[Coral-List] Identifying bleaching resistant reef locations: is it possible?
Ruben van Hooidonk - NOAA Affiliate
ruben.van.hooidonk at noaa.gov
Fri Apr 8 10:07:41 EDT 2016
While I praise any efforts to improve on bleaching projections, when
characterizing the skill or accuracy of those predictions one should never
use percentage correct. Percentage correct is influenced by the frequency
of the event you are trying to predict. It is nicely illustrated with a
story about tornado predictions:
Finley (1884) devised a way to predict tornadoes and claimed an accuracy
expressed in proportion correct of 96.6%. It was shown however that by
predicting no tornadoes at all, proportion correct could be increased to
98.2%. This was because the base rate of tornadoes occurring—the number of
a tornadoes occurring divided by the number of all observations made—was
In coral bleaching it is the same, assuming that no reef ever bleached in
the period 1990–2007 then the global average proportion correct of
predicted versus observed bleaching events is 97.8%. This is because most
reefs have only been observed to bleach two times in the 1990–2007 period.
The prediction that no reefs have bleached is accurate but it has no skill.
To analyze the skill of a prediction you should use a proper skill score,
which could be a simple as calculating the hit rate and subtracting the
false alarm rate. For more information check out this manuscript:
van Hooidonk, R. J., & Huber, M. (2009). Quantifying the quality of coral
bleaching predictions, 28(3), 579–587.
I'm not trying to defend NOAA's products, or be critical of your
improvements, but if we want to compare techniques we need to use the right
tools (skill scores).
Ruben van Hooidonk
On Thu, Apr 7, 2016 at 10:09 PM, Scott Wooldridge <swooldri23 at gmail.com>
> Hi All,
> I have read with great interest the recent discussions on bleaching
> refuges. In contribution to this discussion, I have attached x2
> provisionally accepted articles that expand on these issues. Specifically,
> I provide important new insight into why some reef locations exhibit lower
> (or higher) resistance to thermal bleaching. Such information is crucial,
> since for as great as the NOAA SST products are, they are unable to
> delineate these important/informative sites in their predictions.
> I have developed a new bleaching prediction tool ‘BleachRisk’, which
> utilises the NOAA DHW products, but also includes additional drivers of
> reef-scale bleaching sensitivity (viz. thermal history, water quality and
> solar radiation). For the GBR, BleachRisk predicts with >85% accuracy
> compared with the ~60% of the DHW product alone. Indeed, for the current
> 2016 coral bleaching event on the GBR it predicts at >95%.
> Perhaps of greatest importance, however, I have also utilised the
> BleachRisk modelling framework to identify routinely monitored coral health
> attributes, such as the tissue energy reserves and skeletal growth
> characteristics (viz. density and extension rates) that correlate with
> bleaching resistant reef locations. The x2 manuscripts explain how
> bleaching resistant reefs can be readily identified by characteristic coral
> health signatures that exist outside of thermal stress conditions, thereby
> providing a rational means by which every reef site on the GBR (and
> elsewhere) could be ‘field-ranked’ in terms of coral health and thermal
> bleaching sensitivity *a priori*. Such information is of considerable value
> to local policy-makers and reef managers as it could be used to: (i) guide
> the process of prioritizing enhanced protection of bleaching-resistant reef
> areas, and (ii) monitoring the benefit of catchment-specific water quality
> improvement strategies that seek to increase reef-scale coral health and
> bleaching resistance.
> Please note, it is not my intention to be at all critical of the NOAA heat
> stress products in this research – indeed I think they are excellent
> products. It is just (as we all know) that bleaching sensitivity to thermal
> stress varies across reefs and regions, and that in the end it is
> understanding the reasons for why it varies that is the most crucial issue
> for coral reef management. I hope people find these contributions useful to
> that task.
> Here is the links to the manuscripts:
> Scott Wooldridge
> Catchment to Reef Management Solutions Ltd, Newcastle, Australia 2280
> Coral-List mailing list
> Coral-List at coral.aoml.noaa.gov
More information about the Coral-List