Redefining dry weather flow

Coming soon: The Water Industry Achievement Awards 2011. When: March 29, 2011. Where: ICC, Birmingham. Full details soon.

NEW... The very latest issues are here - get yours free!

WWT Round Table on Asset Management - what do the water companies and the regulators really think? - Read all about it here


	January 2006 Redefining dry weather flow

Ed Bramley of Yorkshire Water and George Heywood of Tynemarch Systems Engineering describe UKWIR's work to develop a new measure of DWF

As a concept, dry weather flow (DWF) has existed for more than 30 years. In the context of sewage treatment, its intended use, in conjunction with quality parameters, is to define the load that can be assimilated by the receiving environment.
By defining the base load, it is also a yardstick by which population growth in a catchment can be identified. It is a key parameter used by the Environment Agency (EA) in setting consents to discharge. And, with recent improvements in flow measurement within the industry, it is likely to attract increased regulator attention for the foreseeable future.
It is therefore important that DWF can be robustly, reliably and consistently measured. Unfortunately, this is not currently the case. This article discusses the work under UKWIR project WW21/D to develop an alternative measure of DWF. Also, it summarises the regulatory progress towards adoption of the new measure, and highlights other flow-related research needs that have recently been identified.

Background
The current definition of DWF was first published in the Institute of Water Pollution Control Glossary in 1970. It is defined as "the average daily flow to the treatment works during seven consecutive days without rain (excluding a period which includes public holidays) following seven days during which the rainfall did not exceed 0.25mm on any one day".
To date, there has been no systematic EA assessment of DWF compliance due to the generally poor availability of validated flow data. Recently imposed requirements for flow monitoring and reporting mean that this data is now becoming available, and EA assessment and reporting of flow compliance is anticipated this year.
DWF is used in the setting and enforcement of effluent discharge consents, for WwTWs design, and to determine the base flow for use in sewerage modelling. In consent setting and enforcement, DWF is used alongside quality limits as a means of controlling the pollutant load discharged to the watercourse. It is also used in determining the flow at which discharges to storm tanks will be permitted by the consent (flow to full treatment, FFT).
In WwTWs design, the main considerations are the incoming BOD load, and the mean and peak flows to treatment. DWF may still be used in the estimation of the latter flow parameters but this would ideally be within the context of a more complete characterisation of wastewater flow and quality.
In sewerage modelling, DWF is used to determine the base flow of domestic and industrial wastewater to be applied at model nodes. Best practice requires the separate estimation of the individual DWF components, with infiltration being estimated either from infiltration surveys or detailed analysis of long-term outfall flow records.

Problems with current definition
There are a number of problems with the current definition of DWF. The opportunities for measuring DWF are limited by the rarity of suitable dry periods in the UK, as illustrated in Figure 1.
Analysis of rainfall data collected within the UKWIR project indicated that only around 62% of works-years contain a qualifying period for calculation of DWF. Seasonal variations in DWF occur, particularly where infiltration is significant. It is of particular concern that DWF values derived from winter dry periods can lead to spurious consent failures where the consented DWF has been set on the basis of summer river needs.
Representative rainfall data is not available for all WwTWs. In addition, for large catchments, spatial variations can mean that no single rain gauge is representative of the catchment.
The current method of determining DWF compliance does not take into account the inherent uncertainties in either the measurement or calculation processes.
The definition does not provide reliable information regarding the statistical distribution of flows for use in consent setting calculations. The current definition makes no specific provision for unusual catchments that have high seasonal variations, for example due to tourism or food processing industries. There is no theoretical link between the current definition and the design formula, PG + I + E. (P = population; G = per capita consumption; I = infiltration; and E = trade effluent.) This formula will continue to be important in forecasting future flows for design and strategic planning purposes.

Selection criteria
To provide a structured approach to evaluating alternative definitions of DWF, the following selection criteria were agreed by the UKWIR group:

Applicability
The method should be such that it can be calculated for a high proportion of (or preferably all) assessment periods and works
Repeatability
It should show low variability between assessment periods at a given works where domestic and industrial flows have not changed significantly, in other words there is little variability due to year-on-year variations in rainfall patterns or the occurrence and timing of dry periods
Statistically based
The method should be defined in statistical terms with reference to the distribution of flows over a defined period
Continuity
It should give results that are close to the best identified interpretation of the traditional method. This best interpretation has been identified by assessing viable interpretations of the current method against the same selection criteria described here, and is referred to as the baseline method
Simplicity
It should be as simple as possible in concept and calculation
Linkage to design formula
As far as practicable, the method should show reasonable agreement with the PG + I + E formula, provided that an appropriate method for calculating the infiltration component is adopted
Manageable implications
The method should have manageable implications for other consented flow parameters (particularly FFT)
Additional information
It should provide additional statistical information to allow flows to be characterised more accurately in consent setting calculations. If this criterion is to be met, more than one output value will be required
Ready for implementation
The method should be ready for implementation without significant further development

Alternative definitions investigated
The following alternative methods were selected for comparison with the current definition of DWF:

Various percentiles of the daily flows, with and without rainfall
conditions
Various percentiles of a moving average of the daily flows, for various averaging periods
The mean, mode and geometric mean of the daily flows
Methods involving fitting a distribution to the daily flows

When evaluating the performance of the current definition of DWF, both an all-year and a summer-only condition have been evaluated

Data availability
Flow and rainfall data were requested from the 11 water and sewerage companies participating in the UKWIR project, together with associated catchment information.
Some 4,447 works-years of flow data and 3,123 works-years of associated rainfall data were received from nine water and sewerage companies (Figure 2). After subjecting both flow and rainfall data to a range of validation tests (described in the following section) complete works-years of data with in excess of 75% of values complying with all validation tests (3,266 works-years of flow data and 1,803 works-years of associated rainfall data) were passed forward for use in the analysis.
Table 1 provides additional details, with one-year denoting any continuous 12-month period. Works with less than 12 months of continuous flow data were not included.

Data validation
The flow data received were subjected to the following validations:

Negative, zero or missing daily flow values were identified and invalidated
n Two tests were performed to identify periods of constant or near-constant flows due to meter or data transmission errors.
Identification of erroneous low flow values, based on the ratio of mean flow to the 20th percentile of valid flows
Data was reviewed manually where outlier values were observed in the results of the various DWF methods, and data exhibiting unrealistic behaviour were invalidated

Similarly, for rainfall data:

Negative or missing daily rainfall values were identified and invalidated
Gauges which are subject to a significant proportion of missed readings were excluded
Long periods without rain were manually reviewed
Checks were carried out for missing values at the beginning or end of the data period
As for flow data, rainfall data were reviewed manually where outlier values were observed
Finally, all summer dry periods were manually reviewed

Analysis tools
All flow data, rainfall data and catchment explanatory factors were imported into a relational database. This database was accessed by bespoke software capable of calculating DWF values using a wide range of alternative methods.
A wide range of method variations were calculated by the software and returned to the database for analysis. Each method was applied to each data set that had passed the validation criteria defined above. Appropriate analysis techniques were applied to these results in order to assess the performance of each method. These included the following:

Review of scatter plots to compare DWF results from two selected methods. These plots highlighted biases related to works size, and whether a method persistently underestimates DWF compared with another method
Review of histograms of DWF results
Review of correlation matrices to identify biases in method results and performance between works with different numeric attributes.
ANOVA analyses were undertaken to identify biases in method results and performance between works with different non-metric (categorical) attributes
Statistical hypothesis tests to identify biases in method results and performance between works in different binary categories.

Further bespoke software was used to calculate various performance indicators from the method results, to provide quantitative means of assessing and comparing the performance of the many method variants considered. These included:

A measure of the success rate of a method, that is the proportion of data sets for which a result could be obtained (termed "success rate")
A measure of the agreement between the results of a method and those of an identified "baseline" method (usually the current definition used in consents, termed "geomean ratio")
A measure of the variation in method results between years at the same works (termed "temporal variability")

Results of analysis
Full details of all the results from the analysis are contained within the UKWIR report. The current DWF definition cannot be calculated for a large proportion of data sets. The current method also gives results with high year-on-year variability when using dry periods from any month of the year. If restricted to summer dry periods only, this variability is significantly reduced.
By relaxing the stringent rainfall conditions imposed in the current method, a variant can be developed which can be calculated for almost all (90%) of data sets and has only a slightly increased (+2.3%) mean. However, 45% of the data sets available for use in the analysis had no corresponding rainfall data.
A wide range of methods have been tested which do not require rainfall data, of which the 20th percentile flow (Q80) gives the best overall performance against the selection criteria. It can be calculated easily for all reasonable data sets without use of rainfall data, agrees well on average with the summer DWF calculated by the existing method (Figure 3), and exhibits significantly reduced year-on-year variability.
The fifth percentile of daily flows (Q95), which was previously a favoured statistic, gave results which were on average 14% lower than the summer DWF calculated by the existing method (and Q80). Q95 has somewhat higher year-on-year variability than Q80 (Figure 4), but in other respects has similar characteristics.
Moving average methods can be developed that give similar performance to Q80 (Figure 5) but these are more sensitive to missing or invalid data and are more complex to understand and apply.
Distribution fitting methods require a more complex calculation algorithm, but have the potential to provide a more effective check on whether effluent flows are exceeding the assumptions made when setting consents. However, these require further development and have more far-reaching implications for current practice.

Regulatory overview
From the analysis of results, based on the agreed selection criteria, the measure that emerged as the most appropriate replacement for DWF was the 20th percentile (Q80). This is in the process of being discussed formally at a national level between water companies and the Environment Agency, and will form the basis of a replacement for assessing DWF compliance. It is expected that an agreement will be reached on the regulatory principles imminently.
Key to completing this transition successfully are a number
of issues:

Changes needed to DWF values in current consents Previous experience in the early 1990s with the national consent translation programme revealed problems that would arise if there needed to be wholesale changes in numeric values in consents. As the 20th percentile has close agreement with DWF values that would be generated by the current methodology, current consents can be treated as if DWF had been set as 20th percentiles
The compliance testing window
It is expected that flow compliance testing will take place on an annual basis, either on a calendar or rolling-year basis
Incorporation of uncertainty into the compliance testing approach
Unlike compliance testing of quality parameters, compliance testing of flow is currently on a pass or fail basis. To create the equivalent of a confidence interval, it is intended that annual compliance would be assessed on the basis of a tenth percentile basis - in other words, a works would be deemed to fail if the tenth percentile flow exceeded the consented DWF in any given year
Strategy for works that fail consent
The analyses carried out showed that currently nearly one third of works-years fail their DWF consent. While this is reduced by use of the 20th percentile, there is still a significant proportion of works that would fail even the new measure. As part of the ongoing national discussion, an appropriate approach is being developed which will include identifying whether such breaches are trivial or would lead to breach of environmental standards, and tying in associated improvements to investment cycles. This will also involve developing an agreed change procedure with Ofwat, as well as the water companies and the Environment Agency.

Future developments

Examination of the concept of (FFT), including the relationship between DWF and FFT, the environmental basis for FFT, and its use as a works design parameter
The relationship between theoretical and measured values of DWF. In the field of natural hydrology, such low-flow links have already been made for over 20 years
How flow values, and particularly flow distributions, are used in the modelling of discharges for consent setting, including the evaluation of "no deterioration"
How flow records are used to identify changes in catchment characteristics, particularly infiltration and populations

Conclusions

Acknowledgements

References

This article first appeared in Water & Wastewater Treatment January 2006 page 13

This story is tagged as: