230 likes | 337 Vues
This report explores user requirements for sea surface temperature (SST) products, focusing on a comprehensive survey of 108 respondents. Methods include literature review, web-based discussions, and a questionnaire to derive user expectations. Key findings emphasize the need for reliable, high-resolution SST data, proper documentation of uncertainties, and compatibility with existing standards such as GHRSST. The study also addresses the systematic characterization of uncertainty in SST retrievals and outlines specifications for future product development, aiming to meet established climate observation goals.
E N D
sst_cci Chris Merchant The University of Edinburgh
User requirements survey • Methods • literature review • lessons learned review • web-based discussions / interviews • questionnaire • Analysis of 108 completed questionnaire respondents
Approach to analysis of user requirements: e.g., spatial resolution Threshold Breakthrough Objective
User Requirements Summary • SST records longer than 30 years (breakthrough) • Phase 1 will cover 1991 – 2010 • L4 SSTs available within 1 week, 99% reliable • Homogeneous record always available, upgrades • Will carry into system specification for operations • Proper uncertainties and simple quality information • Pixel/cell flags • NetCDF available by ftp, CF compliant • Yes + GHRSST compatibility • Simple documentation … that describes all steps in product development (!) • Certainly algorithm and uncertainty information readily obtainable
User Requirements Summary • SST records longer than 30 years (breakthrough) • Phase 1 will cover 1991 – 2010 • L4 SSTs available within 1 week, 99% reliable • Homogeneous record always available, upgrades • Will carry into system specification for operations • Proper uncertainties and simple quality information • NetCDF available by ftp, CF compliant • Yes + GHRSST compatibility • Simple documentation … that describes all steps in product development (!) • Certainly algorithm and uncertainty information readily obtainable
User Requirements for SST • Skin SST retrievals and buoy-depth SST estimates • As planned • GCOS (2006) supports blending skin and “bulk”/in situ • 3 hourly analyses at 10 km resolution or better • Daily at 0.05 deg • Fundamental research for sub-daily analyses proposed as option • Bias: 0.01 K over 100 km scales • SST CCI target is to demonstrate 0.1 K over 1000 km scales • GCOS (2006) states 0.25 K with no indication of applicable scale • Stability 0.01 K, per decade, seasonally, diurnally • Our aim is 0.05 K • GCOS (2006) presents only 0.1 K per decade • Mix of L4 (analyses), L3 (regridded) and L2 (native)
Product Specification Process • Prepared by someone with EO experience within the CRG, advised by Science Team • Covering • file metadata • discovery metadata • document revision control • file format • file naming • Input constraints: GHRSST, CMIP5, CF and Guidance • “Data and Metadata Requirements for CMIP5 Observational Datasets“ • GDS2.0 takes precedence over CMIP5 where in conflict • Such conflicts will be debated within GHRSST • GHRSST community for international review
Consistency of ECVs: two aspects • Spatio-temporal consistency • Compatibility with • CLOUDS at L1B/L2 levels from same sensors • SEA ICE at L2/L3/L4 • COLOUR at L3/L4 – want to be able to co-analyse • SEA-LEVEL? • Estimation consistency • Use compatible auxiliary info: aerosol, winds … • Mutual benefit from joint retrieval (in principle) • CLOUDS (e.g., thin and/or subpixel allowing SST) • AEROSOL (correlations in geophysics and errors)
Starting point • Uncertainty estimation is part of retrieval • (Some) users need to know about variability of uncertainty – need an uncertainty for every SST • Components of uncertainty have different correlation properties. Propagation of uncertainty from L2 to L3 and L4 needs to address each component appropriately.
Uncertainty Characterisation • Six components to uncertainty • Random (precision / uncorrelated) • E.g., Radiometric noise: ~Gaussian NEDT, uncorrelated • Estimate by propagation through retrieval • Pseudo-random (precision / corr. sub-synoptic) • Algorithmic inadequacy • Correlated on synoptic space-time scales • Can simulate • Systematic (accuracy / correlated) • Forward model bias, calibration bias… • Prior error Merchant C J, Horrocks L A, Eyre J and O'Carroll A G (2006), Retrievals of SST from infra-red imagery: origin and form of systematic errors, Quart. J. Royal Met. Soc., 132, 1205-1223.
Uncertainty Characterisation • Contaminant (precision, accuracy) • Non-Gaussian, asymmetric, sporadic • E.g., Failure to detect cloud; retrieval error from aerosol • Various space-time scales • Sampling • Random: scattered gaps because of cloud • Systematic: clear-sky effect?, biased false cloud detection • Stability • Time variation of any systematic effect • Approach: model / quantify each element • Aim: reconcile modelled and observed uncertainty
Uncertainty estimation in Round Robin • SST uncertainty estimation is the reasoned attribution of uncertainty information to an estimate of SST • Algorithms for SST to include SST uncertainty • SST uncertainty estimates will be assessed for • BIAS • INDEPENDENCE • GENERALITY • IMPROVABILITY • DIFFICULTY