SIST ISO 5667-20:2010
Water quality - Sampling - Part 20: Guidance on the use of sampling data for decision making - Compliance with thresholds and classification systems
Water quality - Sampling - Part 20: Guidance on the use of sampling data for decision making - Compliance with thresholds and classification systems
This part of ISO 5667 establishes principles, basic requirements, and illustrative methods for dealing with the use of sample data for decision making based on the assessment of the confidence that water quality:
a) meets targets and complies with thresholds;
b) has changed; and/or
c) lies in a particular grade in a classification system.
This part of ISO 5667 also specifies methods for preliminary examination of the sensitivity of decisions to error and uncertainty, although it does not cover the full range of statistical techniques. This part of ISO 5667 provides general advice on decision making related to constraint formulation for expression of thresholds and targets and the form and scale of sampling programmes. NOTE 1 In the water industry, “standard” is commonly used to indicate the value or limit of a parameter of interest. However, in this part of ISO 5667, the term “threshold” is used to avoid confusion with published national, regional, and International Standards. NOTE 2 This document is framed in terms of sampling and measurement of chemical concentrations, in particular those subject to strong day-to-day temporal variations. The principles apply, however, to any item estimated by sampling which is subject to random error, including microbiological and biological data, and data subject to strong spatial variations.
Qualité de l'eau - Échantillonnage - Partie 20: Lignes directrices relatives à l'utilisation des données d'échantillonnage pour la prise de décision - Conformité avec les limites et systèmes de classification
Kakovost vode - Vzorčenje - 20. del: Navodilo o uporabi podatkov o vzorčenju za odločanje - Skladnost z mejnimi vrednostmi in klasifikacijskimi sistemi
Ta del ISO 5667 vzpostavlja načela, osnovne zahteve in razlagalne metode, ki obravnavajo uporabo podatkov o vzorčenju za odločanje, osnovano na vrednotenju zaupanja, da je kakovost vode:
a) skladna z mejnimi vrednostmi in izpolnjuje cilje;
b) spremenjena; in/ali
c) v določeni meri odvisna od klasifikacijskega sistema.
Ta del ISO 5667 prav tako določa metode za predhodni pregled občutljivosti odločitev na napake in negotovost, čeprav ne zajema celotnega obsega statističnih tehnik. Ta del ISO 5667 zagotavlja splošno navodilo glede odločanja, povezanega z formulacijo omejitev za izražanja mejnih vrednosti, ciljev in oblike ter stopnje programov vzorčenja. OPOMBA 1 V vodni industriji se beseda »standard« pogosto uporablja za nakazovanje vrednosti ali meje aktualnega parametra. Vendar je v tem delu ISO 5667 izraz »mejna vrednost« uporabljen v izogib zamenjavi z objavljenimi nacionalnimi, regionalnimi in mednarodnimi standardi. OPOMBA 2 Ta dokument je zasnovan ob upoštevanju vzorčenja in merjenja kemijskih koncentracij, zlasti tistih, ki so predmet močnih vsakodnevnih začasnih sprememb. Vendar ta načela veljajo za vsako postavko, ki jo predvidi vzorčenje ter je predmet naključne napake, vključno z mikrobiološkimi in biološkimi podatki ter podatki, ki so podvrženi močnim prostorskim spremembam.
General Information
Buy Standard
Standards Content (Sample)
МЕЖДУНАРОДНЫЙ ISO
СТАНДАРТ
5667-20
Первое издание
2008-03-15
Качество воды. Отбор проб.
Часть 20.
Руководство по использованию
выборочных данных для принятия
решения. Соответствие порогам и
классификационным системам
Water quality — Sampling —
Part 20: Guidance on the use of sampling data for decision making —
Compliance with thresholds and classification systems
Ответственность за подготовку русской версии несёт GOST R
(Российская Федерация) в соответствии со статьёй 18.1 Устава
Ссылочный номер
ISO 5667-20:2008(R)
©
ISO 2008
---------------------- Page: 1 ----------------------
ISO 5667-20:2008(R)
Отказ от ответственности при работе в PDF
Настоящий файл PDF может содержать интегрированные шрифты. В соответствии с условиями лицензирования, принятыми
фирмой Adobe, этот файл можно распечатать или вывести на экран, но его нельзя изменить, пока не будет получена
лицензия на загрузку интегрированных шрифтов в компьютер, на котором ведется редактирование. В случае загрузки
настоящего файла заинтересованные стороны принимают на себя ответственность за соблюдение лицензионных условий
фирмы Adobe. Центральный секретариат ISO не несет никакой ответственности в этом отношении.
Adobe торговый знак фирмы Adobe Systems Incorporated.
Подробности, относящиеся к программным продуктам, использованным для создания настоящего файла PDF, можно найти в
рубрике General info файла; параметры создания PDF были оптимизированы для печати. Были приняты во внимание все
меры предосторожности с тем, чтобы обеспечить пригодность настоящего файла для использования комитетами-членами
ISO. В редких случаях возникновения проблемы, связанной со сказанным выше, просьба проинформировать Центральный
секретариат по адресу, приведенному ниже.
ДОКУМЕНТ ЗАЩИЩЕН АВТОРСКИМ ПРАВОМ
© ISO 2008
Все права сохраняются. Если не указано иное, никакую часть настоящей публикации нельзя копировать или использовать в
какой-либо форме или каким-либо электронным или механическим способом, включая фотокопии и микрофильмы, без
предварительного письменного согласия ISO по соответствующему адресу, указанному ниже, или комитета-члена ISO в стране
заявителя.
ISO copyright office
Case postale 56 • CH-1211 Geneva 20
Tel. + 41 22 749 01 11
Fax + 41 22 749 09 47
E-mail copyright@iso.org
Web www.iso.org
Опубликовано в Швейцарии
ii © ISO 2008 – Все права сохраняются
---------------------- Page: 2 ----------------------
ISO 5667-20:2008(R)
Содержание Страница
Предисловие .iv
Введение .vi
1 Область применения .1
2 Обзор ключевых пунктов .1
3 Типы ошибки и изменчивость .2
3.1 Общие положения .2
3.2 Аналитическая ошибка.3
3.3 Общая неопределенность .4
4 Действия.5
4.1 Оценка суммарных статистик .5
4.2 Соответствие порогам для качества воды .6
4.3 Достоверность отказа .8
4.4 Методы для порогов, выраженных в процентилях.9
4.5 Непараметрические методы.11
4.6 Обзорные таблицы .15
5 Определение порогов.16
5.1 Общие положения .16
5.2 Идеальные пороги .16
5.3 Абсолютные пределы .17
5.4 Процент отказавших выборок .20
5.5 Пределы вычисления для сбросов сточных вод.21
6 Заявление об обнаружении какого-либо вещества.22
7 Обнаружение изменения.23
8 Классификация .26
8.1 Общие положения .26
8.2 Достоверность изменения класса .28
Приложение А (информативное) Вычисление пределов достоверности .31
Приложение В (информативное) Вычисление для биномиального распределения .33
Приложение С (информативное) Результаты выборки с высокой ошибкой или
представленные как результаты ниже предела обнаружения .36
Библиография.38
© ISO 2008 – Все права сохраняются iii
---------------------- Page: 3 ----------------------
ISO 5667-20:2008(R)
Предисловие
Международная организация по стандартизации (ISO) является всемирной федерацией национальных
организаций по стандартизации (комитетов-членов ISO). Разработка международных стандартов
обычно осуществляется техническими комитетами ISO. Каждый комитет-член, заинтересованный в
деятельности, для которой был создан технический комитет, имеет право быть представленным в этом
комитете. Международные правительственные и неправительственные организации, имеющие связи с
ISO, также принимают участие в работах. ISO осуществляет тесное сотрудничество с международной
электротехнической комиссией (IEC) по всем вопросам стандартизации в области электротехники.
Проекты международных стандартов разрабатываются по правилам, указанным в Директивах ISO/IEC,
Часть 2.
Главная задача технических комитетов состоит в разработке международных стандартов. Проекты
международных стандартов, принятые техническими комитетами, рассылаются комитетам-членам на
голосование. Их опубликование в качестве международных стандартов требует одобрения, по
меньшей мере, 75 % комитетов-членов, принимающих участие в голосовании.
Обращается внимание на возможность патентования некоторых элементов данного международного
стандарта. ISO не несет ответственности за идентификацию какого-либо или всех таких патентных
прав.
ISO 5667-20 был подготовлен Техническим комитетом ISO/TC 147, Качество воды, Подкомитетом SC 6,
Отбор проб (общие методы).
ISO 5667 состоит из следующих частей под общим заглавием Качество воды. Отбор проб:
⎯ Часть 1. Качество воды. Отбор проб. Часть 1. Руководство по составлению программ и
методикам отбора проб
⎯ Часть 3. Руководство по хранению и обращению с пробами воды
⎯ Часть 4. Руководство по отбору проб из естественных и искусственных озер
⎯ Часть 5. Руководство по отбору проб питьевой воды из очистных сооружений и
трубопроводных распределительных систем
⎯ Часть 6. Руководство по отбору проб из рек и потоков
⎯ Часть 7. Руководство по отбору проб воды и пара из котельных установок
⎯ Часть 8. Руководство по отбору проб влажных осаждений
⎯ Часть 9. Руководство по отбору проб морской воды
⎯ Часть 10. Руководство по отбору проб из сточных вод
⎯ Часть 11. Руководство по отбору проб грунтовых вод
⎯ Часть 12. Руководство по отбору проб из донных отложений
⎯ Часть 13. Рекомендации по отбору проб шлама сточных вод и на сооружениях водоочистки
iv © ISO 2008 – Все права сохраняются
---------------------- Page: 4 ----------------------
ISO 5667-20:2008(R)
⎯ Часть 14. Руководство по обеспечению качества при отборе проб природных вод и обращении
с ними
⎯ Часть 15. Руководство по консервированию и обработке проб осадка и отложений
⎯ Часть 16. Руководство по биотестированию проб
⎯ Часть 17. Руководство по отбору валовых проб взвешенных твердых частиц
⎯ Часть 18. Руководство по отбору проб подземных вод на загрязненных участках
⎯ Часть 19. Руководство по отбору проб в морских отложениях
⎯ Часть 20. Руководство по использованию выборочных данных для принятия решения.
Соответствие порогам и классификационным системам
Следующие части находятся в стадии разработки:
⎯ Часть 21. Руководство по отбору проб питьевой воды, распределяемой регулируемыми
нетрадиционными средствами
⎯ Часть 22. Руководство по проектированию и установке пунктов отбора проб грунтовой воды
⎯ Часть. 23. Определение значительных загрязнений в поверхностных водах методом
пассивного отбора проб
© ISO 2008 – Все права сохраняются v
---------------------- Page: 5 ----------------------
ISO 5667-20:2008(R)
Введение
В этой части ISO 5667 рассматривается использование информации о качестве воды, полученной при
отборе проб, в принятии решений — в измерении успеха, отказа или изменения в контексте
неизбежных неопределенностей, связанных с выборкой. Эта часть ISO 5667 обеспечивает руководство
по управлению риском таких неопределенностей, приводящих к неоптимальным решениям.
Неоптимальные решения могут часто вытекать из того, как формулируются и устанавливаются в
регламентах и разрешениях пороги (предельно допустимые концентрации загрязняющих примесей)
для водосливов и цели использования окружающей водной среды. В этой части ISO 5667 также
исследуются проблемы, возникающие при оценке соответствия этим порогам с использованием
данных, полученных при выборке.
Цель данной части ISO 5667 состоит в том, чтобы в будущих законах, регламентах и руководствах
устанавливалось требование оценивать и представлять в отчетах статистически значимые результаты.
ПРИМЕЧАНИЕ 1 Разные решения могут приводить к одобрению или критике людей, мест, компаний,
хозяйственных секторов или государств. Решения могут приводить к судебному иску и/или к дорогостоящим и
длительным мероприятиям для улучшения качества воды.
На Рисунке 1 показаны связи между следующими проблемами:
a) установление порогов для принятия решений о необходимости улучшения качества воды,
возможно, включая критерии для сведения к минимуму ухудшение качества;
b) разработка программ для отбора проб, удовлетворяющих требованиям этих порогов, и
необходимости оценивать результаты согласно этим порогам;
c) использование результатов программ отбора проб для принятия решений.
vi © ISO 2008 – Все права сохраняются
---------------------- Page: 6 ----------------------
ISO 5667-20:2008(R)
Рисунок 1 — Связи между проблемами, относящимися к отбору проб и принятию решений
В этой части ISO 5667 рассматривается проблема c). В проблемы a) и b) включают широкий и
разнообразный круг вопросов по собственному праву, и их подробное рассмотрение не относится к
области применения данной части ISO 5667. Тем не менее, в этой части ISO 5667 все-таки даются
рекомендации для выражения целей и порогов для качества воды [проблема a)], которые являются
важными при использовании выборочных данных для принятия решений. В этой части ISO 5667 также
дается совет по программам отбора проб [проблема b)], чтобы обеспечить их соответствие способу
определения порогов и не внести ненужные трудности и ошибки в процесс принятия решения.
К другим областям, которые лежат вне сферы применения этой части ISO 5667, относятся следующие:
детальная механика отбора проб и обращения с ними; обеспечение репрезентативности проб для
водного объекта, из которого они отбираются, по прошествии времени; выполнение химических
анализов проб. Все они рассматриваются в других документах. Тем не менее, некорректные
результаты, полученные в этих областях, могут существенно повлиять на общие неопределенности
выборки и вызвать дополнительные трудности в принятии решений. В этой части ISO 5667
описываются некоторые из этих дополнительных трудностей.
Эта часть ISO 5667 не охватывает весь диапазон статистических методов, которые можно применять,
и обстоятельств, в которых их следует использовать. Главная цель состоит в том, чтобы установить
принцип, согласно которому неопределенность, возникающую при отборе проб и анализе, (и вообще
ошибки) следует всегда оценивать и учитывать как часть процесса принятия решений. Если это не
делается, то могут быть приняты неправильные решения, например, относительно того, какое
необходимо действие и масштаб этого действия.
ПРИМЕЧАНИЕ 2 Некоторые статистические методы используются как иллюстративные примеры. Эти методы
уже имели обычное применение в некоторых нормативных режимах, которые правильно учитывают
статистические неопределенности. Они годятся для использования в ситуациях, которые аналогичны
рассматриваемым отработанным примеры.
Целью этой части ISO 5667 не является разработка нормативных условий. В ней устанавливаются
принципы и инструменты для поддержки менеджмента, включая регулирование. Считается, что
нормативные пороги разработаны с использованием диапазона стратегий, которые включают
технические, социальные и законодательные обоснования. Также полагают, что и другие средства
помимо статистического анализа данных, вероятно, должны использоваться для интерпретации и
применения порогов.
© ISO 2008 – Все права сохраняются vii
---------------------- Page: 7 ----------------------
МЕЖДУНАРОДНЫЙ СТАНДАРТ ISO 5667-20:2008(R)
Качество воды. Отбор проб.
Часть 20.
Руководство по использованию выборочных данных для
принятия решения. Соответствие порогам и
классификационным системам
1 Область применения
В этой части ISO 5667 устанавливаются принципы, основные требования и иллюстративные методы
для использования данных выборочного исследования, чтобы принять решение на основе оценки
достоверности, что качество воды:
a) удовлетворяет целям и соответствует порогам;
b) изменилось; и/или
c) относится к конкретному классу в системе классификации.
В этой части ISO 5667 также устанавливаются методы для предварительного исследования
восприимчивости решений к ошибке и неопределенности, хотя и не рассматривается полный диапазон
статистических методов.
В этой части ISO 5667 дается общий совет по принятию решений в связи с формулировкой
ограничений для выражения порогов и целей, а также формы и масштаба программ отбора проб.
ПРИМЕЧАНИЕ 1 В водном хозяйстве термин “эталон” (“standard”) обычно используется, чтобы указать значение
или предел рассматриваемого параметра. Однако в этой части ISO 5667 термин “порог” используется для
избежания путаницы с опубликованными национальными, региональными и международными стандартами.
ПРИМЕЧАНИЕE 2 Этот документ построен на основе отбора проб и измерения химических концентраций, в
особенности тех, которые подвергаются сильным повседневным временным изменениям. Устанавливаемые
принципы применяются, однако, к любому параметру, оцениваемому при отборе проб, который подвержен
случайной ошибке, включая микробиологические и биологические данные и данные, испытывающие сильные
пространственные изменения.
2 Обзор ключевых пунктов
Качество воды часто оценивается по результатам химического анализа некоторого количества проб,
взятых за какой-то период времени.
Неопределенность вводится за счет случайного действия при отборе проб. Она может присутствовать
в любой серии измерений качества воды, отбираемой за какой-то период времени. Значения для
результатов химического анализа этих проб зависят от качества конкретных небольших объемов воды,
которые извлечены или измерены. Если качество воды меняется в пространстве или времени, то
вторая серия проб, взятых за тот же самый период, будет иметь другие значения, потому что эти
пробы будут состоять из других небольших объемов воды, взятых в другое время. Каждая серия проб
позволяет оценить истинное качество воды. Эти оценки будут различаться: они будут иметь разное
© ISO 2008 – All rights reserved 1
---------------------- Page: 8 ----------------------
ISO 5667-20:2008(R)
среднее и охватывать разный диапазон. По номинальному значению можно сделать различные
предположения о соответствии порогам и целям.
Неопределенность выборки (или погрешность) является термином, который часто приписывается
этому эффекту. Неопределенность выборки включает неопределенности и ошибки, связанные с
химическим анализом, и возникает даже в случае тривиальных ошибок в химическом анализе и даже
при отсутствии всяких ошибок в методах, которые применялись для отбора проб и обращения с ними.
Неопределенность выборки уменьшается, если отбирать больше проб, но масштаб неопределенности
является часто недооцененным.
В этой части ISO 5667, “общая (суммарная) неопределенность” включает эти случайные эффекты
отбора проб и все другие источники изменчивости в серии проб. Эта изменчивость отражает основные
сигналы, создаваемые естественными или, может быть, неестественными процессами; она включает
влияние ошибок химического анализа и обращения с пробами. Она может содержать систематические
изменчивости, обусловленные трендами и суточными, недельными и сезонными циклами. В этом
контексте наиболее подходящим термином является “общая неопределенность”, “общая ошибка” или
[5]
“общая аналитическая ошибка” (ISO/IEC Guide 99:1993 ).
Общую неопределенность следует квантифицировать по крайней мере приблизительно и учитывать
во всех случаях, где качество воды изменяется и выборочное исследование используется для оценки
информации, используемой для принятия решения. Сюда относятся оценивание соответствия порогам
(см. Раздел 5), решения относительно изменения качества воды (см. Раздел 7) и отнесение
анализируемой воды к классу классификационной системы (см. Раздел 8). В этой части ISO 5667
рекомендуется, чтобы:
a) пороги, соответствие которым оценивается посредством выборки, определялись или
использовались так, чтобы общую неопределенность можно было оценивать и рассматривать
надлежащим образом (см. 5.2);
b) пороги, определенные как абсолютные пределы, рассматривались в процентилях при оценивании
соответствия посредством выборки (см. 5.3);
c) пороги, определенные как пределы, которым должен соответствовать процентный состав проб,
определялись или использовались как соответствующие процентили (см. 5.4);
d) при оценивании соответствия порогам оценивалась степень достоверности (см. Раздел 4); и,
e) при намерении продемонстрировать изменение или отсутствие изменения оценивалась степень
достоверности изменений или различий (см. 8.2).
3 Типы ошибки и изменчивость
3.1 Общие положения
Во многих процедурах используются для принятия решений различные результаты, полученные для
проб, взятых за некоторый период времени (например, за год). Эту информацию можно использовать,
чтобы сделать заключение:
a) о соответствии качества воды в реке требуемым порогам;
b) о работе водоочистной станции в этом году по сравнению с прошлым;
c) о необходимости улучшения качества воды в озере;
d) о сравнительной оценке соответствия сброса сточных вод в разных компаниях; или
2 © ISO 2008 – Все права сохраняются
---------------------- Page: 9 ----------------------
ISO 5667-20:2008(R)
e) о степени риска при воздействии на окружающую среду сброса сточных вод конкретного типа.
Маловероятно, чтобы много значительных изменений качества воды происходило в течение года
каждую секунду, но ежедневные изменения являются обычным явлением. Они могут быть
обусловлены суточными циклами, влиянием внутрилабораторных случайных и систематических
ошибок, погодными условиями, ступенчатыми, ежедневными и ежечасными изменениями (возможно,
изменениями природных процессов в воде, водосливом и водозабором), сезонными и экономическими
циклами и несколькими базовыми и перекрывающимися долгосрочными трендами и циклами.
ПРИМЕЧАНИЕ 1 Иногда лаборатория представляет несколько или большинство данных, которые меньше
заданного предела обнаружения. Такие данные называются цензурированными. В зависимости от типов решений,
которые зависят от этих данных, применяются статистические методы для оценивания значений суммарных
статистик и их неопределенностей.
Кроме того полный набор проб должен быть репрезентативным для среднего качества масс воды, из
которых они были взяты, например, за период времени исследования. При оценивании среднего
годового неприемлемо отбирать все пробы, например, в апреле. Эти требования должны быть
установлены при составлении программы отбора проб.
[1]
ПРИМЕЧАНИЕ 2 Руководство по всем этим аспектам дано более подробно в ISO 5667-1 .
3.2 Аналитическая ошибка
Аналитическими ошибками являются ошибки, введенные в процессе химического анализа, и они
показывают, что измерения не свободны от погрешности. Возможно, что результат для единичной
пробы может быть установлен с точностью до заданного предела, например, ±15 %.
ПРИМЕР 1 Фактическое значение аналитической ошибки зависит от возможностей оборудования и лаборатории,
которые используются для проведения анализа. Обсуждение в этой части ISO 5667 сосредоточено на случайной
ошибке, но всегда есть риск и для неслучайной ошибки, например, при замене приборов или методов, при
сильном расхождении матрицы пробы и калибровочных материалов и для результатов, которые выше пределов
[5]
обнаружения (ISO/IEC Guide 99:1993 ).
ПРИМЕЧАНИЕ 2 Результаты химического анализа в настоящее время представляются со значениями
[6]
неопределенности по ISO/IEC 17025 .
Когда вычисляется среднее из n выборок, влияние на неопределенность в оценке среднего для
случайных ошибок в химическом анализе имеет тенденцию к усреднению с понижением согласно n.
Например, если бы аналитическая ошибка, связанная с единичной пробой, была ±15 %, тогда ошибка
в оценке среднего для нескольких химических анализов имела бы тенденцию к уменьшению
приблизительно до ±4 % для 12 выборок или до ±2,5 % для 36 выборок.
При использовании проб для принятия решения этот вид ошибки, обусловленный химическим
анализом, увеличивается, но часто он бывает меньше, чем другие вклады в общую неопределенность,
особенно в неопределенность, связанную со случайностью при отборе ограниченного числа проб.
Ошибка химического анализа проявляется как добавление к ошибке, связанной со случайностью при
отборе ограниченного количества проб, но, возможно, это небольшое добавление. {Тем не менее, в
некоторых исследованиях нужно отделять изменчивость отбора проб от разнородности локального
окружения (см. Работу [7]).}
ПРИМЕЧАНИЕ 3 Общее непонимание состоит в том, что данные, полученные полностью методом
внутрилабораторного статистического анализа, могут не годиться для конкретных интерпретаций из-за ошибок,
связанных с отбором малого числа проб.
ПРИМЕЧАНИЕ 4 Это изучение относительной важности аналитической ошибки обычно относится к тем типам
вопросов, которые рассматриваются в этой части ISO 5667, но оно следует из оценивания аналитической ошибки
в таких случаях и сравнения ее с другими ошибками. Аналитическую ошибку всегда следует оценивать.
© ISO 2008 – Все права сохраняются 3
---------------------- Page: 10 ----------------------
ISO 5667-20:2008(R)
Аналогичные пункты могут быть установлены для обеспечения репрезентативности проб и проверки изменений
методов отбора проб.
Когда результаты выборки используются для оценки других суммарных статистик, таких как
процентили (например, 95-й процентиль является значением, превышенным для 5 % времени), то эта
картина аналогична картине для среднего, т.е. ошибки обратно пропорциональны n, но они больше,
чем для среднего.
3.3 Общая неопределенность
Неопределенность возникает из-за изменений качества воды, отбираемой для проб, и из-за
способности процесса отбора проб точно отражать эти изменения. В серии проб, взятых за какой-то
период времени, результаты подвержены действию законов вероятности при сборе конкретных проб.
Это создает неопределенность, даже если:
1)
⎯ аналитические ошибки близки к нулю ;
⎯ программа отбора проб гарантирует, что пробы действительно репрезентативны для времени и
пространства;
⎯ нет никаких ошибок в обращении с пробами и в записи результатов анализа.
При использовании выборочного исследования основной источник неопределенности обычно связан с
числом, отобранных проб. В решениях о действиях, приведенных ниже в пунктах от a) до f), этот
источник неопределенности обычно представляет более важную проблему, чем, например,
неопределенность, связанная с ошибками химического анализа. Общую неопределенность следует
оценивать и использовать, чтобы квантифицировать неопределенность в тех случаях, когда качество
воды меняется, и решение принимают, исходя из следующих действий:
a) использование выборочного исследования для измерения и представления отчета о качестве
воды;
b) использование выборок для оценки суммарных статистик, например среднемесячного значения,
годового процентиля или годового максимума;
c) составление заявлений относительно годовых суммарных статистик, выше или ниже они по
[6]
сравнению с прошлым годом (см. в ISO/IEC 17025 более широкое представление вопроса,
связанного с поиском изменения);
d) установление, превышает ли качество воды порог;
e) использование суммарных статистик, чтобы отнести качество воды к соответствующему классу
системы классификации; или
f) оценивание, произошло ли изменение класса.
Во всех этих ситуациях цель состоит в том, чтобы оценить, является ли изменение или статус
статистически значимым, и потребовать, чтобы будущие законы, регламенты и руководства выдвигали
требования относительно оценки и представления статистически значимых результатов.
1) Почти всегда есть такая гипотетическая возможность. Многие следы элементов измерены около их пределов
обнаружения и имеют аналитическую неопределенность почти ±100 %. Многие органические химикаты могут
иметь регенерацию ±50 %.
4 © ISO 2008 – Все права сохраняются
---------------------- Page: 11 ----------------------
ISO 5667-20:2008(R)
4 Действия
2)
4.1 Оценка суммарных статистик
Оценка суммарных статистик зависит от значений качества небольших объемов воды, которые
захватываются при отборе проб, и от того, точно ли измерены эти значения. Оценка, обусловленная
общей неопределенностью, почти наверняка будет отличаться от истинного значения суммарной
статистики — значения, которое было бы получено, если бы можно было обеспечить непрерывный
безошибочный мониторинг за весь период, для которого применяется эта суммарная статистика.
Неопределенностью можно управлять, вычисляя доверительные пределы. Доверительные пределы
определяют интервал, в котором ожидается истинное значение оценки суммарной статистики. В
примере в Таблице 1 оценка среднего из восьми выборок составляет 101 мг/л и имеются два 95-%
доверительных предела, 46 мг/л и156 мг/л. 95-% доверительные пределы означают, что истинное
среднее выше нижнего доверительного предела 46 мг/л и ниже верхнего доверительного предела
156 мг/л. В общем, истинное среднее значение находится в интервале между 46 мг/л и 156 мг/л со
3)
степенью достоверности 90 % .
Этот интервал в оценке среднего представляет большие ошибки, но эти ошибки редко оцениваются
или используются для принятия решений на основе этих данных. К тому же в этом обсуждении
подразумевается нормально распределенная случайная ошибка. Нужно сделать следующие
допущения. Случайная ошибка, возможно, не была нормально распределенной; она могла быть
неслучайной и обусловленной грубыми ошибками и просчетами. Как правило, их воздействие будет
увеличивать масштаб ошибки. Ошибки следует всегда оценивать, даже если это делается путем
допущения, что они следуют нормальному распределению.
ПРИМЕЧАНИЕ 1 В этом примере используется реднее, потому что суммарная статистика обычно является
законодательным требованием. В других случаях могут быть основания и возможность для использования других
статистических характеристик, таких как медиана, например, чтобы объяснить различия между большими и
малыми выборками. Медиана полезна для набора данных, подверженных выбросам, и для медианы можно
вычислять доверительные пределы.
ПРИМЕЧАНИЕ 2 Всегда следует сообщать о принимаемых допущениях. В данном случае делается допущение о
нормальном распределении ошибок.
ПРИМЕЧАНИЕ 3 Существуют различия между большими и маленькими выборками, т.е. использование t-
статистики в зависимости от стандартного нормального отклонения,
...
INTERNATIONAL ISO
STANDARD 5667-20
First edition
2008-03-15
Water quality — Sampling —
Part 20:
Guidance on the use of sampling data for
decision making — Compliance with
thresholds and classification systems
Qualité de l'eau — Échantillonnage —
Partie 20: Lignes directrices relatives à l'utilisation des données
d'échantillonnage pour la prise de décision — Conformité avec les
limites et systèmes de classification
Reference number
ISO 5667-20:2008(E)
©
ISO 2008
---------------------- Page: 1 ----------------------
ISO 5667-20:2008(E)
PDF disclaimer
This PDF file may contain embedded typefaces. In accordance with Adobe's licensing policy, this file may be printed or viewed but
shall not be edited unless the typefaces which are embedded are licensed to and installed on the computer performing the editing. In
downloading this file, parties accept therein the responsibility of not infringing Adobe's licensing policy. The ISO Central Secretariat
accepts no liability in this area.
Adobe is a trademark of Adobe Systems Incorporated.
Details of the software products used to create this PDF file can be found in the General Info relative to the file; the PDF-creation
parameters were optimized for printing. Every care has been taken to ensure that the file is suitable for use by ISO member bodies. In
the unlikely event that a problem relating to it is found, please inform the Central Secretariat at the address given below.
COPYRIGHT PROTECTED DOCUMENT
© ISO 2008
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized in any form or by any means,
electronic or mechanical, including photocopying and microfilm, without permission in writing from either ISO at the address below or
ISO's member body in the country of the requester.
ISO copyright office
Case postale 56 • CH-1211 Geneva 20
Tel. + 41 22 749 01 11
Fax + 41 22 749 09 47
E-mail copyright@iso.org
Web www.iso.org
Published in Switzerland
ii © ISO 2008 – All rights reserved
---------------------- Page: 2 ----------------------
ISO 5667-20:2008(E)
Contents Page
Foreword. iv
Introduction . vi
1 Scope . 1
2 Summary of key points . 1
3 Types of error and variation . 2
3.1 General. 2
3.2 Analytical error. 3
3.3 Overall uncertainty . 3
4 Activities . 4
4.1 Estimation of summary statistics . 4
4.2 Thresholds for water quality and compliance . 6
4.3 Confidence of failure . 7
4.4 Methods for thresholds expressed as percentiles. 7
4.5 Non-parametric methods . 10
4.6 Look-up tables . 13
5 Definition of thresholds. 14
5.1 General. 14
5.2 Ideal thresholds . 14
5.3 Absolute limits . 15
5.4 Percentage of failed samples . 18
5.5 Calculating limits for effluent discharges . 18
6 Declaring that a substance has been detected . 19
7 Detecting change. 20
8 Classification. 23
8.1 General. 23
8.2 Confidence that class has changed. 25
Annex A (informative) Calculation of confidence limits. 27
Annex B (informative) Calculation for the binomial distribution. 29
Annex C (informative) Sample results with high error or reported as less than a limit of detection . 32
Bibliography . 34
© ISO 2008 – All rights reserved iii
---------------------- Page: 3 ----------------------
ISO 5667-20:2008(E)
Foreword
ISO (the International Organization for Standardization) is a worldwide federation of national standards bodies
(ISO member bodies). The work of preparing International Standards is normally carried out through ISO
technical committees. Each member body interested in a subject for which a technical committee has been
established has the right to be represented on that committee. International organizations, governmental and
non-governmental, in liaison with ISO, also take part in the work. ISO collaborates closely with the
International Electrotechnical Commission (IEC) on all matters of electrotechnical standardization.
International Standards are drafted in accordance with the rules given in the ISO/IEC Directives, Part 2.
The main task of technical committees is to prepare International Standards. Draft International Standards
adopted by the technical committees are circulated to the member bodies for voting. Publication as an
International Standard requires approval by at least 75 % of the member bodies casting a vote.
Attention is drawn to the possibility that some of the elements of this document may be the subject of patent
rights. ISO shall not be held responsible for identifying any or all such patent rights.
ISO 5667-20 was prepared by Technical Committee ISO/TC 147, Water quality, Subcommittee SC 6,
Sampling (general methods).
ISO 5667 consists of the following parts, under the general title Water quality — Sampling:
⎯ Part 1: Guidance on the design of sampling programmes and sampling techniques
⎯ Part 3: Guidance on the preservation and handling of water samples
⎯ Part 4: Guidance on sampling from lakes, natural and man-made
⎯ Part 5: Guidance on sampling of drinking water from treatment works and piped distribution systems
⎯ Part 6: Guidance on sampling of rivers and streams
⎯ Part 7: Guidance on sampling of water and steam in boiler plants
⎯ Part 8: Guidance on the sampling of wet deposition
⎯ Part 9: Guidance on sampling from marine waters
⎯ Part 10: Guidance on sampling of waste waters
⎯ Part 11: Guidance on sampling of groundwaters
⎯ Part 12: Guidance on sampling of bottom sediments
⎯ Part 13: Guidance on sampling of sludges from sewage and water treatment works
⎯ Part 14: Guidance on quality assurance of environmental water sampling and handling
⎯ Part 15: Guidance on preservation and handling of sludge and sediment samples
⎯ Part 16: Guidance on biotesting of samples
iv © ISO 2008 – All rights reserved
---------------------- Page: 4 ----------------------
ISO 5667-20:2008(E)
⎯ Part 17: Guidance on sampling of bulk suspended solids
⎯ Part 18: Guidance on sampling of groundwater at contaminated sites
⎯ Part 19: Guidance on sampling of marine sediments
⎯ Part 20: Guidance on the use of sampling data for decision making — Compliance with thresholds and
classification systems
The following parts are under preparation:
⎯ Part 21: Guidance on sampling of drinking water distributed by non-continuous, non-conventional means
⎯ Part 22: Guidance on design and installation of groundwater sample points
⎯ Part 23: Determination of significant pollutants in surface waters using passive sampling
© ISO 2008 – All rights reserved v
---------------------- Page: 5 ----------------------
ISO 5667-20:2008(E)
Introduction
This part of ISO 5667 concerns the use of information on water quality obtained by taking samples in taking
decisions — in measuring success, failure or change, in the context of the inevitable uncertainties associated
with sampling. This part of ISO 5667 provides guidance on controlling the risk of such uncertainties leading to
non-optimal decisions.
Non-optimal decisions can also stem from the way in which thresholds for discharges and targets for
environmental waters are formulated or set out in regulations and permits. This part of ISO 5667 also
examines the problems caused when compliance with these thresholds is assessed using data obtained by
sampling.
This part of ISO 5667 aims to ensure that future laws, regulations, and guidance assert the requirement to
assess and report statistical significance.
NOTE 1 Decisions might result in the commendation or criticism of people, sites, companies, sectors or nations.
Decisions can give rise to legal action and/or expensive and time-consuming remedial actions to improve water quality.
Figure 1 shows the links between the following topics:
a) the setting up of thresholds for taking decisions on the need to improve water quality, possibly including
criteria to minimize water quality deterioration;
b) the establishment of sampling programmes to satisfy the requirements of these thresholds and the need
to assess performance against them;
c) making use of the outcome of sampling programmes to take decisions.
Figure 1 — Links between topics associated with sampling and taking decisions
vi © ISO 2008 – All rights reserved
---------------------- Page: 6 ----------------------
ISO 5667-20:2008(E)
This part of ISO 5667 deals with topic c). Topics a) and b) are huge and wide ranging in their own right, and
their detailed treatment lies outside the scope of this part of ISO 5667. Nevertheless, this part of ISO 5667
does make recommendations for the expression of targets and thresholds for water quality [topic a)], which
are important when using sample data to take decisions. This part of ISO 5667 also gives advice on what is
required for sampling programmes [topic b)] in order that they be compatible with the way thresholds are
defined, and so as to place no unnecessary difficulties and errors in the process of taking decisions.
Other areas which lie outside the scope of this part of ISO 5667 are: the detailed mechanics of taking and
handling samples; assurance that samples are representative over time of the body of water being sampled;
and performance of chemical analyses on samples. These are all covered in other documents. Nonetheless, if
poorly obtained results from these areas can add substantially to overall sampling uncertainties and impose
extra difficulties in taking decisions. This part of ISO 5667 describes some of these extra difficulties.
This part of ISO 5667 does not cover the full range of statistical techniques that may be applied and the
circumstances in which they should be used. The main purpose is to establish the principle that uncertainty
from sampling and analysis (and errors generally) should always be assessed and taken into account as part
of the process of taking decisions. If this is not done, incorrect decisions can result, for example, on where
action is needed, and the scale of that action.
NOTE 2 Some statistical techniques are used as illustrative examples. These are techniques that have seen routine
use in some regulatory regimes that take proper account of statistical uncertainties. They are suitable for use in situations
that resemble the worked examples discussed.
It is not the purpose of this part of ISO 5667 to direct the development of regulatory conditions. This part of
ISO 5667 provides principles and tools to support management, including regulation. It is recognised that
regulatory thresholds are developed using a range of strategies that incorporate technical, social and legal
considerations. It is also recognised that tools other than statistical data analysis are likely to be used in
interpreting and applying thresholds.
© ISO 2008 – All rights reserved vii
---------------------- Page: 7 ----------------------
INTERNATIONAL STANDARD ISO 5667-20:2008(E)
Water quality — Sampling —
Part 20:
Guidance on the use of sampling data for decision making —
Compliance with thresholds and classification systems
1 Scope
This part of ISO 5667 establishes principles, basic requirements, and illustrative methods for dealing with the
use of sample data for decision making based on the assessment of the confidence that water quality:
a) meets targets and complies with thresholds;
b) has changed; and/or
c) lies in a particular grade in a classification system.
This part of ISO 5667 also specifies methods for preliminary examination of the sensitivity of decisions to error
and uncertainty, although it does not cover the full range of statistical techniques.
This part of ISO 5667 provides general advice on decision making related to constraint formulation for
expression of thresholds and targets and the form and scale of sampling programmes.
NOTE 1 In the water industry, “standard” is commonly used to indicate the value or limit of a parameter of interest.
However, in this part of ISO 5667, the term “threshold” is used to avoid confusion with published national, regional, and
International Standards.
NOTE 2 This document is framed in terms of sampling and measurement of chemical concentrations, in particular
those subject to strong day-to-day temporal variations. The principles apply, however, to any item estimated by sampling
which is subject to random error, including microbiological and biological data, and data subject to strong spatial variations.
2 Summary of key points
Water quality is often assessed by the results of chemical analysis of a number of samples taken over a
period of time.
Uncertainty is introduced by the action of random chance in taking samples. It can be present in any set of
measurements of water quality taken over a period of time. The values for chemical analysis of these samples
depend on the quality of the particular small volumes of water that are extracted or measured. If water quality
varies in space or time, a second set of samples taken over the same period will have different values
because these samples are made up of different small volumes of water taken at different times. Each set of
samples allows an estimate of the true water quality. These estimates will differ: they will have a different
mean and span a different range. They have the potential, if taken at face value, to suggest different
conclusions about compliance with thresholds and targets.
Sampling uncertainty (or sampling error) is the term often given to this effect. Sampling uncertainty includes
uncertainties and errors associated with chemical analysis, and occurs even in the case of trivial errors in
chemical analysis and if there are no mistakes in the methods by which samples are taken and handled.
© ISO 2008 – All rights reserved 1
---------------------- Page: 8 ----------------------
ISO 5667-20:2008(E)
Sampling uncertainty is reduced if more samples are taken, but the scale of the uncertainty is often
unappreciated.
In this part of ISO 5667, “overall uncertainty” includes these chance sampling effects and all the other sources
of variation in a set of samples. This variability reflects the underlying signals generated by natural or perhaps
unnatural processes; it includes the effects of errors in chemical analysis and the handling of samples. It might
contain systematic variations from trends and diurnal, weekly, and seasonal cycles. In this context, the more
[5]
appropriate term is “overall uncertainty”, “overall error” or “total assay error” (ISO/IEC Guide 99:1993 ).
Overall uncertainty should be quantified, at least approximately, and taken into account in all cases where
water quality varies and sampling is used to estimate information used in decision making. This includes
assessing compliance with thresholds (see Clause 5), deciding whether water quality has changed (see
Clause 7), and putting waters into grades in classification systems (see Clause 8). This part of ISO 5667
recommends that:
a) thresholds for which compliance is assessed by sampling should be defined or used so that the overall
uncertainty can be estimated and dealt with appropriately (see 5.2);
b) thresholds defined as absolute limits should be treated as percentiles when assessing compliance using
sampling (see 5.3);
c) thresholds defined as limits to be met by a percentage of samples should be defined or used as the
corresponding percentiles (see 5.4);
d) the degree of confidence should be estimated when assessing compliance with thresholds (see
Clause 4); and,
e) the degree of confidence in changes or differences should be estimated when aiming to demonstrate
change or no change (see 8.2).
3 Types of error and variation
3.1 General
In many procedures by which sample data are used to take decisions, there is a set of results taken over a
period of time (e.g. a year). This information might be used to make such judgements as whether:
a) water quality in a river failed to meet required thresholds;
b) a treatment works performed better this year than last;
c) water quality in a lake needs improvement;
d) one company has better effluent discharge compliance than another; or
e) most of the risk of environmental impact is from a particular type of effluent discharge.
There are unlikely to be many significant changes in water quality from second to second throughout a year,
but variations from day to day are common. These can be due to diurnal cycles, the play of random errors and
bias from the laboratory, the weather, step changes, day-to-day and hour-by-hour variations (perhaps in the
natural processes in water or caused by discharges and abstractions and changes in these), seasonal and
economic cycles, and several underlying and overlapping long-term trends and cycles.
NOTE 1 Sometimes several or most of the data are reported by a laboratory as being less than a specified limit of
detection. Such data are called censored data. Depending on the types of decisions that depend on the data, special
statistical techniques are available for estimating the values of summary statistics and their uncertainties.
2 © ISO 2008 – All rights reserved
---------------------- Page: 9 ----------------------
ISO 5667-20:2008(E)
In addition, the total set of samples shall be representative of the average quality of the masses of water from
which they were taken, e.g. over a period of time under review. In estimating an annual mean, it is not
acceptable for all samples to be taken in April, for example. These requirements should be set up in the
design of the sampling programme.
[1]
NOTE 2 Guidance on all these aspects is given in more detail in ISO 5667-1 .
3.2 Analytical error
Analytical errors are those introduced by the process of chemical analysis and reflect that these
measurements are not error free. It might be that the result for a single sample can be specified to within a
specific range, e.g. ±15 %.
NOTE 1 The actual value of the analytical error depends on the capabilities of the equipment and the laboratory that
has been used to perform the analysis. The discussion in this part of ISO 5667 focuses on random error, but there is
always a risk of non-random error, e.g. when there is a change of instrument or methods, when the sample matrix varies
[5]
greatly from the calibration materials, and for results just above the detection limit (ISO/IEC Guide 99:1993 ).
NOTE 2 The results of chemical analysis are nowadays reported with uncertainty values in accordance with
[6]
ISO/IEC 17025 .
When a mean is calculated from n samples, the effect on the uncertainty in the estimate of the mean of
random errors in chemical analysis tends to average down according to n. For example, if the analytical
error associated with a single sample were ±15 %, then the error in the estimate of the mean of a set of
chemical analyses would tend to reduce to something like ±4 % for 12 samples or to ±2,5 % for 36 samples.
In using samples to take decisions, this kind of error from chemical analysis augments but is often smaller
than other contributions to the overall uncertainty, especially that associated with chance in the taking of a
limited number of samples. Chemical analysis error comes through as an addition to that associated with
chance in the taking of a limited number of samples, but it might be a small addition. {Nevertheless, some
studies need to separate sampling variance from local environmental heterogeneity (see Reference [7]).}
NOTE 3 It is not commonly understood that data fully within the statistical control of a laboratory might be unsuitable for
particular interpretations because of errors associated with taking a small number of samples.
NOTE 4 This observation on the relative importance of analytical error applies generally to the types of issues
considered in this part of ISO 5667, but it follows from estimating the analytical error in such cases, and comparing it with
other errors. The analytical error should always be estimated. Similar points can be made about making sure samples are
representative, and about checking changes to methods of sampling.
When the sample results are used to estimate the value of other summary statistics such as percentiles (e.g.
the 95-percentile, which is the value exceeded for 5 % of the time), the picture is similar to that for the mean,
i.e. the errors are inversely proportional to n, but are larger than for the mean.
3.3 Overall uncertainty
Uncertainty occurs because of variations in the quality of the water being sampled, and the ability of the
sampling process to accurately reflect these variations. In a set of samples taken over a period of time, the
results are affected by the operation of the laws of chance in the way the particular samples came to be
collected. This produces uncertainty even if:
1)
⎯ analytical errors are close to zero ;
⎯ the sampling programme guarantees samples that are truly representative in time and space;
⎯ there are no mistakes in handling the samples and recording the results of analysis.
1) Nearly always this is a hypothetical possibility. Many trace elements are measured near their detection limits and have
analytical uncertainty of about ±100 %. Many organic chemicals can have recoveries of ±50 %.
© ISO 2008 – All rights reserved 3
---------------------- Page: 10 ----------------------
ISO 5667-20:2008(E)
In using sampling, the main source of uncertainty is usually associated with the number of samples taken. In
the types of decision on activities listed in items a) to f) below, this source of uncertainty is usually a bigger
issue than, for example, that associated with errors of chemical analysis. Overall uncertainty should be
assessed and used to quantify uncertainty in cases where water quality varies and decisions are taken as a
consequence of the following types of activities:
a) using sampling to measure and report on water quality;
b) using samples to estimate summary statistics, e.g. the monthly mean, the annual percentile or the annual
maximum;
c) making statements about whether this year’s summary statistics are higher or lower than last year’s (see
[6]
ISO/IEC 17025 for a wider view of the issue of looking for change);
d) establishing whether water quality exceeds a threshold;
e) using summary statistics to place water quality in a particular class within a classification system; or
f) assessing whether a change in class has occurred.
In all these situations, the aim is to assess whether the change or the status is statistically significant and to
require that future laws, regulations and guidance assert the requirement to assess and report statistical
significance.
4 Activities
2)
4.1 Estimation of summary statistics
An estimate of a summary statistic depends on the values of water quality in the small volumes of water that
happen to be captured by sampling and whether these values are measured accurately. The estimate, due to
the overall uncertainty, is almost certain to differ from the true value of the summary statistic — the value that
would be obtained if it were possible to achieve continuous error-free monitoring over the entire period for
which the summary statistic applies.
Uncertainty can be managed by calculating confidence limits. Confidence limits define the range within which
the true value of the estimate of the summary statistic is expected to lie. In the example in Table 1, the
estimate of the mean from eight samples is 101 mg/l and there is a pair of 95 % confidence limits, 46 mg/l and
156 mg/l. There is 95 % confidence that the true mean exceeds the lower 95 % confidence limit of 46 mg/l and
95 % confidence that the true mean is less than the upper 95 % confidence limit of 156 mg/l. Overall there is
3)
90 % confidence that the true value of the mean falls in the range between 46 mg/l and 156 mg/l .
This range in the estimate of the mean represents large errors but these errors are seldom estimated or used
to help take decisions based on the data. Also this discussion is for normally distributed random error. Such
assumptions should be stated. Random error might not be normally distributed; it could be non-random and
subject to mistakes and blunders. As a rule, the effect of these will be to increase the scale of the error. Errors
should always be estimated even if this is done by making an assumption that they follow a normal distribution.
NOTE 1 The mean is used in this example because this summary statistic is commonly required by legislation. In other
cases, there may grounds and opportunity to use other statistics like the median, e.g. to explain differences between large
and small samples. The median is useful for data sets affected by outliers, and confidence limits can be calculated for the
median.
2) Some documents use the concept of “sampling target”. The sampling target could be the annual water quality, and a
mean value over 1 year, or a 95-percentile over 1 year, is what is estimated.
3) This range is sometimes called the 90 % confidence interval, calculated from: Xt± σ , where X is the mean; t is
X
derived from the t-distribution with n − 1 degrees of freedom, used instead of the normal standard deviation for low rates of
[5]
sampling (ISO/IEC Guide 99:1993 ); and σ is the “standard error” derived from the standard deviation divided by the
X
square root of the number of samples, σ n .
4 © ISO 2008 – All rights reserved
---------------------- Page: 11 ----------------------
ISO 5667-20:2008(E)
NOTE 2 Assumptions should always be stated. In this case, a normal distribution of errors is assumed.
NOTE 3 There are differences between large and small samples, i.e. use of the t-statistic versus the standard normal
deviate, z.
Figure 2 illustrates the range of uncertainty. This range is an estimate of the distribution of errors in the
estimate of the mean. The confidence limits are shown as points that mark 5 % and 95 % of the area of this
distribution.
Table 1 — Example of confidence limits for the mean
Parameter Value
Estimate of the mean 101 mg/l
Standard deviation 82 mg/l
No. samples 8
Lower confidence limit 46 mg/l
Upper confidence limit 156 mg/l
Key
p probability
X value of the mean
1 distribution of errors in the estimate of the mean
2 lower confidence limit
3 upper confidence limit
Figure 2 — Conf
...
SLOVENSKI STANDARD
SIST ISO 5667-20:2010
01-september-2010
.DNRYRVWYRGH9]RUþHQMHGHO 1DYRGLORRXSRUDELSRGDWNRYRY]RUþHQMX]D
RGORþDQMH6NODGQRVW]PHMQLPLYUHGQRVWPLLQNODVLILNDFLMVNLPLVLVWHPL
Water quality - Sampling - Part 20: Guidance on the use of sampling data for decision
making - Compliance with thresholds and classification systems
Qualité de l'eau - Échantillonnage - Partie 20: Lignes directrices relatives à l'utilisation
des données d'échantillonnage pour la prise de décision - Conformité avec les limites et
systèmes de classification
Ta slovenski standard je istoveten z: ISO 5667-20:2008
ICS:
13.060.45 Preiskava vode na splošno Examination of water in
general
SIST ISO 5667-20:2010 en
2003-01.Slovenski inštitut za standardizacijo. Razmnoževanje celote ali delov tega standarda ni dovoljeno.
---------------------- Page: 1 ----------------------
SIST ISO 5667-20:2010
---------------------- Page: 2 ----------------------
SIST ISO 5667-20:2010
INTERNATIONAL ISO
STANDARD 5667-20
First edition
2008-03-15
Water quality — Sampling —
Part 20:
Guidance on the use of sampling data for
decision making — Compliance with
thresholds and classification systems
Qualité de l'eau — Échantillonnage —
Partie 20: Lignes directrices relatives à l'utilisation des données
d'échantillonnage pour la prise de décision — Conformité avec les
limites et systèmes de classification
Reference number
ISO 5667-20:2008(E)
©
ISO 2008
---------------------- Page: 3 ----------------------
SIST ISO 5667-20:2010
ISO 5667-20:2008(E)
PDF disclaimer
This PDF file may contain embedded typefaces. In accordance with Adobe's licensing policy, this file may be printed or viewed but
shall not be edited unless the typefaces which are embedded are licensed to and installed on the computer performing the editing. In
downloading this file, parties accept therein the responsibility of not infringing Adobe's licensing policy. The ISO Central Secretariat
accepts no liability in this area.
Adobe is a trademark of Adobe Systems Incorporated.
Details of the software products used to create this PDF file can be found in the General Info relative to the file; the PDF-creation
parameters were optimized for printing. Every care has been taken to ensure that the file is suitable for use by ISO member bodies. In
the unlikely event that a problem relating to it is found, please inform the Central Secretariat at the address given below.
COPYRIGHT PROTECTED DOCUMENT
© ISO 2008
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized in any form or by any means,
electronic or mechanical, including photocopying and microfilm, without permission in writing from either ISO at the address below or
ISO's member body in the country of the requester.
ISO copyright office
Case postale 56 • CH-1211 Geneva 20
Tel. + 41 22 749 01 11
Fax + 41 22 749 09 47
E-mail copyright@iso.org
Web www.iso.org
Published in Switzerland
ii © ISO 2008 – All rights reserved
---------------------- Page: 4 ----------------------
SIST ISO 5667-20:2010
ISO 5667-20:2008(E)
Contents Page
Foreword. iv
Introduction . vi
1 Scope . 1
2 Summary of key points . 1
3 Types of error and variation . 2
3.1 General. 2
3.2 Analytical error. 3
3.3 Overall uncertainty . 3
4 Activities . 4
4.1 Estimation of summary statistics . 4
4.2 Thresholds for water quality and compliance . 6
4.3 Confidence of failure . 7
4.4 Methods for thresholds expressed as percentiles. 7
4.5 Non-parametric methods . 10
4.6 Look-up tables . 13
5 Definition of thresholds. 14
5.1 General. 14
5.2 Ideal thresholds . 14
5.3 Absolute limits . 15
5.4 Percentage of failed samples . 18
5.5 Calculating limits for effluent discharges . 18
6 Declaring that a substance has been detected . 19
7 Detecting change. 20
8 Classification. 23
8.1 General. 23
8.2 Confidence that class has changed. 25
Annex A (informative) Calculation of confidence limits. 27
Annex B (informative) Calculation for the binomial distribution. 29
Annex C (informative) Sample results with high error or reported as less than a limit of detection . 32
Bibliography . 34
© ISO 2008 – All rights reserved iii
---------------------- Page: 5 ----------------------
SIST ISO 5667-20:2010
ISO 5667-20:2008(E)
Foreword
ISO (the International Organization for Standardization) is a worldwide federation of national standards bodies
(ISO member bodies). The work of preparing International Standards is normally carried out through ISO
technical committees. Each member body interested in a subject for which a technical committee has been
established has the right to be represented on that committee. International organizations, governmental and
non-governmental, in liaison with ISO, also take part in the work. ISO collaborates closely with the
International Electrotechnical Commission (IEC) on all matters of electrotechnical standardization.
International Standards are drafted in accordance with the rules given in the ISO/IEC Directives, Part 2.
The main task of technical committees is to prepare International Standards. Draft International Standards
adopted by the technical committees are circulated to the member bodies for voting. Publication as an
International Standard requires approval by at least 75 % of the member bodies casting a vote.
Attention is drawn to the possibility that some of the elements of this document may be the subject of patent
rights. ISO shall not be held responsible for identifying any or all such patent rights.
ISO 5667-20 was prepared by Technical Committee ISO/TC 147, Water quality, Subcommittee SC 6,
Sampling (general methods).
ISO 5667 consists of the following parts, under the general title Water quality — Sampling:
⎯ Part 1: Guidance on the design of sampling programmes and sampling techniques
⎯ Part 3: Guidance on the preservation and handling of water samples
⎯ Part 4: Guidance on sampling from lakes, natural and man-made
⎯ Part 5: Guidance on sampling of drinking water from treatment works and piped distribution systems
⎯ Part 6: Guidance on sampling of rivers and streams
⎯ Part 7: Guidance on sampling of water and steam in boiler plants
⎯ Part 8: Guidance on the sampling of wet deposition
⎯ Part 9: Guidance on sampling from marine waters
⎯ Part 10: Guidance on sampling of waste waters
⎯ Part 11: Guidance on sampling of groundwaters
⎯ Part 12: Guidance on sampling of bottom sediments
⎯ Part 13: Guidance on sampling of sludges from sewage and water treatment works
⎯ Part 14: Guidance on quality assurance of environmental water sampling and handling
⎯ Part 15: Guidance on preservation and handling of sludge and sediment samples
⎯ Part 16: Guidance on biotesting of samples
iv © ISO 2008 – All rights reserved
---------------------- Page: 6 ----------------------
SIST ISO 5667-20:2010
ISO 5667-20:2008(E)
⎯ Part 17: Guidance on sampling of bulk suspended solids
⎯ Part 18: Guidance on sampling of groundwater at contaminated sites
⎯ Part 19: Guidance on sampling of marine sediments
⎯ Part 20: Guidance on the use of sampling data for decision making — Compliance with thresholds and
classification systems
The following parts are under preparation:
⎯ Part 21: Guidance on sampling of drinking water distributed by non-continuous, non-conventional means
⎯ Part 22: Guidance on design and installation of groundwater sample points
⎯ Part 23: Determination of significant pollutants in surface waters using passive sampling
© ISO 2008 – All rights reserved v
---------------------- Page: 7 ----------------------
SIST ISO 5667-20:2010
ISO 5667-20:2008(E)
Introduction
This part of ISO 5667 concerns the use of information on water quality obtained by taking samples in taking
decisions — in measuring success, failure or change, in the context of the inevitable uncertainties associated
with sampling. This part of ISO 5667 provides guidance on controlling the risk of such uncertainties leading to
non-optimal decisions.
Non-optimal decisions can also stem from the way in which thresholds for discharges and targets for
environmental waters are formulated or set out in regulations and permits. This part of ISO 5667 also
examines the problems caused when compliance with these thresholds is assessed using data obtained by
sampling.
This part of ISO 5667 aims to ensure that future laws, regulations, and guidance assert the requirement to
assess and report statistical significance.
NOTE 1 Decisions might result in the commendation or criticism of people, sites, companies, sectors or nations.
Decisions can give rise to legal action and/or expensive and time-consuming remedial actions to improve water quality.
Figure 1 shows the links between the following topics:
a) the setting up of thresholds for taking decisions on the need to improve water quality, possibly including
criteria to minimize water quality deterioration;
b) the establishment of sampling programmes to satisfy the requirements of these thresholds and the need
to assess performance against them;
c) making use of the outcome of sampling programmes to take decisions.
Figure 1 — Links between topics associated with sampling and taking decisions
vi © ISO 2008 – All rights reserved
---------------------- Page: 8 ----------------------
SIST ISO 5667-20:2010
ISO 5667-20:2008(E)
This part of ISO 5667 deals with topic c). Topics a) and b) are huge and wide ranging in their own right, and
their detailed treatment lies outside the scope of this part of ISO 5667. Nevertheless, this part of ISO 5667
does make recommendations for the expression of targets and thresholds for water quality [topic a)], which
are important when using sample data to take decisions. This part of ISO 5667 also gives advice on what is
required for sampling programmes [topic b)] in order that they be compatible with the way thresholds are
defined, and so as to place no unnecessary difficulties and errors in the process of taking decisions.
Other areas which lie outside the scope of this part of ISO 5667 are: the detailed mechanics of taking and
handling samples; assurance that samples are representative over time of the body of water being sampled;
and performance of chemical analyses on samples. These are all covered in other documents. Nonetheless, if
poorly obtained results from these areas can add substantially to overall sampling uncertainties and impose
extra difficulties in taking decisions. This part of ISO 5667 describes some of these extra difficulties.
This part of ISO 5667 does not cover the full range of statistical techniques that may be applied and the
circumstances in which they should be used. The main purpose is to establish the principle that uncertainty
from sampling and analysis (and errors generally) should always be assessed and taken into account as part
of the process of taking decisions. If this is not done, incorrect decisions can result, for example, on where
action is needed, and the scale of that action.
NOTE 2 Some statistical techniques are used as illustrative examples. These are techniques that have seen routine
use in some regulatory regimes that take proper account of statistical uncertainties. They are suitable for use in situations
that resemble the worked examples discussed.
It is not the purpose of this part of ISO 5667 to direct the development of regulatory conditions. This part of
ISO 5667 provides principles and tools to support management, including regulation. It is recognised that
regulatory thresholds are developed using a range of strategies that incorporate technical, social and legal
considerations. It is also recognised that tools other than statistical data analysis are likely to be used in
interpreting and applying thresholds.
© ISO 2008 – All rights reserved vii
---------------------- Page: 9 ----------------------
SIST ISO 5667-20:2010
---------------------- Page: 10 ----------------------
SIST ISO 5667-20:2010
INTERNATIONAL STANDARD ISO 5667-20:2008(E)
Water quality — Sampling —
Part 20:
Guidance on the use of sampling data for decision making —
Compliance with thresholds and classification systems
1 Scope
This part of ISO 5667 establishes principles, basic requirements, and illustrative methods for dealing with the
use of sample data for decision making based on the assessment of the confidence that water quality:
a) meets targets and complies with thresholds;
b) has changed; and/or
c) lies in a particular grade in a classification system.
This part of ISO 5667 also specifies methods for preliminary examination of the sensitivity of decisions to error
and uncertainty, although it does not cover the full range of statistical techniques.
This part of ISO 5667 provides general advice on decision making related to constraint formulation for
expression of thresholds and targets and the form and scale of sampling programmes.
NOTE 1 In the water industry, “standard” is commonly used to indicate the value or limit of a parameter of interest.
However, in this part of ISO 5667, the term “threshold” is used to avoid confusion with published national, regional, and
International Standards.
NOTE 2 This document is framed in terms of sampling and measurement of chemical concentrations, in particular
those subject to strong day-to-day temporal variations. The principles apply, however, to any item estimated by sampling
which is subject to random error, including microbiological and biological data, and data subject to strong spatial variations.
2 Summary of key points
Water quality is often assessed by the results of chemical analysis of a number of samples taken over a
period of time.
Uncertainty is introduced by the action of random chance in taking samples. It can be present in any set of
measurements of water quality taken over a period of time. The values for chemical analysis of these samples
depend on the quality of the particular small volumes of water that are extracted or measured. If water quality
varies in space or time, a second set of samples taken over the same period will have different values
because these samples are made up of different small volumes of water taken at different times. Each set of
samples allows an estimate of the true water quality. These estimates will differ: they will have a different
mean and span a different range. They have the potential, if taken at face value, to suggest different
conclusions about compliance with thresholds and targets.
Sampling uncertainty (or sampling error) is the term often given to this effect. Sampling uncertainty includes
uncertainties and errors associated with chemical analysis, and occurs even in the case of trivial errors in
chemical analysis and if there are no mistakes in the methods by which samples are taken and handled.
© ISO 2008 – All rights reserved 1
---------------------- Page: 11 ----------------------
SIST ISO 5667-20:2010
ISO 5667-20:2008(E)
Sampling uncertainty is reduced if more samples are taken, but the scale of the uncertainty is often
unappreciated.
In this part of ISO 5667, “overall uncertainty” includes these chance sampling effects and all the other sources
of variation in a set of samples. This variability reflects the underlying signals generated by natural or perhaps
unnatural processes; it includes the effects of errors in chemical analysis and the handling of samples. It might
contain systematic variations from trends and diurnal, weekly, and seasonal cycles. In this context, the more
[5]
appropriate term is “overall uncertainty”, “overall error” or “total assay error” (ISO/IEC Guide 99:1993 ).
Overall uncertainty should be quantified, at least approximately, and taken into account in all cases where
water quality varies and sampling is used to estimate information used in decision making. This includes
assessing compliance with thresholds (see Clause 5), deciding whether water quality has changed (see
Clause 7), and putting waters into grades in classification systems (see Clause 8). This part of ISO 5667
recommends that:
a) thresholds for which compliance is assessed by sampling should be defined or used so that the overall
uncertainty can be estimated and dealt with appropriately (see 5.2);
b) thresholds defined as absolute limits should be treated as percentiles when assessing compliance using
sampling (see 5.3);
c) thresholds defined as limits to be met by a percentage of samples should be defined or used as the
corresponding percentiles (see 5.4);
d) the degree of confidence should be estimated when assessing compliance with thresholds (see
Clause 4); and,
e) the degree of confidence in changes or differences should be estimated when aiming to demonstrate
change or no change (see 8.2).
3 Types of error and variation
3.1 General
In many procedures by which sample data are used to take decisions, there is a set of results taken over a
period of time (e.g. a year). This information might be used to make such judgements as whether:
a) water quality in a river failed to meet required thresholds;
b) a treatment works performed better this year than last;
c) water quality in a lake needs improvement;
d) one company has better effluent discharge compliance than another; or
e) most of the risk of environmental impact is from a particular type of effluent discharge.
There are unlikely to be many significant changes in water quality from second to second throughout a year,
but variations from day to day are common. These can be due to diurnal cycles, the play of random errors and
bias from the laboratory, the weather, step changes, day-to-day and hour-by-hour variations (perhaps in the
natural processes in water or caused by discharges and abstractions and changes in these), seasonal and
economic cycles, and several underlying and overlapping long-term trends and cycles.
NOTE 1 Sometimes several or most of the data are reported by a laboratory as being less than a specified limit of
detection. Such data are called censored data. Depending on the types of decisions that depend on the data, special
statistical techniques are available for estimating the values of summary statistics and their uncertainties.
2 © ISO 2008 – All rights reserved
---------------------- Page: 12 ----------------------
SIST ISO 5667-20:2010
ISO 5667-20:2008(E)
In addition, the total set of samples shall be representative of the average quality of the masses of water from
which they were taken, e.g. over a period of time under review. In estimating an annual mean, it is not
acceptable for all samples to be taken in April, for example. These requirements should be set up in the
design of the sampling programme.
[1]
NOTE 2 Guidance on all these aspects is given in more detail in ISO 5667-1 .
3.2 Analytical error
Analytical errors are those introduced by the process of chemical analysis and reflect that these
measurements are not error free. It might be that the result for a single sample can be specified to within a
specific range, e.g. ±15 %.
NOTE 1 The actual value of the analytical error depends on the capabilities of the equipment and the laboratory that
has been used to perform the analysis. The discussion in this part of ISO 5667 focuses on random error, but there is
always a risk of non-random error, e.g. when there is a change of instrument or methods, when the sample matrix varies
[5]
greatly from the calibration materials, and for results just above the detection limit (ISO/IEC Guide 99:1993 ).
NOTE 2 The results of chemical analysis are nowadays reported with uncertainty values in accordance with
[6]
ISO/IEC 17025 .
When a mean is calculated from n samples, the effect on the uncertainty in the estimate of the mean of
random errors in chemical analysis tends to average down according to n. For example, if the analytical
error associated with a single sample were ±15 %, then the error in the estimate of the mean of a set of
chemical analyses would tend to reduce to something like ±4 % for 12 samples or to ±2,5 % for 36 samples.
In using samples to take decisions, this kind of error from chemical analysis augments but is often smaller
than other contributions to the overall uncertainty, especially that associated with chance in the taking of a
limited number of samples. Chemical analysis error comes through as an addition to that associated with
chance in the taking of a limited number of samples, but it might be a small addition. {Nevertheless, some
studies need to separate sampling variance from local environmental heterogeneity (see Reference [7]).}
NOTE 3 It is not commonly understood that data fully within the statistical control of a laboratory might be unsuitable for
particular interpretations because of errors associated with taking a small number of samples.
NOTE 4 This observation on the relative importance of analytical error applies generally to the types of issues
considered in this part of ISO 5667, but it follows from estimating the analytical error in such cases, and comparing it with
other errors. The analytical error should always be estimated. Similar points can be made about making sure samples are
representative, and about checking changes to methods of sampling.
When the sample results are used to estimate the value of other summary statistics such as percentiles (e.g.
the 95-percentile, which is the value exceeded for 5 % of the time), the picture is similar to that for the mean,
i.e. the errors are inversely proportional to n, but are larger than for the mean.
3.3 Overall uncertainty
Uncertainty occurs because of variations in the quality of the water being sampled, and the ability of the
sampling process to accurately reflect these variations. In a set of samples taken over a period of time, the
results are affected by the operation of the laws of chance in the way the particular samples came to be
collected. This produces uncertainty even if:
1)
⎯ analytical errors are close to zero ;
⎯ the sampling programme guarantees samples that are truly representative in time and space;
⎯ there are no mistakes in handling the samples and recording the results of analysis.
1) Nearly always this is a hypothetical possibility. Many trace elements are measured near their detection limits and have
analytical uncertainty of about ±100 %. Many organic chemicals can have recoveries of ±50 %.
© ISO 2008 – All rights reserved 3
---------------------- Page: 13 ----------------------
SIST ISO 5667-20:2010
ISO 5667-20:2008(E)
In using sampling, the main source of uncertainty is usually associated with the number of samples taken. In
the types of decision on activities listed in items a) to f) below, this source of uncertainty is usually a bigger
issue than, for example, that associated with errors of chemical analysis. Overall uncertainty should be
assessed and used to quantify uncertainty in cases where water quality varies and decisions are taken as a
consequence of the following types of activities:
a) using sampling to measure and report on water quality;
b) using samples to estimate summary statistics, e.g. the monthly mean, the annual percentile or the annual
maximum;
c) making statements about whether this year’s summary statistics are higher or lower than last year’s (see
[6]
ISO/IEC 17025 for a wider view of the issue of looking for change);
d) establishing whether water quality exceeds a threshold;
e) using summary statistics to place water quality in a particular class within a classification system; or
f) assessing whether a change in class has occurred.
In all these situations, the aim is to assess whether the change or the status is statistically significant and to
require that future laws, regulations and guidance assert the requirement to assess and report statistical
significance.
4 Activities
2)
4.1 Estimation of summary statistics
An estimate of a summary statistic depends on the values of water quality in the small volumes of water that
happen to be captured by sampling and whether these values are measured accurately. The estimate, due to
the overall uncertainty, is almost certain to differ from the true value of the summary statistic — the value that
would be obtained if it were possible to achieve continuous error-free monitoring over the entire period for
which the summary statistic applies.
Uncertainty can be managed by calculating confidence limits. Confidence limits define the range within which
the true value of the estimate of the summary statistic is expected to lie. In the example in Table 1, the
estimate of the mean from eight samples is 101 mg/l and there is a pair of 95 % confidence limits, 46 mg/l and
156 mg/l. There is 95 % confidence that the true mean exceeds the lower 95 % confidence limit of 46 mg/l and
95 % confidence that the true mean is less than the upper 95 % confidence limit of 156 mg/l. Overall there is
3)
90 % confidence that the true value of the mean falls in the range between 46 mg/l and 156 mg/l .
This range in the estimate of the mean represents large errors but these errors are seldom estimated or used
to help take decisions based on the data. Also this discussion is for normally distributed random error. Such
assumptions should be stated. Random error might not be normally distributed; it could be non-random and
subject to mistakes and blunders. As a rule, the effect of these will be to increase the scale of the error. Errors
should always be estimated even if this is done by making an assumption that they follow a normal distribution.
NOTE 1 The mean is used in this example because this summary statistic is commonly required by legislation. In other
cases, there may grounds and opportunity to use other statistics like the median, e.g. to explain differences between large
and small samples. The median is useful for data sets affected by outliers, and confidence limits can be calculated for the
median.
2) Some documents use the concept of “sampling target”. The sampling target could be the annual water quality, and a
mean value over 1 year, or a 95-percentile over 1 year, is what is estimated.
3) This range is sometimes called the 90 % con
...
SLOVENSKI STANDARD
oSIST ISO 5667-20:2009
01-junij-2009
.DNRYRVWYRGH9]RUþHQMHGHO 1DYRGLORRXSRUDELSRGDWNRYRY]RUþHQMX]D
RGORþDQMH6NODGQRVW]PHMQLPLYUHGQRVWPLLQNODVLILNDFLMVNLPLVLVWHPL
Water quality - Sampling - Part 20: Guidance on the use of sampling data for decision
making - Compliance with thresholds and classification systems
Qualité de l'eau - Échantillonnage - Partie 20: Lignes directrices relatives à l'utilisation
des données d'échantillonnage pour la prise de décision - Conformité avec les limites et
systèmes de classification
Ta slovenski standard je istoveten z: ISO 5667-20:2008
ICS:
13.060.45 Preiskava vode na splošno Examination of water in
general
oSIST ISO 5667-20:2009 en
2003-01.Slovenski inštitut za standardizacijo. Razmnoževanje celote ali delov tega standarda ni dovoljeno.
---------------------- Page: 1 ----------------------
oSIST ISO 5667-20:2009
---------------------- Page: 2 ----------------------
oSIST ISO 5667-20:2009
INTERNATIONAL ISO
STANDARD 5667-20
First edition
2008-03-15
Water quality — Sampling —
Part 20:
Guidance on the use of sampling data for
decision making — Compliance with
thresholds and classification systems
Qualité de l'eau — Échantillonnage —
Partie 20: Lignes directrices relatives à l'utilisation des données
d'échantillonnage pour la prise de décision — Conformité avec les
limites et systèmes de classification
Reference number
ISO 5667-20:2008(E)
©
ISO 2008
---------------------- Page: 3 ----------------------
oSIST ISO 5667-20:2009
ISO 5667-20:2008(E)
PDF disclaimer
This PDF file may contain embedded typefaces. In accordance with Adobe's licensing policy, this file may be printed or viewed but
shall not be edited unless the typefaces which are embedded are licensed to and installed on the computer performing the editing. In
downloading this file, parties accept therein the responsibility of not infringing Adobe's licensing policy. The ISO Central Secretariat
accepts no liability in this area.
Adobe is a trademark of Adobe Systems Incorporated.
Details of the software products used to create this PDF file can be found in the General Info relative to the file; the PDF-creation
parameters were optimized for printing. Every care has been taken to ensure that the file is suitable for use by ISO member bodies. In
the unlikely event that a problem relating to it is found, please inform the Central Secretariat at the address given below.
COPYRIGHT PROTECTED DOCUMENT
© ISO 2008
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized in any form or by any means,
electronic or mechanical, including photocopying and microfilm, without permission in writing from either ISO at the address below or
ISO's member body in the country of the requester.
ISO copyright office
Case postale 56 • CH-1211 Geneva 20
Tel. + 41 22 749 01 11
Fax + 41 22 749 09 47
E-mail copyright@iso.org
Web www.iso.org
Published in Switzerland
ii © ISO 2008 – All rights reserved
---------------------- Page: 4 ----------------------
oSIST ISO 5667-20:2009
ISO 5667-20:2008(E)
Contents Page
Foreword. iv
Introduction . vi
1 Scope . 1
2 Summary of key points . 1
3 Types of error and variation . 2
3.1 General. 2
3.2 Analytical error. 3
3.3 Overall uncertainty . 3
4 Activities . 4
4.1 Estimation of summary statistics . 4
4.2 Thresholds for water quality and compliance . 6
4.3 Confidence of failure . 7
4.4 Methods for thresholds expressed as percentiles. 7
4.5 Non-parametric methods . 10
4.6 Look-up tables . 13
5 Definition of thresholds. 14
5.1 General. 14
5.2 Ideal thresholds . 14
5.3 Absolute limits . 15
5.4 Percentage of failed samples . 18
5.5 Calculating limits for effluent discharges . 18
6 Declaring that a substance has been detected . 19
7 Detecting change. 20
8 Classification. 23
8.1 General. 23
8.2 Confidence that class has changed. 25
Annex A (informative) Calculation of confidence limits. 27
Annex B (informative) Calculation for the binomial distribution. 29
Annex C (informative) Sample results with high error or reported as less than a limit of detection . 32
Bibliography . 34
© ISO 2008 – All rights reserved iii
---------------------- Page: 5 ----------------------
oSIST ISO 5667-20:2009
ISO 5667-20:2008(E)
Foreword
ISO (the International Organization for Standardization) is a worldwide federation of national standards bodies
(ISO member bodies). The work of preparing International Standards is normally carried out through ISO
technical committees. Each member body interested in a subject for which a technical committee has been
established has the right to be represented on that committee. International organizations, governmental and
non-governmental, in liaison with ISO, also take part in the work. ISO collaborates closely with the
International Electrotechnical Commission (IEC) on all matters of electrotechnical standardization.
International Standards are drafted in accordance with the rules given in the ISO/IEC Directives, Part 2.
The main task of technical committees is to prepare International Standards. Draft International Standards
adopted by the technical committees are circulated to the member bodies for voting. Publication as an
International Standard requires approval by at least 75 % of the member bodies casting a vote.
Attention is drawn to the possibility that some of the elements of this document may be the subject of patent
rights. ISO shall not be held responsible for identifying any or all such patent rights.
ISO 5667-20 was prepared by Technical Committee ISO/TC 147, Water quality, Subcommittee SC 6,
Sampling (general methods).
ISO 5667 consists of the following parts, under the general title Water quality — Sampling:
⎯ Part 1: Guidance on the design of sampling programmes and sampling techniques
⎯ Part 3: Guidance on the preservation and handling of water samples
⎯ Part 4: Guidance on sampling from lakes, natural and man-made
⎯ Part 5: Guidance on sampling of drinking water from treatment works and piped distribution systems
⎯ Part 6: Guidance on sampling of rivers and streams
⎯ Part 7: Guidance on sampling of water and steam in boiler plants
⎯ Part 8: Guidance on the sampling of wet deposition
⎯ Part 9: Guidance on sampling from marine waters
⎯ Part 10: Guidance on sampling of waste waters
⎯ Part 11: Guidance on sampling of groundwaters
⎯ Part 12: Guidance on sampling of bottom sediments
⎯ Part 13: Guidance on sampling of sludges from sewage and water treatment works
⎯ Part 14: Guidance on quality assurance of environmental water sampling and handling
⎯ Part 15: Guidance on preservation and handling of sludge and sediment samples
⎯ Part 16: Guidance on biotesting of samples
iv © ISO 2008 – All rights reserved
---------------------- Page: 6 ----------------------
oSIST ISO 5667-20:2009
ISO 5667-20:2008(E)
⎯ Part 17: Guidance on sampling of bulk suspended solids
⎯ Part 18: Guidance on sampling of groundwater at contaminated sites
⎯ Part 19: Guidance on sampling of marine sediments
⎯ Part 20: Guidance on the use of sampling data for decision making — Compliance with thresholds and
classification systems
The following parts are under preparation:
⎯ Part 21: Guidance on sampling of drinking water distributed by non-continuous, non-conventional means
⎯ Part 22: Guidance on design and installation of groundwater sample points
⎯ Part 23: Determination of significant pollutants in surface waters using passive sampling
© ISO 2008 – All rights reserved v
---------------------- Page: 7 ----------------------
oSIST ISO 5667-20:2009
ISO 5667-20:2008(E)
Introduction
This part of ISO 5667 concerns the use of information on water quality obtained by taking samples in taking
decisions — in measuring success, failure or change, in the context of the inevitable uncertainties associated
with sampling. This part of ISO 5667 provides guidance on controlling the risk of such uncertainties leading to
non-optimal decisions.
Non-optimal decisions can also stem from the way in which thresholds for discharges and targets for
environmental waters are formulated or set out in regulations and permits. This part of ISO 5667 also
examines the problems caused when compliance with these thresholds is assessed using data obtained by
sampling.
This part of ISO 5667 aims to ensure that future laws, regulations, and guidance assert the requirement to
assess and report statistical significance.
NOTE 1 Decisions might result in the commendation or criticism of people, sites, companies, sectors or nations.
Decisions can give rise to legal action and/or expensive and time-consuming remedial actions to improve water quality.
Figure 1 shows the links between the following topics:
a) the setting up of thresholds for taking decisions on the need to improve water quality, possibly including
criteria to minimize water quality deterioration;
b) the establishment of sampling programmes to satisfy the requirements of these thresholds and the need
to assess performance against them;
c) making use of the outcome of sampling programmes to take decisions.
Figure 1 — Links between topics associated with sampling and taking decisions
vi © ISO 2008 – All rights reserved
---------------------- Page: 8 ----------------------
oSIST ISO 5667-20:2009
ISO 5667-20:2008(E)
This part of ISO 5667 deals with topic c). Topics a) and b) are huge and wide ranging in their own right, and
their detailed treatment lies outside the scope of this part of ISO 5667. Nevertheless, this part of ISO 5667
does make recommendations for the expression of targets and thresholds for water quality [topic a)], which
are important when using sample data to take decisions. This part of ISO 5667 also gives advice on what is
required for sampling programmes [topic b)] in order that they be compatible with the way thresholds are
defined, and so as to place no unnecessary difficulties and errors in the process of taking decisions.
Other areas which lie outside the scope of this part of ISO 5667 are: the detailed mechanics of taking and
handling samples; assurance that samples are representative over time of the body of water being sampled;
and performance of chemical analyses on samples. These are all covered in other documents. Nonetheless, if
poorly obtained results from these areas can add substantially to overall sampling uncertainties and impose
extra difficulties in taking decisions. This part of ISO 5667 describes some of these extra difficulties.
This part of ISO 5667 does not cover the full range of statistical techniques that may be applied and the
circumstances in which they should be used. The main purpose is to establish the principle that uncertainty
from sampling and analysis (and errors generally) should always be assessed and taken into account as part
of the process of taking decisions. If this is not done, incorrect decisions can result, for example, on where
action is needed, and the scale of that action.
NOTE 2 Some statistical techniques are used as illustrative examples. These are techniques that have seen routine
use in some regulatory regimes that take proper account of statistical uncertainties. They are suitable for use in situations
that resemble the worked examples discussed.
It is not the purpose of this part of ISO 5667 to direct the development of regulatory conditions. This part of
ISO 5667 provides principles and tools to support management, including regulation. It is recognised that
regulatory thresholds are developed using a range of strategies that incorporate technical, social and legal
considerations. It is also recognised that tools other than statistical data analysis are likely to be used in
interpreting and applying thresholds.
© ISO 2008 – All rights reserved vii
---------------------- Page: 9 ----------------------
oSIST ISO 5667-20:2009
---------------------- Page: 10 ----------------------
oSIST ISO 5667-20:2009
INTERNATIONAL STANDARD ISO 5667-20:2008(E)
Water quality — Sampling —
Part 20:
Guidance on the use of sampling data for decision making —
Compliance with thresholds and classification systems
1 Scope
This part of ISO 5667 establishes principles, basic requirements, and illustrative methods for dealing with the
use of sample data for decision making based on the assessment of the confidence that water quality:
a) meets targets and complies with thresholds;
b) has changed; and/or
c) lies in a particular grade in a classification system.
This part of ISO 5667 also specifies methods for preliminary examination of the sensitivity of decisions to error
and uncertainty, although it does not cover the full range of statistical techniques.
This part of ISO 5667 provides general advice on decision making related to constraint formulation for
expression of thresholds and targets and the form and scale of sampling programmes.
NOTE 1 In the water industry, “standard” is commonly used to indicate the value or limit of a parameter of interest.
However, in this part of ISO 5667, the term “threshold” is used to avoid confusion with published national, regional, and
International Standards.
NOTE 2 This document is framed in terms of sampling and measurement of chemical concentrations, in particular
those subject to strong day-to-day temporal variations. The principles apply, however, to any item estimated by sampling
which is subject to random error, including microbiological and biological data, and data subject to strong spatial variations.
2 Summary of key points
Water quality is often assessed by the results of chemical analysis of a number of samples taken over a
period of time.
Uncertainty is introduced by the action of random chance in taking samples. It can be present in any set of
measurements of water quality taken over a period of time. The values for chemical analysis of these samples
depend on the quality of the particular small volumes of water that are extracted or measured. If water quality
varies in space or time, a second set of samples taken over the same period will have different values
because these samples are made up of different small volumes of water taken at different times. Each set of
samples allows an estimate of the true water quality. These estimates will differ: they will have a different
mean and span a different range. They have the potential, if taken at face value, to suggest different
conclusions about compliance with thresholds and targets.
Sampling uncertainty (or sampling error) is the term often given to this effect. Sampling uncertainty includes
uncertainties and errors associated with chemical analysis, and occurs even in the case of trivial errors in
chemical analysis and if there are no mistakes in the methods by which samples are taken and handled.
© ISO 2008 – All rights reserved 1
---------------------- Page: 11 ----------------------
oSIST ISO 5667-20:2009
ISO 5667-20:2008(E)
Sampling uncertainty is reduced if more samples are taken, but the scale of the uncertainty is often
unappreciated.
In this part of ISO 5667, “overall uncertainty” includes these chance sampling effects and all the other sources
of variation in a set of samples. This variability reflects the underlying signals generated by natural or perhaps
unnatural processes; it includes the effects of errors in chemical analysis and the handling of samples. It might
contain systematic variations from trends and diurnal, weekly, and seasonal cycles. In this context, the more
[5]
appropriate term is “overall uncertainty”, “overall error” or “total assay error” (ISO/IEC Guide 99:1993 ).
Overall uncertainty should be quantified, at least approximately, and taken into account in all cases where
water quality varies and sampling is used to estimate information used in decision making. This includes
assessing compliance with thresholds (see Clause 5), deciding whether water quality has changed (see
Clause 7), and putting waters into grades in classification systems (see Clause 8). This part of ISO 5667
recommends that:
a) thresholds for which compliance is assessed by sampling should be defined or used so that the overall
uncertainty can be estimated and dealt with appropriately (see 5.2);
b) thresholds defined as absolute limits should be treated as percentiles when assessing compliance using
sampling (see 5.3);
c) thresholds defined as limits to be met by a percentage of samples should be defined or used as the
corresponding percentiles (see 5.4);
d) the degree of confidence should be estimated when assessing compliance with thresholds (see
Clause 4); and,
e) the degree of confidence in changes or differences should be estimated when aiming to demonstrate
change or no change (see 8.2).
3 Types of error and variation
3.1 General
In many procedures by which sample data are used to take decisions, there is a set of results taken over a
period of time (e.g. a year). This information might be used to make such judgements as whether:
a) water quality in a river failed to meet required thresholds;
b) a treatment works performed better this year than last;
c) water quality in a lake needs improvement;
d) one company has better effluent discharge compliance than another; or
e) most of the risk of environmental impact is from a particular type of effluent discharge.
There are unlikely to be many significant changes in water quality from second to second throughout a year,
but variations from day to day are common. These can be due to diurnal cycles, the play of random errors and
bias from the laboratory, the weather, step changes, day-to-day and hour-by-hour variations (perhaps in the
natural processes in water or caused by discharges and abstractions and changes in these), seasonal and
economic cycles, and several underlying and overlapping long-term trends and cycles.
NOTE 1 Sometimes several or most of the data are reported by a laboratory as being less than a specified limit of
detection. Such data are called censored data. Depending on the types of decisions that depend on the data, special
statistical techniques are available for estimating the values of summary statistics and their uncertainties.
2 © ISO 2008 – All rights reserved
---------------------- Page: 12 ----------------------
oSIST ISO 5667-20:2009
ISO 5667-20:2008(E)
In addition, the total set of samples shall be representative of the average quality of the masses of water from
which they were taken, e.g. over a period of time under review. In estimating an annual mean, it is not
acceptable for all samples to be taken in April, for example. These requirements should be set up in the
design of the sampling programme.
[1]
NOTE 2 Guidance on all these aspects is given in more detail in ISO 5667-1 .
3.2 Analytical error
Analytical errors are those introduced by the process of chemical analysis and reflect that these
measurements are not error free. It might be that the result for a single sample can be specified to within a
specific range, e.g. ±15 %.
NOTE 1 The actual value of the analytical error depends on the capabilities of the equipment and the laboratory that
has been used to perform the analysis. The discussion in this part of ISO 5667 focuses on random error, but there is
always a risk of non-random error, e.g. when there is a change of instrument or methods, when the sample matrix varies
[5]
greatly from the calibration materials, and for results just above the detection limit (ISO/IEC Guide 99:1993 ).
NOTE 2 The results of chemical analysis are nowadays reported with uncertainty values in accordance with
[6]
ISO/IEC 17025 .
When a mean is calculated from n samples, the effect on the uncertainty in the estimate of the mean of
random errors in chemical analysis tends to average down according to n. For example, if the analytical
error associated with a single sample were ±15 %, then the error in the estimate of the mean of a set of
chemical analyses would tend to reduce to something like ±4 % for 12 samples or to ±2,5 % for 36 samples.
In using samples to take decisions, this kind of error from chemical analysis augments but is often smaller
than other contributions to the overall uncertainty, especially that associated with chance in the taking of a
limited number of samples. Chemical analysis error comes through as an addition to that associated with
chance in the taking of a limited number of samples, but it might be a small addition. {Nevertheless, some
studies need to separate sampling variance from local environmental heterogeneity (see Reference [7]).}
NOTE 3 It is not commonly understood that data fully within the statistical control of a laboratory might be unsuitable for
particular interpretations because of errors associated with taking a small number of samples.
NOTE 4 This observation on the relative importance of analytical error applies generally to the types of issues
considered in this part of ISO 5667, but it follows from estimating the analytical error in such cases, and comparing it with
other errors. The analytical error should always be estimated. Similar points can be made about making sure samples are
representative, and about checking changes to methods of sampling.
When the sample results are used to estimate the value of other summary statistics such as percentiles (e.g.
the 95-percentile, which is the value exceeded for 5 % of the time), the picture is similar to that for the mean,
i.e. the errors are inversely proportional to n, but are larger than for the mean.
3.3 Overall uncertainty
Uncertainty occurs because of variations in the quality of the water being sampled, and the ability of the
sampling process to accurately reflect these variations. In a set of samples taken over a period of time, the
results are affected by the operation of the laws of chance in the way the particular samples came to be
collected. This produces uncertainty even if:
1)
⎯ analytical errors are close to zero ;
⎯ the sampling programme guarantees samples that are truly representative in time and space;
⎯ there are no mistakes in handling the samples and recording the results of analysis.
1) Nearly always this is a hypothetical possibility. Many trace elements are measured near their detection limits and have
analytical uncertainty of about ±100 %. Many organic chemicals can have recoveries of ±50 %.
© ISO 2008 – All rights reserved 3
---------------------- Page: 13 ----------------------
oSIST ISO 5667-20:2009
ISO 5667-20:2008(E)
In using sampling, the main source of uncertainty is usually associated with the number of samples taken. In
the types of decision on activities listed in items a) to f) below, this source of uncertainty is usually a bigger
issue than, for example, that associated with errors of chemical analysis. Overall uncertainty should be
assessed and used to quantify uncertainty in cases where water quality varies and decisions are taken as a
consequence of the following types of activities:
a) using sampling to measure and report on water quality;
b) using samples to estimate summary statistics, e.g. the monthly mean, the annual percentile or the annual
maximum;
c) making statements about whether this year’s summary statistics are higher or lower than last year’s (see
[6]
ISO/IEC 17025 for a wider view of the issue of looking for change);
d) establishing whether water quality exceeds a threshold;
e) using summary statistics to place water quality in a particular class within a classification system; or
f) assessing whether a change in class has occurred.
In all these situations, the aim is to assess whether the change or the status is statistically significant and to
require that future laws, regulations and guidance assert the requirement to assess and report statistical
significance.
4 Activities
2)
4.1 Estimation of summary statistics
An estimate of a summary statistic depends on the values of water quality in the small volumes of water that
happen to be captured by sampling and whether these values are measured accurately. The estimate, due to
the overall uncertainty, is almost certain to differ from the true value of the summary statistic — the value that
would be obtained if it were possible to achieve continuous error-free monitoring over the entire period for
which the summary statistic applies.
Uncertainty can be managed by calculating confidence limits. Confidence limits define the range within which
the true value of the estimate of the summary statistic is expected to lie. In the example in Table 1, the
estimate of the mean from eight samples is 101 mg/l and there is a pair of 95 % confidence limits, 46 mg/l and
156 mg/l. There is 95 % confidence that the true mean exceeds the lower 95 % confidence limit of 46 mg/l and
95 % confidence that the true mean is less than the upper 95 % confidence limit of 156 mg/l. Overall there is
3)
90 % confidence that the true value of the mean falls in the range between 46 mg/l and 156 mg/l .
This range in the estimate of the mean represents large errors but these errors are seldom estimated or used
to help take decisions based on the data. Also this discussion is for normally distributed random error. Such
assumptions should be stated. Random error might not be normally distributed; it could be non-random and
subject to mistakes and blunders. As a rule, the effect of these will be to increase the scale of the error. Errors
should always be estimated even if this is done by making an assumption that they follow a normal distribution.
NOTE 1 The mean is used in this example because this summary statistic is commonly required by legislation. In other
cases, there may grounds and opportunity to use other statistics like the median, e.g. to explain differences between large
and small samples. The median is useful for data sets affected by outliers, and confidence limits can be calculated for the
median.
2) Some documents use the concept of “sampling target”. The sampling target could be the annual water quality, and a
mean value over 1 year, or a 95-percentile over 1 year, is what is estimated.
3) This range is sometimes called the 90 % confidence interva
...
Questions, Comments and Discussion
Ask us and Technical Secretary will try to provide an answer. You can facilitate discussion about the standard in here.