IN SITU OBSERVATIONS: SYSTEMS AND MANAGEMENT
223
purpose, the models need better QC’d data for which methods employing
scientific expertise are used to correct the data (drift and offset) and to
provide error estimates of the corrections. ARGO quality control procedures
will be discussed to highlight the different aspects.
5.2.1
Real-time quality control procedures for ARGO
Because of the requirement for delivering data to users within 24 hours
of the float reaching the surface, the quality control procedures on the realtime data are limited and automatic. 16 automatic tests divided in 4
categories:
¾ Gross error tests: date, position, float speed at drift, temperature,
Salinity
¾ Profile coherence: decrease of the pressure, spike detection, excess
gradient between two points, density inversion, constant value or
overflow for T or S
¾ Coherence between profiles: jump or big drift in temperature or
salinity between two cycles (see figure 7)
¾ Grey List: For the float in this list, all profiles must be checked by
an operator because their behaviour is "strange"
5.2.2
Delayed mode quality procedure for ARGO
The free-moving nature of profiling floats means that most float
measurements are without accompanying in situ “ground truth” values for
absolute calibration (such as those afforded by shipboard CTD
measurements). In general pressure sensors are regarded as good even if
time drift may be possible; no agreed method exist yet for ARGO but the
impact of pressure drift is not negligible: 5 dbar will result in a salinity drift
of 0.003psu. Temperature sensors perform pretty well and similar method
could be applied to detect temperature drifts.
ARGO salinity delayed-mode procedures rely on statistical methods for
detecting artificial trends in float salinity measurements. However, since the
ocean has inherent spatial and temporal variability, ARGO delayed-mode
quality control is accurate only to within the associated statistical
uncertainties.
Using 2-stage objective mapping methods, salinity data mapped from a
historical database of existing profiles can be compared to float
measurements. Careful analysis of the spatial and temporal scales of the
mapping gives realistic confidence levels for the mapped values. A weighted
average in the vertical (giving more weight to stable water masses) results in
a single salinity offset for each float profile, as compared with the mapped
223
purpose, the models need better QC’d data for which methods employing
scientific expertise are used to correct the data (drift and offset) and to
provide error estimates of the corrections. ARGO quality control procedures
will be discussed to highlight the different aspects.
5.2.1
Real-time quality control procedures for ARGO
Because of the requirement for delivering data to users within 24 hours
of the float reaching the surface, the quality control procedures on the realtime data are limited and automatic. 16 automatic tests divided in 4
categories:
¾ Gross error tests: date, position, float speed at drift, temperature,
Salinity
¾ Profile coherence: decrease of the pressure, spike detection, excess
gradient between two points, density inversion, constant value or
overflow for T or S
¾ Coherence between profiles: jump or big drift in temperature or
salinity between two cycles (see figure 7)
¾ Grey List: For the float in this list, all profiles must be checked by
an operator because their behaviour is "strange"
5.2.2
Delayed mode quality procedure for ARGO
The free-moving nature of profiling floats means that most float
measurements are without accompanying in situ “ground truth” values for
absolute calibration (such as those afforded by shipboard CTD
measurements). In general pressure sensors are regarded as good even if
time drift may be possible; no agreed method exist yet for ARGO but the
impact of pressure drift is not negligible: 5 dbar will result in a salinity drift
of 0.003psu. Temperature sensors perform pretty well and similar method
could be applied to detect temperature drifts.
ARGO salinity delayed-mode procedures rely on statistical methods for
detecting artificial trends in float salinity measurements. However, since the
ocean has inherent spatial and temporal variability, ARGO delayed-mode
quality control is accurate only to within the associated statistical
uncertainties.
Using 2-stage objective mapping methods, salinity data mapped from a
historical database of existing profiles can be compared to float
measurements. Careful analysis of the spatial and temporal scales of the
mapping gives realistic confidence levels for the mapped values. A weighted
average in the vertical (giving more weight to stable water masses) results in
a single salinity offset for each float profile, as compared with the mapped