Transcript Setting targets

Obtaining data and setting targets:
methods and limitations
Bob Smith
Durrell Institute of Conservation & Ecology
• Problems with data quality (focussing on
presence/absence data)
• Suggestions on data requirements at the fine-scale
• How to develop cost and threat datasets
• How to set targets for species and landcover types
A planning system will only be useful if its results
are implemented and there are several ways to
increase the likelihood of this, which include:
• Conduct the analysis at a relevant spatial scale
• Include data on relevant conservation features
• Use up-to-date information
• Set justifiable representation targets
• Include relevant socio-economic & political data
One common source of distributional data comes
from atlas projects, which generally show the
distribution of a range of species as
presence/absence maps.
These type of data are commonly analysed in the
scientific literature.
I will discuss the problems with using such
datasets, both to identify specific limitations and
illustrate broader issues.
This paper uses a
dataset that shows
the distribution of
3882 vertebrate
species in 1957 1º
grid squares in subSaharan Africa.
Each of these grid squares is approximately 105 km x 105 km
and the presence or absence of each species in each grid
square is recorded. They set a target of at least one
representation for each species.
The authors then used a
complementarity-based
algorithm to identify the 50
sites that, when combined,
would represent the largest
number of species.
They also used the WWF
ecoregion map to label each
site according to the
ecoregion that it falls within.
By failing to involve stakeholders they increase the chances
of mis-naming areas or choosing unsuitable areas.
This is not
montane
grassland so
reduces
credibility of
output
Problems of scale: Implementation
Problems of scale: Measuring representation
Data quality: distribution errors
Data quality: sampling bias
Reddy & Davalos (2003). J. of Biogeography 30, 1719-1727
Arbitrary targets
There is no way of knowing whether protecting one
representation of each species will be sufficient for their
conservation or whether each population in each grid
square is viable.
Basing an analysis on complementarity may maximise the
efficiency of the final PA system but it might not necessarily
protect viable populations when using presence/absence
data.
Beetle
Beetle
Beetle
Butterfly
Butterfly
Butterfly
Lizard
Lizard
Lizard
Toad
Toad
Toad
Tortoise
Tortoise
Tortoise
Beetle
Beetle
Beetle
Butterfly
Butterfly
Butterfly
Lizard
Lizard
Lizard
Toad
Toad
Toad
Tortoise
Tortoise
Tortoise
This phenomenon may have other serious implications
Beetle
Beetle
Beetle
Butterfly
Butterfly
Butterfly
Lizard
Lizard
Lizard
Toad
Toad
Toad
Tortoise
Tortoise
Tortoise
SWAZILAND
Lesotho
South
Africa
1501 - 1800
1301 - 1500
1001 - 1300
801 - 1000
501 - 800
301 - 500
0 - 300
8 PAs
3.8% of country
Protected area
Natural = 59.5%
Degraded = 12.8%
Transformed = 27.7%
Natural
Degraded
Transformed
No of
records
Birds
Mammals
No of
species
Units
sampled
Range
of
distrib.
18,255
476
101
1-101
905
122
50
1-43
No of
Species
spp with richness
1 unit
range
range
23 108-307
27
2-65
45
10
8
Mammal species richness
Bird species richness
40
35
30
25
6
4
2
20
15
0
0.2
0.4
0.6
0.8
1.0
Proportion of natural vegetation
Birds
N = 101, rs = 0.09, P = 0.353
0.2
0.4
0.6
0.8
1.0
Proportion of natural vegetation
Mammals
N = 50, rs = -0.05, P = 0.971
Irreplaceable
Irreplaceable
Flexible
Flexible
Birds
17 irreplaceable + 4 flexible
Mammals
13 irreplaceable + 1 flexible
Proportion of natural vegetation (+1 SD)
Proportion of natural vegetation (+1 SD)
1.00
Outside PAs
Contains PA
0.75
0.50
0.25
0.00
1.00
Outside PAs
Contains PA
0.75
0.50
0.25
0.00
Not selected
Selected
Birds
df = 3, Χ2 = 0.677, P = 0.879
Not selected
Mammals
Selected
df = 3, Χ2 = 1.226, P = 0.747
The selected planning units were still not
significantly less transformed than other units.
One reason for this might be the number of species
that were only recorded in one planning unit.
No of
records
Birds
Mammals
No of
species
Units
sampled
Range
of
distrib.
18,255
476
101
1-101
905
122
50
1-43
No of
Species
spp with richness
1 unit
range
range
23 108-307
27
2-65
67 % transformed
28 mammal species recorded
Only record of Kuhl’s pipistrelle
Conclusions
•
Swaziland species distribution data were not sensitive to
levels of agricultural and urban transformation.
•
This was partly driven by the number of species that
were only recorded in one planning unit, which may
have been exacerbated by under-sampling.
•
Species list data should ideally only be used for coarsescale planning exercises, whereas finer scale exercises
should include data that relates to ecological viability.
Other effects of sampling bias
The recorded data came from the Southern African Bird
Atlas Project (SABAP), which describes the distribution of
the region’s bird species in a series of ¼ degree grid
squares. The data were collected by a series of expert
volunteers. Each square was visited a number of times and
a list of recorded species was compiled on each occasion.
Each record of each
species in each grid
square was then
compiled and stored in
a central database.
Point 8 – The data is affected by sampling bias
Record number
300
200
100
0
1.5
2.5
3.5
Log10 record number
4.5
Species number
Number of species
400
The land-cover map had an
overall accuracy of 86.9 % and a
resolution of 30 m.
It contained 29 natural and 5
transformed land-cover types.
Recorded
distribution
Cloud cisticola
Number of recorded
squares = 6
Associated with woody
& hygrophilous
grassland
Recorded
distribution
Cloud cisticola
Number of recorded
squares = 6
Associated with woody
& hygrophilous
grassland
Recorded
distribution
Modelled
distribution
Cloud cisticola
Number of recorded
squares = 6
Number of modelled
squares = 13
Recording success = 0.462
Distinctive species were classified on the basis of their
appearance and/or song.
Distinctive appearance:
Plumage, bills or legs that contained red, yellow,
pink or purple coloration.
Bills or tails that were more than 50 % of their
body length.
Body length of > 80 cm
Distinctive song:
Described as “loud”, “characteristic”, “penetrating”,
“far-carrying”, “raucous”, “strident”, “booming” or
“piercing”.
The relationship between recording success and
distinctiveness was tested and there was a significant
difference between distinct and nondescript species.
n = 429
t = -2.825
p = 0.005
Using the
distinctive and
complete data
sets, only 7 of
the 17 grid
squares selected
were the same.
100 km
All & distinctive
All
Distinctive
This shows that it
might not always
be best to include
all available data.
• This suggests that species distributions should be
modelled to produce fine-scale data, rather than
using raw presence/absence data.
• Vegetation/geology/soil etc maps can provide much
more reliable data.
• Recent landcover data is also important.
• Point location data reduces flexibility in the system,
helping to anchor larger PAs.
MOST PLANNING EXERCISES USE WHATEVER
ADEQUATE DISTRIBUTION DATA ARE AVAILABLE.
Conservation value
High
Low
Vegetation types
Forest types
Threatened tree species
Threatened vertebrate species
MARXAN acts to minimise the planning unit costs
and these can be based on:
• Area
• Financial value
• Human population density
• Risk
• Opportunity costs
• Etc
Risk of agricultural transformation
Probability of being cleared related
to elevation, slope, geology and
distance to agriculture
1
Transformation
probability
0
20 km
Protected Area
Agricultural land
or water
The spatial distribution
of bark stripping
Game ranching profitability in Maputaland
1.0
0.9
0.8
DETECTION PROBABILITY
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0
20
40
60
80
PERPENDICULAR DISTANCE (m)
100
120
140
Targets
• 40% original extent of threatened and endemic
landcover types
• 20% original extent of other natural landcover
types
• 25% of natural landcover in each communal area
Cost of each planning unit =
US$10,000 – Potential profitability from game ranching
Revenue
'000s US$
Non-tribal areas
1967
Mathenjwa
409
Tembe
1852
Nyawo
670
Mngomezulu
285
Mashabane
256
Mabaso
699
Siqakatha
27
Zikhali-Mbila
87
Myeni-Ntsinde
67
Manukuza-Jobe
513
Myeni-Ngwenya
301
Mnqobokazi
238
Qwabe-Makhasa
32
Nibela
59
Mpukunyoni-Mkhwanazi
13
Total revenue
7476
Setting targets
Setting targets is a vital part of systematic
conservation planning and the target values have a
profound affect on the final conservation portfolio.
Setting targets
Setting targets is a vital part of systematic
conservation planning and the target values have a
profound affect on the final conservation portfolio.
Original targets were often political, eg
10% of the planet
12% of original cover
These have been criticised for a lack of biological
relevance, with targets of 50% having been
suggested.
Also problematic because they assume all elements
are equal.
Instead, it might be preferable to set individual
targets based on criteria such as:
• Endemism
• Red list status or other measures of threat
• Life-history characteristics
• Original extent
• Non-biodiversity value (eg watersheds)
• Genetic diversity
Setting targets
Setting targets is a vital part of systematic
conservation planning and the target values have a
profound affect on the final conservation portfolio.
• Expert workshops
• Estimates of minimum viable populations
• Species/area curves for habitat targets
1.00
Proportion of species
Work developed
by Desmet &
Cowling (2004)
has used
species/area
curves to
estimate the
amount of a
landcover type
that is needed to
represent 90% of
plant species.
0.75
0.50
0.25
Type A
Type B
0.00
0
50
100
Area sampled
150
MARXAN and clumps
Another feature of MARXAN is that it can include
information on patch and population size of the
conservation features. MARXAN uses the term “clump” to
describe these characteristics and defines clumps using the
following parameters:
1) Clump distance
2) Clump size
3) Clump target
Clump distance
This is the maximum distance between planning units,
below which it is assumed that units belong to the same
clump. These distances are measured from the centre of
each planning unit.
Clump size
This is the minimum size for a viable clump. The size of
each clump is measured by combining the amount of the
conservation feature found in each of the associated
planning units.



A possible priority setting exercise for
CWRs in Europe
• Use existing atlas, landcover and PA data
• Use clump option in MARXAN to identify
different populations
• Set targets based on minimum number of
populations
A possible priority setting exercise for
CWRs in Europe
Analyses:
• Identify populations in grid cells within no
associated PAs
• Identify irreplaceable grid cells based on
minimising the amount of unsuitable land found
within the grid cells
• Carry out field work to check the presence of
viable populations of CWRs in PAs in
irreplaceable cells
Systematic conservation planning should be
considered as a form of adaptive management.
Outputs will change as more information is added and
refined.
However, there is already sufficient data to produce a
useful initial product and these outputs would be the
most effective at identifying where CWRs should be
conserved.