Persistent elevated mortality framework in salmon farming data: EIR release
Information request and response under the Environmental Information (Scotland) Regulations 2024
Information requested
I am writing about my further request of 4 September 2026, reference 202600535957, which followed your decision 202600515523 of 21 May 2026 on the persistent elevated mortality framework.
Since I wrote, the peer-reviewed methods paper has been published: Moriarty et al. (2026), Decoding persistent elevated mortality in farmed marine Scottish Atlantic salmon (Salmo salar L.) using artificial intelligence and logical modelling, Preventive Veterinary Medicine, 107012, DOI 10.1016/j.prevetmed.2026.107012. I have obtained the article and have also located the screening code at https://github.com/MeadhbhMoriarty/PEM. Item 1 of my request is therefore met, and I would like to narrow the remainder under Regulation 9 so that it can be answered efficiently. I ask for the following, to the extent held:
1. Appendix 1 (process maps) and Appendix 2 (data preparation) of the paper, if they are not published as supplementary material with the article, and in particular Section A2.4, the structured inference guide, definitions and keyword dictionaries provided to Microsoft 365 Copilot.
2. The 500-record Fish Health Inspectorate expert-validated reference dataset used to evaluate the AI categorisation (Section 3.1 of the paper), with the model predictions for each of the 13 runs, in anonymised form if site identification is a concern. Site names and numbers may be removed or replaced with the FS site numbers already published in Annex B.
3. The categorised SGMD mortality-events file that the published code reads (named in the repository as "FHI 066 Mortality events 16092025_grouped_causes.csv"), that is, the three ranked mortality-driver categories and the production-stage flag per mortality event, in anonymised form if necessary.
4. The lists of sites at each stage of the framework: the 28 sites flagged for recurrent elevated mortality, the 25 possible persistent sites, the 20 reviewed sites and the 10 confirmed sites (Table 7 of the paper), by FS site number.
5. The outputs of the sensitivity analysis described in Section 2.5 and Table 4 (the percentage of sites classified at each combination of threshold, streak and cycles), in tabular form.
Response
As the information you have requested is ‘environmental information’ for the Purposes of the Environmental Information (Scotland) Regulations 2004 (EIRs), we are required to deal with your request under those Regulations. We are applying the exemption at section 39(2) of the Freedom of Information (Scotland) Act 2002 (FOISA), so that we do not also have to deal with your request under FOISA.
This exemption is subject to the ‘public interest test’. Therefore, taking account of all the circumstances of this case, we have considered if the public interest in disclosing the information outweighs the public interest in applying the exemption. We have found that, on balance, the public interest lies in favour of upholding the exemption, because there is no public interest in dealing with the same request under two different regimes.
This is essentially a technical point and has no material effect on the outcome of your request.
Question 1 - Appendix 1 (process maps) and Appendix 2 (data preparation) of the paper, if they are not published as supplementary material with the article, and in particular Section A2.4, the structured inference guide, definitions and keyword dictionaries provided to Microsoft 365 Copilot.
Under regulation 6(1)(b) of the EIRs, we do not have to give you information which is already publicly available and easily accessible to you in another form or format. Where information is already publicly available, I have provided you with the links to the relevant websites. If, however, you do not have internet access to obtain this information from the websites listed, then please contact me again and I will send you a paper copy.
The information you have requested is included in the link that was sent to you by the author of the paper.
The Journal has called the full supplemental material ‘Appendix A’ and it can be downloaded directly this includes the Appendix 1 and 2.
Question 2 - The 500-record Fish Health Inspectorate expert-validated reference dataset used to evaluate the AI categorisation (Section 3.1 of the paper), with the model predictions for each of the 13 runs, in anonymised form if site identification is a concern. Site names and numbers may be removed or replaced with the FS site numbers already published in Annex B.
Attached are the 14 CSV files containing the outputs from the 13 model runs and the human expert check (Document 002a-Document 002n). For this, the author uses the Mortality Event Number rather than the FS number or site number. The model predictions can be reproduced using the original code; however, only the final publication plots have been retained.
Question 3 - The categorised SGMD mortality-events file that the published code reads (named in the repository as "FHI 066 Mortality events 16092025_grouped_causes.csv"), that is, the three ranked mortality-driver categories and the production-stage flag per mortality event, in anonymised form if necessary.
Attached is ‘Document 003a’ which is the categorised SGMD mortality-events file (named in the repository as "FHI 066 Mortality events 16092025_grouped_causes.csv”). Within this file, you will find the event number, the mortality driver categories and the production stage file that you have requested.
Question 4 - The lists of sites at each stage of the framework: the 28 sites flagged for recurrent elevated mortality, the 25 possible persistent sites, the 20 reviewed sites and the 10 confirmed sites
(Table 7 of the paper), by FS site number.
Attached is ‘Document 004a’ which is the lists of sites at each stage of the framework: the 28 sites flagged for recurrent elevated mortality, the 25 possible persistent sites, the 20 reviewed sites and the 10 confirmed sites.
Question 5 - The outputs of the sensitivity analysis described in Section 2.5 and Table 4 (the percentage of sites classified at each combination of threshold, streak and cycles), in tabular form.
While our aim is to provide information whenever possible, in this instance the Scottish Government does not hold the information you have requested as the author has coded the analysis and printed the summary plots that were required for the paper. Therefore, no file is held with the requested information in a tabular form.
Therefore, with respect to your request for the outputs of the sensitivity analysis described in Section 2.5 and Table 4 (the percentage of sites classified at each combination of threshold, streak and cycles), in tabular form, this is a formal notice under Regulation 10(4)(a) of the EIRs confirming that the Scottish Government does not hold this information.
This exception is subject to the ‘public interest test’. It is important to note that although we do not hold the information and have applied Regulation 10(4)(a) – information not held, it is a requirement that we have to apply the public interest test.
Wider guidance on the Environmental Information regime can be found in the Key Concepts section under ‘Information Not Held’ and ‘The Public Interest Test’ in the attached Scottish Information Commissioner Briefings and Guidance document. Further guidance on the public interest test can also be found on page 8 of the Scottish Information Commissioner guidance.
About FOI
The Scottish Government is committed to publishing all information released in response to Freedom of Information requests. View all FOI responses at https://www.gov.scot/foi-responses.
Contact
Please quote the FOI reference
Central Correspondence Unit
Email: contactus@gov.scot
Phone: 0300 244 4000
The Scottish Government
St Andrew's House
Regent Road
Edinburgh
EH1 3DG