The choice of the statistical representation of "normal" levels and minimum number of samples that can be used did invoke a lot of discussion when our project was trying to decide what to use. Different interests favoured different percentiles - we were certainly keen to capture the essence of "normal and widespread" by using a high percentage of the data population.
In reply to Chris Dainton's comments, yes the uncertainty captured by using confidence limits does increase with declining sample numbers, though for most domains, especially lead, we do have more than statistically adequate number of samples. The strength of this work is the sheer number of compatible systematically sampled soil results we have access to. The work did highlight gaps in knowledge e.g. number of BaP results but the NBCs can be improved as more data becomes available. Don't get too hung up on the actual values - it is the methodology that is important. I copy below the last comments from the end of our final project report:
"8. Normal background concentrations for the contaminant domains are our best effort to define what is the upper limit of “normal” levels of contaminants in soil as described by the Part 2A contaminated land Statutory Guidance. They are not a planning or risk assessment tool and must be used in the context of the SG in the manner described in the TGSs [Technical Guidance Sheets].
9. We define the NBC as the upper 95% confidence limit of the 95th percentile. Other percentiles and their confidence limits are listed in the TGS supplementary information, should others wish to consider our definition of “normal” levels in the context of other statistical information."
I know people like to have a single value with which to screen things, and our use of confidence limits on the percentiles does introduce an element of fuzziness, but we believe it is important that we move away from single national values and use domain NBCs that do capture the potential errors that could be caused by low number of sample results.
The BGS web page for NBCs is now live at: www.bgs.ac.uk/gbase/NBCDefraProject.html . We soon hope to make the GIS resources used by the project (e.g. domain polygons) available from this page as WMS (web map service) delivered files. We will also put all the project pdf files into the NERC online Research archive (NORA) with links from the project publications list to make it easier to find these reports.
Chris Johnson
Dr C C Johnson
Team Leader
Geochemistry Baselines and Medical Geology
British Geological Survey (BGS) Keyworth
-----Original Message-----
From: Contaminated Land Management Discussion List [mailto:[log in to unmask]] On Behalf Of Chris Dainton
Sent: 10 October 2012 12:05
To: [log in to unmask]
Subject: Re: Technical Guidance Sheets on normal background levels of contaminants in English Soils are published by Defra.
Hello List
So have looked at Lead (which is of interest for us in urban areas and locally in Derbyshire).
The use of UCL95 for Lead Mineralisation & BaP Urban highlights the potential concerns raised earlier and it would be great if BGS could comment on here (hello to Chris Johnson from the BGS who posted today).
The CL (Lower, Middle & Upper) numbers on the 95th Percentile of the data (all mg/kg):
BaP Principal Domain:
LCL: 0.31
MCL: 0.39
UCL: 0.5
Very small sample size (71) and 1.6x increase from LCL to UCL
BaP Urban:
LCL: 1.2
MCL: 2.2
UCL: 3.6
Tiny sample size (32) and 3x increase from LCL to UCL
Lead Principal Domain:
LCL: 170
MCL: 170
UCL: 180
Huge sample size (>30,000) and minimal increase from LCL to UCL
Lead Urban Domain:
LCL: 770
MCL: 790
UCL: 820
Large sample size (7,500) and very small increase from LCL to UCL
Lead Mineralisation Domain:
LCL: 1600
MCL: 1,900
UCL: 2,400
Reasonable sample size (347) and 1.5x increase from LCL to UCL
This seems to illustrate well how the UCL95 decreases relative to the sample size. So the smaller the data set, the higher the NBC would be? Which could mean that sites screened using an NBC could be poorly categorised as having levels consistent with background concentrations.
Of course I'm assuming (incorrectly?), that as the sample size increases (all things being equal etc), that the LCL and UCL tend towards the MCL.
Perhaps another approach could be to consider the level of variance of LCL to UCL and adopt MCL (or even LCL) values, if the difference between LCL to UCL is considered to be too high.
Kind regards
Chris Dainton
Peak Environmental Solutions Limited
Unit 10, Aston Ind Estate, Parsons Lane, Hope, Hope Valley, Derbyshire, S33 6RB
http://peakenvironmentalsolutions.com/
http://www.linkedin.com/in/chrisdainton
--
This message (and any attachments) is for the recipient only. NERC
is subject to the Freedom of Information Act 2000 and the contents
of this email and any reply you make may be disclosed by NERC unless
it is exempt from release under the Act. Any material supplied to
NERC may be stored in an electronic records management system.
|
