* * * D R A F T * * *

Estimating globally averaged coastal sea-level rise from high-quality tide gauges

By Dave Burton

NOAA has done linear regression analysis of sea-level measurements from coastal tide gauges at (currently) 375 locations around the world, which they've divided into "U.S." and "global" (i.e., non-U.S.) spreadsheets. On August 30, 2015, I downloaded the two files (U.S. and global) from NOAA's web page, and combined them into a single Excel spreadsheet. For ease of sorting, I changed the U.S. station ID numbers by adding an "A-" prefix. I also added "average" and "median" lines at the end of the spreadsheet.

The average rate of sea-level rise from all 375 NOAA-analyzed tide stations is 1.28 mm/yr, and the median is 1.71 mm/yr:
http://www.sealevel.info/NOAA_AllStationsLinearSeaLevelTrends_2015-08.xls or

NOAA estimates that the global average rate of mean sea-level (MSL) rise is 1.7-1.8 mm/yr, which is faster than most of the best long-term tide gauges have measured. Some of the difference is probably due to NOAA's addition of model-derived GIA adjustments to the measured rates when calculating the average, to attempt to account for Post-Glacial Rebound (PGR). My guess is that they're using Prof. Richard Peltier's figures.

Prof. Peltier also estimates that meltwater load from the melting of the great ice sheets (~10k years ago) is causing the ocean floors to sink at a rate fast enough to cause a 0.3 mm/yr fall in sea-level, absent other factors. That number (0.3 mm/yr) is usually added to calculated "global average" sea-level rise rates, inflating the reported average, even though the resulting sum is not truly sea-level rise, and is not useful for projecting sea-level for coastal planning. It's an attempt to calculate what the rate of sea-level rise would be, were it not for the hypothesized sinking of the ocean floor.

Unfortunately, many of the tide station records in NOAA's expanded list of 375 are much too short to be appropriate for measuring long-term sea-level trends. The literature indicates that at least 50-60 years of data are needed to establish a robust sea-level trend from a tide station record. The shortest record in NOAA's list is Apra Harbor, Guam, with just 21 years of data, and 40% of the stations they've analyzed have less than 50 years of data. (The text at the top of NOAA's page says, "Trends with the widest confidence intervals are based on only 30-40 years of data," but 14 locations actually have less than 30 years of data.)

So I also made a version of this spreadsheet in which the 150 stations with records shorter than 50 years are omitted. Considering only tide stations with records of at least 50 years, the average and median rates of MSL rise (of the 225 remaining stations) are 0.90 mm/yr and 1.41 mm/yr, respectively:
http://sealevel.info/NOAA_AllStationsLinearSeaLevelTrends_2015-08_50yr.xls or

(I also tried limiting it to stations with records of at least 60 years, with very similar results: average 0.77 mm/yr, and median 1.37 mm/yr.)

The average (0.90 mm/yr) is probably unrealistically low, due to the disproportionate number of stations in northern Europe which see low or negative rates of measured sea-level rise due to PGR. The fact that the average is less than the median also suggests that there are a disproportionate number of low-end outliers.

I also tried another approach, in which I excluded the most extreme latitudes. I started with just the "50+ year" stations, and included only stations within a latitude range of 45 (i.e., I excluded stations above 45 north or below 45 south). The resulting average and median for 137 stations were 2.22 mm/y and 2.02 mm/yr, respectively:
http://www.sealevel.info/NOAA_AllStationsLinearSeaLevelTrends_2015-08_50yr_lowLat.xls or

That approach largely solves the problem of low-side bias introduced by stations which are affected by PGR (which lowers the calculated average), but it doesn't solve the problem of high-side bias introduced by stations affected by subsidence (which raises the calculated average). So the average (2.22 mm/yr) is probably unrealistically high. The fact that the average is greater than the median also suggests that there are a disproportionate number of high-end outliers.

So I tried another approach, this time explicitly eliminating "outliers." I started with just the "50+ year" stations, but excluded the 40 stations with the lowest rate of sea-level rise (including most of those experiencing falling sea-level), and the 30 stations with the highest rate of sea-level rise (including most of those experiencing severe land subsidence, like Galveston, which is built on sinking fill dirt). The resulting average and median rates of sea-level rise (calculated from 155 stations) are both 1.48 mm/yr:
http://www.sealevel.info/NOAA_AllStationsLinearSeaLevelTrends_2015-08_50yr_less_high30_and_low40.xls or

That figure, 1.48 mm/yr, is my current best estimate of globally averaged coastal sea-level rise. At first glance, excluding more low outliers than high outliers might seem to bias the result to the high end. But I think it is justifiable, because of the disproportionate number of northern European and North American stations at locations where the land is rising due to PGR. The fact that the median and average are equal suggests that there aren't disproportionate numbers of either high or low outliers. (I also tried excluding the low and high 35 stations, and the result was an average MSL rise of 1.36 mm/yr, and median 1.41 mm/yr, which suggests that it included slightly more low outliers than high outliers.)

Note that if you add Peltier's 0.3 mm/yr GIA to the calculated 1.48 mm/yr global average rate of MSL rise, the sum is within NOAA's 1.7-1.8 mm/yr range.