500) • ANALYSIS OF NOVEMBER/ DECEMBER 2011 DATA (using 2 weeks of LO=2min)
ICM QWG-4 STUDY TBH_10min-TBH_2min TBV_10min-TBV_2min DecimationStudy (10min, 6min and 2min) 57 orbitsfrom 24th – 25th March 2010 TB_10min-TB_2min 20100325T040821 descending
ICM QWG-4 STUDY SSS LO 6min -2min SSS LO 10min -2min SSS LO 10min - 2min
UPC LO ANALYSIS data with decimated calibration. (Tx+Ty)/2 averaging 0 to 40º incidence angles. 6min-2min 10min-2min Spatial structures appear both cross-track and along-track The structures are still present after averaging
UPC LO ANALYSIS The LO-phase error induces structure in xi-eta 10min-2min Average of 126 snapshots of TBn-TB2 6min-2min
CONCLUSIONS FROM QWG-4 & 5 • Clear impact on TB (larger in V-pol) and SSS. • TB differences between LO-10min and LO-2min have a global STD of 0.47 K. • LO phase drift produces non-random visibilities errors -> TB spatial structures in xi-eta. • Relatively small phase error (low STD) may have impact on the image due to this systematic behaviour. • Spatial structures are expected in SSS retrievals due to imperfect LO phase error cancellation. • Differences in SSS global STD 0.48 psu.
LOCal Freq Analysis in QWG-7 • 2) DATA SET: DECIMATION STUDY • 24th-26th March 2010 (81 semi-orbits) with MIRAS at LO=2 min • Repeatedthedecimitationwiththenewestversions of processors: • L1PP5.00 (no land-sea contamination bug) • L2PP5.00
2) DECIMATION MARCH 2010 Differentimpactacrosstrack : Spatialstructures, as observed in UPC study. SSS 10 min- SSS 2min SSS 10 min- SSS 2min Averaging X-Swath (Km) X-Swath (Km) One half-orbit from Pacific (similar with many)
2) DECIMATION MARCH 2010 Comparing calibration at 10 min vs. 6 min TB 10 min- TB 2min TB 6 min- TB 2min 0.55 0.15 3 days, 81 semi-orbits Strong filtering (optimal conditions) World except N. Atlantic & Med. Different OTTs for 2, 6, 10 min. • Large Reduction of STD in TB • Reduction of tails
2) DECIMATION MARCH 2010 SSS L3 map 10 min – SSS L3 map 2 min -> 3 days, 0.5obinned SSS 10min – SSS 2min Latitudinal average Lat=[45S, 30N] Bias = 0.05 psu Std = 0.11 psu Filtered by: X-Swath < 400 km Poor_geo & poor_ret
2) DECIMATION MARCH 2010 WORLD LAT=[45S, 30N] SSS maps-> done - averaging 81 s-orb - 4H9 - Filteredbypoor_geophysicalpoor_retrieval SSS 10 min- SSS 2min 0.28 0.11 Comparing 10 and 6 min. 0.07 0.16 SSS 6 min- SSS 2min • Significantreduction of bias & Std SSS bygoingto6min, reductionof tails • Accuracyreq. SSS maps 0.1 psu. -> low cal. freqimpliesadditionalnon negligible error!
3) DATA FROM NOVEMBER – DECEMBER 2011: • 3 periods of 10 days have been selected (from DPGS) • 20th – 29th November : LO=10 min • 6th – 15th December : LO=2 min • 20th - 29th December : LO=10 min • Maps of 10 days at 4H9 have been constructed.
3) DATA FROM NOV. – DEC. 2011 Mean AVERAGE (SMOSminus ARGO) Periods of 10 days -> oneevery 3 days L2PP550 01-10 July 2011 <SSS-ARGO> L1PP500 L2PP500 31 Dec, 2011 10 Jan 2012 Mixed LO data 2min & 10 min 5-19 Dec LO =2min high variation over time LO effect between different periods of time is masked
3) DATA FROM NOV. – DEC. 2011 SSS Nov 21 to 30 (LO = 10min)- SSS Dec 6 to 15 (LO = 2min) Latitudinal average Lat=[45S, 30N] Bias = 0.25 psu -> higherthandecimated
3) DATA FROM NOV. – DEC. 2011 To compare data from different periods it is necessary to consider all the factors that contribute to the differences in SSS. Period A Period B Period C (SSS10min – SSS2min)_AB = Δ_impact_LO_AB + Δ_SSSvariability_AB + Δ_other_AB 20-29 Dec LO =10min 20-29 Nov LO =10min 6-15 Dec LO =2min TO BE ESTIMATED Estimated with ARGO measurements Includes TEC, Sun, galaxy, instrument … Hardtoestimate; wehaveused a linear interpolationfrom AC period
3) DATA FROM NOV. – DEC. 2011 Δ_impact_LO_AB = (SSS10min – SSS2min)_AB - Δ_SSSvariability_AB - Δ_other_AB (SSS10min – SSS2min)_AB highly noisy highly noisy low noise: candidate to detect LO contribution Lat: 0N-30S Bias = -0.21 Lon: 180W-90W STD = 0.34 4H9 L3 binned map averaged to 1x1 degree
3) DATA FROM NOV. – DEC. 2011 Δ_impact_LO_AB = (SSS10min – SSS2min)_AB - Δ_SSSvariability_AB - Δ_other_AB Δ_SSSvariability_AB Optimal interpolation 1x1 degree Differences between OI maps of ARGO measurements
3) DATA FROM NOV. – DEC. 2011 Δ_impact_LO_AB = (SSS10min – SSS2min)_AB - Δ_SSSvariability_AB - Δ_other_AB Δ_other_AB high noise: LO effect will be masked low noise: changes in other factors low enough to allow assessing LO contribution Latitude: 0N-30S Longitude: 180W-90W
3) DATA FROM NOV. – DEC. 2011 Δ_impact_LO_AB = (SSS10min – SSS2min)_AB - Δ_SSSvariability_AB - Δ_other_AB Δ_impact_LO_AB Δ_impact_LO_AB = -0.03 bias, 0.3 std in themost favorable case LO cal. freq. seemstoimpactonlandcontamination
3) DATA FROM NOV. – DEC. 2011 Δ_impact_LO_AB in optimalsituation Lat=[30S , 0N] Lon=[180W , 90W] Mean: -0.0320 psu Median: -0.0318 psu Std dev.: 0.2940 psu ∆_SSSvariability mean = 0.0183 median = 0.0206 std = 0.1126 ∆_other mean = -0.2579 median = -0.2559 std = 0.1831 Δ_impact_LO_AB
CONCLUSIONS (I) • In previous QWG it was shown that LO impact is not • negligible nor at TB neither at SSS level. • TB spatial structures in xi-eta and across track are observed • Study on decimated data (3 days, March 2010) using v500: • - TB 10min – 2min: • mean= 0.07 K & std=0.54 K • - SSS L3 0.5o binned maps 10min – 2min: • bias = 0.08 psu & Std = 0.28 psu • Study of two different periods (Nov-Dec 2011) • - Instrumental factors are much larger than LO contribution (thus hard to estimate). However, our analysis points to a significant LO impact on SSS maps (under optimal conditions: box in S Pacific) • bias = -0.03 psu & std = 0.29 psu
CONCLUSIONS (II) • Evidence is limited as the data for the study are limited. However, now as before, the issue seems serious enough to be taken into account. • Incremental studies on incremental samples of data will probably lead to more or less the same conclusions. • Processing improvements (CW and so on) are unlikely to change this situation. • Definitively: with 10 min LO calibration we will never meet the mission requirements over ocean
In summary: • 10 min kills SMOS salinity requirements • Processing options will not help • Time to turn on Guillermo’s proposal