short wavelength correlated errors of altimetry ... tryandmigatethehump!arfact!with!beer...
TRANSCRIPT
Short wavelength correlated errors of altimetry: implications for SWOT
G.Dibarboure, C.Dufau , J.Poisson (CLS) M.Orsztynowicz, R.Morrow (LEGOS) PY.Le Traon (Ifremer), N.Picot, F.Boy (CNES) Washington, January 2014
The spectral hump: a short wavelength correlated error
1 Hz / 7km record
20 Hz / 300m record
• Most oceanographers use al?metry at 1 Hz (i.e. 6-‐7 km resolu?on)
• To analyze scales shorter than 100 km the limit the 1 Hz « noise floor »
• Bigger picture @ 20Hz: it is not instrument noise floor (visible only below 3km)
• There is a hump-‐shaped ar?fact on most al?metry spectra
Goal of this talk: to discuss implica?ons for (Air)SWOT and the SDT community
Jason-‐2 Al?Ka
• The footprint of Jason-‐class al?mers is a large disc (e.g. 10km radius) and the hump ar?fact appears when the al?meter footprint samples inhomogenei?es in sigma0 or SWH
• Likely triggered by geophysical causes (e.g. atmosphere and sea state transi?ons) • Even small events are sampled repeatedly è smoothed / correlated error
• Implica1on: Similar phenomenons might happen with KaRIN LR (λ< 1km) • Implica1on: Need realis?c simula?ons of what will happen when KaRIN’s OBP
averaging has to handle unusual and non-‐Gaussian situa?ons
What is the origin of the hump ar?fact?
WF leading edge
WF trailingedge
• 1 Hz floor is actually the sum of two components: true instrumental noise (that increases with SWH), and the hump that exists almost everywhere but is stronger wherever major backscacer events (blooms, low winds, rain) are likely to happen
Illustra?on on maps of 1 Hz « spectral floor »
WINTER
SUMMER
1 Hz « noise floor» SWH
Instrumental noise dominates
Instrumental noise dominates
Backscacer events dominate
• The hump actually has a flat spectrum from 10 to 50 km on Jason-‐2
• Implica1on: The methodology used for the SSH requirements in SWOT’s SRD (derived from Xu & Fu, 2012) uses valid assump?ons
Link with spectral studies and SWOT’s SRD
Jason-2 AltiKa
Al?Ka VS. Jason-‐2: spectral slope
Spectral slope
• Spectral slopes from Jason-‐2, Al?Ka and Cryosat are very consistent
• Also consistent with observa?ons from Xu and Fu
• Implica1on: Al?Ka and Cryosat agree with the SRD (Jason spectra)
Al?Ka VS. Jason-‐2: spectral « 1 Hz noise » floor
1 Hz floor on Jason-2 1 Hz floor on AltiKa
• Same type of distribu?on (hump and instrument noise)
• But the 1 Hz floor is lower on Al?Ka (1.5-‐3cm) than on Jason-‐2 (2.3-‐5cm)
• Implica1on: SWOT’s nadir al?meter would benefit from the 40 Hz and small footprint of Al?Ka è Lower « noise floor » for becer comparisons with KaRIN
What scales can be well observed at 1 Hz for each mission?
• Following the SRD methodology, we define the « observability limit » as the intersec?on between the SSH slope and the 1Hz spectral floor (hump & noise)
1 Hz spectrum
Influence of the high wavenumber errors on observability
Jason-2
AltiKa Jason-2 slope and SWOT noise floor
Observable scales
• Observability limit slightly smaller for Al?Ka (35-‐75km) than for Jason-‐2 (50-‐90km)
• Implica1on: Al?Ka should be preferred to Jason-‐2 to analyze small scales in SSH profiles (not spectra)
Ar?fact mi?ga?on with 20 Hz records and with becer processing
• The ar?fact can be mi?gated on exis?ng GDR products if one uses the 20 Hz rate and with becer edi?ng schemes
• Comparison between a 1Hz edi?ng procedure (CalVal) and two 20 Hz edi?ng schemes: Labroue et al 2011 (blue) & Tournadre (red) à The hump is reduced by 30% or more
• Implica1on: Becer processing and 20 Hz are recommended for small scale oceanography analyses
The value of SARM al?metry for SWOT
• No hump observed on Synthe?c Aperture Radar mode from Cryosat-‐2 (fewer inhomogenei?es in very thin and small SARM footprint)
• SARM observability limit of the order of 30-‐50 km (Jason-‐2: 50-‐90km)
• Implica1on: Recommended to use the SARM dataset from CNES: 1 year processed in all SARM zones and extensions are possible (zone & period)
• Implica1on: the « interleaved SARM » discussed for Jason-‐CS is very acrac?ve for CalVal comparisons with KaRIN: it produces becer LRM + SARM concurrently
Using KaRIN imagery to understand 1D al?metry
• Karin will provide a 2D image of what tradi?onal al?meters have been integra?ng in their footprint and waveforms è Higher KaRIN resolu?on is desirable for such studies
• KaRIN imagery may help inves?gate what happens in « hump genera?on events » and improve decades of small scale al?metry records
• Implica1on: We need to collect reference datasets from KaRIN or AirSWOT with high resolu?on and nadir overlaps
LRM
SARM
Karin image (1 km)
LRM
SARM
Karin image (250 m)
Summary
• The hump ar?fact is a correlated SSH error visible only at full resolu?on
• Implica?ons for CalVal and Instrument Processing communi?es: – Need to understand the impact of OBP averaging with very realis?c simula?ons
and unusual contexts (inc. sea state and atmosphere) – Need to collect high-‐resolu?on datasets from KaRIN (to validate small scales of
KaRIN and to try and improve 20-‐30 years of al?metry) – Need to get overlapping records from Al?Ka and AirSWOT
• Implica?ons for oceanographers interested in small scales: – Prefer Al?Ka or SARM (CryoSat-‐2 and Sen?nel-‐3) to Jason-‐2 – Use the full 20 Hz resolu?on (or 40 Hz on Al?Ka) – Try and mi?gate the hump ar?fact with becer processing
(e.g. retracker) and post-‐processing (edi?ng, non-‐linear filters)
• Implica?ons for the design of future nadir al?meters: – SWOT’s nadir would be even becer with smaller footprint (Al?Ka-‐like or SARM)
and with 40 Hz rate (current baseline: copy of Jason-‐3) – Having an « interleaved SARM » on Jason-‐CS should be recommended by the SDT
Thank you for your attention
Reference: Dibarboure et al , 2014 (submiced to JTECH, revised drap available) Inves?ga?ng short wavelength correlated errors in low resolu?on mode al?metry
Noise removal in two arbitrary boxes
TOPEX spectra and Two-‐pass retracker
Benefit of SARM to inves?gate the hump ar?fact
SARM observability limit
Cryosat-2 SARM
Jason-2
1 Hz « noise » floor Observable scales
SWOT SWOT + Jason-CS + Sentinel-3Type: SWOT JasonCS Sentinel-‐3
XOVER dt in daysXOVER dt in days
Type: SWOT
Number of concurrent XOVER zonesNumber of concurrent XOVER zones
Arbitrary stacking of
calibratio
n segm
ents
Arbitrary stacking of
calibratio
n segm
ents
Arbitrary stacking of
calibratio
n segm
ents
Arbitrary stacking of
calibratio
n segm
ents