1 / 7

Introduction to delayed mode workshop for PIs

Introduction to delayed mode workshop for PIs.  In the 12 months since AST-5 in Hangzhou, lots have happened in the Argo delayed-mode realm: Argo netcdf Version 2, regional improvements, etc … It is therefore timely for AST-6 to discuss the PI’s role in the delayed-mode process.

inge
Télécharger la présentation

Introduction to delayed mode workshop for PIs

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Introduction to delayed mode workshop for PIs  In the 12 months since AST-5 in Hangzhou, lots have happened in the Argo delayed-mode realm: Argo netcdf Version 2, regional improvements, etc … It is therefore timely for AST-6 to discuss the PI’s role in the delayed-mode process.  Bear in mind that there should be a delayed-mode procedure for each parameter measured by the floats: PRES, TEMP, PSAL, ….  For salinity, there are two parts to the Argo delayed-mode process: (1). A semi-automatic part for identifying artificial drifts and offsets, where the methods used are repeatable, documented, and have quantified uncertainties. (2). A subjective part for identifying more subtle errors, that involves inspection of individual profiles by PI/expert.  The purpose of this workshop is to attempt to introduce some uniformity to the subjective PI/expert part.

  2. Argo salinity drift & offset QC procedures Float salinity that are deemed by RT to have a “good” relative vertical profile (ie. PSAL_QC = 1, 2, 3) are put through this process. Float salinity that are deemed “bad” by RT (ie. PSAL_QC = 4) are unadjustable. PSAL_QC = 4 PSAL_ADJUSTED = PSAL (original value) PSAL_ADJUSTED_ERROR = FillValue PSAL_ADJUSTED_QC = 4 Manual evaluation to detect anomalies on the relative profile, such as spikes, that are not detected in RT. Remove anomalies that may skew the drift/offset correction. Use in least squares fit. Do not use in least squares fit. PSAL_ADJUSTED = adjust as rest of profile PSAL_ADJUSTED_ERROR = FillValue PSAL_ADJUSTED_QC = 4 Compare with statistical recommendation obtained by accepted methods and reference database. Drift and/or offset is not significant < max [ 2 x statistical uncertainty, instrument precision ] Drift and/or offset is significant > max [ 2 x statistical uncertainty, instrument precision ]

  3. Drift and/or offset is not significant < max [ 2 x statistical uncertainty, instrument precision ] Drift and/or offsetis significant > max [ 2 x statistical uncertainty, instrument precision ] PI evaluation Accept float salinity as having no sensor drift or offset. No adjustment needed. Cannot accept float salinity as stable. Cannot accept statistical recommendation. Sensor drift and/or offset detected. Accept statistical recommendation. PSAL_ADJUSTED = value recommended by statistical analysis PSAL_ADJUSTED_ERROR = max [ statistical uncertainty, instrument precision ] PSAL_ADJUSTED_QC = 1 PSAL_ADJUSTED = PSAL (original value) PSAL_ADJUSTED_ERROR = max [ statistical uncertainty, instrument precision ] PSAL_ADJUSTED_QC = 1 PSAL_ADJUSTED = correction provided by PI (this can be original value PSAL) PSAL_ADJUSTED_ERROR = correction uncertainty provided by PI PSAL_ADJUSTED_QC = 1, 2 or 3 PSAL_ADJUSTED = PSAL (original value) PSAL_ADJUSTED_ERROR = FillValue PSAL_ADJUSTED_QC = 4 OR

  4. What is the role of the PI in the delayed-mode process? To determine the stability of the float data.  Determine that the reference database is sufficient for where your floats are.  Determine that the statistical method used is appropriate for where your floats are.  Determine that the statistical uncertainty levels are realistic. Determine that the drifts and offsets identified by the semi-automatic part are truly artificial and not due to ocean events, or determine that the float is stable.  Determine other instrument errors, other than artificial drifts and offsets.  Determine a correction, or determine that the float measurements are good. In both cases, determine an error bound.

  5. Why is it necessary to have this subjective PI part? Because there is no “absolute reference”, and because new instrument errors are still being discovered.  The semi-automatic part cannot distinguish water mass boundaries, fronts, etc., and most importantly, signatures of ocean events such as eddies, interannual variability, and decadal changes.  Other than artificial drifts and offsets, there are more subtle instrument errors to be identified, e.g. salinity spikes associated with sharp thermal gradients.  PI brings in expert info on the float instrument type, local oceanography knowledge of float sampling area, and other recent and close-by data.  Departure from traditional calibration: float delayed-mode calibration has no “absolute reference”. Hence it requires a new kind of thinking: one that synthesizes climatological analysis and climate change analysis.

  6. Hopeful outcomes from this workshop … Some basic guidelines for PIs So far the only agreed guideline we have is not to correct any drift or offset that is less than 2 x statistical uncertainty or instrument precision, whichever is greater. This means that Argo considers that a float is stable if its measurements deviate from climatology by less than 2 x statistical uncertainty or instrument precision. Do we want to put a ceiling on the statistical uncertainty? Also, there are ocean events that can deviate from climatology by more than 2 x statistical uncertainty. If a float sampled these events, it doesn’t mean that it wasn’t stable. What are some examples of these ocean events? How do we identify them? What do we quote as error bars for these events? If the float data are good but the local statistical uncertainty is large and we quote that as the data accuracy, we will turn the signal into noise. Means for exchanges of delayed-mode experience An email subscription list? Regular features in AST and ADT meetings? The Argo Delayed-Mode Manual I hope to have Version 1 on the GDAC sites by end of March. We need to agree on V1.

  7. Unresolved technical issues that need AST input …  Is the criterion max [2 x statistical uncertainty, instrument precision] ok? Do we want to put a ceiling on the statistical uncertainty? What do we put down for PSAL_ADJUSTED_ERROR when the data are good?  Correct a series (trend) versus correct individual profiles?  A sliding window for looking at series if correcting series.  A set of conventions for assigning salinity delayed-mode qc flags PSAL_ADJUSTED_QC = 1, 2, or 3? Do we want to use these to denote quality of dm adjustments, or do we want to use these to flag measurements that have been through real-time qc and delayed-mode qc, but are suspicious according to PI? Future … Very soon Argo profiles will become the primary large-scale ocean observing system within the global observing network. Users will treat delayed-mode data as true values. Calibration by a static climatological reference database will become increasingly inadequate, especially in the Southern Ocean. We need research and development of methods of calibrations that use nearby good floats in a semi-automatic manner.

More Related