Addressing Ecological Objectives through the Setting of Targets

4 Addressing Ecological Set t ing of Target s Obj ect ives t hrough t he Louise Lieberknecht,1* Jeff A. Ardron,2 Ralph Wells,3 Natalie C. Ban,4 Mervyn Lötter,5 Jose L. Gerhartz,6 David J. Nicolson7 1Finding Sanctuary 2Pacific Marine Analysis and Research Association 3University of British Columbia, Centre for Applied Conservation Research 4University of British Columbia, Fisheries Centre 5Mpumalanga Tourism & Parks Agency; Scientific Services 6WWF‐Canada 7Nature Conservancy of Canada *Correspondence: E‐mail: Louise.Lieberknecht southwestfoodanddrink.com A BSTRACT Central to meeting the ecological objectives of a reserve network is that its spatial conservation targets are adequately determined and met. There are a number of approaches to setting conservation feature targets in Marxan, dependent on the ecological objectives and the available information. Often, it is appropriate to set broad scale representation targets for habitat or biotope classifications that cover the entire region (the coarse filter), and then set additional targets for spatially discreet individual features (the fine filter). Proportional targets for coarse filter features, such as habitat classifications, can be set the same for all classes for the feature (e.g., 10%), or can be scaled depending on the overall abundance of each feature class, with rarer ones given higher proportional targets than more abundant ones (see Box 4.1). For individual fine filter features, minimum viable population sizes and species/area curves can help define targets, when such information is available (see Section 4.3.2). Higher targets should be set for features of particular conservation concern (defined using explicit criteria such as rarity, decline, and threat). Existing protection measures (spatial and non‐spatial) should influence how targets are set, and existing legal frameworks and political commitments may contain general targets as a required starting point. A trans‐regional perspective will often influence targets, e.g., a feature that is rare in one planning region might be common in an adjacent region, and therefore be treated differently from a feature that is globally rare. In almost all cases, expert knowledge and opinion will be valuable in helping to define those targets or target ranges that are most likely to achieve ecological objectives (see Section 4.3.3 and Box 4.2). Where there is some uncertainty on this matter, it can be helpful to explore a range of targets to develop different scenarios. 4 .1 I N TROD UCTI ON Good systematic conservation planning requires explicit objectives formulated into a well‐defined problem (see Chapter 1: Introduction). In the context of Marxan, this chapter discusses the development of feature targets based on such ecological objectives. This is but one, albeit very important, aspect of translating Ecological goals into objectives which can then be quantified. In subsequent chapters, there is an outline of how other settings can be used in Marxan to address other design considerations (see Chapter 5: Chapter 4: Addressing Ecological Objectives through the setting of Targets 24 Reserve Design Considerations); and, how to incorporate socio‐economic objectives (see Chapter 6: Addressing Socioeconomic Objectives). Broad goals and specific objectives 4 of a project should be stated at the outset, and then used to formulate the conservation planning problem. Moving from general goals to quantifiable specific objectives is an important step in using Marxan, which can deal with some specific objectives explicitly and exactly, but others require interpretation, or require some imagination. Central to this problem specification is the determination of feature targets. 4 .2 CON SERVATI ON F EATURES AN D T ARGETS A conservation feature is a measurable, spatially definable component of biodiversity that is to be conserved within a reserve network. Conservation features can be defined at different levels of ecological scale, e.g., it is possible to protect species, communities, habitat types, populations, and genetic subtypes. In a Marxan analysis, each conservation feature is given a target, which is the amount of the conservation feature to be included within the reserve network, e.g., 10 000 ha of a habitat, or 30% of its original extent, or one occurrence. 5 How conservation features and targets are incorporated into a Marxan analysis is a reflection of the ecological objectives of the reserve network. For example, if there is a representation objective to include the full regional range of habitats within the network, then a regional habitat classification layer would be included in the analysis with each class incorporated as a specific conservation feature, for which targets would be set (e.g., 10% of each habitat type (see Box 4.1). If another objective is to adequately protect a particularly endangered mammal, then the species itself might be included as an additional conservation feature, with the minimum viable population size as a target. Alternatively, it might be more appropriate to use a habitat of importance to the species as a conservation feature (e.g., foraging habitat), and set a target of a given number of hectares. How conservation features are selected, and their targets set, will depend on the type, scale, quality and quantity of the available ecological datasets that relate to them. In practice, the availability of good quality spatial data will often limit what conservation features and targets can be used. Ecological datasets come in many forms: point samples of species occurrences, observation records, abundances, species distribution maps (binary maps or probability of occurrence maps), habitat maps, habitat suitability maps, Note that there is some semantic confusion in the literature about goals and objectives. In this handbook, goals will be considered as broader and more general, while objectives are more specific and quantifiable. 4 These terms are not to be confused with terminology used by The Nature Conservancy, which uses the word target to mean the term conservation feature as used in this handbook. 5 Chapter 4: Addressing Ecological Objectives through the setting of Targets 25 numbers of individuals or numbers of species recorded on grid squares, probability of occurrence, etc. Perhaps one of the greatest challenges is integrating and pre‐processing the various available data, so each conservation feature is summarised in a single unified dataset, associated with the planning units (see Chapter 7: Assessing and Managing Data and Chapter 8: Ensuring Robust Analysis). If a probability surface is used, the probability values can be used as a surrogate for abundance, with total “abundance” being the sum of the probability surface in the study area. A good practice is to document not just the conservation features and targets that are used, but also the rationale(s) behind their selection. 4 .3 S ETTI N G M EAN I N GFUL T ARGETS 4 .3 .1 H ow coa r se a nd fin e filt e r t a r ge t s w or k t oge t h e r Coarse filter features are those that cover most or all of the planning area and usually represent habitats, biomes, or higher level species communities. For example, a representation target of, say, 25% for a vegetation class may protect an estimated 75% of all species found within this vegetation class. This is a coarse filter, as it is not considering any single species, per se, but rather a general grouping that usually occur together. The further inclusion of targets for fine scale or point locality data for selected species or habitats, refines this coarse filter approach, to include those critical areas where taxa of particular conservation concern are known (or likely) to occur. Criteria to select fine filter conservation features are various, and can include conservation features with special habitats not adequately represented through a coarse filter, such as rare, threatened or endangered species; keystone or umbrella species; endemic species; or species which have a disproportionate influence on their surrounding environment. In some cases, coarse filter features are distributed across a study area with a broad range of sizes. For example, a seabed habitat classification might include large swathes of sandy areas, punctuated by small patches of rocky reef. In such instances, if targets of equal proportions are applied to all features, then the network can become dominated by vast swaths of common, likely less threatened features, and protecting such large common features may not be the best use of limited conservation resources (see Box 4.1). Chapter 4: Addressing Ecological Objectives through the setting of Targets 26 Box 4.1: One approach for scaling coarse filter targets Protecting 30% of a habitat covering 1 000 000 hectares is a considerably larger undertaking than protecting 30% of a more unusual habitat covering 1000 hectares, Ideally, adequacy data (e.g., species‐habitat curves – see below) should be used to select appropriate percentages. However, for coarse filter features, these data are seldom if ever available, particularly in the marine environment. In such cases, other approaches should be explored to scale proportional targets based on the overall abundance of the conservation features. One approach to contending with multi‐scalar features is to normalise the spatial data using a square root transformation (just as species populations can often be normalised using a logarithmic transformation), and then scaling representation targets roughly in proportion to the square‐root of the ratio of representative features’ overall areas. Thus, within a given feature class (e.g., benthic habitats, or marine biomes), for any two features (x & y), protection would be such that: (xp / yp) ≈ (xt / yt)0.5 ...where the subscript “p” represents the area protected of a given feature and the subscript “t” represents the total area of a given feature in the network. Put another way, the distribution of targets for multiple representative features of the same general kind should fall within a continuum roughly proportional to the square root of their respective total areas. In the above example, if 30% of the 1000 hectare feature is protected (i.e., 300 ha) then according to the formula, we would expect about 9500 hectares of the common one million hectare feature to be protected, 6 which works out to be about 9.5%. So, perhaps a 10% target would be set for the larger feature and a 30% target for the smaller one. Statistical assumptions behind this concept are discussed by Ardron (2008). Whether it is appropriate to scale representation targets, such as suggested here, will depend on the ecological objectives of the network. For example, Johnson et al. (2008) point out that marine species associated with more common habitats will likely recruit from protected as well as unprotected sites, but that those associated with less common habitats will be more reliant on the dispersed “stepping stones” of protected areas, and thus proportionally more of those less‐common habitats should be protected. Likewise, if a greater emphasis is put on protecting rare or unusual features, or if it is pragmatically unrealistic to protect very large areas, then scaling the targets could be appropriate. On the other hand, if the representation objective is to faithfully reflect the natural relative abundances of all representative features across the network, then it may not be appropriate to include a higher proportion of rarer features. 6 That is, 300 ha * (1 000 000/1000)0.5 = 9487 ha. Chapter 4: Addressing Ecological Objectives through the setting of Targets 27 4 .3 .2 H ow m u ch is e n ough? Via bilit y a sse ssm e nt s, spe cie s- a r e a cur ve s, a n d e x pe r t opin ion Ideally, we would inform minimum targets for individual species using detailed viability assessments – e.g., a reserve system should include enough habitat for 1000 Mountain Zebras (Ferrar and Lötter 2007). For individual species, the absolute values of targets may be guided by knowledge on minimum viable population sizes (MVPs). Where available, species‐area curves can be helpful guidance in the setting of targets for the required areas of different communities / habitats / biomes to be placed under protection (Desmet and Cowling 2004, Pryce et al. 2006). The principle is that enough area should be protected to ensure that the characteristic species of a community or biome are likely to be included; however, the details of implementation can be complicated (Tjørve 2003), and the way that a curve was developed will influence how it should be used (Scheiner 2003). Generally, as more area is set aside, the rate of increasing ecological benefits for the given species community or biome will begin to flatten (see Figure 4.1), and somewhere in this flattening section is where a target should be set. One “rule of thumb” is to locate the region of the curve where 1/10 increase in area gives 1/10 increase in species (Cain 1938). Other related curves consider larval dispersal, pollination distances, species range, and so forth. Figure 4.1: Species area curve (from Pryce et al. 2006) Chapter 4: Addressing Ecological Objectives through the setting of Targets 28 Unfortunately, especially in the marine realm, such data are usually unavailable and expert opinion is often the only substitute. In discussions with experts, it is helpful to understand what is considered to be a minimum viable patch size, and what would be considered a minimum network size to ensure ecological goals such as genetic diversity. Spatially relevant issues such as patch separation, terrestrial corridors / marine larval dispersal distances, and life history stages should also be discussed and translated into spatial targets, if known and when appropriate (see Section 4.3 ‐ Setting Meaningful Targets and Box 4.1). 4 .3 .3 Ex pe r t a dvice a nd pe e r r e vie w The choice of ecological targets used in an analysis could have far‐reaching implications, and will have to be defended, perhaps in a court of law. The initial selection of ecological targets by the analysis team should incorporate expert and sometimes also stakeholder input (see Chapter 10: Using Marxan in Multi‐Stakeholder Planning Processes). Box 4.1 shows the questionnaire that was used in a series of expert workshops to inform marine conservation planning processes in British Columbia (BC), Canada. Good practice is to aim for agreement on a range of plausible target values. However, many experts are not comfortable with the use of numerical target values, and/or tend to overvalue their own particular areas of research. Thus the task of balancing the numeric values for all ecological targets in the analysis may ultimately reside with the core analysis team. During refinements, it can be very helpful to consider the relative target values of conservation features as a related set rather than absolute values for individual features. In earlier BC analyses, protection targets for features were first ranked relatively using quantitative terms (low, mod‐low, moderate, mod‐high, high, very high) and then afterwards various numerical targets were applied to these terms in different scenarios (Ardron et al. 2000, Ardron 2003, 2008). Box 4.2: Expert workshops to assist in setting targets The British Columbia Marine Conservation Analysis (BCMCA, www.bcmca.ca) has taken an expert‐based approach to selecting features and setting targets. To do this, the project team organized one‐day themed expert workshops (for ecological themes including seabirds, marine plants, fishes, invertebrates, and marine mammals). After an introduction to the project, each workshop was dedicated to filling out worksheets based on the questions listed below. Chapter 4: Addressing Ecological Objectives through the setting of Targets 29 SECTION 1 - FEATURES Marine Feature Rationale List the unique species/ecological features from this dataset (e.g., species, families, groupings of species or of species habitats) that require individual consideration in the BCMCA. You may also wish to delineate features by season/ region or both. Justification for classifying features or treating them separately. SECTION 2 - ECOLOGICAL TARGETS Comments/ Justifications Measure Target (range) The type of measure that will be used to capture the marine feature (e.g., Percent of current extent of feature in study area, percent of current population, number of occurrences). The amount of the feature required for meeting the BCMCA's 4 ecological objectives: (1) Represent the diversity of BC's marine ecosystems (2) maintain viable populations of native species; (3) sustain ecological and evolutionary processes; (4) build a conservation network that is resilient to environmental change. Ranges should span minimum to preferred amounts. SECTION 3 - ECOLOGICAL CONSIDERATIONS Minimum Patch Size Minimum size of patch/population needed to ensure population viability. Replication Separation Distance Other Ecological Comments Considerations How many unique patches are needed to ensure long-term population persistence/to safeguard against disturbances? The minimum distance that distinct patches of a feature should be from one another (consider dispersal distances). e.g., connectivity, ecosystem linkages, dynamics, special management considerations. SECTION 4 - SOURCES OF FLORA DATA AND PRE-PROCESSING Dataset/Layer Description Geometry Spatially georeferenced data that captures the location of the features. Preference will be given to digital data. This list need not be inclusive but should represent the best available data for science-driven analyses. Brief description of dataset. Geometry type (point line or polygon) Provider, Custodian Data provider/ reference Extent Geographi c Extent of Database Key Fields/ Attributes Descriptive information stored with the spatial data. SECTION 5 - PRE-PROCESSING Pre-Processing How should this dataset (or combined datasets) be processed/prepared for use in Marxan? Chapter 4: Addressing Ecological Objectives through the setting of Targets 30 Box 4.3: Lessons from the expert workshops By Karin Bodtker, BCMCA The BCMCA found that completing worksheets (see Box 4.2) worked better in some workshops than others. In general, they had greater success completing the worksheet under these conditions: • Features were relatively easy to itemize on a species–by‐species basis. • The range of experts in attendance covered the full range of species groups being discussed. • The experts in attendance either held the data they were recommending for the BCMCA project or they had good knowledge of them. • There were no prior misconceptions about Marxan. The BCMCA found that even though they had developed clear ecological objectives for Marxan scenarios, many experts were uncomfortable recommending target ranges for features because they usually had little or no evidence to support their recommendations. In hindsight, these three suggestions may help to solicit clearer responses from experts: • Develop materials on examples of real‐world Marxan analyses. From these, discuss a range of scenario objectives, itemized features, targets and results. • Acknowledge that peer reviewed science that prescribes targets based on specific objectives largely does not exist and in order to move forward the project is trying to ascertain reasonable target ranges based on expert knowledge of the relevant ecological features. • Emphasize that a range of targets will be explored, acknowledging that a single “right” number probably does not exist. Furthermore, BCMCA held workshops at different times with different attendees and facilitators over a ten month period. While the format for the workshops was the same (large group plenary and small group break‐out sessions with 4‐6 experts in each small group), there were different approaches taken by different groups for identifying features or targets, possibly the result of “group‐think.” Chapter 4: Addressing Ecological Objectives through the setting of Targets 31 Box 4.4: An alternative to expert workshops to assist in setting targets Dave Nicolson, Black Coffee Consulting An alternative approach to workshops with worksheets would be to hold a workshop focused on identifying features and data sources to populate those features, and introducing the topic of targets, followed by a Delphi survey/questionnaire to help set targets for each of the features. Invited experts would independently assign targets for all identified features, then be shown the average target and range of targets from all experts and be given an opportunity to revise their responses. Benefits of this approach include: • reduced group‐think bias; • experts know the metrics that the data support prior to assigning targets; • efficient use of participants time; and • all experts have equal opportunity to contribute. Difficulties of this approach include: • low response rate when soliciting expert feedback by questionnaire; • time lapse between explanation of targets and request for target recommendations (i.e., experts forget or are unsure and do not respond as a result); and • Opinions on features and targets beyond participant expertise. 4 .3 .4 Con se r va t ion st a t u s a s a pr ox y for t a r ge t - se t t in g The conservation status of species and habitats (e.g., IUCN, or national red lists, globally or regionally at risk) can be used to inform priorities and set targets for individual features. Criteria that are commonly used to develop these lists include threat, recent or historical declines, rarity, endemism, or proportional importance, and features that fall into these criteria sometimes receive higher percentage targets. Where species or habitats have declined in extent, data on historic distributions may be available. These can be used to guide setting targets for what remains, as a proportion of its historic abundance. Knowledge on the causes of declines can also help inform target setting. For example, where estuary fish populations are threatened by river obstructions, the ecological objective may be to protect one whole un‐dammed river length from catchment to sea, and the target would be a specified length of unaltered aquatic habitat containing headwater, tributary and mainstream elements. The use of conservation status to direct fine filter targets risks that the emphasis of fine filter protection could rest on rare and threatened features, and these alone are very unlikely to ensure a healthy functioning ecosystem. Therefore, good practice dictates that other considerations such as ecological significance (e.g., keystone species for fine filter targets) and representativity (for coarse filter targets) are also quantified. Chapter 4: Addressing Ecological Objectives through the setting of Targets 32 Box 4.5: Tracking “optional features” in the analysis Sometimes there may be features of possible secondary interest that are not directly relevant to the declared ecological objectives of the analysis. These can still be included in the analysis by setting their target and/or penalty factors to zero. Marxan will not try to collect them, but the tabular outputs will allow the user to track how much was included. Alternatively, giving a feature a much lower than usual (but non‐zero) penalty factor and/or target will allow it to “tip the scales” in situations when two planning units are otherwise calculated as being equal. However, good practice dictates that such fine‐ tuning only occurs after the more important features and their targets have already been sorted out (see Section 4.4 ‐ Targets and Trade‐Offs). 4 .3 .5 Ex ist ing pr ot e ct ion le ve ls If a user “locks in” existing spatially protected areas, current protection levels for features within those areas will be accounted for in an analysis. However, when a particular feature is already protected through non‐spatial measures, such as stringent quotas, this cannot be directly accounted for in a Marxan analysis. If such a feature is to be given spatial protection, then the benefit it is already deriving from existing non‐spatial protection can mean that a lower spatial target will provide sufficient overall protection. This will depend on the ecological objective, bearing in mind it can be hard to accurately quantify the effectiveness of non‐spatial measures and translate them into a spatial equivalent. 4 .3 .6 Le ga l fr a m e w or k or m a n da t e Many countries are signatories to conventions or subject to legislation that require a certain proportion of particular species / habitats to be preserved (e.g., the European Community’s Birds and Habitats Directives). This will provide a starting point for setting ecological targets, as well as making them defensible publicly and in a court of law; however, caution is advised, as a legal target may not match what is ecologically required. If there appears to be a discrepancy in this regard, running two sets of scenarios (the “legal” and the “ecological”) can help visually demonstrate the differences in the possible solutions to stakeholders and decision‐makers. Likewise, some organisations have mission statements that commit them to certain targets. Again, these can be run alongside targets based on project‐specific ecological assessments, as well as those created by other stakeholders, government policy, or legal requirements (see Box 4.5). Chapter 4: Addressing Ecological Objectives through the setting of Targets 33 Box 4.6: The policy side of setting targets By George F. Wilhere, Washington Department of Fish and Wildlife Target development should be informed by science, but it should not be freed from economic concerns or removed from political discourse (Wilhere 2008). 7 Those who develop conservation targets within the context of a planning process without acknowledging the nexus with economics and ethics are possibly guilty of “stealth” policy advocacy (Lackey 2007). 8 To avoid inadvertent stealth policy advocacy, Wilhere (2008) makes the following recommendations: Scientists should understand that (1) conservation targets are ultimately expressions of acceptable risk; i.e., how much of a gamble are you willing to make? (2) attitudes toward anthropogenic extinction risk are based on ethical values, and (3) ethical value judgments are outside the realm of science. Therefore, when scientists base a conservation assessment on a set of subjective targets, they should clearly state that it represents just one policy option from a wider range of potential options. Whenever practical, scientists should do conservation assessments for a range of targets, even including targets that might make conservation biologists uncomfortable. Scientists should refrain from favouring a particular set of a priori targets within the published assessment, except for indicating how well these targets would be expected to meet the conservation objectives of the process. When the data are available, conservation biologists should ideally work with economists to estimate the relative costs of different risk levels, including information about absolute and marginal costs. Extinction risk could be framed in terms of cost–benefit trade‐offs, while still recognizing the decision as ultimately an ethical dilemma. 4 .3 .7 Tr a n s- r e giona l pla n nin g Noting the proportional importance or degree of endemism of a species within a study region may assist in setting targets and in edge‐matching of conservation plans, particularly if the regions follow administrative rather than ecological boundaries, or in ecological regions where it is not possible to run one single assessment. For example, a species that is rare in a particular region may be at the edge of its geographical range, and well represented (and protected) in an adjacent region: such a species should generally be given a lower target than a globally rare species. Usually it is good practice to preserve species in the core of their geographical range, i.e., their “strongholds;” however in some cases it might be appropriate to consider targets for species at the edge of their geographical range, if it is known that the distribution is shifting in that direction; e.g., due to climate change. 7 Wilhere, G.F. 2008. The how‐much‐is‐enough myth. Conservation Biology 22: 514‐517. 8 Lackey, R. T. 2007. Science, scientists, and policy advocacy. Conservation Biology 21: 12–17. Chapter 4: Addressing Ecological Objectives through the setting of Targets 34 Where the geographical range of feature classes spans multiple planning regions, and it is not possible to run a single assessment for all regions together, setting the same representativity targets for the different feature classes in adjacent regions, irrespective of their relative abundance within each region, will ensure some degree of consistency between the plans covering adjacent areas. Such consistency can be important to stakeholders, whereby all regions are seen to be treated equally, which is interpreted to mean “fairly.” Ecologically, however, this is usually not ideal, as it disregards ecological differences between regions and good practice would dictate attempting to take these differences into account, when practicable. 4 .4 T ARGETS AN D T RAD E- O FFS 4 .4 .1 I t e r a t ive pla n n in g Achieving broad ecosystem goals, and thus all ecological objectives and Marxan targets that flow from them, should be considered central to any ecosystem‐based analysis. However, pragmatic considerations often require trade‐offs to be made. In exploring trade‐offs, it is a good practice to iteratively explore a range of plausible targets, to document the pros and cons, and the reasoning behind the decisions ultimately made. For example, if experts are unable to agree on a target for a conservation feature, it can be helpful to run different scenarios, exploring these differences. It may be found that they do not make much difference to the reserve design. Or, if they do change the solutions dramatically, then this clearly indicates an area where more research or additional advice is required (see Section 8.4 ‐ Sensitivity Analysis). Likewise, it can be important for stakeholders to understand that changing some targets may not impact outcomes significantly, because targets for other objectives are driving overall outcomes. Since it is often not possible to be certain if a particular target will meet the criteria of adequacy (of conserving viable populations, for example), it is important to communicate other (higher or lower risk) options may exist, so that stakeholders understand the tradeoffs (see Chapter 10: Using Marxan in Multi‐Stakeholder Planning Processes). 4 .4 .2 W e igh t ing t a r ge t s t h r ou gh t h e spe cie s pe na lt y fa ct or Promising Marxan solutions should be evaluated in light of meeting their ecological targets, and initially the two most likely issues will be: • Under‐representation: When not all targets were achieved, what was the shortfall? Was it ecologically / statistically significant? Does this indicate that the target was unrealistic? Should the target be lowered, or alternatively, does increasing the Chapter 4: Addressing Ecological Objectives through the setting of Targets 35 species penalty factor 9 correct this shortfall? What impact does making those adjustments have on the new Marxan solutions overall? • Over‐representation: Were targets overachieved? This can mean that the solution is not spatially efficient and the associated species penalty factor could be adjusted downwards. What impact does making those adjustments have on the new Marxan solutions overall? As indicated above, the user can decide how important it is to meet the target for a specific conservation feature through adjustment of the species penalty factor. Initially, it ought to be set the same for all features in an analysis. If some targets are not being met for some conservation features, these may be iteratively given a higher species penalty factor than others, in order for all targets to be met (see Section 8.3.1 ‐ Iterations). Some practitioners may hesitate to do this, reasoning that each conservation feature and its ecological objective (hence target) is considered equally important. However, it may be that some conservation features are more costly / difficult to obtain than others, and with a flat penalty factor their targets will not be met. If including these features is a necessary objective, then their species penalty factor will have to be increased (or, their costs decreased; or, the factors for other features decreased). Philosophically, the use of equal penalty factors assumes a “flat” ecological hierarchy, which can be difficult to defend, since it goes against the commonly accepted notions of the heightened ecological importance of keystones species, the intrinsic value in protecting rarity, etc. Regardless of which decision is taken, good practice dictates that the underlying reasoning for such decisions be clearly explained. Further guidance on setting the species penalty factor is given in the Marxan User Manual and in Chapter 8: Ensuring Robust Analysis. 4 .4 .3 Adj u st in g t a r ge t s ba se d on pr a gm a t ic conside r a t ions A decision may be taken to lower some targets. For example, if targets are set to represent a percentage of each broad habitat within a study area, even if that percentage is low, it may only be possible for Marxan to meet those targets by selecting very large areas ‐ especially if additional constraints are included in the analysis (e.g., locked in areas or fine filter targets for individual species, as discussed above). In such circumstances, the outcome may not be politically or practically achievable, and such targets that drive selection towards large swaths of area may have to be lowered. For the purposes of decision‐making, a final product may include a number of reserve scenarios created based on the same conservation features but a variety of different targets for the features (e.g., 10%, 20%, 30%, etc.), reflecting differing levels of protection, differing conservation costs, and differing risks to species viability and ecological integrity. The species penalty factor (SFP) can be applied to any kind of feature, including species, but also habitats, biomes, etc. To clarify this, some users instead use the term conservation penalty factor. 9 Chapter 4: Addressing Ecological Objectives through the setting of Targets 36 It may be impractical to achieve all ecological objectives. Difficult trade‐offs often have to be made by decision makers and planners, which are ultimately borne by the wider stakeholders and public; i.e., society. That is not to say that scientists and conservationists should not continue to argue their case for ecological objectives. Presenting to planners and decision makers a variety of output solutions illustrating trade‐offs with other activities or interests is good practice. Marxan outputs can be compared with solutions from other conservation planning tools, such as C‐Plan (see Carwardine et al. 2006) or Zonation (http://www.helsinki.fi/ bioscience/consplan/). These can suggest other solutions to difficult trade‐offs. It may be enlightening to compare solutions to random surfaces, in order to measure the ratio of efficiency (e.g., 30 times more efficient than random chance), which can also help build confidence that the tool is working as it should despite the difficult trade‐offs. As always, the rationale behind the Marxan trade‐offs and the ultimate decisions taken should be transparent and documented, ideally with the relevant authorities accountable for their decisions. More detailed guidance on evaluating Marxan outputs is provided in Chapter 9: Interpreting and Communicating Outputs. 4 .5 CH ALLEN GES 4 .5 .1 Ga ps in qua lit y a nd cove r a ge of spa t ia l da t a The ecological importance of a feature has to be balanced with the quality of its data. Weak or incomplete data should not be “driving” the analysis. It may be tempting to include all conservation features for which there is some data, but if particular datasets are very weak, it may be preferable not to include them at all. However, if a feature is rare or otherwise important, then perhaps including incomplete or weak data will be judged as being better than using none at all. In such cases, it is generally good practice to assign the weaker dataset a lower than normal species penalty factor. It’s a balancing act. Whatever the case, analysts and planners should record the decision and its rationale. Sampling bias is a common problem; the algorithm will gravitate towards data‐rich areas, so that even those features that are more widely distributed and recorded will be chosen in these data‐rich areas, if possible (see Chapter 7: Assessing and Managing Data). If there are comprehensive surveys for a particular feature in one part of the region, and not another, and it is decided to include it, then mitigating strategies will have to be employed, such as breaking the study area into sub‐regions, and setting targets for conservation features for both the entire study area and each of the regions in which data are found (e.g., Pryce et al. 2006). If significantly different data collection methods have been used in different places, then these might be better treated as separate features. Sometimes, it is easier to run a separate analysis for smaller data‐rich regions. Chapter 4: Addressing Ecological Objectives through the setting of Targets 37 Where data are even but patchy then statistical models can be used to extrapolate distributional data for features evenly across the region (Rondini et al. 2005). When there are no data for conservation features to inform targets for a particular ecological objective, then a surrogate or modelled surface should be used if possible. In terms of good practice, it is important to understand the actual effects of using incomplete or varying datasets, make an informed decision, and to communicate the trade‐offs clearly, especially if important ecological objectives cannot be addressed. More advice on data preparation is given in Chapter 7: Assessing and Managing Data and Chapter 8: Ensuring Robust Analysis. 4 .5 .2 Ga ps in scie n t ific k n ow le dge Even with good knowledge on the distribution of features, it is often not known what targets are necessary in order to achieve ecological objectives. In some cases, there may be very specific scientific evidence that can be used, e.g., minimum viable population sizes, or the minimum area of a habitat required for foraging for individuals of a particular species. However, such cases are the exception. It is usually hard to come up with definite figures such as a percentage of the total area of a habitat that should be placed under protection in order to ensure its integrity and persistence. No matter what value is chosen, some species inevitably will fare better than others. In such cases, the exploration of plausible ranges of values may be more meaningful, and often can highlight likely compromises. Plotting the cost of the overall network versus the upper and lower range values of the targets can indicate if there are non‐linear relationships to be considered, where the network is relatively “cheap” up to a certain point, and then becomes costly. Chapter 4: Addressing Ecological Objectives through the setting of Targets 38 5 Reserve Design Considerat ions Mervyn Lötter,1* Louise Lieberknecht,2 Jeff A. Ardron,3 Ralph Wells,4 Natalie C. Ban,5 David J. Nicolson,6 Jose L. Gerhartz7 1Mpumalanga Tourism & Parks Agency; Scientific Services 2Finding Sanctuary 3Pacific Marine Analysis and Research Association 5University 4University of British Columbia, Centre for Applied Conservation Research of British Columbia, Fisheries Centre 6Nature Conservancy of Canada 7WWF‐Canada *Correspondence: E‐mail: mervyn.lotter gmail.com A BSTRACT In addition to setting ecological targets (see Chapter 4: Addressing Ecological Objectives through the Setting of Targets), there are a number of spatial reserve design considerations that can be addressed using Marxan. These include options to set minimum patch sizes for specific features (to allow for the capture of ecological processes that operate at known spatial scales), and to specify a minimum distance between and number of replications of patches for specific features (to allow the incorporation of an “insurance factor” against local catastrophic events). One of the most complex issues in conservation planning is ecological connectivity; this can be partially addressed by choosing appropriate boundary length modifier (BLM) values and planning unit shapes, as well as modifying the boundary file to bias the algorithm towards selecting specific sets of spatially separated planning units together. The main obstacle to the adequate incorporation of ecological connectivity into the planning process, however, remains the lack of spatially explicit knowledge about connectivity at broad ecological scales. No single tool can consider all ecological aspects, and for those that are not easily delineated spatially, other tools could be used in conjunction with Marxan. 5 .1 I N TROD UCTI ON In addition to the setting of ecological targets (see Chapter 4: Addressing Ecological Objectives through the Setting of Targets), Marxan allows for several other reserve network design considerations to be incorporated. This chapter outlines these considerations and how they relate to Marxan functionalities, which together with well chosen targets, can help to achieve ecological objectives. Readers are also directed to the Marxan User Manual, which discusses the operation of these features. Within the framework of systematic conservation planning there are at least four general classes of ecological objectives: representation, adequacy, efficiency and design Chapter 5: Reserve Design Considerations 39 (Possingham et al. 2006). 10 Representation objectives are about “getting a bit of everything”. For example, the specific objective to conserve every habitat type is primarily about representation, with little thought about adequacy and design. Adequacy objectives focus on creating a reserve network that is sufficient to conserve the conservation features in perpetuity. For example, the specific objective of conserving a viable population of mammals emphasises adequacy. Ecological uncertainty in these issues speaks to the need to understand trade‐offs (costs versus precaution) when defining such objectives. The threat to particular areas, and allowing that threat to modify decisions, can also be considered as part of adequacy objectives; e.g., to conserve 15% of the distribution of all types of coral reef, using those sites that are least threatened by coral bleaching. Efficiency objectives focus on achieving ecological objectives while still keeping the cost of the whole system small, where cost is often measured in socio‐economic terms (see Chapter 6: Addressing Socioeconomic Objectives). Finally, reserve design objectives address issues of spatial position, size and shape. For example, we may wish to ensure all reserves are at least 1000 ha, or the whole system is relatively compact (has a low edge to area ratio). Reserve design considerations may also include issues such as replication and connectivity, as discussed below. 5 .2 CON N ECTI VI TY Connectivity (within the sea/landscape) is defined as “the degree to which the landscape facilitates or impedes movement among resource patches” (Taylor et al. 1993). It is a complicated issue, varying from feature to feature, and cannot be fully incorporated using reserve selection tools currently available, including Marxan. However, some aspects of it can be addressed. Brooks (2003) recognises two components to landscape connectivity: structural and functional connectivity. Structural connectivity is the spatial structure of a landscape and can be described from map elements (e.g., vegetation units). Ensuring clusters and corridors connecting clusters across the landscape, assists in maintaining structural connectivity. In Marxan changing the “boundary length modifier” (BLM) alters the relative importance of maintaining structural connectivity. The shape of planning units will also have an effect on structural connectivity. Multi‐ facetted (edged) planning units (e.g., hexagons) are often more efficient than a square grid in creating reserves with low edge to area ratios. If unevenly sized planning units are chosen, larger units will naturally be internally better connected than smaller ones, but the chances of external connectivity (between planning units) is greater with smaller units because they generally “cost” less to join together than larger units. On land, 10 There are several variations on this theme. For example, the OSPAR and HELCOM regional seas conventions consider four network criteria: representation, adaquacy, connectivity, and replication (OSPAR 2007). The Convention on Biological Diversity is considering the four OSPAR/HELCOM criteria, plus a fifth: Ecologically or Biologically Significant Areas (CBD 2008) Chapter 5: Reserve Design Considerations 40 planning units are sometimes chosen to represent watersheds, which are inherently internally connected. In the marine realm, however, such delineations are much less clear, and regularly sized units are most often selected. In the Great Barrier Reef Marine Park rezoning a decision was made that each coral reef would be a single planning unit so, in general, whole reefs are either inside, or outside, the reserve system. Functional connectivity recognises the response of individuals to landscape features (e.g., certain fish require continuous stretches of river for feeding and spawning). Functional connectivity considers the biology and life history of the features concerned, and the reality is that there may be very limited data available. Furthermore, connectivity distances vary widely between species, and a “sink” for one may be a “source” for another, making it difficult to address connectivity at a landscape or ecosystem scale (as opposed to for individual species or populations). In the marine environment, ocean currents are often used as a proxy for connectivity, based on the fact that many species have a larval development phase, during which they may disperse “passively” in currents. However, the actual dispersal patterns of marine species can be strongly affected by larval behaviour, often resulting in shorter dispersal distances than expected based on current strengths and direction (e.g., Leis 2002). In the terrestrial environment, it may be possible to spatially define and map corridors for individual features, such as migration routes along mountain passes. It may also be possible to define probabilistic corridors in the marine environment. Spatial models can be used to assist in predicting corridors; a least‐cost path analysis (such as available in several GIS) is one approach. If there are known critical corridors or “bottlenecks” they can be locked into the Marxan reserve system. If there is reliable evidence of functional connectivity between two spatially separated areas for a given feature, they can be given a common boundary in the boundary input file. It does not matter if planning units do not actually share a boundary – if they share a high boundary value in the input file, Marxan is effectively ʺtrickedʺ into treating them as neighbours, and it will act to reduce boundary costs by selecting them together (see the CLUZ website at www.mosaic‐conservation.org/cluz/). New versions of Marxan will make this more explicit by calling the boundary length a “connectivity cost”. In effect, the “connectivity cost” allows for the fact that the whole network is more than the sum of its parts, and that spatial adjacency is just one form of connectivity, amongst others. 5 .3 M I N I M UM CLUM P OR P ATCH SI ZE In Marxan, it is possible to define how large a patch of a particular feature must be in order for it to count towards meeting its target. This can help ensure persistence and integrity of the feature. Species area curves (Figure 4.1), or population viability analyses, can be helpful in calculating minimum clump targets, if such data are available. Minimum clump size targets can also ensure that dynamic ecological processes are captured, which are otherwise difficult to add as features. For example, if we know that most fires burn to around a certain size, the user may wish to make all reserves several Chapter 5: Reserve Design Considerations 41 times that size so the chance that a whole park burns is very small, and each park maintains a mosaic of successional stages. Similarly we may wish to set a minimum clump size large enough to allow for plant‐pollinator interactions to continue. Within the marine environment, it has been suggested that if clump sizes are too small, most larvae will disperse beyond the protected areas, making populations within the clusters unable to sustain themselves (Halpern and Warner 2003). However, there is evidence for high levels of larval retention mechanisms in many marine environments, such as coral reefs (e.g., Swearer 1999, Cowen et al. 2000, Leis 2002, Cowen et al. 2006). Where there is good data available for the spatial scales of larval dispersal/retention, this can help inform minimum clump size targets, as well as design considerations relating to wider connectivity (see above). In general, however, marine organisms have a wide variety of dispersal distances, and thus it is important in designating patch sizes to consider whether there are neighbouring sites that can supply recruits (Johnson et al. 2008). It is worth bearing in mind that increasing the BLM setting will increase the clump size throughout the reserve network. This is of course different from specifying the minimum clump size for individual conservation features, but in practice is often sufficient. The effect of the BLM can readily be calculated in a GIS by merging (“dissolving”) all of the selected reserve network units and calculating the average patch size and other statistics, which can be compared with the required patch sizes to support viable populations of conservation features. Setting minimum clump targets for individual features increases Marxan processing time considerably, and it may also result in some representation targets not being achieved (one solution may then be to increase the species penalty values for those features not adequately represented). It is important to understand such trade‐offs, and how they can affect final solutions, when using minimum clump size targets. 5 .4 R EPLI CATI ON AN D S EPARATI ON To ensure long‐term persistence of some conservation features, a conservation area network may require those features to be protected in multiple separate patches, spaced apart from one another. Replication of features can: • spread risk against damaging events and long term change affecting individual sites; • ensure that natural variation in the feature is covered (either at a genetic level within species or within habitat types); • increase the number of connections between sites and enhance connectivity in the network; • allow the establishment of replicate scientific reference areas; and, • allow for uncertainty in the identification of features, such that the greater the uncertainty, the more replication is required to ensure the feature is likely being protected (OSPAR 2007, p27). Chapter 5: Reserve Design Considerations 42 To achieve this, the practitioner can set minimum distances between different protected patches containing a particular feature. Additionally, the design can also ask for a given number of replications of features within a reserve network. For example, in Mpumalanga, South Africa, at least three replications of large montane grassland patches (>15 000 hectares each) were required, with a minimum distance of 20 km between these patches, so that they could not be contiguous (Ferrar and Lötter 2007). Where separation distance and replication are used to provide an insurance against disaster, areas containing the same features should be separated by sufficiently large distances to offset the chance of a catastrophic event affecting more than one site. The distances and number of replications required will vary depending on the nature of the main threat(s), and the vulnerability of the features to those threats. Applying the precautionary principle means increasing the number replicates when there is uncertainty about data, for features that are particularly vulnerable, and in areas / regions that are particularly threatened (e.g., for seabird feeding areas near major shipping routes, where oil spills are more likely than elsewhere). When using separation distance, it is important to be aware of the distance units being used. If units and boundaries are measured in metres, then distance (separation) needs to be measured in metres. Use of the separation distance option can also considerably increase processing time. Planners need to weigh the importance of insurance against catastrophic events against considerations that may require distances between patches to be kept small, e.g., facilitating recolonisations in a metapopulation and ecological connectivity in general (see above). There are possible alternatives to the use of replication settings. In the Colombian Caribbean the names (codes) of single conservation features were altered depending on what part of the planning region they occurred in (e.g., feature1area1, feature1area2 etc.), and then targets were set for each “different” feature. That ensured a spread of the feature across the protected area network, rather than it being represented in a single clump (Alonso et al. 2008). In Cuba, a planning region was split into sub‐regions, each of which was included as a conservation feature, assigning it a high penalty factor and setting specific targets for each of the sub‐regions. This way reserves were replicated across the entire planning region (Halidina et al. 2004). However, these approaches limit to some degree the algorithm’s ability to spatially optimise solutions. Because of the computational costs of the two Marxan parameters, clumping and separation distance, we recommend that the user first try alternative methods such as discussed in the previous paragraph. 5 .5 S H APE ( ED GE TO A REA R ATI O ) Although Marxan does not select optimal portfolios according to specific reserve shapes, the shape of the reserve network can be influenced by the boundary length modifier and Chapter 5: Reserve Design Considerations 43 boundary cost values. The BLM acts to cluster units, so the higher the BLM, the more Marxan tries to cluster them. If the boundary cost values are altered for certain planning units, the algorithm will tend to cluster those sets which share higher cost boundaries. By modifying boundary cost, it is possible to create solutions with different degrees of fragmentation in different parts of a study region (see Box 5.1). Box 5.1: Accounting for varying degrees of fragmentation in the landscape In British Columbia, Canada, where there are a variety of open and constrained marine water bodies, Ardron (2003, 2008) used boundary cost to fine‐tune the relative clumping of hexagons in the analysis’ four Ecological Regions (inlets, passages, shelf, slope). To determine this value the edge to area ratio of each of these regions was calculated to inform an appropriate scalar. The non‐dimensional measure used was: (P2/A)0.5, where P = total perimeter of region, and A = total area of the region. Altering the boundary costs per region allowed for more fragmented solutions in areas constrained by geography, such as inlets, but encouraged more clumped solutions in open waters, such as over the continental slope. In trying to select reserves with low perimeter to area ratios, it may be expedient to first undertake the calculations and select suitable sites that qualify in a GIS, before incorporating them as features within Marxan. For example, within the Mpumalanga Biodiversity Conservation Plan (Ferrar and Lötter 2007), suitable grassland areas were selected by first removing transformed land, then selecting grassland patches with low perimeter to area ratios, and only incorporating these patches as a feature layer within the Marxan assessment. Following from that, minimum clump size, clump distance, and number of clumps were set for that feature. 5 .6 COST Every planning unit in Marxan has a cost. Marxan tries to meet all the biodiversity constraints for minimum total cost (ignoring for now design issues). Hence the cost setting can be used to favour selection of planning units in certain areas, over other areas of equal size, e.g., to favour the selection of planning units in areas of high biological integrity. Usually cost is calculated either as simple reflection of area, or as an economic cost; however there is no reason why the cost of each planning unit cannot reflect an ecological issue where high cost sites are ones we wish to avoid, all else being equal. Here we discuss some possible ecological applications of the cost variable to influence reserve design. As mentioned in Section 5.2 ‐ Connectivity, the cost surface can effectively be used to reflect corridors that connect protected areas with one another, or species with protected areas, by lowering the cost value of the planning units within these identified corridors. Several options exist to create a cost surface, such as least‐cost path analysis or friction Chapter 5: Reserve Design Considerations 44 surfaces (reduced movement through the landscape). One can use a friction surface (cost raster in Idrisi or ArcGIS) directly as a Marxan cost surface where increased friction areas, or areas unsuitable for the movement or migration of species through the landscape, are more expensive and so will be generally avoided during Marxan’s selection of planning units. In order to favour areas with high ecological integrity, planning units in a healthy (less disturbed) state can be given a lower cost (relative to their size) than planning units in unhealthy areas. The Nature Conservancy of Canada developed a cost surface (called the suitability index) by weighting between and within a number of factors using a pairwise comparison and expert evaluation (see Box 5.2) (Pryce et al., 2006). The Nature Conservancy routinely uses external threats to a site as a surrogate of cost. Threat can be a good proxy for economic cost, and if the threat cannot be abated then making threatened sites high costs will mean the chosen reserve system is less likely to be influenced by these external forces. The cost of planning units can be increased in areas that are important for economic activities, such as fishing, relative to areas that are less important for fishing, as was tested in the Irish Sea Pilot project in the UK (Lieberknecht et al. 2004). That way, it is possible to explore ways of meeting ecological targets whilst minimising impacts on ongoing human activities (see Chapter 6: Addressing Socioeconomic Objectives). Marxan can, however, only use one cost surface within each analysis. It is possible to combine different spatial surfaces into one cost layer, but care should be taken not to combine too many different themes into one layer where consideration may need to be given to transformation, scaling, standardisation, weighting, etc. If one needs to merge different layers into one cost surface, it is advisable to use a more rigorous and defensible method of layer integration, such as the use of a multi criteria analysis (MCA) methods or software (see Chapter 6: Addressing Socioeconomic Objectives). Good practice suggests keeping the costs as straight‐forward and interpretable, as possible. If data depicting the economic cost of implementing conservation measures is available, it is a good practice to use this as the cost layer as it will produce cost‐efficient solutions. The use of a new cost layer may require adjusting some species penalty factors (up or down) in order to efficiently achieve ecological targets (see Chapter 4: Addressing Ecological Objectives through the Setting of Targets and Chapter 8: Ensuring Robust Analysis). 5 .7 A D APTI VE R ESERV E N ETW ORK P LAN N I N G Reserve network design should be adaptive to changing environments and priorities. While Marxan is principally designed to develop a network based on a static “snapshot” of the way features are spatially distributed, options such as those outlined above illustrate that the tool is flexible, and thus it can be used iteratively to adapt to changing situations. Marxan can also incorporate regular data updates, and be used for re‐ analyses as the environment or priorities change. The F‐TRAC Florida Forever program is a land acquisition program that was able to update the plan every 6 months (Oeting et al. 2006). Furthermore, if a reserve system designed with a low conservation target, e.g., Chapter 5: Reserve Design Considerations 45 5% of every feature, needs to be expanded, then using a higher target with the original 5% system locked in is reasonably efficient (Stewart et al. 2007). Box 5.2: Developing a cost surface using multiple factors The Nature Conservancy of Canada wanted to include a number of ecological suitability factors into its cost surface (Pryce et al. 2006). However, they recognised that factors such as road density and land development did not have equal impacts on suitability for conservation or chance of conservation success. To overcome this limitation they developed a suitability index using a linear combination of factors thought to affect suitability. Each factor was represented by a separate term in the equation, and each term multiplied by a weighting factor. The weighting factors were obtained through a technique known as pair‐wise comparisons where expert (local knowledge and subject matter) opinion is solicited for the rank and relative importance of each term in the equation, comparing two terms at a time. The cost (suitability) was defined by the following equation: Terrestrial Suitability = A * management_status + B * land_use + C * road_density + D * future_urban_potential + E * fire_condition_class Where A, B, C, D and E are weighting factors, calculated from expert input and pairwise comparison, which collectively sum to 100%. Sub‐weights, summing to 100%, were also applied to sub‐factors within the management status, land use and fire condition classes. For example: land_use = q * % urban + r * % agriculture + s * % mine Values for each factor (or sub‐factor) are based on the percent area of that factor in the planning unit. Values for each factor are standardised prior to applying the weights according to the following equation: Standardised score = (score for that PU / highest score for all PU)*100 This standardisation has the advantage of creating equal score ranges, which are easily comparable. As a drawback, it does not account for varying data variability, and will tend to over‐emphasise factors that come with little variability. Thus, it is best applied in situations where factors have similar variabilities, or where these variabilities have also been standardised. Although the simple index used in this assessment cannot account for the many complex local situations which influence successful conservation, the study concluded that some reasonable generalities such as this were useful for assessing conservation opportunities across an entire study area. Chapter 5: Reserve Design Considerations 46 5 .8 O TH ER CH ALLEN GES 5 .8 .1 D ifficu lt e cologica l issu e s Many context‐specific design considerations arise when striving to meet ecological objectives. It is impossible to cover all eventualities, but here are three common examples: • There is sometimes strong political pressure to account for the effects of climate change, even though it is often very difficult to do so. Simple proxies can start to address some climate change considerations, though. For example, if it is expected that taxa may migrate to specific areas, such as the cooler slopes of a mountain in an area where they cannot migrate higher up the slopes, then it is possible to include the cool slopes as a conservation feature and set a target for them, or to bias the algorithm towards selection of these areas by lowering the relative cost of planning units within them. Features on the periphery of their range may become more important under a climate change scenario. Some areas may become more prone to extreme weather events and thus could have higher associated costs applied to them. In all cases, though, the use of such proxies assumes a level of scientific certainty. Good practice would suggest that these proxies be applied or emphasised (e.g., though the Species Penalty Factory) commensurate with their certainty. • Not all configurations of planning units are suited to all types of ecological objectives. For example, dividing up a planning region into regular hexagons might not be suitable to freshwater conservation plans, which could accrue greater benefit using (modelled) sub‐catchments. Some features, such as wetlands or reefs, are ideally treated as whole, functional, planning units. If they are sub‐divided by smaller planning units, however, then strategies will have to be employed to keep them together (e.g., high internal boundary costs). Be aware that units grouped together to best represent one feature will preclude grouping them together for other features. In cases where the correct choice of planning units is not clear, a sensitivity analysis on the effects of different planning unit choices is good practice, and can help eliminate options that appear to be heavily skew results as compared to other choices. Starting with a basic grid is a good way to get a sense of what baseline solutions might look like; then, other more sophisticated planning unit shapes can be explored. In addition to above (see Section 5.2 ‐ Connectivity), planning units are also discussed in Chapter 7: Assessing and Managing Data). • Integrating planning realms, such as freshwater and marine, or terrestrial and aquatic, is an ongoing area of research. The Mpumalanga Biodiversity Conservation Plan (Ferrar and Lötter 2007) combined the output from a freshwater Marxan analysis as input files (cost surface) for a finer‐scale terrestrial Marxan analysis. This was an attempt to combine both freshwater and terrestrial conservation planning into one holistic plan. Chapter 5: Reserve Design Considerations 47 5 .8 .2 Lim it a t ions of M a r x a n in a ddr e ssin g e cologica l obj e ct ive s Marxan is primarily designed to consider objectives that translate into static spatial targets. For example, Marxan is very good at achieving targets related to objectives such as representativity, or incorporating specific sites important to certain life history stages of a feature. However, persistence of a habitat or species is often influenced by ecological processes that are hard to represent spatially and are difficult to incorporate into a spatial “snapshot” analysis. Thus, spatial tools like Marxan are just one tool in the toolbox. Likewise, spatial planning is just one toolbox, and other approaches (such as economic tools) will likely be necessary to ensure overall sustainable resource use and conservation. With Marxan it is difficult to consider: • objectives for which there are no or few spatial data; • ecological objectives that are not persistent in space and/or time; • resilience; • connectivity (other than straight‐line distances, and using boundary costs as described above); and • ecological functions that are not spatially defined or persistent. In summary, no single analysis tool can address all aspects of ecology, or incorporate all kinds of ecological objectives. Using different tools (spatial and non‐spatial) as a suite can be more powerful than one at a time, where the output from one tool may help to inform the input to another (see Box 2.1). For example, habitat suitability models can produce input into Marxan when complete data coverage is not available. Likewise Marxan outputs can be used to as input options for non‐spatial analyses. For example, the size and relative protections of species in a given site can be fed into a trophic model to indicate the possible effects on the food web of a site. Or, a model of larval dispersal can shed light on network level objectives, such as connectivity and gene flow, amongst various Marxan scenarios under consideration. Chapter 5: Reserve Design Considerations 48