Conservation assessments using CLUZ: steps 9 to 12

Step 9: Set your boundary length modifier value

Choosing your boundary length modifier (BLM) value is not an exact science. The normal process is to choose an initial number, run MARXAN and then based on the results, re-run the analysis with an adjusted BLM value to increase or decrease the fragmentation levels of the portfolio. However, there is a technique for setting the initial BLM value and this is explained below, together with the underlying theory.

The total portfolio cost tends to be a trade-off between the boundary and planning unit costs, as reducing the total boundary cost inevitably increases the total planning unit cost. This is illustrated in the diagram on the right: reducing the boundary length (shown in red) involves adding the central planning unit to the portfolio and so increasing the combined planning unit cost.

The BLM is used to change the relative importance of the boundary cost in the total portfolio cost. If you set a high BLM value then the total boundary cost becomes the major factor in determining the total portfolio cost. Thus, MARXAN acts to reduce this total cost by identifying portfolios that contain more planning units but are less fragmented.

When it comes to setting the initial BLM value, you need to know the relative values of the combined planning unit costs and the combined boundary lengths, and this will depend on how these two factors are measured. For example, the boundary length could be measured as being 5000 metres or 5 kilometres but, to have the same effect, the initial BLM would have to be 1000 times larger if the lengths were measured in metres instead of kilometres.

So, the best way to set the initial BLM value is to first compare the mean planning unit cost with the boundary length of a typical planning unit. If the MARXAN boundary file was produced in CLUZ then it will be measured in the original units of the planning unit shapefile (which is usually metres). Then choose a BLM value so that the BLM value * planning unit boundary length = the mean planning unit cost.

Remember, you will probably need to adjust this BLM value at a later point, so don't worry too much about setting an exact value - the nearest integer will probably be sufficient.

To find out the mean planning unit cost value, first open the planning unit theme table. Then select the Cost field and go to the 'Statistics...' function in ArcView's Field menu. ArcView will produce an output that includes the mean of all the values in the field.

Step 10: Conduct a trial run in MARXAN

Your first MARXAN assessment should aim to test whether you have formatted the data correctly and chosen suitable BLM and target values. Therefore, you should choose parameters that ensure that this assessment is completed quickly.

To undertake a quick assessment then choose a small number of iterations (eg 1,000,000) and runs (eg 10) and do not choose the Use two step selection process option.

Step 11: Inspect the results are modify the parameters

MARXAN generally produces a different near-optimal portfolio at the end of each of the specified runs. It then identifies which of these portfolios has the lowest total cost and this best portfolio is displayed in CLUZ. It also counts the number of times each planning unit appears in each of the runs and this summed score map is also displayed in CLUZ. You should use both of these outputs to investigate the following:

a) Did the best portfolio meet all the targets?

Click on the Open MARXAN Target Results button to check whether the best portfolio met all of the targets. This will display the results from the previous MARXAN assessment.

If the best portfolio didn't meet all the targets then first go to the Target table and check whether the target amount for any of the conservation features is more than the total amount in the planning region.

If any of the targets weren't met then repeat the trial assessment but set a higher number of runs and iterations, as this increases the likelihood that MARXAN will identify a portfolio that meets all the targets.

If this does not solve the problem, then it might be that the only way for the targets to be met is by including some of the Excluded planning units. So, check how much of each feature is found in the Available, Conserved and Earmarked planning units and whether this is higher than the target value.

MARXAN will tell you if any of your targets cannot be met, so check the dialog box as it runs.

b) Does the portfolio include unsuitable planning units?

MARXAN can select planning units solely for their connectivity value and these units may be highly transformed and not suitable for conservation (unless restored). If MARXAN has included any of these planning units, or units that are unsuitable for other reasons, then it is worth setting their status as being Excluded. However, it is important to document the reason for excluding these units so that the assessment process remains transparent.

c) Is the portfolio too fragmented?

If the best portfolio is too fragmented then repeat the assessment and increase the BLM value. Similarly, if the portfolio only contains large patches of planning units then it is worth experimenting with a lower BLM value.

The standard way that CLUZ runs MARXAN means that all of the Conserved planning units are included at the beginning of each run, together with a randomly chosen set of units. The conserved units will always appear in the final portfolio, whereas the random set might be removed as part of the simulated annealing process.

In some situations, you might want to replace this randomly chosen set of units with a specified set of units. For example, you might want the analysis to initially include all of the Earmarked planning units that you have chosen, so that MARXAN can use this as a starting point in the simulated annealing process and identify related but potentially more effective portfolios. In this case, use CLUZ's Modify Earmarked units using MARXAN module.

Step 12: Produce the final output

You should have now decided on all the final parameter settings and so can undertake the final assessment. You will obviously want this assessment to produce good results, so it will be worth running MARXAN for a long period to ensure that you identify lots of near-optimal portfolios.

There are three ways to increase the chances of identifying lots of good portfolios when you run MARXAN:

1) Choose a large number of runs - at least 50 and ideally 200 or more.
2) Increase the number of iterations per run (although see below).
3) Choose the Use two step selection process option.

Increasing the number of runs and iterations both increase MARXAN's processing time. In my experience, increasing the number of runs produces better results than increasing the number of iterations per run, so I wouldn't select more than 2,000,000 iterations unless you are using tens of thousands of planning units.

MARXAN will now produce two outputs - the best portfolio map and the summed solution map - and you will need to decide which of these is the most appropriate for your situation.

When to use the summed solution map
Systematic conservation planning requires input from a range of groups at a number of stages and the best way to explain the process to these people is to use maps. However, you need to be very careful that this information isn't misinterpreted. This is especially important when displaying results from preliminary assessments that don't include all of the relevant information, such as socioeconomic or risk data (Smith et al., 2006).

In these situations it is better to show the summed solution map, which gives an irreplaceability score for each planning unit. This map is helpful because it identifies important areas but it does not specify where new conservation areas should be located. This reduces the likelihood that land-owners will object to the map because their property has been highlighted before anyone has discussed future conservation plans with them. It also means that your map will not specifically identify areas that could not be conserved, based on local conditions, making it less likely that local experts will doubt the value of the systematic conservation planning approach.

So, it is generally best to use the summed solution map when displaying the results from an analysis that is likely to be broadly modified in the future. However, you will still need to explain the following to avoid confusion:

1. You should specify which conservation features were used in the analysis, together with how the targets were set and what cost data were used.
   
   
2. You should emphasise that it is likely that some areas with relatively low irreplaceability scores will be needed to meet the targets.
   
   
3. You should emphasise that high irreplaceability scores are often dependent on proximity to existing conservation areas. Therefore, it is likely that if you add a new conservation area to the assessment then the irreplaceability scores of neighbouring planning units will increase dramatically.

When to use the best portfolio map
The final conservation assessment process involves choosing priority areas, so if you are happy that the assessment included all of the relevant data then you should illustrate your results with the best portfolio map. However, it is likely that you will still need to modify the best portfolio map to remove any patches of planning units that are too small to make viable conservation areas.

More importantly, it is also likely that you will need to modify your portfolio based on information from implementation agencies and relevant stakeholders. Such data should ideally be included in the MARXAN analysis but this may not always be possible. CLUZ has been designed to allow on-screen modification of the portfolio, so the final stage in the assessment process should involve working with a group of experts to produce the final portfolio.

CLUZ has a number of tools that you can use to clean-up and modify the best portfolio. These include:

The Change Status panel. This lets you change the conservation status of planning units that you have selected manually or by querying the planning unit shapefile table.

The Similarity Tool. This lets you click on an Earmarked planning unit and identify other Available units that contain a similar amount of each conservation feature. This is useful for identifying alternatives to small patches of planning units that need to be removed from the portfolio.