Download What is Normal Variation?

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

History of statistics wikipedia , lookup

Misuse of statistics wikipedia , lookup

Time series wikipedia , lookup

Transcript
What is Normal Variation?
Setting Baselines for Amphibian Monitoring
in the Credit River Watershed
Kirk Bowers
Introduction
• Baselines….
– Monitoring results must have an ecological context
– How do we know if an indicator is changing?
– What are the reference conditions?
• Thresholds….
– What amount of change in indicator values is
acceptable?
– After what amount of change should flags be raised?
• Widely used thresholds and established
baseline data sets are not always available
Amphibian Monitoring at CVC
• 26 Monitoring Sites
• Marsh monitoring
protocols
• Auditory system based
on call level
• 7 years of data (20032009)
• Includes frogs and
toads (no salamanders)
• 10 species recorded
CVC Monitoring Thresholds
• 3 Standard Deviations - Critical Threshold
– Values exceeding this threshold are associated with
significant increasing or decreasing trends and should
always be investigated thoroughly
• 2 Standard Deviations – Warning Threshold
– Implemented for the early detection of changing
trends or significant decline in a parameter; recognize
problems before they become irreversible
• A time series with values falling within the 2 SD warning
limits (in the “Green Zone”) is exhibiting natural
variability and is considered to be stable and not of
concern based on variance alone
Statistical Process Control
• Uses ecological time series to identify baselines and
develop monitoring thresholds
• Identifies instances when a time series exhibits nonrandom behaviour
• Series exhibiting natural variability around a reference
point (e.g. mean) is considered to be “In-Control”
• Series demonstrating non-random behaviour is
considered to be out of control
• Only a time series that is “In-Control” can be used
as a monitoring baseline
Statistical Process Control
• Control charts divided into 6 zones based on mean and
standard deviation of time series
• Parameter values on x-axis and time on y-axis
Statistical Process Control
• Distribution of data points is arranged around the time
series mean (green line)
Statistical Process Control
• Upper and lower critical limits, defined by +/- 3 SD,
encompass the “in-control” range of the series
(red lines)
Statistical Process Control
• Upper and lower warning limits are defined by +/- 2
standard deviations
Example 1: Mean species per site
Question: Can data collected from 2003 to 2009 be used
as a baseline for future monitoring of per-site
species richness?
4
Mean species per site
3.5
3
2.5
2
1.5
1
0.5
0
2003
2004
2005
2006
Year
2007
2008
2009
Example 1: Mean species per site
Example 1: Mean species per site
• Data In Control? - YES
• Can Data be used as a Monitoring Baseline – YES
• Can Data be used to set Monitoring Thresholds? - YES
Example 1: Mean species per site
Upper Critical Threshold = 3.45 species per site
Lower Critical Threshold = 1.87 species per site
Example 1: Mean species per site
Upper Warning Threshold = 3.19 species per site
Lower Warning Threshold = 2.14 species per site
Example 2: Green Frog Site Occupancy*
Question: Can data collected from 2003 to 2009 be used
as a baseline for future monitoring of Green
Frog site occupancy?*
* Not actual monitoring data
Example 2: Green Frog Site Occupancy*
Example 2: Green Frog Site Occupancy*
Example 2: Green Frog Site Occupancy*
• Data In Control? - NO
• Can Data be used as a Monitoring Baseline – NO
• Can Data be used to set Monitoring Thresholds? – NO
• Non-random pattern in data could be...
– An isolated anomaly (data will subsequently return to “in-control”
status)
or
– A sign of significant change in the Green Frog population
• This monitoring indicator must be tracked closely
Baseline Length
Question: How many sequential monitoring years should
compose a baseline?
• Answer may be dependant on the needs of the particular
monitoring program and the indicator being considered
• Generally, need a minimum of 5 sequential years of “InControl” data
• “In-Control” sequence should be as current as possible
Baseline Length
Question: How many sequential monitoring years should
compose a baseline?
Seven sequential years of data
Baseline Length
Question: How many sequential monitoring years should
compose a baseline?
Seven-year sequence is not “In-Control”
But...
Baseline Length
Question: How many sequential monitoring years should
compose a baseline?
...Last 5 years of data are “In-Control”
Baselines and Monitoring Design
Question: How can the identified baseline be applied to
monitoring design?
• Properties of the baseline data set (mean, standard
error) can be used in power analysis
• Power analysis can help answer such questions as:
– Does the program have sufficient statistical power to detect
threshold changes in the given indicator?
– How many sampling points (sites, visits, etc.) are required to
detect threshold changes in the given indicator at a desired
level of statistical power?
Review
• Baselines are a key component of monitoring because…
– They describe reference conditions
– They represent randomly variable data against which future
trends can be compared
– They allow for the development of monitoring thresholds
– They aid in program design and assessment
• We would like to extend special thanks to…
– Rob Milne and Lorne Bennet
– Paul Zorn
– Private and Municipal land owners
Any Questions?