**4. Land-use planning for ecosystem services and biodiversity protection in productive landscapes**

#### **4.1. Spatial optimisation of ecosystem services**

Generally, two independent strands contribute to integrated conservation planning – ecosys‐ tem-centred and species-centred prioritisation [52]. An ecosystem-centred approach prioritises efforts that increase the representation of indigenous biodiversity across the full range of environment, ecosystem, and habitat types by enhancing or protection highly modified ecosystem types (thus enhancing or protecting Environmental Representation). A speciescentred approach prioritises species based on their conservation status, or some measure of current vs potential distribution – with conservation efforts benefiting the most severely threatened species receiving the highest priority. Some frameworks also consider existing conservation efforts in prioritising new efforts. For instance, threatened species or environment types that already receive a high degree of protection may be assigned lower priority than

Clearance of indigenous vegetation for agriculture and land-use intensification has severely reduced indigenous biodiversity representation within productive lowland ecosystems (i.e. has reduced environmental representation), so that there is often little or no remnant habitat available for conservation (e.g. [53, 54]. Consequently, ecological restoration is necessary to

In many countries, clearance of indigenous vegetation has been especially severe in environ‐ ments of limited geographic extent, such as coastal and riparian habitats (e.g. [57]), or ecosys‐ tems on unusual substrates (e.g. [58, 59]). Thus areas providing very high gains in environmental representation through restoration or protection will often occupy very small sites. This leads to a right-skewed distribution where most sites provide low environmental representation gains while a few sites provide very large gains. Because environmental representation gain will often be strongly right-skewed, it may be especially vulnerable to trade-offs in multi-objective optimisations of restoration effort. This arises because high values for environmental representation gain are unlikely to co-occur with high values for ecosystem service gains [60]. This means that when ecosystem service benefits are included as criteria for deciding where to apply restoration effort, environmental representation gains will often be much lower than if it were the only criterion. The environmental representation strand of integrated conservation planning thus reveals that a focus on non-biodiversity objectives in designing restoration programmes may result in drastically lower rates of biodiversity gain

Perhaps the most important implication of integrated conservation planning for biodiversity enhancement schemes is that programmes focussed on the farm scale will likely be very inefficient at contributing to national biodiversity objectives. Not all farms will contain significant areas of highly modified environment types. Hence the potential gain in environ‐ mental representation for many farms will often be quite low. Similarly, few farms are likely to contain any threatened species, or have the potential to provide suitable habitat for threat‐ ened species. Therefore, any scheme that operates primarily by incentivising individual landowners to manage for biodiversity will result in relatively low gains in national-level conservation priorities per unit effort. By contrast, schemes focussing on the landscape scale

will be able to target resources to areas where the potential gains are highest.

ensure representation of these ecosystems in conservation networks [53, 55, 56].

those that receive little or no protection.

6 Biodiversity - The Dynamic Balance of the Planet

per unit of restoration effort.

Spatial optimisation is a powerful method to explore the potentials of a given area to improve the spatial coherence of land-use functions. It is suitable for identifying land-use configurations which optimally match with spatially varying ecosystem characteristics as well as stake-holder expectations.

Spatial optimisation models have been successfully used to address complex spatial planning problems [61–65] including forest management and timber harvest [66], agricultural issues [61, 65, 67], general issues of land-use change [68], and habitat suitability [69]. Modelling method‐ ologies range from dynamic models based on difference equations of exponential growth [66, 69] to complex models based on systems of non-linear differential equations [70].

The complexity of an optimisation model depends on the complexity of the ecosystem (number of variables, degree of non-linearity, etc.) and the spatial complexity (size of the study area, grid cell size, number of spatially interacting processes). Within land-use planning linear optimisation methods are often not applicable because of the qualitative character of the relations and the large number of variables and/or relations to be optimised. In this case, heuristic methods such as Genetic Algorithms are applied, given that there are few restrictions regarding the formulation of the variables and their relations [61].

Using spatial optimisation tools that systematically consider a range of scenarios, objectives, constraints, and stakeholder or societal preferences helps decision-makers gain insight into the full spectrum of feasible solutions. The tools also allow them to explore opportunities creatively in relation to the imposed limits. However, such use also can result in a simplified representation of options and trade-offs. The accuracy of the result of a spatial optimisation exercise depends on the quality of the input data and the complexity of the model. The more complex the model and the more spatial relationships considered, the greater the uncertainty in the optimisation. Furthermore, a relatively stable land-use pattern indicates a larger degree of freedom in terms of planning alternatives, whereas a relatively unstable land-use pattern indicates there is little room for trade-offs without significantly changing the expectations (i.e. constraints) [71].

uncertainty associated with input performance scores may lead to an overestimation of the

Prioritising Land-Use Decisions for the Optimal Delivery of Ecosystem Services and Biodiversity Protection in…

http://dx.doi.org/10.5772/58255

9

Modelling patterns of land-use change is a fundamental component of conservation planning in productive landscapes [74–78]. Land-use change models can provide a tool for capturing the essence of where land-use change is likely to take place and what is driving these patterns [75, 79] This information can then be used to assess the vulnerability of remaining indigenous

Assessment of the vulnerability of species and habitats to imminent proximate threats such as land-use change is a fundamental component of conservation management and planning [76, 77, 80]. Spatially explicit models can provide a tool for capturing the essence of where landuse change is likely to take place and what is driving these patterns [79]. This offers the opportunity to develop an understanding of how land-use change responds to different changes in policy or land-management plans. They can also be used to assess the vulnerability of remaining indigenous habitat and help identify the relative urgency of their protection.

Over the past two decades a range of models of land-use change have been developed, including process-based and statistical models [81]. Statistical models typically rely on the implicit assumption that land-use change processes are stationary. Process-based models, on the other hand, are able to deal with temporal changes in driving forces or processes. While process-based models have sound theoretical basis, statistical models can be easier to imple‐ ment [82]. As a result, most land-use change models have relied more on a statistical approach to land-use change modelling. These models include Markov, logistic function, and regression

Each land-use change model adopts different statistical techniques to capture present and future land-use patterns. These techniques are either regression or transition based. The regression-based approach is used to understand historical, current, and future land-use patterns by establishing a relationship between a wide range of environmental or socioeconomic variables [82]. The influence of these locational factors on land-use change is modelled with distance decay functions, where influence decreases with increasing distance from some feature [89]. In comparison, the transition-based models are rooted in a stochastic Markov-chain technique, where the transition probability is calculated by determining how

Most land-use change models are spatial transition models and are designed to be part of a decision support system aimed to capture present and future land-use patterns [81, 84, 85, 90]. These models have relied on user-supplied assumptions about how people actually used the land [91]. These assumptions are based on the 'maximum power principle', i.e. people will use the most economically productive land first [92]. However, recent observations of landuse conversion indicate these patterns are changing [93]. Economically viable land is becoming less available, although demand is rapidly increasing [94, 95]. This suggests current and future

optimisation benefits.

models [83–88].

**4.2. Modelling patterns of land-use change**

habitats and help identify their relative urgency for protection.

often the system moved from one state to another [83].

In recent years spatial optimisation tools have also been used to link supply of ecosystem services to land use, climate and soil information [71]. Spatial land-use optimisation techniques can help to raise awareness on trade-offs and understand how a landscape configuration could be optimally manage for ecosystem services. LUMASS, for example, is a freeware that has been specifically developed to address this situation [71]. It is a multi-objective decision making tool that can be used for spatial optimisation of ecosystem services. It uses linear programming to optimise the spatial allocation of resources to satisfy an objective (single or multiple), subject to some constraints. Objectives and constraints are specified with regard to a set of criteria representing indicators (such as sediment loss, nitrate leaching and carbon sequestration) of ecosystem services (Figure 2). The optimal allocation of (area of) land uses across the land scape is expressed by the decision variables.

**Figure 2.** Diagram of the layers incorporated into a multi-objective decision making tool (adapted from71).

The accuracy of the result of a spatial optimisation model depends critically on the quality of the input data. Performance scores used as input data for an optimisation module can be derived from quantitative process-based landscape models or from expert empirical knowl‐ edge. Little is known about the impact of uncertain input data in terms of performance scores and constraints on the produced land-use pattern. However, Herzig et al. [73] found that uncertainty associated with input performance scores may lead to an overestimation of the optimisation benefits.

#### **4.2. Modelling patterns of land-use change**

indicates there is little room for trade-offs without significantly changing the expectations (i.e.

In recent years spatial optimisation tools have also been used to link supply of ecosystem services to land use, climate and soil information [71]. Spatial land-use optimisation techniques can help to raise awareness on trade-offs and understand how a landscape configuration could be optimally manage for ecosystem services. LUMASS, for example, is a freeware that has been specifically developed to address this situation [71]. It is a multi-objective decision making tool that can be used for spatial optimisation of ecosystem services. It uses linear programming to optimise the spatial allocation of resources to satisfy an objective (single or multiple), subject to some constraints. Objectives and constraints are specified with regard to a set of criteria representing indicators (such as sediment loss, nitrate leaching and carbon sequestration) of ecosystem services (Figure 2). The optimal allocation of (area of) land uses across the land

**Figure 2.** Diagram of the layers incorporated into a multi-objective decision making tool (adapted from71).

The accuracy of the result of a spatial optimisation model depends critically on the quality of the input data. Performance scores used as input data for an optimisation module can be derived from quantitative process-based landscape models or from expert empirical knowl‐ edge. Little is known about the impact of uncertain input data in terms of performance scores and constraints on the produced land-use pattern. However, Herzig et al. [73] found that

constraints) [71].

8 Biodiversity - The Dynamic Balance of the Planet

scape is expressed by the decision variables.

Modelling patterns of land-use change is a fundamental component of conservation planning in productive landscapes [74–78]. Land-use change models can provide a tool for capturing the essence of where land-use change is likely to take place and what is driving these patterns [75, 79] This information can then be used to assess the vulnerability of remaining indigenous habitats and help identify their relative urgency for protection.

Assessment of the vulnerability of species and habitats to imminent proximate threats such as land-use change is a fundamental component of conservation management and planning [76, 77, 80]. Spatially explicit models can provide a tool for capturing the essence of where landuse change is likely to take place and what is driving these patterns [79]. This offers the opportunity to develop an understanding of how land-use change responds to different changes in policy or land-management plans. They can also be used to assess the vulnerability of remaining indigenous habitat and help identify the relative urgency of their protection.

Over the past two decades a range of models of land-use change have been developed, including process-based and statistical models [81]. Statistical models typically rely on the implicit assumption that land-use change processes are stationary. Process-based models, on the other hand, are able to deal with temporal changes in driving forces or processes. While process-based models have sound theoretical basis, statistical models can be easier to imple‐ ment [82]. As a result, most land-use change models have relied more on a statistical approach to land-use change modelling. These models include Markov, logistic function, and regression models [83–88].

Each land-use change model adopts different statistical techniques to capture present and future land-use patterns. These techniques are either regression or transition based. The regression-based approach is used to understand historical, current, and future land-use patterns by establishing a relationship between a wide range of environmental or socioeconomic variables [82]. The influence of these locational factors on land-use change is modelled with distance decay functions, where influence decreases with increasing distance from some feature [89]. In comparison, the transition-based models are rooted in a stochastic Markov-chain technique, where the transition probability is calculated by determining how often the system moved from one state to another [83].

Most land-use change models are spatial transition models and are designed to be part of a decision support system aimed to capture present and future land-use patterns [81, 84, 85, 90]. These models have relied on user-supplied assumptions about how people actually used the land [91]. These assumptions are based on the 'maximum power principle', i.e. people will use the most economically productive land first [92]. However, recent observations of landuse conversion indicate these patterns are changing [93]. Economically viable land is becoming less available, although demand is rapidly increasing [94, 95]. This suggests current and future land-use trends are no longer following traditional decision-making processes supported by the maximum power principle.

best estimates of future land-use change, vulnerability estimates based on too narrow a time range may also provide less accurate forecasts, because they are based on a small sample of

Prioritising Land-Use Decisions for the Optimal Delivery of Ecosystem Services and Biodiversity Protection in…

http://dx.doi.org/10.5772/58255

11

Systematic conservation planning is the process of identifying and configuring complemen‐ tary actions required to achieve conservation goals [73, 76]. Since the 1980s, numerous spatial approaches have been developed for identifying priority areas for conservation [103]. These approaches rely on information on the distribution of biodiversity (e.g. [104]); the distribution and effects of threatening processes or 'pressures' on biodiversity (such as pest, weeds, pollution and habitat conversion) and consequent vulnerability (the likelihood or imminence of biodiversity loss to current or impending threatening processes, in the sense of Pressey et al [51] and Wilson et al. [77]; and the effects, and costs, of potential management of pressures

A number of software packages have been developed for conservation planning and resource allocation. Among these are Marxan [107], C-Plan [108], ResNet [109] and Zonation [112, 113]. The two most widely used tools for conservation planning that have integrated ecosystem

Marxan is popular with conservation practitioners worldwide and has been applied to both terrestrial and marine ecosystems [107]. It is designed to identify a set of planning units that meets a number of targets for a minimum cost. It can be applied to a variety of conservation features considered in conservation planning and can incorporate ecological processes, site condition, or socio-political influences [110]. It addresses most objectives typically considered in conservation planning and also uses a flexible algorithm that has a variety of applications. Chan et al. [111] were the first to publish the integration of ecosystem services into Marxan. They compared differences in service provisions between conservation and development in the central coast of California. Marxan has also been used to examine the spatial congruence between biodiversity and ecosystem services in South Africa [21]. In both cases, Marxan produces a map of reserve networks that captured all biodiversity and economic targets at an optimal cost. However, the integration of ecosystem services into the Marxan framework still requires further development and the application should be used with caution. The tool lacks

Zonation [112, 113] differs from Marxan and other conservation planning approaches in that it primarily produces problem ranking rather than meeting targets at a minimum cost. Zonation can produce a priority ranking across an entire landscape using large data sets, while also identifying fine-scale prioritisation of biodiversity. Zonation provides priority ranking that balances the needs of biodiversity and competing land uses [114]. This capability, along with its ecologically based model of conservation value, makes it possible for Zonation to

Both Marxan and Zonation address the basic principles of conservation planning. However, neither account for the dynamic processes such as on-going habitat loss, site availability rates,

several features that are required for ecosystem service planning [16, 111].

incorporate the ecosystem services provision into the prioritisation process

**4.3. The application of conservation planning tools for ecosystem services**

series into their configuration are Marxan and Zonation.

conversion events.

[105, 106].

Spatial transition models are useful for land-use change modelling because of their robustness but they do not account for temporal heterogeneity. Land-use decisions are often triggered by single events such as economic crises and/or are often remote in space and time and operate at a higher hierarchical level [93]. In these cases, process-based models or models using an economic framework provide better representations of the decision making process. However, for systematic conservation planning these different models may complement each other – logistic regression models effectively identifying areas vulnerable to change, while processbased models can be used to better understand the drivers of these changes.

Most models attempt to illustrate the temporal patterns of land-use change [12, 79, 88, 96]. Quantifying the extent of future land-use change is difficult in itself, given the numerous social, political, and economic factors that drive such change [84, 97–99]. As a result, decision makers often prefer to use a Boolean map (change/no change) to illustrate the temporal and spatial patterns of land-use change. Boolean maps are easier to interpret and decipher and allow decision makers to compare scenarios of land-use change and make management decisions based on the interpolations of the various scenarios.

Because land-use systems respond to a combination of proximate (biophysical) and ultimate (socio-economic) drivers, modelling land-use change ideally requires a multidisciplinary approach [81, 100, 101]. It is often useful to incorporate a wide range of socio-economic and environmental predictors of change; particularly because the individual importance of factors in explaining patterns of land use for a past period will not necessarily reflect their ability to predict a future landscape. The predictive strength of empirical (observed) patterns of landuse change can be enhanced or diminished based on the combination of different. This also underlies the importance of validation in the modelling process [75].

Because land-use change is a dynamic threat, it is important that practitioners keep abreast of change and regularly validate the utility of their vulnerability assumptions and models [80, 102]. A common weakness of land-use modelling is the use of the same data for both calibration (making the model as consistent as possible with the data from which the parameters were estimated) and validation (assessment of the predictive power of the model) [87]. Lack of consideration of model uncertainty through rigorous validation has been shown to result in inaccurate and over-confident predictions. Validation therefore requires testing the predic‐ tions of independent data (i.e. not those used in model parameterisation) to ensure the relationships inferred by a model are robust and the predictions are reliable [75].

Spatial statistical models provide a tool for predicting where land-use change is most likely to take place [75, 79, 91]. However, models based on patterns of past change will not necessarily provide reliable predictions of future change because over time exhaustion of formerly suitable areas and changes in global markets, technology, and crops can alter both the distribution and rate of habitat conversion [77, 103]. The reliability of land-use change predictions is therefore likely to decrease as they are projected further into the future, risking misallocation of scarce conservation resources [75]. However, though recent land-use change data will provide the best estimates of future land-use change, vulnerability estimates based on too narrow a time range may also provide less accurate forecasts, because they are based on a small sample of conversion events.

#### **4.3. The application of conservation planning tools for ecosystem services**

land-use trends are no longer following traditional decision-making processes supported by

Spatial transition models are useful for land-use change modelling because of their robustness but they do not account for temporal heterogeneity. Land-use decisions are often triggered by single events such as economic crises and/or are often remote in space and time and operate at a higher hierarchical level [93]. In these cases, process-based models or models using an economic framework provide better representations of the decision making process. However, for systematic conservation planning these different models may complement each other – logistic regression models effectively identifying areas vulnerable to change, while process-

Most models attempt to illustrate the temporal patterns of land-use change [12, 79, 88, 96]. Quantifying the extent of future land-use change is difficult in itself, given the numerous social, political, and economic factors that drive such change [84, 97–99]. As a result, decision makers often prefer to use a Boolean map (change/no change) to illustrate the temporal and spatial patterns of land-use change. Boolean maps are easier to interpret and decipher and allow decision makers to compare scenarios of land-use change and make management decisions

Because land-use systems respond to a combination of proximate (biophysical) and ultimate (socio-economic) drivers, modelling land-use change ideally requires a multidisciplinary approach [81, 100, 101]. It is often useful to incorporate a wide range of socio-economic and environmental predictors of change; particularly because the individual importance of factors in explaining patterns of land use for a past period will not necessarily reflect their ability to predict a future landscape. The predictive strength of empirical (observed) patterns of landuse change can be enhanced or diminished based on the combination of different. This also

Because land-use change is a dynamic threat, it is important that practitioners keep abreast of change and regularly validate the utility of their vulnerability assumptions and models [80, 102]. A common weakness of land-use modelling is the use of the same data for both calibration (making the model as consistent as possible with the data from which the parameters were estimated) and validation (assessment of the predictive power of the model) [87]. Lack of consideration of model uncertainty through rigorous validation has been shown to result in inaccurate and over-confident predictions. Validation therefore requires testing the predic‐ tions of independent data (i.e. not those used in model parameterisation) to ensure the

Spatial statistical models provide a tool for predicting where land-use change is most likely to take place [75, 79, 91]. However, models based on patterns of past change will not necessarily provide reliable predictions of future change because over time exhaustion of formerly suitable areas and changes in global markets, technology, and crops can alter both the distribution and rate of habitat conversion [77, 103]. The reliability of land-use change predictions is therefore likely to decrease as they are projected further into the future, risking misallocation of scarce conservation resources [75]. However, though recent land-use change data will provide the

relationships inferred by a model are robust and the predictions are reliable [75].

based models can be used to better understand the drivers of these changes.

based on the interpolations of the various scenarios.

underlies the importance of validation in the modelling process [75].

the maximum power principle.

10 Biodiversity - The Dynamic Balance of the Planet

Systematic conservation planning is the process of identifying and configuring complemen‐ tary actions required to achieve conservation goals [73, 76]. Since the 1980s, numerous spatial approaches have been developed for identifying priority areas for conservation [103]. These approaches rely on information on the distribution of biodiversity (e.g. [104]); the distribution and effects of threatening processes or 'pressures' on biodiversity (such as pest, weeds, pollution and habitat conversion) and consequent vulnerability (the likelihood or imminence of biodiversity loss to current or impending threatening processes, in the sense of Pressey et al [51] and Wilson et al. [77]; and the effects, and costs, of potential management of pressures [105, 106].

A number of software packages have been developed for conservation planning and resource allocation. Among these are Marxan [107], C-Plan [108], ResNet [109] and Zonation [112, 113]. The two most widely used tools for conservation planning that have integrated ecosystem series into their configuration are Marxan and Zonation.

Marxan is popular with conservation practitioners worldwide and has been applied to both terrestrial and marine ecosystems [107]. It is designed to identify a set of planning units that meets a number of targets for a minimum cost. It can be applied to a variety of conservation features considered in conservation planning and can incorporate ecological processes, site condition, or socio-political influences [110]. It addresses most objectives typically considered in conservation planning and also uses a flexible algorithm that has a variety of applications.

Chan et al. [111] were the first to publish the integration of ecosystem services into Marxan. They compared differences in service provisions between conservation and development in the central coast of California. Marxan has also been used to examine the spatial congruence between biodiversity and ecosystem services in South Africa [21]. In both cases, Marxan produces a map of reserve networks that captured all biodiversity and economic targets at an optimal cost. However, the integration of ecosystem services into the Marxan framework still requires further development and the application should be used with caution. The tool lacks several features that are required for ecosystem service planning [16, 111].

Zonation [112, 113] differs from Marxan and other conservation planning approaches in that it primarily produces problem ranking rather than meeting targets at a minimum cost. Zonation can produce a priority ranking across an entire landscape using large data sets, while also identifying fine-scale prioritisation of biodiversity. Zonation provides priority ranking that balances the needs of biodiversity and competing land uses [114]. This capability, along with its ecologically based model of conservation value, makes it possible for Zonation to incorporate the ecosystem services provision into the prioritisation process

Both Marxan and Zonation address the basic principles of conservation planning. However, neither account for the dynamic processes such as on-going habitat loss, site availability rates, changing or unknown acquisition costs, species-specific connectivity requirements, or temporally varying distributions of features [115]. Nor do they formally incorporate multiple conservation actions such as land acquisition, restoration or easements. Furthermore, both software packages allow for detailed non-linear process descriptions and/or account for sophisticated spatial neighbourhood relationships. However, they are predominantly focused on conservation biology and hence offer only limited flexibility to configure the number and type (i.e. minimisation or maximisation) of objective functions as well as the specification of constraints. It would therefore appear they are less applicable to general land-use pattern optimisation for maximising ecosystem services.

four times as much funding and are more likely to expand opportunities for conservation [119, 120]. Given that ecosystems services projects are engaging a wider set of funders and becoming increasingly popular around the globe, there is a great need to continue to build alignment between biodiversity protection, human well-being and the delivery of ecosystem services.

Prioritising Land-Use Decisions for the Optimal Delivery of Ecosystem Services and Biodiversity Protection in…

http://dx.doi.org/10.5772/58255

13

There is growing support for understanding the economic costs and benefits of conserving ecosystems, particularly if it will help allocate scarce dollars more efficiently. Investments in biodiversity conservation may be strategically aligned to ecological services of high economic value, and vice versa. By explicitly valuing the costs and benefits associated with services, it may be possible to achieve meaningful biodiversity conservation at lower costs with greater co-benefits [111]. Cost-benefit analyses are widely used in other fields to inform policy decision making (e.g. health, safety, and transport), however conservation planning tools have been slow to integrate this into their framework [121–123]. Spatial cost-benefit analysis could prove invaluable for informing conservation planning, even when relevant data may be limited [122]. There is increased awareness of the economic value of ecosystem services (including biodi‐ versity) and quantifying these values can help decision makers best allocate scarce resources

Provision of habitat is a necessary but not sufficient condition for threatened species popula‐ tion increases in productive landscapes. Threatened species may be completely absent from the landscape so that translocations will be required for them to occupy habitat made available through restoration or preserved through protection. Further management actions may be required, such as exclusion of domestic livestock, control of invasive predators, herbivores, and weeds (e.g. 124, 125). Consequently, ecological restoration activities required to improve ecosystem services such as water quality or carbon storage may often be insufficient to enhance

Obviously, any attempt to enhance threatened species requires an understanding of the primary factors limiting threatened species populations in productive landscapes. We also need to know whether or not threatened species populations are likely to increase in response to management interventions before they are applied on large scales. Obtaining this informa‐ tion for all groups of the indigenous biota is challenging. It may be impracticable to document responses of all high priority species to the pressures imposed by productive landscapes. Similarly, it may not possible to document the response of all high priority species to man‐ agement interventions aimed at mitigating pressures. Indeed, many studies focus on demon‐ strating the effect of pressures and management interventions on community-level changes in species composition, without considering implications for high-priority species. This is understandable since rare species are often poorly captured by objective sampling designs. A shift towards studies focussed on capturing variation in high-priority species might help improve our understanding of how pressures and management interventions harm and benefit national conservation goals. However, it might be more efficient to find a way of using existing studies on changes in species composition to predict responses of high-priority species to pressures and management interventions. Functional traits provide such a means of

to various policy objectives.

biodiversity (Figure 3;60).

**5.2. Thinking beyond habitat provision**
