07.03.2014 Views

standardization of environmental data and information - International ...

standardization of environmental data and information - International ...

standardization of environmental data and information - International ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

3. What did we learn from DISCOL?<br />

3.1. Restrictions, number <strong>of</strong> samples <strong>and</strong> processing efforts<br />

A critical point <strong>of</strong> a proposed experimental design is that the<br />

samples will be segregated in different physical locations <strong>and</strong> that it is<br />

impossible to r<strong>and</strong>omise <strong>and</strong> intersperse samples receiving different<br />

treatment. Disregard <strong>of</strong> r<strong>and</strong>omisation <strong>and</strong> interspersion results in<br />

uncontrolled conditions in an experimental design because differences<br />

between the localities <strong>of</strong> two samples may already have existed before the<br />

experiment. This may result in unrecognised locality effects that would bias<br />

the treatment effects. Such a design can lead to misinterpretation <strong>of</strong> the<br />

results <strong>and</strong> would be regarded as pseudoreplication 27 .<br />

Although interspersion in both impacted <strong>and</strong> unimpacted areas<br />

certainly would be preferred, this feature is difficult to achieve in the type <strong>of</strong><br />

large-scale experiments we propose. The DISCOL experiment demonstrated<br />

that true replication could not be achieved within the experimental field<br />

because it was impossible to designate sub-areas not affected by<br />

resettlement <strong>of</strong> resuspended material that could serve as truly unimpacted<br />

control sites. Reliable interspersion <strong>of</strong> treatments would thus require a<br />

number <strong>of</strong> spatially well-separated experimental locations (at least four 28 )<br />

plus a corresponding number <strong>of</strong> reference areas. The resultant effort, both<br />

at sea <strong>and</strong> during <strong>data</strong> analyses over the entire programme, would be many<br />

times greater than under the proposed scheme <strong>and</strong> would go far beyond<br />

any reasonable or practical limits.<br />

In the present case, there is only a limited risk <strong>of</strong> biasing effects due<br />

to the pre-existence or later intrusion <strong>of</strong> differences between locations,<br />

because <strong>of</strong> the relative homogeneity <strong>of</strong> deep ocean basin environments.<br />

The proposed treatment area <strong>of</strong> approximately 2 nmi² is relatively large, <strong>and</strong><br />

dispersion <strong>of</strong> samples over that area should compensate for the potential <strong>of</strong><br />

pseudoreplication. Accordingly, care should also be taken to have similar<br />

inter-sample distances at the two reference sites. However, a certain<br />

<strong>environmental</strong> variability within the entire proposed target area cannot be<br />

excluded a priori <strong>and</strong> should be properly evaluated in the pre-baseline <strong>and</strong><br />

baseline studies. It is obvious that <strong>data</strong> analyses <strong>and</strong> interpretations <strong>of</strong><br />

these studies must be completed before the start <strong>of</strong> the experiment,<br />

because they constitute the basis for selection <strong>of</strong> the three experimental<br />

<strong>and</strong> reference localities (out <strong>of</strong> five baseline localities), which should be as<br />

similar as possible.<br />

INTERNATIONAL SEABED AUTHORITY 343

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!