NATIONAL CONSENT TRANSLATION PROJECT PHASE II FINAL REPORT

March 1998

EXECUTIVE SUMMARY

1. INTRODUCTION

1.1 Background

The National Consent Translation Project (NCTP) Phase I was established in 1992 in response to industry concerns over the compliance implications of the then forthcoming Urban Waste Water Treatment Directive (UWWTD). One particular problem was that the UWWTD effluent limits related to 24-hour composite sampling rather than the traditionally adopted spot-sampling monitoring regime, which meant that there was a shortage of effluent quality data relating to the proposed new regime.

The Phase I project was carried out during 1993-94, and its main aim was to establish reliable estimates of the relationships between spot-based (S95) and composite-based (C95) 95percentile effluent quality for the determinands BOD, suspended solids (SS), ammoniacal nitrogen (AmmN) and COD. The exercise generated an unprecedented volume of data. In all, 46 sewage treatment works (STWs) were involved, and at most of these, an extensive spot and composite-based monitoring programme was carried out over a full 12-month period. The monitoring focused primarily on effluent quality, but the influent was monitored also at the majority of the STWs.

There were two additional objectives. One was to examine the relationship (for any particular determinand and sample type) between 95%ile quality and higher percentiles in order to give a more objective and quantitative insight into the implications of various numerical factors for setting absolute limits (so-called 'upper-tiers'). The project also examined the behaviour of percent removal for the determinands BOD and COD.

A comprehensive account of the Phase I project can be found in the FWR publication FR/CL 0003 (December 1994), and its companion volume of appendices.

1.2 Objectives of NCTP Phase II project

The objectives of NCTP Phase II were as follows:

1. Factors affecting percent reduction

Investigate the factors affecting percent reductions during treatment, using the same methodology as that used to investigate the S95 to C95 translation.

2. Correlation between determinands

Examine further the relationship between COD and BOD to establish whether between-determinand correlations can be used to explain more of the variations between STWs. Particular relationships to be studied are:

*COD % removal v. BOD % removal; *effluent 95%ile COD v. effluent 95%ile BOD; *individual relationships for the 30 STWs for which NCIP data sets are available.

3. Upper tier limits

Evaluate the implications of different percentiles selected for the upper tier. Evaluate the sensitivity of the upper tier:95%ile ratios to outliers. Establish which of the extreme outliers in the NCTP data can reasonably be attributed to (i) exceptional circumstances, and (ii) operational error, and hence be disregarded in the upper-tier calculations which should then be repeated.

1.3 Organisational arrangements and responsibilities

These can be summarised as follows:

Foundation for Water Research (FWR): The FWR was responsible for the management of the overall programme through the appointment of a Programme Manager whose roles were to liaise with all participants and undertake report production and dissemination.

Water Research Centre (WRc): WRc was appointed as the Specialist Sub-Contractor to undertake the statistical analysis and interpretation of the data and to prepare the report on the Project.

Regulators: The Regulators, comprising the Environment Agency, the Scottish Environment Protection Agency, and the Department of the Environment for Northern Ireland represented by SNIFFER (the Scotland and Northern Ireland Forum for Environment Research) funded the programme management by FWR and the work of the Specialist Sub-Contractor, WRc.

Operators: The Operators, comprising the Water Services Companies of England and Wales, together with the Water and Sewage Treatment Operators in Scotland and Northern Ireland, permitted the data generated during Phase I to be used for this Project, and provided additional information where this was available.

Steering Group: A Steering Group representing all participants in the Project was formed to oversee the study and approve outputs.

1.4 Structure of report

There are six main sections following this introduction:

• Section 2 describes the investigation of factors related to percent removal;
  
• Section 3 addresses the estimation of within-STW relationships between BOD and COD;
  
• Section 4 describes the investigation of factors related to variation in the average BOD v. COD relationship from works to works;
  
• Section 5 describes the efforts made to reappraise the status of the outliers identified in NCIP Phase I, and presents revised estimates for upper-tier:95%ile values;
  
• Section 6 addresses the compliance implications of different percentile upper-tier definitions; and finally
  
• Section 7 presents the project conclusions.

The report contains eight Appendices. These provide detailed sets of graphical and tabular output (Appces A and B); technical background material (Appces C, D, E and I;), and various data tabulations (Appces G and H).

2 RESULTS

2.1 Factors related to percent removal

Percent removal relationships were developed that accounted for between 60% and 90% of the works-to-works variation. The models can be summarised as follows:

For BOD, higher percent removals are associated with:

• lower values of the ratio of influent COD to BOD;
  
• lower values of the actual-to-consent ratio for 95%ile effluent BOD; and
  
• locations other than Welsh, South West and Wessex.

For SS, higher percent removals are associated with:

• locations other than North West, Yorkshire, Welsh, South West and Wessex.

For AmmN, higher percent removals are associated with:

• nitrifying works; and
  
• (for non-nitrifying works) lower trade effluent proportions in the influent.

For COD, higher percent removals are associated with:

• higher influent COD; and
  
• locations other than North West, Yorkshire, Welsh, South West and Wessex.

2.2 Relationships between COD and BOD

Works-specific relationships between COD and BOD effluent quality were estimated for 42 STWs. The non-parametric statistical procedure that was used to cope with the problem of measurement error in both determinands worked very well, and strong associations were found for the majority of works, for both spots and composites data.

For the 28 STWs with influent data, a similar exercise was conducted to estimate works-specific relationships between BOD and COD percent removal. Again, the models provided a good fit in the majority of cases. As expected, there is a strong association between COD percent removal and BOD percent removal There is also a discernible tendency for works with higher COD:BOD ratios in the influent to have lower COD percent removals.

2.3 Factors related to BOD v. COD slopes

The search for factors explaining the variations in BOD v. COD slopes was less successful than the percent removal investigation, with the models accounting for only about 30-40% of the works-to-works variation. Increases in the BOD v. COD slope were as sociated with higher values of the actual-to-consent ratio for 95%ile effluent BOD, and with Filter (rather than Activated Sludge) treatment works. One unexpected finding was the presence of the factor 'Sampler type' in all three models. It was concluded that this may well be a surrogate for differences in operational practices between one Operator and another.

Despite the largely unexplained differences from works to works, the individual composite BOD v. COD relationships did allow one general conclusion to be drawn of relevance to the UWWT Directive. For all but four of the 42 STWs, a composite COD level of 125 mg/l was found to equate to a composite BOD concentration of less than 25 mg/l - in most cases substantially so. For most works, therefore, an effluent composite sample that is borderline in relation to the Directive's COD limit will tend to be well below the Directive's BOD limit. It follows that, for marginal STWs, compliance with the BOD effluent limit will tend to be easier to achieve than compliance with the COD effluent limit.

2.4 Relationships between high percentile concentrations

The principal message from NCTP Phase I was that, for upper-tier definitions beyond the 99%ile, the upper-tier:95%ile ratio apparently needed was far larger than had previously been supposed. This was especially so if not just 50% but 95% of STWs were to have a realistic chance of meeting it. However, it was also recognised that the calculations were very sensitive to, and probably biased by, the presence of outliers in the data; and it was to investigate this problem further that the present study was undertaken.

Essentially two types of approach were pursued, namely:

1. Consultation with Operators The original NCTP participants were invited to comment on the outliers found for their specific STWs in the hope that they could identify cases of 'excusable' or 'bad' outliers that could be excluded in a re-analysis of the data.
   
2. Robust statistical methods The percentile estimation method was revised so that it was more 'robust'- that is, was less sensitive to the presence of outliers.

The outcome of the outlier review was understandably limited. Three of the Operators did succeed in uncovering interesting evidence showing that some outliers could be attributed to analytical or transcription mistakes. In all other cases, however, the respondents were unable to add any new information - primarily because of the length of time that had elapsed since the Phase I exercise.

The revision of the statistical methodology proved more successful, and three main features were introduced:

• a more robust method of estimating high percentiles: this was less sensitive to extreme data values (whether outliers or not), and so produced more stable estimates of the uppertier:95%ile ratio for as far as the method allowed - typically up to the 99.3 or 99.4%ile;
  
• a more robust definition of 'achievable' - through switching from the criterion '95% of STWs' to '90% of STWs'; and
  
• a smoothed extrapolation method - allowing plausible estimates to be obtained for upper-tier definitions up to at least the 99.5%ile.

Following these revisions, the resulting values for the 99.5%ile-to-95%ile ratios (achievable by 90% of STWs) were as follows:

• Detd Sample type Upper tier defined as 99.5%ile Value of U-tier:95%ile ratio achieved by: 90% of works
  
• BOD Spot 3.7 Comp 3.2
  
• SS Spot 7.7 Comp 6.9
  
• AmmN Spot 4.3 Comp 6.7
  
• COD Spot 7.6 Comp 3.0

These values should be contrasted with the absolute upper-tier:95%ile ratios prescribed in the UWWT Directive, namely 2.0 for BOD and COD and 2.5 for Suspended Solids.

2.5 Compliance implications of upper-tier definition

The statistical simulation results demonstrated that the upper-tier limit should not be set at a percentile as low as the 99%ile. This would generate far too many false alarms - even at sampling frequencies as low as monthly.

In contrast, a 99.9%ile definition would be satisfactory, but cannot in practice be implemented (even from the large amount of data generated by NCTP Phase I). Uppertier:95%ile ratios cannot reliably be estimated beyond the 99.5%ile definition of upper-tier.

The false-alarm risk for a strict 99.5%ile definition would be unsatisfactorily high particularly at sampling frequencies of fortnightly or greater. However, extra protection is built in via the '90% of STWs' criterion. This ensures that, for the great majority of works, a notional 99.5%ile upper-tier will in practice be more extreme than a 99.5%ile.

Overall, therefore, it is considered that an acceptable balance is struck by the '99.5%ile for 90% of STWs' definition of the upper-tier:95%ile ratio. On the one hand, it can be estimated with reasonable precision from the NCTP data; and on the other, its use will ensure that a tolerable level of protection is provided against 'statistical' upper-tier failures.

Copies of the report are available from FWR, price £35.00, less 20% to FWR Members.