Australia's "Poster Children" updated and analysed

I read a great deal of absolute rubbish on the 'net about sea-levels. Someone speaking at the American Geophysical Union meeting recently claimed that Miami was suffering more frequent flooding due to sea-level rise. Ignoring for a moment that much of Miami was already below tidal high-water before it was built, proper examination reveals that the rate of subsidence is greater than the rate of sea-level rise. At Cape Canaveral, further north, the measured rate is just over 5mm/year, but CGPS (Continuous GPS) data shows the land close to the tide gauge to be sinking at 2.8mm/year - the greater part of the rise is due to subsidence along the coast.

A couple of years ago, I read what I considered to be an outrageous claim - that the Tidal Unit (formerly National Tidal Centre) of the BOM was publishing "exaggerated data" in the ABSLMP (Australian Baseline Sea Level Monitoring Project) series. Worse, that the data was "homogenised". Well, it's impossible to "homogenise" tide-gauge data. Each gauge uses a different base-line for measurements, referred to as "Tide Gauge Zero". It's a "virtual" reference level, referenced to a physical benchmark (the TGBM or "Tide gauge Bench Mark") at the gauge site, one of a set, the remainder being on the adjacent dockside or land.

I'll relate the full story in a later post, but for now just let's say it concerned the record for Sydney, and that for Port Kembla, some 64Km to the south. Here's the full chart for Sydney, to October 2016.

Sydney 1914-2016.   Source: BOM

While the long-term rate for Sydney is 1.02 mm/year, it's easy to see that any short-term rate depends entirely on the start point. Start in 1997 or 1998, when there was a "dip" due to the 1997-8 El Niño, and you get a very high trend; start in 1990, a much lower rate. I've plotted the rate for a 20-year sliding window, using annual average data:

Rate (mm/year) for 20-year sliding window; end year on X-axis.

Port Kembla (an ABSLMP station), installed in 1991:

Port Kembla 1991-2016   Source: BOM

Sydney 1990-2016 for comparison.

Sydney 1990-2016  Source: BOM

Both together, after Port Kembla data adjusted by +58mm - the offset due to the different "Tide Gauge Zero" benchmarks.

Sydney & Port Kembla compared; 1991-2016

The claim of "homogenistation" is egrarious bullshit, invented by the morally, evidentially and statistically challenged. Put even more forcibly, it's a lie.

On to Fremantle, another source of claims and misinterpretation. The full plot first:

Fremantle, WA 1897-2016

This is the chart for Hillarys (Boat Yard), on the far side of Perth from Fremantle, to the north.

Hillarys, WA 1992-2016

Fremantle 1992-2016 for comparison:

Fremantle, WA 1992-2016

And both together, with Fremantle adjusted down by 64mm:

Hillarys/Fremantle over-plotted; Fremantle adjusted down by 64mm

Hillarys starts somewhat lower than Fremantle, but they plot together from 2002 onwards. It's easy to show such correlation between adjacent stations around Oz, even if they're several hundred Km apart. Even if it were possible to "homogenise" gauge data, the tides do a perfectly good job already.

Data is sourced from The main BOM Tidal Unit for Sydney and Fremantle, and from the AMSLMP project for Port Kembla and Hillarys.

NOTE: I've just discovered that SONEL have updated with data for many Oz stations. The CGPS pillars are often very close to tide gauges. At Hillarys they're co-located, and the latest plot shows a drop of 2.78 mm/year:

Source: SONEL

Updates include data for many of the Pacific islands, including Sceptic poster-child Tuvalu. Watch this space.

The effect of ENSO on sea-level in the South-western Pacific and Australia

In a previous post, I explored the connection between ENSO ((El Niño/La Niña Southern Oscillation) and sea-levels at Darwin and Fremantle on the west coast of Australia, in particular, the remarkably close correlation between smoothed monthly average levels and a smoothed, trended multiple of the SOI (Southern Oscillation Index).

It had already occurred to me that if the correlation was so convincing, that perhaps it might be possible to adjust monthly average data by the multiple of SOI to remove its effect and reveal the underlying pattern and trend of sea-level change at such strongly-affected locations. First attempts showed that the tenfold multiple I'd used to show correlation was too high, and that a multiple of 7 was "just right" for ENSO adjustment. Lower multipliers didn't reduce the variability ("lumpiness") sufficiently, and too high a value increased variability in the opposite direction. Lucky 7 turned out to be the "Goldilocks" factor. Here's my revised chart for Darwin:

... and with 7 times SOI subtracted from monthly values

The revised chart for Fremantle for the same period

... and with 7 times SOI subtracted from monthly values

Those two removals are quite convincing, I'd say - "extreme ironing" indeed. Note that the trends for both removals are slightly higher than the originals, despite the post-1990s upticks having been removed. It's because the earlier trend from the mid-1970s was down, which effectively pulled the trend-lines down.

High rates of rise in the western Pacific (as shown by satellite sea-level maps) have been a thorn in the side of some sceptics for some time. They conveniently ignore the fact that sea-levels along the Pacific coasts of the Americas show low or negative rates on the maps, supported by tide-gauge data, and that the high rates in the west are also supported by tide-gauge data, when exactly the same time-spans as the satellite maps are compared. They also ignore the reasoned, researched and informed voices which explain that both phenomena are effects of ENSO.

The island of Pohnpei (Federated States of Micronesia) is in that western "hotspot", and using PSMSL data for the two tide-gauges covering the period I've been able to recreate the record from 1974-2012.

The "ENSO profile" being clear, I went ahead and adjusted the monthly data as before.

Majuro atoll is in the Marshall Islands group, and I've extended my previous reconstruction to end 2012, and added the SOI plot. Note that sea-level is lagging SOI on the extreme right.

Kwajelein is also in the Marshall Islands, not that far from Majuro. Sea level is clearly leading SOI on the right.

The uptick is the subject of a couple of recent posts on wattsupwiththat.com, ENSO not being very high on the list of possible reasons under discussion. Nils-Axel Mörner thinks it's due to subsidence because of recent building, but then he would, he doesn't understand ENSO and the magnitude of its effects. Here's the latest data to June 2013 for Majuro from the South Pacific Sea Level and Climate Monitoring Project.

As you can see, the uptick has now reversed, following the SOI back to zero.

Pago Pago, American Samoa, shows a less-satisfactory correlation overall, but it's still reasonably convincing. It also shows  the sharp recent uptick.

I've had a look at correlation on the eastern side in California, but it's less clear. California is well north of the equator, and ENSO is the Southern Oscillation after all. I'll see if I can find an SOI widget (or make one) for my sidebar. I'll add captions with source data links very soon.

Sea-level in Australia and the Southern Oscillation Index

I've read a fair bit about the Southern Oscillation Index (SOI) over the years, about how much ENSO (El Niño/La Niña Southern Oscillation) affects sea-level in the Pacific. I've noted effects, particularly the El Niño "dip" in the western and central Pacific, and the corresponding "spike" in the east (especially the US Pacific shore). However, my impression was that it was just the extremes, El Niño and the less-well defined La Niña that had any real effect on Pacific sea-level. Before I continue, it's worthwhile quoting what Australia's Bureau of Meteorology has to say about the SOI; it's succinct and informative:

The Southern Oscillation Index, or SOI, gives an indication of the development and intensity of El Niño or La Niña events in the Pacific Ocean. The SOI is calculated using the pressure differences between Tahiti and Darwin.

Sustained negative values of the SOI below −8 often indicate El Niño episodes. These negative values are usually accompanied by sustained warming of the central and eastern tropical Pacific Ocean, a decrease in the strength of the Pacific Trade Winds, and a reduction in winter and spring rainfall over much of eastern Australia and the Top End. You can read more about historical El Niño events and their effect on Australia in the Detailed analysis of past El Niño events.

Sustainted [sic] positive values of the SOI above +8 are typical of a La Niña episode. They are associated with stronger Pacific trade winds and warmer sea temperatures to the north of Australia. Waters in the central and eastern tropical Pacific Ocean become cooler during this time. Together these give an increased probability that eastern and northern Australia will be wetter than normal. You can read more about historical La Niña events and their effect on Australia in the Detailed analysis of past La Niña events. 

The ENSO Wrap-Up includes the latest 30-day SOI value, as well as other information on indicators of El Niño and La Niña events.

 The graph below shows monthly values of the SOI in recent years.

  Source:BOM

On the following page they have links to data tables; Wanting to create a spreadsheet of monthly SOI values (from 1876!) I was dismayed to find the table to be structured as years down and months across. However, using Excel's Copy and Paste-Special/Transpose functions I was able to do it, laboriously year by year. Here's the result for 1959 to May 2013. The reason for the not-so-obvious start year will become obvious very soon.

SOI index 1959-2013    Data source: BOM

I've added a 25-month (2 year) centred moving average to smooth out the spikes without suppressing the signal. A 13-month MA would seem to be more appropriate, but gives too "lumpy" a trace; a 37-month MA smooths just too much. Like Goldilocks' porridge the 25-month MA is "just right".

Comparing the SOI plot with a sea-level plot is easy, but I wondered if I could add the SOI to a sea-level chart in some way. One problem is that the signal is relatively small, and the other is that it varies around a (flat, obviously) zero value. I hit on the wheeze of "magnifying" the SOI signal, and normalising the start of the SOI moving average with the start of the sea-level moving average, incrementing the magnified SOI by the monthly sea-level-trend increment. I tried a factor of 10 for magnification, on the basis that the SOI signal is based on air pressure at sea level, and one hPa change leads to a 10 mm sea-level change (in the opposite direction). My assumption may not have had much maths behind it, but the "porridge effect" operated and it was "just right". Here's the result for Darwin:

Darwin complete sea-level record 1959-2012       Data source: BOM/NTC

I don't now what you think, but I'd call that a rather good correlation, especially over the right-hand half of the chart. Here's one for Fremantle over the same period.

Fremantle sea-level 1959-2013  Data source: BOM/NTC

For both Darwin and Fremantle, note that the recent (since 1994) sharp upward trend has begun a downward reverse, more clearly predicted on the SOI chart which extends to last month (May 2013). The large and broad downward bulge around 1983 corresponds with the intense (some would say most intense on record) El Niño of 1982-3. It shows up well on the 25m MA sea-level plot for Darwin (and most Australian stations), but sea-level rose at Fremantle during that event. All the other El Niños show up on both charts; the broad low during the early 1990s (1991-2 and 1994-5 El Niños, with just 1993 between) and the intense but shorter 1997-8 El Niño. The 2010 El Niño was a more subdued affair.

I think it's clear that the SOI doesn't just affect variations in sea-level, it drives them, at least on the west coast of Australia. In a future post I'll look at other Oz stations for correlation, and where correlation is poor, explore possible reasons.

Damp Data from Down Under - Australian Sea Level Update

Australia's Bureau of Meteorology finally go their act together and published sea-level data (to the end of 2012) at the end of May. Prior to that, the latest data (other than for the newer ABSLMP stations) was to the end of 2010. I've been hard at work updating and expanding my database, and will update my reference page soon. In the meantime, I present charts for the two long-term "poster-children", Fremantle and Sydney. Both much analysed, much discussed, and often misrepresented. A certain Andrea Boretti spent many hours, produced many spreadsheets and charts, wrote many pages, and tortured the Sydney data at length until it confessed that there wasn't much of note going on there - in other words, there was no sign of any significant acceleration in the rate of rise.

In my humble opinion, if your intention is to analyse the rate to identify any significant change, then do just that. There's no need for "sliding windows", spectral analysis, polynomial curve-fitting or anything else. For example, calculate the rate from the start year to a succession of years, e.g. 1900-1910, 1900-1911, and so on to the last year of data. I've made it a standard analysis in almost all of my spreadsheets, and shown examples in a number of posts. While the amount of variation in the long-term rate reduces with the data length, significant year-to-year changes are still clearly represented in addition to longer-period change. Here's the chart for Sydney - I've used a 121-month (10-year) centred running mean. The rate has increased a little from the previous chart's 0.89 mm/year (to 2010).

Sydney, NSW - Sea level 1914-2012  (Data source: BOM)

The plot of the annual rate shows a break in slope at 1997, and a small increase thereafter.

Sydney, NSW - trend in mm/year from date on lower axis.

Fremantle has seen a definite change after 1994 - the running mean shows an uptick after then. This is not surprising, all West and North-West Australia stations show similar upticks, an acceleration in fact; in the Fremantle case after around 40 years of little change, even a slight reduction after the middle 70's.

Fremantle. WA - Sea-level 1897-2012   (Data source: BOM)

The long-term trend plot reflects the change, with a break in slope at 1994, and an increasing upward trend after 1998.

Fremantle. WA - trend in mm/year from date on lower axis.

There are those who would have you believe that nothing of the sort has happened; "situation normal, no change", but the evidence is clear. More on this in a post in preparation. However, the fairly sharp increase in rate over the last couple of decades around the west of Oz and the far-western Pacific in general is matched by virtually no change on the eastern shores - North and South America, and little or no change in the central Pacific. It's quite clear to me that what the satellites have recorded over the last 20 years is fairly accurate - those who would have you believe otherwise never actually compare tide-gauge data for exactly the same period as the satellite timespan. If you've seen such claims, check out what was actually compared - nothing, just a few longer-term charts produced. "See, there's no comparison!" they say, when they've done no comparison whatsoever. More (in detail, with real comparisons) on that in a future post too.

The Nodal Tidal Cycle - more evidence

You've probably never heard of Port Pirie. It's a mining town and port (really?) in South Australia, north of Adelaide, though Australians might call it a city. In common with many ports in Australia it has a tide gauge.

Nothing very spectacular to see there, just a pronounced dip caused by the intense 1997/8 El Niño, followed by a step rise and a relative stasis, and a low overall rate of rise. Apart from the lesser dips going back to 1969, all caused by previous El Niños, there's no obvious signs of any cycles in the data. That data provided by the BOM's National Tidal Centre, has a standard deviation included for the hourly data which is averaged to create the monthly averages plotted on the chart. The standard deviation reflects the variability of the hourly data through the tidal cycles, and that variability is increased or reduced by the effect of the 18.61-year period of the nodal tidal cycle (NTC). If the standard deviation is plotted something quite spectacular appears.

Proof if any is still needed, of the lack of any great influence of the NTC on mean sea-level. Using the standard deviation to show the effect of the NTC on tidal variability is a technique I'd like to be able to claim as my own invention, but I have to give all the credit to one David Pugh, author of several excellent books on things tidal. It's obvious when you think about it - I just didn't think about it - doh! In common with other regions worldwide, some Australian stations show the cycle clearly (Port Pirie is just the clearest example) and some don't. Fremantle in Western Australia is another Oz station which clearly shows the cycle; it has a much longer record than Port Pirie, and the cycle has an evident upward trend. First the gauge record -

..... and the standard deviation showing the NTC -

Note the obvious upward trend - this doesn't indicate any marked effect on mean sea-level, but shows that the tidal range, the difference between low and high water, is increasing somewhat. I don't know whether this is a feature of west-coast sites or not, but Sydney on the east coast shows a much steeper downward trend. What the cause of these trends might be, I don't now, and I can't find any literature which might throw some light on it. Perhaps there's something here waiting to be discovered.