QA/QC for Water Quality Data DIVE INTO DATA November 26 | 12-1pm - - PowerPoint PPT Presentation

With

DIVE INTO DATA Webinar Series

QA/QC for Water Quality Data

November 26 | 12-1pm Eastern

Mary Kruk Water Data Specialist The Gordon Foundation Megan Thompson Aquatic Ecologist, Limnologist Thompson Aquatic Consulting

The Gordon Foundation is a 55- year old charitable organization with a long history of protecting Canada’s waters and promoting citizen engagement in policy- making and action.

Overview

• Error: precision and accuracy
• Definition of QA/QC
• Quality Assurance
• Importance of study design
• Standard procedures
• Quality Control
• Measurement Limits
• Blanks
• Replicates
• Tips for Data Inspection

Precision and Accuracy

Taken from “Scientia Plus Conscientia” blog, 2020

Causes

Sampling Design errors:

• Samples not representative of system

Systematic errors:

• Instrument is not accurate
• Incorrect measurement techniques
• Lack of observation

Random errors:

• Unexpected shifts in sampling

environment

• Personal bias

What causes errors in water quality data?

Solutions

Sampling Design errors:

• Properly define your background

Systematic errors:

• Calibrate and care for equipment
• Follow standard procedures
• Take field notes

Random errors:

• Replication
• Can never eliminate, but can reduce by a

thoughtful study design

https://manoa.hawaii.edu/exploringourfluidearth/physical/world-ocean/map-distortion/practices-science-scientific-error

How Things Go Wrong

Public Domain, https://commons.wikimedia.org/w/index.php?curid=2740821 https://universe.byu.edu/2018/03/07/byu-students-community- participate-in-citizen-science-project-to-improve-utah-lake1/ https://agriculture.vic.gov.au/livestock-and- animals/livestock-health-and-welfare/caring-for-animals- during-extreme-heat

How Things Go Wrong

Goldberg and Strickler 2014 http://www.soest.hawaii.edu/S-LAB/equipment/slab_autoanalyzer.htm

Quality Assurance (QA) = Prevent error

Precautions, standard protocols, sensor calibration Are you measuring what you think you’re measuring?

Quality Control (QC) = Detect error

Test blanks, duplicates, and standards, examine data Is the data you’re looking at reporting what it says it’s reporting?

What is QA/QC?

Quality Assurance

Importance of sampling program design

Carefully plan monitoring programs – address bias and improve accuracy

• Are control sites really unaffected? Are there other impacts that you are not

accounting for?

• Are impacted sites impacted in ways other than what you are monitoring for? What

are those impacts, and how might they be reflected in your data?

• How will conditions change over time, with seasons and over the years? How could

this affect your sampling plan? And how might these variations be reflected in your data? Many additional, important considerations….

Standard procedures

In the field...

PEI ELJ Environment Division 2011 ALS Limited https://3.bp.blogspot.com/_YLEBrXdXjB4/S9MtWoKDn9I/AAAAAAAAAQ4/_DHa4KMsCw8/ s1600/scan0002.jpg

Standard procedures

In the lab...

Quality Control

Measurement limits in water quality

Detection limits refer to the ability to DETECT a constituent in water. Quantification limits refer to the ability to QUANTIFY a constituent in water. Reporting limits are often developed by each laboratory.

Detection, quantification and reporting limits can be different for the same parameter using the same method.

Marine Pollution Studies Laboratory at the Moss Landing Marine Laboratories, 2017

Non-detects in water data (<DL, <MDL, <RL)

What does it mean that a water quality parameter concentration is below detection limit?

• It does not necessarily mean zero or not present, although it could mean that.
• It represents a range of unknown concentration.

Why is this important?

• Very high measurement limits (especially historical) can reduce the usefulness of

data.

• Differences in measurement limits can reduce the ability to compare different

datasets.

Quality Control Samples - Blanks

For water samples, blanks are used to account for potential contamination.

https://extension.usu.edu/utahwat erwatch/monitoring/field- instructions/ecoli/idexxmethod https://clu- in.org/conf/tio/rcraexpert_100516/slides/Data- Review-Manual-110114.pdf

Blank Type How is it collected? What does it identify? Field Blank Container filled with analyte-free water in the field Contamination in transport, storage, field handling Equipment/Rinsate Blank Container filled with analyte-free water that has been passed through collection equipment Contaminated field equipment Trip Blank Container filled with analyte-free water in the lab, taken to sample site, and returned to lab un-

• pened

Contamination during transport Lab Blank Analyte-free solution prepared in lab and analyzed with other samples Contamination in lab

Quality Control Samples – Replicates

For field measures and water samples, replicates are used to assess reproducibility and variability.

Replicate Type How is it collected? What does it identify? Field Replicate- Measurement/Observation Multiple field measurements or

• bservations taken at the

same time, location, and with same controlled variables. Used to assess the reproducibility and variability of the sampling technique. Field Replicate – Data Logger A field measurement or

• bservation verified by

alternate method. Used to assess the quality

• f data taken by data

logger. Field Replicate – Grab Sample Multiple samples collected at the same time, location, and with same controlled variables. Used to assess the reproducibility and variability of the sampling technique and lab analysis. Lab Duplicate A sample that is split into subsamples in the lab. Used to assess the quality

• f data obtained by lab

analysis.

Tips for Quality Control

Plotting Data

• Helps you compare:

○ Monitoring locations → spatial variation ○ Dates → temporal variation ○ Months and seasons → seasonal variation ○ Time of day → diurnal variation

• Also can detect outliers

It is important to plot water quality data before you use it! DataStream can help with this.

Scatter Plots

Scatter Plots

Box Plots

Q3 (75th Percentile) Median (50th Percentile) Q1 (25th Percentile) Outlier (Max) 95th Percentile 5th Percentile

Multiple Box Plots

Outliers

What does it mean when there are

• utliers in your data?
• Rare event (e.g. flood, contaminant

spill)

• Contamination during sample

collection or analysis

• Measurement error
• Recording or data handling error

“Does it represent a reasonably accurate observation of an unusual situation?” – Helsel et al. 2020, Statistical Methods in Water Resources

Outliers – Box Plots

Outliers – Scatter Plots

Unrealistic outliers indicate human error

20 40 60 80 100 120 140 160

2002-01-01 2004-09-27 2007-06-24 2010-03-20 2012-12-14 2015-09-10 2018-06-06

Temperature, water (°C )

Identify Data Below Detection Limits

Check Probe vs Lab Results

3 4 5 6 7 8 9 3 4 5 6 7 8 pH (lab) pH (field)

Sample Fraction

• Describes the portion of the

characteristic being analyzed

• Metals or Ions: Total or Dissolved
• Nutrients: Filtered or Unfiltered

Method Speciation

• Identifies the chemical speciation,

where applicable

• Mainly important for nutrients
• E.g. Ammonia as N or as NH4 will have

a different concentration value This is important to distinguish if you are cross-comparing datasets.

Sample Fraction: Total vs Dissolved

Dissolved Fraction Total Fraction Non- Filterable (Suspended Solids)

Sample Fraction: Check Dissolved vs Total Concentrations

5 10 15 20 25 30 35 40 45 2002-09-01 2004-01-14 2005-05-28 2006-10-10 2008-02-22 2009-07-06 CONCENTRATION (UG/L) Total Copper Dissolved Copper

Check Method Speciation prior to comparison

Quality Control Data – Duplicates/Replicates

○

How different are they?

○

What type of replicate is it?

○

Rule out data handling error. Is it a mislabeled sample?

Data Tables

RPD for Duplicates

RPD = Relative Percent Difference The difference (%) between duplicate samples 𝑆𝑄𝐸 = 𝐸1 − 𝐸2 (𝐸1 + 𝐸2)/2 x 100 Large RPD indicates low reproducibility and precision in your results.

Quality Control Data - Blanks

Data Tables

○

How many are above measurement limit?

○

What type of blank is it? Field, travel, lab

○

Rule out data handling error. Is it a mislabeled sample?

Quality Assurance Quality Control

Thank you

DataStream.org

• MackenzieDataStream.ca
• Atlantic DataStream.ca
• LakeWinnipegDataStream.ca

Subscribe to our newsletter: bit.ly/TGFnewsletter

@DataStreamH20 @DataStreamH20 Accessed through DataStream.org/webinars datastream@gordonfn.org