2015 - the Sheppard analysis

Diego Garcia Tide Gauge Analysis -a critique of Sheppard (2015)

In two recent newsletters (Chagos Conservation Trust News 45; and Ocean Digest Vol 2 pp 2-6) which can be found on the CCT website (http://chagos-trust.org/), Charles Sheppard presented data from the Diego Garcia tide gauge (DGTG) with his analysis and interpretation of sea-level rise using those records.

He concluded that:

1. between 1988 and 2000 sea level rose 5.4 mm yr^-1

2. between 2003 and 2014 sea level rose 5.96 mm yr^-1

3. that in both cases “the simple straight line fits are highly significant statistically”

4. that “the rate seems to be accelerating”

This brief critique shows why those results are flawed and emphasises the need for careful and rigorous science. An up-to-date and accurate assessment of sea level at Diego Garcia and in the Chagos sea area is available on the previous page: 2015 - March update

SUMMARY OF ERRORS

1. Sheppard’s datasets include months where there are substantial missing data, thereby producing an incorrect value.

2. Inclusion of these months does not meet Global Sea Level Observing System (GLOSS) quality control standards and overestimates the Ordinary Least Squares (OLS) trend in the time series by 24% for 1988-2000 and 6% for 2003-2014.

3. Sheppard incorrectly assumed that both of his monthly time series did not have correlated errors (autocorrelation).

Monthly Time series 1988-2000

1. The correct OLS trend is 4.35 (± 1.7 SE) mm yr^-1 (p=0.012).

2. After autocorrelation is accounted for the trend is 3.51 (± 3.68 SE) mm yr^-1 and is not statistically significant (p = 0.3416).

Monthly Time series 2003-2014

1. The correct OLS trend is 5.63 (± 2.62 SE) mm yr^-1 (p=0.034).

2. After autocorrelation is accounted for the trend is 5.198 (± 5.00 SE) mm yr^-1 and is not statistically significant (p=0.3019).

Charles also encourages everyone to plot the data using the Microsoft Excel spreadsheet program as he has done. I do not recommend this, nor do I recommend the use of Excel for regression. The correct use of regression techniques to analyse this data is complex and requires specialist knowledge and the proper scientific software if errors are to be avoided.

Additionally, Sheppard uses the term "eustatic" without further definition. It has been agreed by most of the sea level community that the term ‘eustatic’ not be used, or at least it be defined clearly when used (see Rickards et al [2013] Journal of Coastal Conservation 17:691-693.

The full detailed analysis is available in PDF format and can be downloaded from the link below. Since R is open source software I have also placed copies of the R script for running the analyses together with the data for 2003 - 2014 below.