The Global Drifter Program Observations of Sea Surface Temperature in the World s Ocean Luca Centurioni Lance Braasch Verena Hormannn Global Drifter Program Scripps Institution of Oceanography La Jolla, California, USA lcenturioni@ucsd.edu A scientific and technical workshop on traceability of drifter SST measurements, SO, October 13-14, 2016
OUTLINE What is the Global Drifter Program Historical SST measurements Modern SST measurements GCOS requirements, technology trends GHRSST requirements Summary
What is a Surface Lagrangian Drifter? THE SIO MINI SVP. VELOCITY and SST. DESIGNED TO FOLLOW OCEAN CURRENTS AT 15 m DEPTH Mini SVP technical specifications: Spherical ABS float, 35 cm diameter, or larger; Tether made of polyurethane impregnated wire; Holey sock drogue (length 6-8 m); Strain relief of urethane/sanoprene; SST (thermistor +- 0.05 C); Depth of SST sensor ~17 cm Drogue on/off sensor ARGOS telemetry and localization (accuracy: 150 1000 m) Iridium telemetry with GPS localization (accuracy: 50 m rms) Drag area ratio (=C Ddr Ø dr h dr /CD oth A oth ) ~ 40; Water following accuracy: ~1cm/s slip in 10 m/s wind Over 20,000 drifters have been deployed globally since 1979, most of them with temperature sensors just below the water line
What is the Global Drifter Program and drifter s distribution The GDP is a global array of Lagrangian surface drifters that measure ocean currents at 15 m depth, Sea Surface Temperature and Sea-Level air Pressure, which are both Essential Climate Variables (ECV). Over 20,000 drifters of various types have been deployed to date
Impact of In-Situ Air Pressure Data from Drifters on Marine Weather Forecasting SLP SSV SST The OSE study RMS error differences (denialcontrol), Sea level pressure In-situ sea level air pressure data from drifters were denied during assimilation in the ECMWF forecast model. The plots show normalised differences of mean sea level pressure root mean-squared errors between the control and denial experiment for November-December 2010. Red (blue) colours indicate degradations (improvements) in the denial experiment. Forecast ranges: 12h, 24h, 48h, 72h, 96h and 120h. The positive effect of the drifters data is felt most in the southern ocean, polar regions and mid latitudes. The fact that the beneficial effect of the drifter data is less apparent in the tropical regions is likely because there are few observations to deny in that latitude band. Centurioni et al. 2016, BAMS. In press
Impact of In-Situ Air Pressure Data from Drifters on Marine Weather Forecasting The adjoint model approach provides the contribution of each observation to some metric, the dry energy norm in the case of NASA and ECMWF. This metrics are maintained operationally and can be found at gmao.gsfc.nasa.gov/products/forecasts/systems/fp/obs_impact/ SLP SSV SST Drifters have the largest or nearly largest fraction of beneficial observations in most locations (globe, NH, SH) and in both seasons. Both forecasting and climate studies benefit from drifter data, especially in the southern ocean where the drifters are essentially the only source of in-situ SLP data. Centurioni et al. 2016, BAMS, in press
Use of Velocity from Drifters at Low Frequency: the Mean Dynamic Topography The men dynamic topography is computed by minimizing a cost function that contains the large scale sea level from GRACE (to constrain the planetary scales), wind, and the drifter data (to constrain the smaller scales) SLP SSV SST Maximenko, N., P. Niiler, L. Centurioni, M.-H. Rio, O. Melnichenko, D. Chambers, V. Zlotnicki and B. Galperin (2009). "Mean Dynamic Topography of the Ocean Derived from Satellite and Drifting Buoy Data Using Three Different Techniques*." Journal of Atmospheric and Oceanic Technology 26(9): 1910-1919.
Use of Velocity from Drifters at High Frequency: Tides Mean amplitude (a), rotary coefficient (b), inclination (c) and Greenwich phase (d) of M2 tidal currents averaged in 2º x2º bins. Amplitudes in excess of 5 cm/s are saturated. For the inclination, bins with small (semi-major axis < 3 cm/s) and near-circular (rotary coefficient larger than 0.9 in absolute value) tidal currents are excluded. a c SLP SSV SST b d Poulain, P.-M. and L. Centurioni (2015). "Direct measurements of World Ocean tidal currents with surface drifters." Journal of Geophysical Research: Oceans 120(10): 6986-7003.
Why is SST from Drifters Important All drifters from the NOAA funded Global Drifter Program and its partners provide in-situ SST data, returned in real-time, and available from the Global Telecommunication System (~31K SST obs/day). SIO is handling the GTS posting of the new generation of Iridium drifters Other sources of in-situ SST include Argo floats, tropical moored arrays, coastal buoys, and ships of opportunity All other sources combined do not provide the same global global distribution and quantity of observations that the drifters provide
MOTIVATIONS: CAL/VAL OF SATELLITE SST AND IMPACT OF THE IN-SITU SST DATA FROM DRIFTERS. Why 1,250 DRIFTERS ARE NEEDED? SST GPRA Results vs Size of Drifter Array Left: SST satellite bias vs number of drifters in the GDP array. Such bias is reported quarterly as per the Government Performance Result Act. The required number of drifters to keep the SST bias below 0.5C is 1,250. Zhang et al., BAMS, 2009. Right: time series of SST biases. Note the period of increase biases between 2012 and 2104, when the number of drifters in the global array was as low as ~860 due to severe technical issues. The plot to the right shows the dramatic implications of the loss of in situ-data on the quality of SST products from satellite that are important for climate and weather forecast as well as basic research. A larger number of independent observations from drifters will further reduce the SST satellite bias.
Equivalent Buoy Density Courtesy of Huai-min Zhang, NOAA Independent Observations, ships and buoys combined
Potential Satellite Bias Error Courtesy of Huai-min Zhang, NOAA Maximum satellite bias of 2 C is assumed
Limitations of the Historical SST Dataset from Drifters SST sensors was seldom calibrated, we relied on nominal accuracy of 0.1 C Geolocations was only accurate to 0.2-1.5 km due to the use of the Argos satellite system SST was reported whenever an Argos satellite was available (irregular temporal sampling) Possible contaminations of SST data due to overheating of the drifter s hull Possible contaminations of the SST measurements when the probe is exposed to air due to wobbling of the surface float when the drogue is lost Inefficient sampling algorithms (straits averages of less than 30 values) were used The depth of SST observations was never measured
How SST is measured in modern drifters Example of the SIO drifter We use high quality thermistors, with an accuracy of 0.05 C, replaceable Accuracy: Accuracy verified with a 5 point calibration Sensing Range:-5 C to 40 C As in the past, the thermistor bead is encapsulated in stainless steel tip with thermally conductive resin A thermal barrier insulates the bead/resin within the tip and from the inside of the hull Duty cycle: measured for 2.2 s at 45 hz The internal temperature of the drifter, as well as other parameters, is also measured and reported GPS sampled concurrently and the measurement time is taken from GPS fix, which is very accurate The SST and location data are transmitted at the top of the hour, reported to shore within 10 min The nominal drift of the temperature sensor is 0.002C over ~5 years @ 25C. For every batch of thermistors we purchase we keep a subset on the shelf for verification The cost of adding SST measurement is only $50/drifter
Opportunities for improving in-situ SST measurements from GDP drifters The GDP concurs that a better understanding of the real (vs nominal) accuracy of the drifter measurements, as well of the environmental variability of SST within the satellite footprint should be be obtained The improvement of satellite s SST retrieval algorithms may change the way in-situ data are used. Rather than for calibration, in situ SST data are sometimes used to validate the retrieved SST products Therefore there is a real need to understand how in-situ data can be compared to remotely sensed data in a meaningful way The GDP is transitioning to a 100% Iridium array. This implies that hourly observations can be returned and bigger messages can be sent in a timely manner At minimum, regardless of the need to evaluate and correct satellite SST, there is also an obvious need to assign realistic error bars to the drifter SST dataset The technology for real time subsurface measurements to 12 m also exists, an the stratification associate with temperature (and salinity) vertical gradients can be measured
Subsurface Temperature Measurements
The current in-situ observations is partly compliant with GCOS requirements GCOS SST requirements is 0.2 C 0.5 C on a 500-km grid for weekly time scales The current drifter array (1,250 drifters with current accuracy level) is partly compliant with the requirements to keep the seasonal and regional potential satellite bias error (Zhang et a. 2009) below 0.5 C (monthly, not weekly) A discussion on the real needs of drifter s SST accuracy is underway and density is beyond the scope of this presentation, but should quickly converge to a set of guidelines ranging from NWP to climate (optimization) Different requirements, such as the one outlined by the GHRSST group are more stringent. The PSBE (GCOS) metric assumes a buoy random error of 0.5 C. GHRSST indicates the need for 0.05 C accuracy
Issues with Drifter s SST (by Way of Comparison with Satellite Data) In general, globally average IR and MW SST from satellite have low biases, but regional and seasonal bias can be large, and therefore in-situ data are needed for calibration and/or validation Three way error analysis has indicated that the standard deviation of the random error of drifter s SST is 0.21 C, a factor 2 larger than the nominal specifications (e.g. O Carrol et al. 2008), however Several assumptions were used in the analysis, such as 16 km/3 hours matches so environmental & temporal variability matters Different sampling methods also matters: pixel average (satellite) vs point measurement (drifters) Clearly sub-optimal sampling schemes and inaccuracies of of the older drifters may also play a role
Issues with combining in-situ and satellite SST to deliver the best product, include: Climate requirements (from GHRSST terms of reference): given a global surface temperature change signal of 0.1K/decade, global average temperature time series should be stable to much better than 0.1K/decade to be able to distinguish the signal from the instability of the time series. Assuming an error of 10% of the signal to be negligible, it is prudent to aim for a stability of 0.01K/decade in records of global mean temperature to enable the monitoring of evolving climate change Required SST resolutions: 10 km global, 1 km regional This requires that drifters improve 1. Geolocation: this is already underway, GPS accurate to 50 m rms as part of the transition to Iridium to be completed by 2018. No more Argos drifters are being deployed. The problem of matching subpixel variability remains 2. Calibration of temperature sensors to 0.05 C. Several drifters are already compliant, but other sources of error need to be quantified
Activities that Could Improve the Organization, Implementation or Planning of International Efforts for SST observing The SIO response to the EUMETSAT/GHRSST call for proposal, If funded will deploy SVPB-HRT drifters SST (Standard and HR) High quality Temperature Probe, tank calibrated Accuracy: 0.05 C Sensing Range: -5 C to 40 C Pressure (depth) Strain gauge sensor. Accuracy: 0.1% FSO (~7 cm) Sensing Range: 0 PSI to 100 PSI Sea Level Air Pressure Honeywell Integrated Pressure Transducer (IPT). Accuracy: 0.04% FSO Sensing Range: 0 PSI to 20 PSI Absolute Data telemetry NAL Iridium 9602 SBD modem GPS High quality GPS engine Horizontal Accuracy: < 2.5 m
The GHRSST Drifter Pilot Project Enhanced hardware and sensor suite (previous slide) Better calibration of the SST probe Will lead to measurements of the sub-grid (1 Km) environmental variability of SST Monitoring of drifter s internal temperature to detect possible SST bias due to the overheating of the drifter s Hull will determine if a better thermal barrier is needed Assess and establish the benefit of improved incremental capability of drifting buoys for satellite SST validation, particularly for Copernicus Sentinel- 3 SLSTR
Conclusions The accuracy of the historical SST dataset from drifters may be lower than the nominal value of 0.1 C, probably by a factor 2 Maintaining the status-quo, partly compliant with GCOS requirements, offers room for some improvements at a marginal cost (drastically improved geolocation and more efficient sampling schemes) The drift associated with aging thermistors is not believed to be a major issue Stricter accuracy standards for geolocations and temperature have been proposed by the GHRSST group. The GDP aims at satisfying requirements ranging from NWP to climate application and is already implementing low-cost options that do not impact the drifter s lifetime Improvements of SST accuracy range from low-cost 0.05 C replaceable accurate probes, to more expensive fully calibrated 0.05 C, or better probes with faster time response The GHRSST pilot project should provide indications for the need of improving SST measuring techniques from drifters in about 2 years