Portfolio Selection

1
Portfolio Selection MARXAN Created by Ian Ball
and Hugh Possingham
2
Biodiversity Representation

• Protected areas are now required to be
  representative of biodiversity.
• Selection of protected areas in many places has
  historically been opportunistic.
• Many reserves were originally designated for
  their scenic beauty, cultural significance, lack
  of economic value or to protect a few charismatic
  flagship species.
• These PAs do not adequately represent the
  diversity of ecosystems, leading to duplication
  in the protection of some habitats and species
  and inadequate protection of others.
• Selection techniques improved with the
  understanding that the range of biodiversity
  should be represented.
• These techniques often concentrate on areas rich
  in well-studied habitats and species, and do not
  provide quantitative representation or
  repeatability.

3
Priority Area Selection Methods

• Systematic techniques using algorithms have been
  designed to select priority conservation areas
  both for protected areas and use-zoning.
• These decision support tools are
• - transparent and efficient
• - driven by quantitative reservation goals
• - flexible and
• - repeatable.

4

• It is vital to include expert and stakeholder
  knowledge in the process, whilst allowing
  quantitative representation and repeatability.
• Software such as MARXAN has been designed to
  implement algorithms that allow such
  methodologies.
• They can include many parameters believed to be
  important in biologically meaningful priority
  area design.
• These include multiple representations, patch
  size control and minimum and maximum separation
  distances.
• The techniques can also offer many alternative
  systems that can be negotiated whilst maintaining
  all goals.

5
Reserve Design using Spatially Explicit Annealing

• MARXAN delivers decision support for selecting
  networks of priority conservation sites.
• A region is divided into smaller areas known as
  planning units, to allow comparison between the
  areas through quantification of their
  characteristics.
• The selection of any planning unit over another
  involves evaluating it with regards to all the
  planning units in the area under consideration.
• One unit with several valuable features on its
  own may or may not be the best choice overall,
  depending on the distribution and replication of
  those features in other planning units.

6
Marxan Utilization Worldwide

• Marxan was developed to meet the decision support
  needs of the Great Barrier Reef Marine Planning
  Authority (GBRMPA) in their representative areas
  program that has rezoned the GBR. Other examples
  include
• British Columbia (Canada)
• Galapagos Islands (Ecuador)
• Gulf of California (Mexico)
• Joint Nature Conservancy Council (UK)
• The Florida Keys National Marine Sanctuary (USA)
• Channel Islands National Marine Sanctuary (USA)
• South Australia, University of Queensland
• Northern Gulf of Mexico (USA)
• Trough-Georgia Basin (USA/ Canada)
• North East Atlantic (USA / Canada)

7
Designing a Portfolio

• Marxan can offer decision support for teams of
  experts choosing between thousands of planning
  units and many biodiversity targets.
• It selects a portfolio of spatially cohesive
  units that meet a suite of biodiversity goals
  whilst minimizing the cost.
• The cost of the portfolio consists of a weighted
  sum of planning unit cost, boundary length and
  penalties for not representing biodiversity
  targets to their user defined goal.
• A portfolio consists of a network of planning
  units, some of which are clustered into potential
  sites, with others serving to connect isolated
  areas of existing or intended conservation
  management.

8
(No Transcript)
9
The MARXAN algorithmObjective Function
The algorithm attempts to minimize the total cost
of a portfolio
Or Total Portfolio Cost (cost of selected
sites) (penalty cost for not meeting
conservation goals)
(cost of spatial distribution of the selected
sites).
10
Simulated Annealing

• MARXAN uses an simulated annealing optimization
  algorithm to select a portfolio.
• The algorithm is based on iterative improvement
  with stochastic acceptance of bad moves.
• This allows the algorithm to choose less than
  optimal planning units early in the process that
  may allow for better choices and overall
  portfolio later.
• As the program progresses, the criteria for a
  good selection gets progressively stricter, until
  finally the portfolio is built.

11
Planning Units

• Planning units can be any shape or size, but
  appropriate units should be designed according to
  the available target data and to best facilitate
  conservation efforts in the priority sites
  identified.
• Planning units can be natural, administrative or
  arbitrary sub divisions of the land and seascape.
• The units should be small enough to reflect
  differences between fragmented and non fragmented
  habitats or distributions, but large enough to
  reflect quantitative differences between units.
• Data on distributions within very small units
  becomes presence / absence information and does
  not reflect differences regarding the size of
  patches or the co-existence of biodiversity
  elements or targets between the units.

12
(No Transcript)
13
Factors That Can Be Included In Portfolio Analysis

• Measures can be incorporated by careful setup of
  targets and goals. For example
• Representation quantitative representation of
  all targets.
• Multiple sites a minimum number of sites can be
  stipulated and/or a minimum separation distance
  to lower stochastic occurrence risk.
• Connectivity a maximum separation distance can
  be stipulated and sites thought to be connected
  can be split into separate sub-targets ensuring
  representation within all connected sites.
• Resistance or resilience indicators (if map-able
  eg shaded, well mixed, well flushed areas etc)
  if a high level of confidence is achieved, these
  can be incorporated as separate (fine filter)
  targets.
• Resistant or resilient sites can be
  incorporated as separate targets.

14
Outputs Portfolios and Irreplaceability Maps
15
Outputs

• Information on the number of times a planning
  unit is chosen in a priority area network, and
  the best network can be mapped using GIS.
• Planning units that are chosen more than 50 of
  the time can be thought of as being essential for
  efficiently meeting biodiversity goals. Areas
  with lower irreplaceability are not unimportant
  but are more interchangeable with other similar
  planning units.
• Many design scenarios can be explored, and
  flexible units can be removed and alternatives
  found.

16
(No Transcript)
17
(No Transcript)
18
(No Transcript)
19
(No Transcript)
20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
(No Transcript)
24
(No Transcript)
25
Portfolio Selection Marxan Inputs
26
Marxan Inputs

• Target abundance per planning unit
• Goals
• Cost per planning unit
• Planning unit boundary lengths (optional)
• Biological constraints (optional)
• Spatial clustering (optional)
• Species penalty factors (optional but extremely
  important)

27
1) Target Abundance per Planning Unit
28
2) Goals
20
30
29
3) Planning Unit Cost

• The cost is a relative value applied to planning
  units such that some may be more difficult or
  expensive to set aside than others.
• Marxan attempts to minimize the total cost of
  the portfolio. This consists of cost, boundary
  length and penalties for not representing the
  targets to the goals.
• Cost can represent
• Actual or modeled cost of planning unit area
• Cost of lost opportunity (e.g. fishing yield etc)
• Threat
• Inverted resilience indicators
• Any other measure to minimize in the portfolio as
  a whole.

30
(No Transcript)
31
4) Biological Constraints

• Measures can be introduced to assure the
  portfolio contains targets that have
• a minimum target patch size
• minimum separation distance between patches
  (avoidance of stochastic disasters)
• maximum separation distance (connectivity)
• minimum number of patches.

32
5) Species Penalty Factors

• The species penalty factor (SPF) sets the
  relative importance of target representation
  when selecting areas.
• A spf value should be chosen that allows an
  acceptable number of targets to reach goal
  representation.
• Testing is required to calculate this value.

33
6) Spatial Clustering of Planning Units

• Marxan facilitates the choice of a portfolio with
  increased spatial clustering of planning units
  (PUs).
• It can be set to minimize the boundary length of
  the portfolio, which clusters the planning units
  together.
• This effect can be set to have a strong or a only
  slight effect.
• Clustering the sites can require an increase in
  the number of PUs necessary to meet all
  representation goals, but is thought to increase
  manageability of sites and likelihood of
  persistence of biodiversity targets.

34
Increasing PU Clustering
0.001
0
0.01
0.0001
0.1
0.0005
35
Pre MARXAN Target Data Preparation
36
Pre MARXAN Target Data Preparation

• Data compatibility Data should ideally be of
  the same scale or resolution, of a similar age
  and of similar accuracy.
• Screening eg patch size, health, threat etc.
• Stratification biologically diverse parts of
  targets should be stratified.
• Target weighting fine and coarse filter targets
  should be weighted carefully and have goals
  appropriate to the aims of the portfolio.

37
(No Transcript)
38
Caribbean Ecoregional Assessment
Dominica
39
(No Transcript)
40
Exercise 1 An introduction to MARXAN
41
Exercise 1 An introduction to MARXAN

• This first exercise describes how to use marxan
  to design a portfolio using case study data.
• The first section examines the target
  distribution and planning unit shapefiles in
  ArcView, and the marxan input files in notepad.
• The second section describes setting up a marxan
  run using input files that have been prepared
  from this data, running the algorithm and mapping
  the results.

42
Follow the instructions A1-6 to copy MARXAN and
the tutorial data onto your computer and then
view the data in ArcView.
43
Using Tutorial Input Files

• A7 Up to 5 text files are needed for MARXAN to
  run. (c/marxan/inputs)
• They contain data concerning
• Target abundance
• Target details (name, goal, spf etc)
• Planning unit information
• Boundary information (optional)
• Block definitions (optional)

44
Input Tables
Target Abundance puvspr.dat
Target Details File spec.dat
45
Input Tables
Boundary file bound.dat
Planning unit file pu.dat
46
Input Tables
Block definition file (optional but useful for
setting proportional goals)
47
Section B Using Inedit to set up MARXAN
48
B1) MARXAN is set up using the Inedit program
opened from windows explorer
49
B2)
50
B3)
51
B4)
52
B5)
53
B6)
54
B7)
55
B8)
56
B9) MARXAN is opened and run by executing the
marxan.exe file.
57
(No Transcript)
58
B10) Examine the outputs
Best - the units in the best portfolio
Sum - summary of each run, including whether all
goals have been met.
59
Sum-summary of each run
Solution- number of times each unit was selected
60
Section C Mapping MARXAN Results
61
Mapping Irreplaceability
62
C2) Add the text files tnc_tutorial_run1_ssoln.txt
and tnc_tutorial_run1_best.txt
63
C2) Join tables to the planning unit attribute
table.
64
C3) Display the pu shapefile using graduated
color.
Double click
65
C3 cont.) Use the run1_irr field as the
classification field.
66
Irreplaceability based on 100 runs (cost area,
no areas locked in)
67
Mapping the best Portfolio
68
C4) Display the pu shapefile using unique value.
69
C4 cont.) Use the run1_bst field as the
classification field.
70
Best Portfolio based on 100 runs (cost area,
no areas locked in)
71
C5) Run the algorithm again using a boundary
length modifier (BLM) and view the results of
increased clustering. C6) Run the algorithm with
the protected area planning units locked into the
portfolio. Compare the results to identify
whether the present PA system is efficient or
meets all conservation goals. C7) Run the
algorithm for 200 runs and compare the best
portfolio cost with the best run of 100 runs.
Have 200 runs identified a more efficient
portfolio?
72
Exercise 2 Creating Input Files using Tutorial
Data
73
Creating Input Files using Tutorial
Data Exercise 2 describes the GIS processes and
excel methods that can be used to create the
MARXAN input files from target and planning unit
files.
74

• MARXAN Files
• Target Abundance File puvspr.dat
• Species File spec_goals.dat
• Planning Unit File pu.dat
• Boundary File bound.dat
• Block Definition File block.dat (optional)

75
D) Target Abundance File puvspr.dat (Planning
Unit versus Conservation Feature File)
76
D2) A dbf table is created that contains the ids
of the PUs from the planning unit file.
(puvspr.dat)
77
D2-D4) This table is used by the CLUZ abundance
ArcView script to produce an abundance table
using the target shapefiles and the planning unit
shapefile.
(puvspr.dat)
78
D5) View the abundance dbf table.
(puvspr.dat)
79
D6) The CLUZ puvspr ArcView script is used to
convert the abundance table into the MARXAN
puvspr file format.
(puvspr.dat)
80
D7) Resulting puvspr_abun file.
81
E) Species File spec_goals.dat
82
Follow steps E1 to E3 to create the table
containing a target id column.
83
E4) Goals can be calculated using the abundance
table
(spec.dat)
84
Follow steps E5 to E7 to complete the spec_goals
file.
(spec.dat)
85
F) Planning Unit File pu.dat
86
F1) The planning unit shapefile is exported to a
dbf table.
(pu.dat)
87
F1 F4) The table can be manipulated in excel and
saved as a csv file, then renamed to .dat.
(pu.dat)
88
F5) Follow step F5 to create a PU table that
identifies all PUs with over 50 of their area
under a PA.
89
G) Boundary File bound.dat
90
G1-G2) The boundary file extension to ArcView is
used to create the boundary file from the
planning unit shapefile automatically.
91
F) Block Definition File block.dat (optional)
92
Follow steps H1 and H2 to create the block
definition file
93
Exercise 3 Run marxan with new input files.
94

• Follow steps J1 to J4 to run a series of marxan
  analyses with varying parameters.
• Check that marxan will run successfully using the
  new files.
• View the effects of different parameters such as
  locking protected areas into the portfolio and
  increasing clustering.

95
Results
96
Effect of Increasing Boundary Length Modifier
97
(No Transcript)
98
(No Transcript)
99
(No Transcript)
100
(No Transcript)
101
(No Transcript)
102
BLM Number of Units in Best Number of
Targets Portfolio over Goal 0.001 593
16 0.01 579 16 0.03 567 18 0.06 550
18 0.1 547 17 0.5 674 22
103
Proportion of Goal Held in Portfolio 100 Runs
BLM 0.001 16 Targets Over Goal
Portfolio 593 Units
Mangrove
Wet alluvial
Dry Alluvial
Dry Intrusive
Rain alluvial
Dry Extrusive
Wet intrusive
Rain intrusive
Moist Intrusive
Rain extrusive
Wet limestone
LM wet alluvial
Moist limestone
LM wet intrusive
LM rain intrusive
Dry sedimentary
Reef Linear Reef
Reef Linear Reef
LM rain extrusive
Wet sedimentary
Dry ultramafic
LM wet limestone
Wet ultramafic
Moist sedimentary
Wetland Terrestrial
Moist ultramafic
LM wet ultramafic
Reef Scattered Coral-Rock
Reef Colonized Pavement with C
Target
104
Proportion of Goal Held In Portfolio 100 runs
BLM 0.01 16 Targets Over Goal
Portfolio 579 Units
Mangrove
Dry Alluvial
Rain alluvial
Dry Intrusive
Wet intrusive
Rain intrusive
Wet limestone
Moist Intrusive
Rain extrusive
LM wet alluvial
Moist limestone
Dry ultramafic
Wet ultramafic
LM wet intrusive
Dry sedimentary
LM rain intrusive
Reef Linear Reef
LM wet extrusive
Wet sedimentary
Reef Linear Reef
LM rain extrusive
LM wet limestone
Moist sedimentary
LM wet ultramafic
Wetland Terrestrial
Reef Colonized Bedrock
Reef Colonized Pavement
Reef Scattered Coral-Rock
Reef Colonized Pavement with C
Target
105
Proportion Of Goal Held in Portfolio 100 runs,
BLM 0.03 18 Targets over Goal
Portfolio 567 Units
Mangrove
Dry Alluvial
Wet alluvial
Rain alluvial
Dry Intrusive
Wet intrusive
Rain intrusive
Moist Intrusive
Rain extrusive
Wet limestone
LM wet alluvial
Dry ultramafic
Wet ultramafic
Moist limestone
LM wet intrusive
Dry sedimentary
LM wet extrusive
Wet sedimentary
Reef Linear Reef
Moist ultramafic
Reef Linear Reef
LM rain extrusive
LM rain intrusive
LM wet limestone
Moist sedimentary
LM wet ultramafic
Wetland Terrestrial
Reef Scattered Coral-Rock
Reef Colonized Pavement with C
Target
106
Proportion of Goal Held in Portfolio 100 Runs,
BLM 0.06 18 Targets Over Goal
Portfolio 550 Units
Mangrove
Dry Intrusive
Rain alluvial
Wet intrusive
Dry Extrusive
Rain intrusive
Dry Limestone
Wet limestone
Rain extrusive
LM wet alluvial
Moist limestone
LM wet intrusive
LM rain intrusive
Dry sedimentary
LM wet extrusive
Reef Linear Reef
Wet sedimentary
Reef Linear Reef
LM rain extrusive
LM wet limestone
Moist sedimentary
Dry ultramafic
Wet ultramafic
Wetland Terrestrial
Moist ultramafic
LM wet ultramafic
Reef Scattered Coral-Rock
Reef Patch Reef indiv
Reef Colonized Pavement with C
Target
107
Proportion of Goal Held in Portfolio 100 Runs
BLM 0.1 17 Targets Over Goal
Portfolio 547 Units
Dry Alluvial
Rain alluvial
Dry Intrusive
Wet intrusive
Dry Extrusive
Rain intrusive
Wet limestone
LM wet alluvial
Rain extrusive
Dry ultramafic
Wet ultramafic
LM wet intrusive
Moist limestone
Dry sedimentary
LM wet extrusive
LM rain intrusive
Reef Linear Reef
Moist ultramafic
Reef Linear Reef
Wet sedimentary
LM rain extrusive
Reef Linear Reef
LM wet limestone
Moist sedimentary
Wetland Terrestrial
LM wet ultramafic
Reef Patch Reef indiv
Reef Scattered Coral-Rock
Reef Colonized Pavement with C
Target
108
Proportion of Goal Held in Portfolio 100 Runs,
BLM 0.5 22 Targets Over Goal
Portfolio 674 Units
Rain alluvial
Dry Intrusive
Wet intrusive
Dry Extrusive
Rain intrusive
Moist Intrusive
Wet limestone
Rain extrusive
LM wet alluvial
Dry ultramafic
Moist limestone
Wet ultramafic
Dry sedimentary
LM wet intrusive
LM rain intrusive
Moist ultramafic
LM wet extrusive
Reef Linear Reef
Wet sedimentary
Reef Linear Reef
LM rain extrusive
Moist sedimentary
LM wet limestone
Wetland Terrestrial
LM wet ultramafic
Reef Patch Reef indiv
Reef Colonized Pavement
Reef Scattered Coral-Rock
Reef Colonized Pavement with C
Target
109
Analysis of Highly Irreplaceable Areas
110
(No Transcript)
111
(No Transcript)
112
(No Transcript)
113
(No Transcript)
114
Additional information can be found in the MARXAN
manual which can be downloaded with the program
from http//www.ecology.uq.edu.au/?page20882pid

Please register as a user!