RN WW2 Observations index
SST Diagnostics for the new marine obs
We hoped that the new observations would allow us to improve on
HadSST2 in two main ways:
- The greater range and number of observations would give us a greater coverage and lower sampling error.
- The new source of observations would inform the debate on SST biases - observation methods are changing rapidly in the
WW2 period.
Coverage
The figure below shows the improvement in coverage given by the new obs. The plot is the extra fractional coverage provided by the new obs (fraction of months during 1938 and 1941-7 (1938-47 with 1939 and 1940 removed), for each 5x5 degree square, where there are new obs but no ICOADS obs.) The increase is clear over most of the Atlantic and Indian Oceans, and largest around the African coast, in the Western Indian Ocean, the Mediterranean, and the UK coastal waters.
 |
|---|
| Fractional improvement in coverage (1938 and 1941-7). |
|---|
For more detail, see the
month-by month comparisons.
The next figure shows the global number of observations and total fractional coverage both before and after adding the new observations.
 |
|---|
| Coverage and number of observations with and without the new observations. |
|---|
Differences between new obs and HadSST2
Difference fields between HadSST2 and the new observations have been calculated for
each month. The figure below is the mean of the monthly differences over 1938 and 1941-7. (This uses the un-adjusted version of HadSST2, so no bias adjustments have been applied to either dataset.
 |
|---|
| Mean diffs: new-HadSST2 (1938 and 1941-7). |
|---|
In the Atlantic there are no obvious biases between the old and new data; in the Pacific, the new data may be systematically colder than HadSST2; in the Indian Ocean the new data appear systematically warmer than HadSST2
The next figure is a time series comparing monthly global average SST anomaly from HadSST2 (un-bias adjusted version) and the new data (also not bias adjusted).
 |
|---|
| Global mean from HadSST2 and the new observations. |
|---|
The increasing trend in the 1930's (delieved to be caused by the transition from bucket to engine-room-intake measurements) is present in both datasets, but the behaviour in the 1940s is distinctly different. The step reduction in 1945 (possibly an artifact caused by a change from UK to US ships in the ICOADS data) is not present in the new RN data, but the new data does appear to have a downward trend in the late 1940s.