Accuracy and Stability of Sea-Bird s Argo CTD and Work Towards a Better CTD for Deep Argo David Murphy Sea-Bird Electronics
The Sea-Bird Team This work represents efforts of many people over many years: Norge Larson, Kristi Anson, Dick Guenther, Rick Beed, Genevieve Howell, Jonathon Boyle, Carol Janzen, Mike Vorkapich, Joel Reiter, Ken Lawson
The Points of this Presentation To inform the Argo program of Sea-Bird s Argo CTD accuracy When it leaves the factory Lifetime performance of temperature and pressure To discuss conductivity drift in the context of climatological data corrections To inform the Argo program of Sea-Bird s expectations of performance of the Deep Argo CTD
Motivation Greg Johnson told us that 25% of his floats required salinity correction and many were drifting salty Dean Roemmich told us that the Deep Argo program depended on a CTD with better stability
The Data Sets Over 4 Float data sets downloaded from US GODAE site, /dac/aoml Float data was sorted by PI, 4 PIs are included Selected only floats with delayed mode processing for salinity, used profiles that had only 1 designations Ended up with 24 float data sets Effort required: Coding an ftp application Netcdf format read using intrinsic matlab functions About 3 man weeks to understand the data format and render the data into a state appropriate for this analysis Sea-Bird s calibration history for products that share sensor electronics and characteristics with the Argo CTD
Calibration Terminology: Slope (Span) and Offset Error Y = mx + b b is the offset, an error here is constant across the whole measurement range m is the slope or span Error is smaller at the low end of the range Error is larger at the high end of the range Error = Sensor - Truth 6 4 2-2 2 4 6 8 Measurement Range Slope or Span Error Offset Error No Error
Physical Standards are the Foundation of Temperature Accuracy Standards grade platinum thermometer (SPRT) is calibrated against physical standards Each data point represents 1 2 days work and more than 1 measurements 2 µc jump in Ga melt is caused by NIST calibration of standard resistor Temperature (Deg.C.) Temperature (Deg.C.).15.13.11.99.97 SPRT Data at Triple Point of Water.95 1994 1999 25 21 29.7652 29.765 29.7648 29.7646 29.7644 1 mdegree Full Scale SPRT Data at the Gallium Melt Point 1 mdegree Full Scale 29.7642 1994 1999 25 21
Estimating Temperature Stability SBE 38 Proxy: Calibration history of 411 SBE 38s Average error over - 1.5 32.5 degree range Shares thermistor and circuit with Argo CTD Argo CTD: 96 calibrations of Argo CTD 24 213 This one will never see the ocean Residual Error (C).1.8.6.4.2. -.2 -.4 -.6 -.8 -.1 Residual, (Degrees C).4.2. -.2 -.4 SBE 38 Drift (C) 2 4 6 8 1 12 14 Years After Manufacture +.2 -.2 SBE41cp 746-5 5 1 15 2 25 3 35 Temperature, Degrees C
What are We Doing For Deep Argo? (Temperature) Currently meeting requirement.3 Acquisition technique improves temperature noise performance Calibration apparatus Minor Improvement Temperature Noise (Crms).25.2.15.1.5. Single Measurement Multiple Measurement ~ 2 µc Improvement 1 2 3 4 5 Measurement Time ms
Estimating Pressure Stability Past hurdles Electro-static discharge sensitivity Failed glass metal seals Plots below are for an initial 424 floats Most have very small offset after 72 profiles Mode at 4 decibars possibly induced by seal failure Pressure Offset at Profile 1 Pressure Offset at Profile 72 25 25 Number of Floats 2 15 1 5 Number of Floats 2 15 1 5-1 -8-6 -4-2 2 4 6 8 1-1 -8-6 -4-2 2 4 6 8 1 Pressure Offset Decibars Pressure Offset Decibars
Pressure Drift Very difficult to document for Argo CTD No good comparable use case in other SBE products.5 SBE41cp 1511.5 SBE41 2419.5 SBE41cp 157.25.25.25 Residual, (% F S R). -.25 ~1 dbar Drift 5 years Residual, (% F S R). -.25 ~1 dbar Drift 4 years Residual, (% F S R). -.25 <1 dbar Drift 2 years -.5 5 1 15 2 25 3 Pressure (PSIA) -.5 5 1 15 2 25 3 Pressure (PSIA) -.5 5 1 15 2 25 3 Pressure (PSIA).5.25 SBE41 1299 ~2 dbar Drift 5 years.5.25 SBE41 933.5.25 SBE41 1924 Residual, (% F S R). -.25 -.5 5 1 15 2 25 3 Pressure (PSIA) Residual, (% F S R). -.25 -.5 ~5 dbar Drift 4 years 5 1 15 2 25 3 Pressure (PSIA) Residual, (% F S R). -.25 -.5 <1dBar Drift 2 years 5 1 15 2 25 3 Pressure (PSIA)
CTD Pressure Sensor Screening Process Starting in 29 Accelerated testing for seal failure method developed in collaboration with Druck 5 pressure cycles 3 psia 15 hours at 4 psia and 4 C Twice a week pressure vented Offset measured against barometric pressure Returned to 4 psia Hand selected after infant drift Drift must be within +/-.25 decibar +.25 dbar -.25 dbar HOURS - 15
Reducing Pressure Span Error Over Operating Temperature Range Dana Swift reports errors of as much as 3 dbars at 2 dbars Provoked investigation with CTD 2795 Currently sensor span temperature error is corrected with data at -1 and 35 degrees C with data from Druck Residual, (% F S R).1 + 1 dbar.5. -.5-1 dbar SBE41 2795 -.1 5 1 15 2 25 Pressure (PSIA) 3 psia 26-Sep-11 -. 26-Sep-11 -.2 27-Sep-11 -.2 27-Sep-11 -.2 27-Sep-11.1 No Error at Room Temp ~1 dbar Error at 1 Degree
Pressure Sensor Span Sensitivity Not Try multi-temperature Point Pressure Span Correction Improves error for mid range temperatures Linear Residual, (% F S R).1.5. -.5 -.1-1 dbar SBE41 2795 5 1 15 2 25 3 Pressure (PSIA) 26-Sep-11. 26-Sep-11 -.3 27-Sep-11 -.2 27-Sep-11 -.1 27-Sep-11 -.1 ~1 dbar Error at 1 Degree
Normalizing to Calibration Room Reform calibration equation to T Tr where Tr is room temperature at time of sensor calibration Temperature R e s id u a l, (% F S R ).1.5. -.5-1 dbar SBE41 2795 26-Sep-11. 26-Sep-11 -.1 27-Sep-11 -.1 27-Sep-11 -.1 27-Sep-11 -. -.1 5 1 15 2 25 3 Pressure (PSIA)
What are We Doing for Deep Argo? (Pressure) Extend screening to Deep Argo pressure Custom 7 decibar sensor from Kistler Many currently in field Have received second generation for evaluation with improved internal construction Implement improved calibration of pressure sensor span for temperature sensitivity
Estimating Salinity / Conductivity Stability Number of Floats Climatology measured salinity offsets from 24 floats deployed between 2 211 Histograms for 18 floats at profile 72 of salinity error at next to bottom bin 87.5% of floats within +/-.2 PSU at profile 72 14 12 1 8 6 4 2 Salinity Offset.3.2.1 -.1 -.2 -.3 1 2 3 Float Profile Number of Floats 1 8 6 4 2 Salinity Offset from Climatology Salinity Offset PSU Salinity Offset PSU
Salinity / Conductivity Performance in the Laboratory 97 calibrations 24 213 Error less than +/-.5 PSU Residual, (S /m ).1.5 +.5 PSU. -.5 PSU -.5 SBE41cp 746 equivalent -.1 Conductivity Range 7 S/m 1 2 3 4 5 6 7 Conductivity (Siemens/m)
Investigating Conductivity (Salinity) Stability Salinity performance measured against climatology PMEL reports 25% of their floats require salinity correction UW reports 5% require correction Salinity drifting fresh or negative typical of cell fouling Salinity drifting salty or positive not well understood Error at 2. degrees.323 PSU.263 S/m
Salinity Error Sources Error evaluated at potential temperature of 2. degrees, P and C are dependent variables Possible sources of salinity error Pressure error Temperature error Conductivity error Will examine each in turn
Pressure as Source of Salinity Error In our example: At 2. degrees and 2 dbars, climatologic salinity is 34.54 PSU The example CTD has an error of.3 PSU If temperature is correct and conductivity is correct, Pressure would have to have an error of 6 decibars This error would be evident in surface pressure measurement
Temperature as Source of Salinity Error Same example: At 2. degrees and 2 dbars, climatologic salinity is 34.54 PSU The example CTD has an error of.3 PSU If pressure and conductivity are correct, Temperature would have to be 3 milli-degrees off 3 mc drift in the SP6 thermistor is unlikely, SBE 38 has same thermistor and circuit topology Error Degrees C.2.15.1.5 -.5 -.1 -.15 -.2 SBE 38 Drift (C) / Year 5 1 15 Years of Service
Conductivity as Source of Salinity Error Error sources Conductivity Cell Fouling Negative Drift Reduces cell radius Electrode degradation Cell damage Can show positive drift Typically magnitude of error is high Conductivity Sensor Circuit Frequency of oscillation dependent on circuit components (resistors and capacitors) Indeterminate direction of drift
Plausible Mechanism Presume that fouling dominates negative drift and will not discuss further Identifying conductivity circuitry as a likely source for the positive salinity drift does not explain why some investigators are more impacted. Strategies: Investigate difference in climatology and CTD for clues for plausible mechanism for positive drift Use next to the bottom bin for each profile of each float If bin depth greater than 9 decibars Next to bottom bin assures flushed and equilibrated cell
Salinity Correction Criteria All 4 PIs are making corrections based on the same criteria and in the following order Salinity is corrected for pressure error based on the surface pressure of the preceding profile Cell thermal mass correction is applied to the entire profile.1 PSU magnitude is the correction threshold Once this is crossed the float salinity is corrected back to where the drift started Comments: Some subjectivity in correction criteria Quite a few cooks in the kitchen
ARGO Salinity Error Profile 1 Compared to Profile 72 Histograms are scaled to show distribution of corrections, number of floats corrected +/-.2 ranges to 5 25 2 15 1 5 25 2 15 1 5 UW Profile 1 for 747 floats -.2 -.18 -.16 -.14 -.12 -.1 -.8 -.6 -.4 -.2.2.4.6.8.1.12.14.16.18.2 UW Profile 72 for 687 floats -.2 -.18 -.16 -.14 -.12 -.1 -.8 -.6 -.4 -.2.2.4.6.8.1.12.14.16.18.2 25 2 15 1 5 25 2 15 1 5 Scripps Profile 1 for 861 floats -.2 -.18 -.16 -.14 -.12 -.1 -.8 -.6 -.4 -.2.2.4.6.8.1.12.14.16.18.2 Scripps Profile 72 for 531 floats -.2 -.18 -.16 -.14 -.12 -.1 -.8 -.6 -.4 -.2.2.4.6.8.1.12.14.16.18.2
ARGO Salinity Error Profile 1 Compared to Profile 72 Histograms are scaled to show distribution of corrections, number of floats corrected +/-.2 ranges to 3 25 2 15 1 5 WHOI Profile 1 for 397 floats -.2 -.18 -.16 -.14 -.12 -.1 -.8 -.6 -.4 -.2.2.4.6.8.1.12.14.16.18.2 25 2 15 1 5 PMEL Profile 1 for 418 floats -.2 -.18 -.16 -.14 -.12 -.1 -.8 -.6 -.4 -.2.2.4.6.8.1.12.14.16.18.2 25 2 15 1 5 WHOI Profile 72 for 253 floats -.2 -.18 -.16 -.14 -.12 -.1 -.8 -.6 -.4 -.2.2.4.6.8.1.12.14.16.18.2 25 2 15 1 5 PMEL Profile 72 for 376 floats -.2 -.18 -.16 -.14 -.12 -.1 -.8 -.6 -.4 -.2.2.4.6.8.1.12.14.16.18.2
Positive Conductivity Drift Correlates with Low Latitude Deployments More likely, correlation is with mixed layer temperature 3% 25% 2% 15% 1% 5% % Percent of Positive Drifters by Latitude -8-6 -4-2 2 4 6 8 4 Floats, 1 Positive Drifter Number of Floats Number of Floats 2 15 1 5 25 2 15 1 5 All Floats by Latitude -8-6 -4-2 2 4 6 8 Latitude Positive Drifters by Latitude -8-6 -4-2 2 4 6 8 Latitude
Possible Mechanism More water vapor inside floats at warmer latitudes Components on CTD circuit board effected by higher humidity Desiccants hold less water at warmer temperatures Sud-Chemie, Inc 211
What are We Doing for Deep Argo? (Conductivity) Verify mechanism through experiments with desiccant Hermitically enclose conductivity circuit board Test in-situ calibration of circuit excluding cell
Work in Progress Work in progress Humidity experiments to verify mechanism of positive conductivity drift Acquisition of environmental chamber to calibrate pressure sensor span sensitivity to temperature Qualification of Kistler 7 dbar pressure sensors Work planned Redesign of calibration bath to improve thermal noise Comparison of SBE Microcat post deployment calibration to climatological correction
Fini