DMI-OI analysis in the Arctic DMI-OI processing scheme or Arctic Arctic bias correction method Arctic L4 Reanalysis Biases (AATSR – Pathfinder) Validation results Conclusions/Challenges Future work

DMI-OI L4 processing chain Spatial resolution: 0.05 degrees Daily fields

Arctic Bias correction method Bias Standard deviations Dedicated Arctic bias correction method (Myocean R&D project ) Different temporal and spatial averaging scales tested 250 km spatial smoothing 11 days averaging window AATSR and NAVOCEANO-GAC products used for reference Best performance when both reference sensors are available, but also improvements with ATSR reference alone Bias Stddev Høyer et al. Multi sensor validation and error characteristics of Arctic satellite sea surface temperature observations. Rem. Sens Env, Høyer et al. A bias correction method for Arctic satellite sea surface temperature observations. Rem. Sens Env., 2013.

L3 Validation Høyer et al., 2012, in preparation Bias against in situ observations before and after correction

Arctic Reanalysis  Daily SST fields from 1982 to 2010 without gaps, using:  Pathfinder version  AVHRRs on NOAA satellite series  4 km resolution  Daily nighttime observations from  Quality flags 4-7  ARC version 1.0  ATSR 1+2 on ERS 1+2, AATSR on ENVISAT  0.1 degrees resolution  Daily observations from 1991 to 2012  ICOADS In situ observations, version 2.5  From 1800 to 2012  QC by UK Metoffice. Observations passed all quality control procedures.  OSI-SAF sea ice concentration  Based upon the SSMR + SSMI MW observations  1978 – 2012  10 km resolution.

Reanalysis validation, overall results  Daily validation against in situ observations not included in the analysis  3 types of in situ observations  Drifting buoys  Moored buoys  Ship observations  Nighttime:  Overall bias < 0.1 K  Stddev ~0.6 K Bias Standard deviations Number of comparisons

Validation of Reanalysis  Stable bias results  Drifters and Moored bouys better than ship observations.  Temporal changes in observational network:  1980s: Mostly ship obs  2000s: Mostly Drifting buoys

Validation of Reanalysis, Yearly  Year 2002 to 2010  Stable bias results throughout the year  Yearly bias similar to ARC, 10 times more match-ups Bias Pathfinder OI Arc

Spatial pattern of OI errors  Year 1982 to 2010  Largest in Marginal Ice zone  Erroneous SST observations in ice 

L3 validation results, Bias  Mean bias for full record  Elevated biases in East Greenland PathfinderArc NightArc Day

L3 validation results, stddev  Largest stddev for Pathfinder  Largest close to ice edge PathfinderArc NightArc Day

Pathfinder – AATSR differences  Average daily Pathfinder-AATSR differences  Largest differences during summer  Similar corrections , special

Monthly Pathfinder – AATSR differences  Mean bias in July, year 2003 to 2009  Generally largest in Baffin Bay, Barents/Kara Sea  Ice edge effect not seen

Validation in MIZ  Ice edge defined as ice concentration = 30 %  Matchups with buoys year 2002 – 2010  Distance to ice edge calculcated for match- ups  Different behaviour for Pathfinder and ARC  Cold bias in OI product close to ice edge.

Summary/Challenges  Overall performance of the L4 Arctic is good.  L4 Validation (spatial and temporal) resembles L3 validation  Pathfinder –AATSR biases agree with validation results, but region with largest biases can not be confirmed (no in situ obs)  Cold L4 bias very close to ice edge is from L4 processing  Challenging to find an accurate reference sensor in Arctic, when AATSR is dead  Treatment and inclusion of ice information along ice edge is challenging  Persistent cloudiness make MW SST attractive  No MW observations in coastal regions/archipelago

Future work/ideas  Improve performance in Marginal Ice zone  Check on foundation temperature  Test reference sensors  L4 comparisons, same input L2 data + ice mask  Assess effect of diurnal Variability  Include DV models in estimate  Derive new climatology from reanalysis