Interpreting biodiversity under
diverse syndromes of recording
behaviour
Nick Isaac
Biological Records Centre
Centre for Ecology & Hydrology
Extracting trends from biological
recording data
Nick Isaac
Biological Records Centre
Centre for Ecology & Hydrology
Is biological recording fit for purpose?
• What is the purpose?
• What data are available?
• What are the problem issues?
• What tools might provide a solution?
Is biological recording fit for purpose?
• What is the purpose?
•
•
•
Describing species’ distributions
Detecting and attributing change over time
Identifying novelties
Wikipedia Commons
FERA
Mike Majerus
GBNSS
GBNSS
Biological records data
How do we interpret the gaps?
NBN lists 35 data sources:
• Individual records
• Regional recording projects
• Co-ordinated national surveys
Published Atlases
The primary tool for
understanding UK biodiversity
Authoritative summary of the
current state of knowledge
A snapshot of species’
distributions
Perring, F H, & Walters, S M, eds 1962 Atlas of the
British Flora. Thomas Nelson & Sons, London
Published Atlases
Stock & change in distribution
•
Repeat atlases allow an
assessment of change over
time
•
Prickly Lettuce (Lactuca
serriola) has expanded
northwest since 1970
Repeat atlases: plants & birds, butterflies
Biodiversity change using atlases
‘Square counts‘ on repeat
atlases reveal which
species are increasing vs
decreasing
Greatest losses occurred
among butterflies, then
birds
Thomas, JA et al. (2004). Comparative losses of
British butterflies, birds, and plants and the global
extinction crisis. Science, 303(5665), 1879–81
Where are we now?
• Atlases provide a rather static view of biodiversity
• The unstructured nature of the data makes square
counting unreliable
• Increasing demand for quantitative information
• New methods for estimating trends are being
developed
Detecting and attributing change
Trends in the distribution of 8
common ladybirds
A majority show substantial
negative response to arrival of
Harlequin ladybird
Similar patterns in GB &
Belgium
Mike Majerus
Roy, HE, Adriaens, T, Isaac, NJB et al. (2012). Invasive
alien predator causes rapid declines of native European
ladybirds. Diversity and Distributions, 18(7), 717–725
Past, present and future
Describing change
Attributing change
Biodiversity Indicators
Talk outline
• Extracting trends from Biological records data
•
Problems & possible solutions
• Comparison of candidate methods
•
•
Simulations of recording behaviour
Which methods are useful for detecting trends?
• Applications: which species are declining?
•
•
Trends in Odonata 1970-2011
Biodiversity Indicator
Recording intensity varies among taxa
Extracting trends from biological records
120
100
Number of records
80
60
40
20
0
1970
1980
1990
Year
2000
2010
Recording intensity has increased over time
1000000
100000
Butterflies
Bryophyte
Number of records
10000
Orthoptera
Myriapod
Isopods
Coleoptera
1000
Moths
Bees
Wasps
100
Ants
Hoverflies
Odonata
10
1
1970
1975
1980
1985
1990
1995
2000
2005
2010
Telfer’s Change Index
• Compares two time-periods that differ in
recording intensity &/or geographic coverage
Telfer, MG, Preston, CD & Rothery, P (2002). A general method for
measuring relative change in range size from biological atlas data.
Biological Conservation, 107(1), 99–109
Ball’s Visit Rate model
0.06
Proportion of Visits
0.05
0.04
0.03
0.02
0.01
0
1970
1980
1990
Year
2000
2010
Ball, S, Morris, R, Rotheray, G, & Watt, K (2011). Atlas of
the Hoverflies of Great Britain (Diptera, Syrphidae).
Most lists are incomplete
For most groups, ~50% of visits produce ‘incidental records’
100%
90%
80%
Proportion of all visits
70%
60%
50%
> 3 species
3 species
40%
30%
20%
10%
0%
2 species
Single species
Lists lengths are not constant over time
1
0.9
Proportion of incidental records
0.8
0.7
Bryophyte
0.6
Isopods
Coleoptera
0.5
Moths
0.4
Ants
Hoverflies
0.3
0.2
0.1
0
1970
1975
1980
1985
1990
1995
2000
2005
2010
Mixed model
0.09
0.08
0.07
Proportion
0.06
0.05
0.04
0.03
0.02
0.01
0
1970
1980
1990
Year
2000
2010
Most records come from a few recorders
Bryophytes: 18
Myriapods: 11
Moths: 102
Orthoptera: 39
Spatial pattern of recording behaviour
Orthoptera 1970-2011: top 4 recorders made 14% of all visits
Hill’s Frescalo method
Frescalo estimates the
recording intensity of
each grid cell
Red = under-recorded
White = well-recorded
Hill, MO (2011). Local frequency as a key to interpreting
species occurrence data when recording effort is not known.
Methods in Ecology and Evolution, 3(1), 195–205.
Hill’s Frescalo method
By estimating
recording intensity,
Frescalo calculates
the number of
species that ‘should’
be in each grid cell.
Hill’s Frescalo method
Hill, MO (2011). Local frequency as a key to interpreting
species occurrence data when recording effort is not known.
Methods in Ecology and Evolution, 3(1), 195–205.
Occupancy modelling: a panacea?
Gateshead birders
van Strien, A, van Swaay, C, & Kéry, M (2011). Metapopulation dynamics in the butterfly
Hipparchia semele changed decades before occupancy declined in the Netherlands. Ecological
Applications, 21(7), 2510–2520
Talk outline
• Extracting trends from Biological records data
•
Problems & possible solutions
• Comparison of candidate methods
•
•
Simulations of recording behaviour
Which methods are useful for detecting trends?
• Applications: which species are declining?
•
•
Trends in Odonata 1970-2011
Biodiversity Indicator
How can we estimate trends?
Raw data
Estimate
trends
Recorder
behaviour
Simulations
Simulations
Aims:
1. To compare the performance of different
methods for estimating range change under
realistic scenarios of recorder behaviour
2. To discard methods that are inappropriate
3. To derive rules of thumb for when other
methods are appropriate
Simulation overview
• 1000 sites (no spatial information)
• 1 focal species + 25 others
• Focal species occupies 50% sites
• Impose different patterns of recording
• Run for 10 years
• Estimate trends using different methods
Simulation patterns of recording
• A: Control scenario: even recording
•
Equal probability of sites being visited
• B: Increasing recording intensity
•
Growth in number of visits
• C1: Incomplete recording (even)
•
A fixed proportion of Visits produce short lists
• C2: incomplete recording (increasing)
•
Proportion of short lists increases over time
Type I Error Rates
Type I Error Rates
Change Index
A
Even
Recording
0.027
nRecords
0.024
Visit Rate
0.046
MM2sp
0.061
MM3sp
0.058
MM4sp
0.058
Frescalo
0.040
Type I Error Rates
Change Index
A
Even
Recording
0.027
B
Increasing
Intensity
0.026
C1
Incomplete
even
0.033
C2
Incomplete
increasing
0.037
Type I Error Rates
nRecords
A
Even
Recording
B
Increasing
Intensity
C1
Incomplete
even
C2
Incomplete
increasing
0.024
0.993
0.042
0.609
Type I Error Rates
Visit Rate
A
Even
Recording
B
Increasing
Intensity
C1
Incomplete
even
C2
Incomplete
increasing
0.046
0.060
0.059
0.675
Type I Error Rates
A
Even
Recording
B
Increasing
Intensity
C1
Incomplete
even
C2
Incomplete
increasing
MM2sp
0.061
0.079
0.053
0.195
MM3sp
0.058
0.079
0.060
0.089
MM4sp
0.058
0.073
0.066
0.049
Type I Error Rates
Frescalo
A
Even
Recording
B
Increasing
Intensity
C1
Incomplete
even
C2
Incomplete
increasing
0.040
0.164
0.036
0.060
Type I Error Rates
A
Even
Recording
0.027
B
Increasing
Intensity
0.026
C1
Incomplete
even
0.033
C2
Incomplete
increasing
0.037
nRecords
0.024
0.993
0.042
0.609
Visit Rate
0.046
0.060
0.059
0.675
MM2sp
0.061
0.079
0.053
0.195
MM3sp
0.058
0.079
0.060
0.089
MM4sp
0.058
0.073
0.066
0.049
Frescalo
0.040
0.164
0.036
0.060
Change Index
Power to detect a genuine decline
Change Index
A
Even
Recording
0.574
nRecords
0.642
Visit Rate
0.739
MM2sp
0.665
MM3sp
0.649
MM4sp
0.615
Frescalo
0.612
Power to detect a genuine decline
A
Even
Recording
0.574
B
Increasing
Intensity
0.461
C1
Incomplete
even
0.37
C2
Incomplete
increasing
0.316
nRecords
0.642
0
0.449
0.979
Visit Rate
0.739
0.606
0.507
0.985
MM2sp
0.665
0.424
0.319
0.685
MM3sp
0.649
0.408
0.271
0.463
MM4sp
0.615
0.363
0.211
0.208
Frescalo
0.612
0.768
0.34
0.308
Change Index
Simulations: Conclusions
• The simulation provides a framework for
comparing methods under a range of recording
scenarios
• The Mixed model method performs best so far
(Frescalo & Occupancy results pending)
• In the best recording scenario, a decline of 30%
was detected in 60% of simulated datasets
Talk outline
• Extracting trends from Biological records data
•
Problems & possible solutions
• Comparison of candidate methods
•
•
Simulations of recording behaviour
Which methods are useful for detecting trends?
• Applications: which species are declining?
•
•
Trends in Odonata 1970-2011
Biodiversity Indicator
Odonata trends 1970-2011
• Broad agreement
between methods
• 14/32 species show
significant increases
under both methods
• 2/32 show significant
decreases under
both methods
Odonata trends: winners
Small red-eyed Damselfly
(Erythromma viridulum)
Wikipedia Commons
Scarce chaser
(Libellula fulva)
Emperor Dragonfly
(Anax imperator)
Odonata trends: losers
Variable damselfly
(Coenagrion pulchellum)
Blue-tailed Damselfly
(Ischnura elegans)
Common Blue Damselfly
(Enallagma cyathigerum)
Odonata Indicator
Biological Recording for the 21st Century
• We have the tools to model biodiversity change
using unstructured biological records
• This is only possible if records continue to be
submitted to the database!
• We could be smarter about data collection
• We’re only just beginning to exploit the potential
of biological recording data
•
Indicators, Red Listing, ecosystem service provision,
targeting Agri-environment schemes
Acknowledgments
Tom August
Colin Harrower
David Roy, Helen Roy, Michael Pocock, Gary Powney, Chris Preston
Mark Hill
Arco van Strien
Descargar

Nick Isaac`s presentation