NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center H.-M. Zhang R.W. Reynolds R. Lumpkin R. Molinari K. Arzayus M. Johnson T. Smith NOAA NESDIS, OAR, & CPO BAMS, 2009

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center Content Project Goals Design Strategy Simulation Experiments for Needed In-Situ Network Research to Operation Monitoring & Iterative Process A Few Other Slides

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center GOAL Using Observing System Simulation Experiments for a Cost-Benefit conscious design, implementation and operational monitoring of a global ocean observing system consisting of satellite and in-situ observations for a required climate assessment accuracy using sea surface temperature (SST)

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center RationaleRationale  Surface temperature observations, ~ 70% from sea surface, are used for climate change assessment & monitoring

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center RationaleRationale  Surface temperature observations, ~ 70% from sea surface, are used for climate change assessment & monitoring  For meaningful assessment, sufficiently accurate data are need. GCOS requirements for SST accuracy: 0.2 – 0.5ºC for weekly satellite bias correction on 500km scale

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center RationaleRationale  Surface temperature observations, ~ 70% from sea surface, are used for climate change assessment & monitoring  For meaningful assessment, sufficiently accurate data are need. GCOS requirements for SST accuracy: 0.2 – 0.5ºC for weekly satellite bias correction on 500km scale  Q: What is the minimum and/or optimal observing system for the above requirement?

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center RationaleRationale  Surface temperature observations, ~ 70% from sea surface, are used for climate change assessment & monitoring  For meaningful assessment, sufficiently accurate data are need. GCOS requirements for SST accuracy: 0.2 – 0.5ºC for weekly satellite bias correction on 500km scale  Q: What is the minimum and/or optimal observing system for the above requirement?  Approach: OSSEs to find the relationship between sat bias reduction and needed in-situ data density

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center 1) Buoy: ~ 0.5ºC 2) Ship: ~ 1.3ºC (thus ~ 7 ships = 1 buoy) 3) NOAA Operational Satellites - AVHRR Day: ~ 0.5ºC - AVHRR Night: ~ 0.3ºC Major SST Observing Systems & associated Random Errors: (Reynolds & Smith 1994)

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center Complementary Nature of In-Situ & Satellite Observations In-situ observations provide ground truth but are sparse over the global ocean Satellite observations provide superior spatial coverage but may have large scale biases

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center Complementary Nature of In-Situ & Satellite Observations In-situ observations provide ground truth but are sparse over the global ocean Satellite observations provide superior spatial coverage but may have large scale biases  Blended Analysis (combined sampling) Errors (e.g. using OA/OI) are small enough for the required accuracy

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center Complementary Nature of In-Situ & Satellite Observations In-situ observations provide ground truth but are sparse over the global ocean Satellite observations provide superior spatial coverage but may have large scale biases  Blended Analysis (combined sampling) Errors (e.g. using OA/OI) are small enough for the required accuracy  Need dense enough in-situ data to sufficiently reduced the satellite biases to the required accuracy

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center System Design Strategy Res Q: What is the needed in-situ network to efficiently and sufficiently reduce satellite biases to the required accuracy  STEP I: Need to understand the satellite bias characteristics  STEP II: Run simulated satellite bias reduction vs. buoy data density using OSSEs  STEP III: Evaluate combined actual ship and buoy obs to determine where additionally obs are needed

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center EOF Analysis: STEP I: Characterization of Satellite Biases STEP I: Characterization of Satellite Biases  for monthly AVHRR SST from 1990 – 2002  i=1: Pinatubo

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center EOF Analysis: STEP I: Characterization of Satellite Biases STEP I: Characterization of Satellite Biases  for monthly AVHRR SST from 1990 – 2002  i=2: Seasonal clouds & Saharan desert dust

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center STEP II: OSSE Design Needed in-situ data density depends on both the magnitude and spatial patterns/scales of the biases. Needed in-situ data density depends on both the magnitude and spatial patterns/scales of the biases. E.g., if the biases are the same over the global ocean, only 1 accurate in-situ data would be needed to correct the constant bias. Generally, the more complicated the bias patterns, the more in-situ data would be needed

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center STEP II: OSSE Design Needed in-situ data density depends on both the magnitude and spatial patterns/scales of the biases. Needed in-situ data density depends on both the magnitude and spatial patterns/scales of the biases. E.g., if the biases are the same over the global ocean, only 1 accurate in-situ data would be needed to correct the constant bias. Generally, the more complicated the bias patterns, the more in-situ data would be needed Large satellite biases are related to atmospheric phenomena Large satellite biases are related to atmospheric phenomena Characterization of past biases by EOF analysis

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center STEP II: OSSE Design Needed in-situ data density depends on both the magnitude and spatial patterns/scales of the biases. Needed in-situ data density depends on both the magnitude and spatial patterns/scales of the biases. E.g., if the biases are the same over the global ocean, only 1 accurate in-situ data would be needed to correct the constant bias. Generally, the more complicated the bias patterns, the more in-situ data would be needed Large satellite biases are related to atmospheric phenomena Large satellite biases are related to atmospheric phenomena Characterization of past biases by EOF analysis Monte Carlo simulations are used for future biases to find the relationship between in-situ data density and satellite bias reduction Monte Carlo simulations are used for future biases to find the relationship between in-situ data density and satellite bias reduction

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center Simulation Experiments SSTs are simulated by ‘truths’ + typical random noises: –The ‘truths’ are chosen as the climatological monthly means Simulated buoys are placed at regular grids –Experiments on varying grid size to find the corresponding satellite bias reduction –Simulated buoy SST: Tb(x,t) = Tg(x,t) + A * e(t) Tg = ground truth (monthly climatology) A = amplitude of buoy random error = 0.5°C e(t) is a Gaussian random time series with a zero mean and standard deviation of 1 Simulated satellite data are placed at the actual monthly satellite observations –Simulate monthly satellite observations (e.g., no retrievals with clouds), from January 1990 to December 2002 (t=1 to 156) –Simulated satellite SST: Tsi(x,t) = Tg(x,t) + Bi(x)* a(t) i=1 to 6 for six simulated bias regimes, represented by the chosen six EOFs: Bi(x)=EOFi(x) with a global max of 2°C Tg = ground truth (monthly climatology) a(t) is a Gaussian random time series with a zero mean and a standard deviation of 1 Random noises have little effects because of large number of monthly satellite observations

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center Satellite Bias Reduction vs. Simulated Buoy Density (BD) per 10° box. Dash line is an exponential model fit. PSBE reduces rapidly initially and then levels off with increasing buoy density. Optimal BD range is 2-5. A BD of 2 (vertical thin dash line) or more is required to reduce a 2°C bias to below 0.5°C. Satellite Bias Reduction vs. Simulated Buoy Density (BD) per 10° box. Dash line is an exponential model fit. PSBE reduces rapidly initially and then levels off with increasing buoy density. Optimal BD range is 2-5. A BD of 2 (vertical thin dash line) or more is required to reduce a 2°C bias to below 0.5°C. 2°C

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center STEP III In-Situ System Evaluation STEP III In-Situ System Evaluation In-situ (ship + buoy) data density computed on 10° grid boxes as Equivalent-Buoy- Density (EBD):  Typical random errors are 1.3°C for ships and 0.5°C for buoys, thus roughly 7 ships = 1 buoy to achieve the buoy accuracy, n ≈ 7 from

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center EBD maps in traffic light style  EBD >=2: Green – ok  1 EBD < 2: Yellow  EBD < 1: Red STEP III In-Situ System Evaluation STEP III In-Situ System Evaluation

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center EBD maps in traffic light style  EBD >=2: Green – ok  1 EBD < 2: Yellow  EBD < 1: Red Improvement in S. Ocean STEP III In-Situ System Evaluation STEP III In-Situ System Evaluation

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center EBD maps in traffic light style  EBD >=2: Green – ok  1 EBD < 2: Yellow  EBD < 1: Red Improvement in S. Ocean Actual EBD to Bias Error avg (60S, 60N) Actual EBD to Bias Error avg (60S, 60N) STEP III In-Situ System Evaluation STEP III In-Situ System Evaluation

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center YEAR PSBE

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center YEAR PSBE used as a GPRA Performance Measure (PM) OMB Gov Performance Results Act (GPRA 1993):  Achieving program results  Improving program effectiveness  Communicating effectiveness & efficiency OMB Gov Performance Results Act (GPRA 1993):  Achieving program results  Improving program effectiveness  Communicating effectiveness & efficiency PSBE PM for Managers PM for Managers

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center YEAR PSBE used as a GPRA Performance Measure (PM) Major budget commitment in 2005 with fastest PM improvement OMB Gov Performance Results Act (GPRA 1993):  Achieving program results  Improving program effectiveness  Communicating effectiveness & efficiency OMB Gov Performance Results Act (GPRA 1993):  Achieving program results  Improving program effectiveness  Communicating effectiveness & efficiency PSBE PM for Managers PM for Managers

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center YEAR PSBE used as a GPRA Performance Measure (PM) Number of surface drifters OMB Gov Performance Results Act (GPRA 1993):  Achieving program results  Improving program effectiveness  Communicating effectiveness & efficiency OMB Gov Performance Results Act (GPRA 1993):  Achieving program results  Improving program effectiveness  Communicating effectiveness & efficiency PSBE PM for Managers PM for Managers

NOAA OCO 7 th Annual Observing System Review Silver Spring, MD 27 – 29 October NOAA’s National Climatic Data Center Conceptual Model: Research-to-Operation & Decision Making Gov Performance Measure GPRA PM $$$$ Budgeting

Impact on SST PM by Reduced Ship Times Observing System Simulation Experiments (OSSEs): Scenarios related to various TAO buoy maintenance schemes Issue: What are the effects on the SST Performance Measure by reduced TAO buoy- maintenance cruises, particularly by the aging Kaimimoana ship? Related scenarios: –1) Reduced/eliminated (Kaimimoana) cruises also leads to reduced surface drifter deployments in the tropical area –2) No TAO buoy maintenance leads to reduced or unusable TAO data, but drifters are some how maintained as it have been (may be logistically difficult) –3) Is TAO array is enough w/o drifters in this area?

No moored OR Kaimimoana-dropped buoys: Error increase between 0.05 and 0.1°C (by each) No moored AND Kaimimoana-dropped buoys: Error increase between 0.1 to 0.2°C (combined) No surface drifters: Error increase by about 0.5°C SUMMARY: PSBE increases for FOUR Scenarios

Need More In-Situ Data: Different special regimes All seasons (Simulations w/o TAO Buoys) Satellite Retrievals: Surface Air Temp & Humidity etc - Need to reduce the RMS errors

Global Gridded Products Shown is for 6-hourly sea winds; note the simultaneous Typhoon Talim and Hurricane Katrina Conceptual END-TO-END PROCESS From Data Ingest to Blended Products & Services Marine Surface Observations GTS NoaaPort Satellite Ingest NCDCNCDC QC; Blending USERS Climate Research Decision making (observing network design & monitoring) Weather & ocean forecasts Ecosystem Marine transportation Wind/wave energy Outreach (education) Interactive Data Services

SURFA – Surface Flux Analysis Driver: Surface processes are key in improving NWP & climate model forecast skills Objective: SURFA is to develop the evaluation of near real-time NWP, Reanalysis & Climate model fluxes and related fields with high quality reference data Initiation: WCRP WGSF, WGNE, OOPC … Status: –NCDC as central archive & service –Documents: NWP archive specifications; Submission Agreements –Current Participation: NWP Centers: ECMWF, German DWD, Japan (JMA), Meteor-France, UK In-Situ: Ocean Reference Stations, Flux Towers, Climate Reference Network –Data available from NCDC: WCRP SURFA: Surface Flux Analysis

(Ge Peng)