Great Barrier Reef bleaching study; Karoly et al (April 2016), Part A.

Guest essay by Bob Fernley-Jones

Background:

This headline image is for a publically released article (the article) in the Australian university-partnered blog The Conversation in which ‘the study’ cited therein is of major concern particularly in the point highlighted with red underline:

It focussed on the month of March in 2016, when record high Sea Surface Temperatures (SST) in the Coral Sea prevailed and which coincided with an extraordinarily powerful El Niño unlike anything seen since 1998. Because accessible GBR data were not available until recently, it employed Coral Sea data to predict future GBR bleaching events.

The Australian Bureau of Meteorology (BoM) had reported March 2016 at 29.48 ⁰C, which was 0.08 ⁰C warmer than February, a mere 0.02 ⁰C warmer than twelve years ago in February, when March was inharmoniously 0.16 ⁰C cooler. Also at odds was that the SST’s in January and February are typically warmer than in March but only the anomalously cooler month was considered in the study.

Of major concern is that a Google search as illustrated in this cropped screenshot revealed thousands of hits on the coupled exact phrases as seen default bolded (hyphen optional):

It is still going with 4,210 hits on 22/July and 4,560 on 8/September/2016.

I’m advised via email from the University of Melbourne that the study is to be replaced:

“In the more comprehensive study, we analyse observational data for the narrow coastal GBR sea temperatures and the larger Coral Sea region in March and in Jan-March, as well as climate model simulations and palaeoclimate reconstructions”

All other information has been declined and anyway apart from it being effectively an admission of the former inadequacies it is not relevant to the subject of this review. The issue remains that the public domain is still widely misinformed with defunct assertions for three months after its original release. Requests to remedy this have been refused.

Synopsis and methodology:

Part A of this review employs only empirical data and comparisons with authenticated observations. In so doing it deliberately avoids consideration of some arguable theoretical or cause-and-effect arguments and omissions within either the article OR the study it contains. For instance, in their use of monthly SST data it disabled any analysis of the quite abundant findings in the literature that thermal bleaching may not be caused by only a function of SST x time, but also in the rate of

change in temperature^[1]. In order to avoid debate and distraction from the incontrovertible facts, such things will be deferred to Part B.

Importantly, the study was focussed on revelation of mass bleachings as found by aerial surveys in late March 2016 and was associated with reports of the hottest ever March in the Coral Sea, a tiny part of which is the GBR. Then, in late April with remarkable alacrity and before the 2,300 Km (1,400 miles) x ~3,000 reefs could be surveyed by divers, the five co-authors publically released their study without peer review into the public domain. The aerial surveys could not determine when the bleaching occurred (or how severe) but the author’s study assumed it was in March 2016.

Because of that focus on SST’s in the vast Coral Sea, correlations with past authenticated events and non-events on the GBR were explored. Of eight major GBR observations over the past twenty years there was no correlation with seven of them, and the only one that was a fit was arguably accidental among a host of other El Niño global fits.

Eight other major empirical conflicts were found elsewhere to show that even if the study were to pass peer review and be found to be internally correct for projecting future SST’s in the Coral Sea, it was not relevant to the shallow GBR. (For instance, there are significant zonal issues within the ~3,000 reefs and great length of the GBR of 2,300 Km/1,400 miles). That is not to dispute that there has been very high coral mortality reported (with some puzzling distributions) in the far north in 2016 but that the study is misleading and excessively alarming.

Note: [1] Rate of change: e.g. see page 4 in this substantial synthesis of various causes of bleaching.

Empirical Analysis; (in 9 progressive steps):

Step 1) The last twenty years in the Coral Sea:

Figure 1) shows twenty years of monthly SST anomalies^[2] in the period embracing the extraordinary El Niños of 1998 and 2016. The red circled numbers identify recognised GBR mass bleaching events and those circled in grey are other conflicting observations. Although 2016 fits the concurrent observation very well, seven failings are found WRT earlier events.

Note: [2] Anomaly is relative to the standard BoM 1961 -1990 average

Graphics typically use downloaded BoM data archived in Excel spreadsheet supplements

(1) The first recognised mass bleaching on the GBR (and globally) coincided with the 1998 “Super El Niño”. However, SST in the whole Coral Sea fell 0.07 ⁰C below its 20-year average. Unlike (3) and (8) where bleaching was greatest in the north region, it was reportedly most severe in the centre.

(2) The Coral Sea March SST averaged 0.29 ⁰C above (1) and 0.25 ⁰C above (3) mass bleachings and yet no bleaching was reported.

(3) This was an El Niño neutral mass bleaching confined to the GBR. However, the Coral Sea March SST was slightly below average.

(4) Coral Sea March SST was significantly warmer than (1) and (3) mass bleachings yet the Australian Institute of Marine Research (AIMS) does not recognise a significant event.

(5) AIMS reported localized high mortality thermal bleaching only in the far south particularly around the Keppel Islands and yet the Coral Sea March SST was near average.

(6) The recognised American global authority of Coral Reef Watch is in NOAA. They describe the second global mass bleaching per the 2010 El Niño (whereas AIMS reported 2002 as second on the GBR). However, there were no significant reports on the GBR for 2010.

(7) Coral Sea was significantly warmer in 2015 than (1) & (3) and approaching the record high of (8) and yet AIMS reported on 5/April/2015 significant coral recovery over the previous three years, putting aside heavy cyclone mortality (19.7 %) in the north.

(8) This correlates with the 2015/16 El Niño, global mass bleaching and the globally high record temperatures. Although the Coral Sea correlated with GBR observations in this single instance, it was not an exclusive indicator for GBR bleaching.

Thus there are seven years of conflicts versus one OK in twenty years which conflicts with the following reply comment given under the article by Prof Karoly:

“…GBR experts have confirmed that the sea surface temperature variations in the Coral Sea are highly correlated with those in the shallow waters of the GBR. That is why we included Ove Hoegh-Guldberg as a co-author in our study, as he is an expert on climate variations and coral bleaching in the GBR. In addition, the Bureau of Meteorology uses the observed sea temperatures in the Coral Sea to provide warnings of possible coral bleaching conditions to the GBRMPA [Great Barrier Reef Marine Park Authority]…”

Step 2) AIMS daily data & March typically coolest Month in the hot season:

The study held great significance in the record high SST in the whole vast Coral Sea (CS) in March 2016 and predicted future increasingly severe highs there for March with an assertion that it would be applicable to the GBR.

However, the GBR is located in the shallow waters of the Continental Shelf parts of which at low tides are even out of water with a multitude of varying ambient conditions arising. (to be elaborated in Part B). On the other hand, the Coral Sea has an average depth of ~2.4 Km over an area of ~4.8 million Km² (3 million miles²) together with complex bathymetry and erratic circulation; basic flows CSIRO animation here.

At the time of the study, GBR data were inadequate for purpose and only the readily accessible BoM monthly data for the Coral Sea where employed. However, and importantly, AIMS do have some daily water temperature records and in figure 2 their “long term” (LT) records (which vary in length but average at about ten years) reveal several issues in the use of coarse monthly data:

A monthly data plot is available in supplementary spreadsheets, (Fig 2b)

One of the features in past GBR bleaching events is that their distribution has varied inconsistently from severe to nil over the length of its 2,300 Km. The GBRMPA mortality distribution map for 13/June/2016 is quite intriguing especially around Lizard Island, (a proposed later post). The AIMS data also show a wide water temperature range compared with the more equable Jan–Feb–Mar averages for the whole of the Coral Sea. Those single temperatures do not address the higher variability on the GBR.

In addition not only is February warmer than March but March temperatures typically (and logically) drop more rapidly on the GBR compared with the Coral Sea. That observation is unsurprising at least because of the major difference in thermal mass.

Notice that in the Coral Sea 20-year average monthly data that March is typically the coolest month there too and again unsurprisingly it is more equable. Step 9) compellingly quantifies this issue further.

Step 3) Comparing February and March in the Coral Sea:

Step 2) confirmed that March is not typically the hottest month on the GBR or Coral Sea and figure 3) considers the five most significant events out of those first seen in figure 1):

· February in 1998 and 2002 better support the reported mass bleaching on the GBR than does March in the Coral Sea. (High global mortality was reported in 1998 but 2002 was unique to the GBR).

· February in 2001 was less contradictory than March because March was warmer than in 1998 & 2002 but February was cooler. (March 2001 per Coral Sea ideation should have endured bleaching if it were a correct predictor).

· Both February and March 2015 are highly contradictory to a period of good coral recovery from 2013 through 2015 according to AIMS (apart from the previously mentioned 19.7 % cyclone loss in the north).

Step 4) The BoM released SST data for the GBR after public launch of the study.

This new data [13] now enables comparison of anomalies in the typically hotter month of February with the March ideation:

· February GBR data is mostly a better fit to observations than March in the Coral Sea particularly in the big El Niño mass global bleaching of 1998 and the non-event in 2015.

· However, February 2004 is even more paradoxical than March given that AIMS does not recognise significant bleaching that year.

Potential explanations for that paradox are to be presented in Parts B & C, but quickly moving on:

Step 5) Anomalies versus actual temperatures.

The focus of the study was on a record-high SST in March 2016 which is presented by the BoM in their time-series graphs as a deviation from the BoM standard 1961-1990 average for that particular month, or an anomaly. The anomaly for March 2016 was most extraordinary, and outlying of the twenty-year trend. However, each month and site has its own declared average base for those anomalies (per drop-down menu options at the link immediately above). Thus, although an anomaly may be high for a particular month it may actually be cooler than a typically hotter month with a lower anomaly.

The primary cause of bleaching is attributed to high water temperature. Accordingly, figure 5) reviews temperatures versus anomalies for the conflicting years (those first featured in Step 1).

Coming back to the focus of the study on the anomaly for March 2016 in the Coral Sea, when GBR data (from the same BoM source) are considered together with February it can be seen that the dark red bars foretelling the study’s model are rather less significant than when first seen alone in fig. 1. Unfortunately, the GBR data are also single monthly temperatures for the whole of the GBR (the average of nine points) over which there is high regionality in bleaching events and in water temperature range, (Re figure 2), but that aside here are but four big issues:

• There are wide disparities between the GBR and Coral Sea data, particularly with respect to actual temperatures rather than anomalies.
• It seems that there should have been a major bleaching in February 2004 but not so according to AIMS.
• The High SST’s in the Coral Sea in 2001 and 2015 are strongly contradicted by the GBR data which corrects those ‘CS-should-have’ but actually ‘non-happening-events’ on the GBR.
• The data support a major bleaching in February 1998 and 2002 with 1998 being the warmest although AIMS reports that 2002 was more severe. (Whilst Coral Sea temperatures were below the 20-year average in March!)

Step 6) Monthly data versus daily, and regionality of events:

According to AIMS, (and others) coral bleaching and mortality can occur in periods much shorter than a month with higher temperatures.

Figure 6) compares water temperatures in the hot season on the GBR per AIMS together with their longer term (LT) trends.

The big aerial surveys finding mass bleachings were conducted in late March 2016 but they could not determine the timing of the events.

These graphics show that there are volatile short hot-periods which may fall in any of the three hot months or across a cusp, individually or severally. Although the sample in the north is very small, it does support assessments of the greatest coral loss there in March. However, in the centre and south, hot periods predominated in January and February or across cusps and they are contrary and notably bigger anomalies in the south.

• Monthly data are too coarse to reliably predict bleaching (according to AIMS), particularly in March when the shallow waters of the GBR rapidly cool.
• Modelling based on March in the Coral Sea also cannot emulate January and February events on the GBR or their regionality.

Step 7) March 2016 in the Coral Sea….a sample of 1:

Figure 7) shows that the two “Super El Niño” years deviate significantly from the long term averages, predominantly in February but with March 2016 being especially unique versus the past 20 years:

• Figure 7a shows that the two “Super El Niños” had a significant warming effect over the long term underlying trends on the GBR and by logical extension big La Niñas (which typically follow) will have a cooling effect, including that the article gives that La Niñas are also associated with increased cloudiness in the Coral Sea region. (Prof Karoly comment^[3])
• The high values in 2015 seen in figure 7b ^[4] also point to the big El Niño being the main driver in the Coral Sea because early 2016 is but the peak in a two-year oscillation.

Notes:

[3] Regional SST’s are not only typically cooler during La Niñas by virtue of changes in ocean circulation but also from reduced sunshine penetration. E.g. Prof Karoly advised in comments under the article: “El Niño changes the sunshine duration in the region by reducing the cloudiness in the Coral Sea region, not changing the length of day. This is consistent with its impact of reducing rainfall in eastern Australia.”

[4] Figure 7b was constructed from cropped copies of BoM time-series graphs [14]. Figure 7a was found from downloaded BoM data as archived in the supplementary Excel Spreadsheets. (As typically with the preceding graphics)

8) Coincidence of mass bleachings with big El Niños

The 2016 “worst ever” GBR bleaching was associated with a “Super El Niño” year the like of which has not emerged for 19 years and both 1998 and 2016 were also mass global events both in temperature and coral bleaching. Despite these coincidences with mass bleachings on the GBR, the article included these assertions (my bold):

“The decaying El Niño event may also have affected the likelihood of bleaching events. However, we found no substantial influence for the Coral Sea region as a whole…

… Overall, this means that the influence of El Niño on the Coral Sea as a whole is weak.”

Putting aside that the Coral Sea is not the GBR, figure 8a) shows all twelve months of the year in the Coral Sea and the red ellipses frame three years where six or more months are clustered at anomalously high levels. Those years are associated with recognised super or big El Niños and three global mass bleaching events as recognised by the world authority NOAA:

• Clearly the data show that El Niños do affect the Coral Sea SST’s but the seasonal timing of their impact is unpredictable
• The GBR local ENSO neutral mass bleaching in 2002 is not impressively reflected by proportionate Coral Sea SST anomalies.
• The high value in February 2004 appears to be faulty…..a non-event…..Parts B & C will be to offer some potential explanations

Figure 8b) shows a stronger twelve-month correlation in anomalies with El Niños on the GBR but a repeat of the paradox with the local ENSO-neutral mass bleaching in January 2002.

• Another problem WRT predictability from poorly understood phenomena is uncertainty in timing of ENSO events and how they impact on GBR modelling difficulties

9) Use of CS to model GBR also invalidated by big thermo-dynamic differences:

• In terms of monthly data, March is clearly the coolest and the Autumnal onset effects are more rapid on the GBR in March.
• Diurnal variability in atmospheric conditions and solar exposure etcetera are consequently more volatile on the GBR
• A question not explored here is whether SST’s rule supreme in predicting bleaching versus what might happen under greatly varying ambient conditions for corals when exposed during low tides, particularly WRT time of day and weather.
• Also, as with all the temperatures and anomalies reviewed here, they are mean values over day and night and another issue that could be explored is if this is the best metric to assess risk of bleaching

Discussion:

The combined resources of five authors from three Australian universities launched a non-peer-reviewed article into the public domain on 29/April/2016 which subsequently went globally viral, presumably in a widespread acceptance that it was authoritatively true. The writers included the media-popular Prof Ove Hoegh-Guldberg and Prof David Karoly.

Hoegh-Guldberg is described by his University of Queensland as: “…[an] expert for media…” and, as “…the inaugural Director of the Global Change Institute… … is deeply-motivated by a desire to communicate science effectively…”. Amongst many other things he has performed on TV with a scientific experiment to prove that the oceans are acidifying by blowing air from his lungs through a tube into a container of water + pH meter! But, moving on and in brief:

This Part A rebuttal focuses on the University of Melbourne’s (UniMelb) involvement in which Karoly has taken the lead in onsite comments and in email exchanges WRT the study and its headline claim of: “Coral bleaching … was made 175 times more likely thanks to climate change”. Oddly though, it seems that based on advice from the Head of Science at UniMelb, that the true author of the modelled statistic is co-author Andrew King who has remained conspicuously silent throughout.

It is not the first time that Karoly has taken the mantle of co-authorship in a field where he apparently has zero expertise, seemingly in order to advance his mantra of CO2 driven planetary doom. For instance, he was co-author in a UniMelb biology paper ‘Early emergence in a butterfly causally linked to anthropogenic warming’ that found from unnatural hatchings in a lab that when the data were transcribed to the temperature record for Laverton, (a declared UHI affected suburb of Melbourne) the poor creatures would be emerging ten days too early. For your entertainment I quote from his email reply to me on some issues that I raised in his role as a biologist (my underline): “…our study on the temperature trends and the changes in phonology of the common brown butterfly…” In effect he was declaring that increasing levels of CO2 were changing the voices of the butterfly and when I teased him on it he claimed it was a typo, although if he intended e instead of o they are far apart on qwerty keyboards. Oh, and many readers may be aware of the ‘Gergis et al’ hockey-stick saga.

In a series of emails and evasions by Karoly it became futile so I went to higher levels at UniMelb and after more evasions I eventually received in full:

· From Head of Science Prof Phillips on 27/July: I believe that Prof Karoly has addressed the issues raised in your various emails. Consequently, we consider the matter closed and will not be providing further responses.

· From Dean of Science Prof Day on 29/July: I believe that Prof Karoly has addressed the issues raised in your various emails. Consequently, we consider the matter closed and will not be providing further responses.

But, it was precisely that Karoly, Phillips and Day had absolutely not responded over these matters, including presentation of key graphics similar to those above. Also, despite that they claimed to be submitting a very different enhanced study* to an unknown journal, they have not responded to a request to retract the non-peer-reviewed faulty one gone globally viral in the public domain or to consider its review in that domain.

* It allegedly adds consideration of the warmer months of January and February and actual GBR SST’s rather than inappropriate Coral Sea data.

Conclusions:

1). A prime conclusion is that the claim of ‘175 times worse’ is false and it should be openly and effectively retracted from the public domain

2). Contrary to assertions in the article, the 1997/1998 and 2015/2016 El Ninos did significantly affect monthly SST’s in the Coral Sea but differently to the GBR. (One factor is in accumulative lag effects from earlier months)

3). Predictions for bleaching events on the GBR fail if founded on March average SST’s for the whole of the Coral Sea, particularly versus the big north-south distribution of water temperature and observed events along the 2,300 Km length of the GBR.

3). The 20-year averages for January and February are typically hotter than the wrongly chosen month of March.

4). A superior model would use daily SST’s (not monthly) on the GBR that analyses the entire hot season in at least three zones; North, Centre and South. However, the poorly understood ENSO and hot water poolings remain as an obstacle to reliable predictions.