Improving Sea Level Record in Arctic using Envisat Altimeter Measurements

Size: px

Start display at page:

Download "Improving Sea Level Record in Arctic using Envisat Altimeter Measurements"

Logan Douglas
5 years ago
Views:

1 Improving Sea Level Record in Arctic using Envisat Altimeter Measurements P. Thibaut, JC. Poisson, D. Hoang Collecte Localisation Satellite, Toulouse, France, G. Quartly, A. Kurekin Plymouth Marine Laboratory, Plymouth, UK J. Benveniste ESRIN/ESA

Context : Sea Level Climate Change Initiative The Arctic is an important component of the climate system whose exact influence on ocean circulation

The Arctic is also very sensitive to global warming and some direct consequences like ice melting are particularly visible.

the understanding of rapid changes occurring in this region.

2 Context : Sea Level Climate Change Initiative The Arctic is an important component of the climate system whose exact influence on ocean circulation is still poorly understood. The Arctic is also very sensitive to global warming and some direct consequences like ice melting are particularly visible. Page 2 In this context, extending the knowledge of the sea level variability as far as possible in the Arctic Ocean is a valuable contribution to the understanding of rapid changes occurring in this region. In the frame of the ESA Climate Change Initiative (Sea Level ECV), Collecte Localisation Satellite and Plymouth Marine Laboratory have been funded by ESA to improve the Sea Level closure budget in particular estimating the Sea Level Anomaly over arctic iced regions. 10 years of Envisat/RA-2 altimeter data have been reprocessed to deliver an improved long MSL record in Arctic.

3 3 main steps: General Sea Level CCI flowchart for Arctic RA-2 data Page 3 1) Floe / Lead / Ocean discrimination Editing criteria Waveform peakiness NN classification on waveform shapes Validation wrt NSIDC Validation with TerraSARX images 2) Sea Level Height estimation with waveform retracking (over water surfaces leads or open water) 3) Sea Level Editing/Smoothing/Mapping 2 1 Retrack Leads Filter Leads Region Selection General Filtering Specular/Diffuse Discrimination Floe/Lead/Ocean Discrimination Calculate Sea Surface Height Retrack Ocean 3 NSIDC Ice conc. maps Waveform peakiness

Step 1 : Floe / Lead / Ocean discrimination Envisat RA-2 waveforms Page 4 Ocean waveforms latitudes Waveforms from mixed regions leads Ice floes Two competing techniques have been implemented to

4 Step 1 : Floe / Lead / Ocean discrimination Envisat RA-2 waveforms Page 4 Ocean waveforms latitudes Waveforms from mixed regions leads Ice floes Two competing techniques have been implemented to identify leads on which sea level will be computed : Lead filtering (PML) : rise rate of the leading edge strength of echo signal (sigma0) position of tracking bin power in echo tail discrepancy with model waveform Neural Network for waveform classification (CLS) : (supervised classification)

5 Step 1 : Floe / Lead / Ocean discrimination Waveform classes for the Neural Network 12 classes are defined to describe the main observed waveforms. Knowing that the classification is performed on all ENVISAT measurements, it is important to define a class not only for waveform shapes of interest but also for all other waveforms. Page 5 Brownian Brown + pic Multi peak/ noise Distorted Brownian Peaky Shifted 2 Leading Edge Unknown ??

6 Neural Network waveform classification Ocean measurements are identified by class 1. Cycle 12 : Dec2002 Jan2003 Cycle 17 : July 2003 Page 6 Measurements affected by sea ice are mainly identified by classes 4, 6 (and a class 5 at borders). The different classes correspond to different types of surface backscattering. Peaky echo (corresp. to leads) are identified by class 2. Class 1 Cycle 19 : Aug - Sept 2003 The sea ice is detected and discriminated from ocean. Class 2 Class 4 Class 6

7 Evolution of sea ice types and leads with time Page 7 Class 4 sea ice waveforms with high backscattering Class 2 leads / peaky waveforms

Seasonal variations Page 8 80% 0% 2003 2004 2005 2006 2007 2008 2009 2010 Sea ice extent is maximum in February /

The proportion of peaky waveforms strongly increases during the sea ice melting and decreases as the freeze-up begins.

8 Seasonal variations Page 8 80% 0% Sea ice extent is maximum in February / March, the minimum is in September / October. The proportion of peaky waveforms strongly increases during the sea ice melting and decreases as the freeze-up begins. Sep 2007: Minimum Sea Ice Extent for 30 yrs NSIDC Ocean waveforms (class 1) Peaky waveforms / leads (class 2) Sea ice waveforms (classes 4, 5 and 6) Total: 4.3 million sq. km Courtesy Dave McAdoo, 2010

same proportion of leads except one region at very high latitudes

9 Consistency between CLS and PML results Page 9 CLS PML % of leads mesurements % of leads mesurements Very good consistency with the same proportion of leads except one region at very high latitudes (red circle) (All leads identified by PML are identified by CLS)

10 Evolution of lead population with time January 2008 March 2008 June 2008 Page 10 Patterns of leads concentration logically evolve with sea ice melting. The maximum proportion of leads is reached in June/July when melting is maximum. Leads are always found in all seasons but not in all regions. Issues to extend sea level estimates in certain regions/periods

11 Step 2 : Range estimation - retracking Different retrackers have been implemented based on models able to adapt from ocean waveforms (Brown model) to waveforms obtained over leads and ice floes (in addition to the classical sea ice retracker). PML and CLS retrackers are based on the same approach continuity between ocean and leads (no biases) Page 11 Measured waveform Fitted model position of the Epoch Ocean Lead

Bin number Bin number The off-nadir correction Page 12 The echo returned back by leads at some angle can

This is known as off-nadir effect that adversely affects retracking accuracy of the neighbour waveforms.

Hyperbolic signatures Across track 2D waveforms, corrected for tracker range and AGC The off- nadir

12 Bin number Bin number The off-nadir correction Page 12 The echo returned back by leads at some angle can produce hyperbolic signatures. This is known as off-nadir effect that adversely affects retracking accuracy of the neighbour waveforms. A new method for off-nadir correction has been developed at PML The off-nadir effect Waveform anomalies Hyperbolic signatures Across track 2D waveforms, corrected for tracker range and AGC The off- nadir correction Waveforms Correction for AGC Along track Estimation of across track peakiness Detection and flagging waveform anomalies Floe/Lead/Ocean Discrimination The off-nadir filtering

13 SSH above mean sea level, mm Absolute val. Bin number MAX/Avr Along track waveform index The off-nadir correction Waveforms, db scale Waveform MAX Local average MAX/Average Threshold Detected spikes Off-nadir effect Giles retracker, no correction Ext. Brown retracker, off-nadir correction

14 SLA monthly gridded products Page 14 Monthly SLA gridded maps have been produced for the entire period of Envisat They have been computed from Envisat/RA-2 measurements over ocean and leads after editing/filtering (hooking effects for example)

Similar work has been undertaken with Altika measurements (CNES Peachi project) Page 15 Saral/AltiKa altimetry mission has been launched in 2013 Works in Ka band on the historical Envisat tracks

15 Similar work has been undertaken with Altika measurements (CNES Peachi project) Page 15 Saral/AltiKa altimetry mission has been launched in 2013 Works in Ka band on the historical Envisat tracks Better spacial sampling of the AltiKa measurements (40Hz along track, reduced waveform footprint) Similar SLA mapping over the arctic regions has been implemented producing SLA gridded maps. Saral/AltiKa

16 Conclusions A discrimination of ocean / ice floe / leads has been implemented from RA-2 waveforms (based on a Neural Network or editing approach) and run over the entire ENVISAT period. Waveforms identified as coming from leads have been retracked and the results filtered to derive The quality of SSH measurements has been improved by improving: the accuracy of classification into leads, floes and open ocean the properties of retracker applied to each class Monthly gridded Sea Surface Height maps have been produced over the Envisat / RA-2 period ( ) Comparison with existing products (from DTU) will be done in the frame of the CCI A similar processing has been developed from AltiKa data in the frame of the CNES PEACHI project with Ka-band advantages : lower microwave penetration, better precision, higher sampling rate, Page 16

17 Page 17 Thank you for your attention

STM Product Evolution for Processing Baseline 2.24

STM Product Evolution for Processing Baseline 2.24 PREPARATION AND OPERATIONS OF THE MISSION PERFORMANCE CENTRE (MPC) FOR THE COPERNICUS SENTINEL-3 MISSION Contract: 4000111836/14/I-LG Customer: ESA Document Contract No.: 4000111836/14/I-LG Project: PREPARATION