Monday, September 24, 2018

The World According To Measurements - 20

Fig. 1 Latitude Layers
I am now using the new and increased quantities of the The World Ocean Database (WOD).

Each dataset type (APB, CTD, DRB, GLD, DBT, MBT, MRB, OSD, PFL, SUR, UOR, and XBT) has been placed into its own SQL table.
Fig 2a (T) APB, Layer 5
Fig 2b (SP) APB, Layer 5

Fig 3a (T) CTD, Layer 5
Fig 3b (SP) CTD, Layer 5
One reason for that is the variations in the different datasets.

For example, some of them do not have salinity measurements to go along with the temperature measurements.

Fig 4a (T) GLD, Layer 5
Fig 4b (SP) GLD, Layer 5
Today's graphs are made from those that have both types of in situ measurements, i.e. both T and SP.

The DRB, DBT, MBT, MRB, OSD, SUR, UOR, and XBT datasets contain plenty of in situ measurements for temperature (T) but are not adequate for salinity (SP).

At least not in the WOD layer (5) depicted in these graphs (Fig. 1).
Fig 5a (T) PFL, Layer 5
Fig 5b (SP) PFL, Layer 5

Any accurate world-view of the ocean forms in our minds according to measurements we have recorded.

Notice that the measurement streams recorded in the WOD, which I used to make today's graphs, vary according to several factors.

One factor shown in today's graphs is that the type of instruments used can have a huge effect on the results.

Other factors are the time of year, depth, and quantity of measurements.

Putting them all together into one mean average can reveal a trend when the individual ones by themselves can't (more about that below in this post).

The subject is vast (The World Ocean Database) but it can be more easily handled one step at a time (An inventory of Arctic Ocean data in the World Ocean Database).

The first step is to analyze the data types:
"(XBT) data provide one of the longest available records of upper-ocean temperature."
(The Impact of Improved Thermistor Calibration on the Expendable Bathythermograph Profile Data). It may sound good, but read the rest of the story in that paper to see that tons of work has been done to bring a lot of those measurements back to reality.

In the present situation I am going to take some time to move through all of the Latitude Layers (Fig. 1) to check out each data type one dataset at a time.

Just like I did today on Latitude Layer 5 (Fig. 2a - Fig. 5b).

Even though the WOD XBT data have no SP data, that dataset can be used for some temperature research if I figure out the best way to use that paper's suggestions to nullify any bias in that dataset.

When all of the datasets are used without careful consideration and selection (including especially the XBT dataset), they produce an inaccurate picture.

Fig. 6a (T) All layers, all datasets
Fig. 6b (SP) All layers, all datasets
The graphs at Fig. 6a and Fig. 6b
show the data gyrations when all of the datasets are used without any adjustments.

Notice that before the WW II years there is considerable gyration.

After the war years, say mid-forties, the data pattern stabilizes.

A glaring error is the in situ temperatures recorded. [see UPDATE below]

They indicate that the temperature at 250m deep is in the vicinity of 10 degrees C (~50 deg. F).

I suspect the XBT type data (all T, no SP) because the graph at Fig. 6b presents a more accurate picture after the WW II years.

The XBT has no SP measurements, so its influence is not present in the Fig. 6b graph as it is in the Fig. 6a graph.

I am going to do some more research using different combos of the various datasets to reach a best-use scenario.

For instance, what will Fig. 6a look like without the XBT (and other T only , no SP) datasets left out?

My expectation is that it will be more accurate.

Fig. 7a
Fig. 7b
I added logic to the software module to conform the in situ measurements to the maximum and minimum values set forth in the WOD Manual.

Those maximums and minimums apply to every ocean area on Earth, and does so at all depth levels.

Those maximum and minimum values apply to both T (in situ temperature) and SP (in situ conductivity, a.k.a. salinity).

The way it works is if a measurement is above the maximum, it is reset to the maximum for the calculations taking place.

In this context (non TEOS-10) I consider this the better practice over rejecting the value completely.

These value go back a couple of hundred years, so "the more measurements the better" possibilities come to mind (the individual, smaller datasets shown in Fig. 2a - Fig. 5b show the effect that lack of data has on graphs).
Fig. 8a
Fig. 8b

Anyway, as you can see (Fig. 7a - Fig. 8b) the expectation I had about the MBT and XBT datasets was not verified.

That is good news because, by adding the maximum / minimum validation to all datasets, we can safely use every  dataset in the WOD.

The "with and without MBT and XBT" datasets make virtually identical graphs (Fig. 7a - Fig. 8b).

So, now I can move on to the typical Dredd Blog processing of the full WOD datasets with the verification logic in place.

I plan, over the next few months to do graphs that were done previously on Dredd Blog using only the CTD and PFL datasets.

The main result will be a new look at the older data which in some cases goes back a couple hundred years.

The next post in this series is here, the previous post in this series is here.

No comments:

Post a Comment