Skip to main content

Modelling consequences for nature’s benefits to people

The development and implementation of policies and practices that ensure and enhance the flow of ecosystem services to people require the inclusion of ecosystem services in decision making. There are a wide variety of contexts in which decisions are made concerning ecosystem services, and effectively including ecosystem services in these decisions requires different types of models and scenarios that can identify how social and ecological change alters the dynamics of ecosystem services and human well-being.

A variety of approaches and tools have been used to assess and model ecosystem services. They are generally classified into three categories:

  • Correlative models
  • Process-based models
  • Expert-based models

Approaches to modelling ecosystem services vary in their analytical strengths and weaknesses as well as in the time required to apply them (Table 1).

Table 1: Comparing different types of modelling methods of ecosystem services. Many models mix different types of approaches. Participatory social-ecological scenarios often combine expert-based models with other types of models.

Modelling approach

Temporal dynamics

Model type

Ease of use

Time to learn

Beneficiaries

References

Correlative models

No

Correlative

Easy

Medium

Multiple

Schulp et a. 2014

System dynamics models

Yes

Process

Hard

High

Multiple

Boumans et al. 2015

Bayesian belief netowrks

No

Expert

Easy

Medium

Multiple

Haines-Young, 2011

Fuzzy congitive maps

No

Expert

Easy

Medium

Multiple

van Vliet et al. 2011

Matrix models

No

Expert

Easy

Low

Single

Burkhard et al., 2009

Social-ecological scenario analysis

Yes

Expert +

Easy

Medium

Multiple

Oteros-Rozas et al., 2015

 
Correlative models

At the simplest level, these models are approximations of ecosystem service flows at a single point in time. Biodiversity, land use, land cover and/or discrete elements of natural capital are usually used as proxies for ecosystem services. For example, spatial data on perennial vegetation extent can be used to estimate the flow of ecosystem services such as carbon sequestration for climate regulation. Soil and broader land-cover data have also been used in correlative models for other regulating services such as erosion prevention.

Correlative models are simple and easy to apply. Their relative simplicity means that they require fewer resources and less technical expertise. This simplicity makes correlative models useful where ecosystem service data and understanding are lacking, but their extrapolations should be treated as initial assessments. However, their simplicity does make them very amenable to participatory processes.

Correlative models not based on process understanding, which means that they can dramatically fail in novel or data-poor situations. System dynamics such as socio-ecological feedbacks, complex interactions, temporal changes and the inclusion of external drivers of change are typically absent in correlative models. When these dynamics are important or expected to play a strong role, correlative models may produce inaccurate results.

 
Process-based models

Process-based models aim to describe the ecosystem functions and biophysical processes that underlie the supply of services of benefit to people. These models can estimate the flow of ecosystem services from natural capital with more realism than correlative models. Process-based models can include socio-ecological feedbacks and interactions at fine scales, and therefore are highly suitable for assessing the changes to ecosystem services from changes to external drivers under a management, policy or climate scenario. For example, hydrological process models can be used to link changes in land cover and land management to changes in the quantity of fresh water supply and the quality of freshwater.

The main strength of process-based models is that they represent a scientific understanding of key dynamics, which can enable learning and enrich decision assessment. Process-based models are designed to mechanistically represent key system dynamics, which enables them to include key ecological and social feedback processes and to evaluate alternative future management scenarios in complex situations. The main weakness of process-based models is that they require substantial knowledge and time to develop and use.

Examples of process-based models:

 
Expert-based models

Social-ecological dynamics are often complex and poorly understood. Models of social-ecological dynamics often need to integrate disparate types of data and expert knowledge in the absence of mechanistic theory or quantitative data. There is a variety of ‘soft systems’ approaches which can be used to model ecosystem services:

A significant strength of expert-based models is that they allow the relatively easy incorporation of diverse types of expert knowledge into ecosystem service models. This strength is particularly useful for work that seeks to bridge multiple knowledge systems. However, this strength comes with the weakness that expert knowledge is often partial, biased and can be incorrect, especially when applied to novel, complex or highly uncertain situations.

Examples of expert-based modelling tools:

 
Bayesian belief networks

Bayesian probabilistic models can be used to integrate expert knowledge with multiple data sources to model the flow of ecosystem services. Although they do not themselves model biophysical processes, Bayesian models integrate outputs from biophysical models (correlative and/or process-based) with probabilistic qualitative data often derived from expert knowledge about social systems. Their ability to integrate expert and stakeholder knowledge with quantitative data and models makes Bayesian models very useful for comparing alternative scenarios in situations of limited data availability and/or where there are participatory and/ or co-design requirements.

Bayesian models have been proposed as a robust way to bridge the gap between the more accurate but less transferable participatory process models, and the simple and transferable but heavily generalized correlative models.

More information on Bayesian belief networks can be found here.

 
Fuzzy cognitive maps

Fuzzy cognitive maps are similar to Bayesian belief networks because they combine an identification of causal links with probabilistic estimations of their impact. These models aim to capture the interactions among variables in the absence of detailed data. The models are typically developed from discussions with experts, then iteratively revised into a model structure and function which corresponds to shared expert knowledge. Fuzzy cognitive maps can be used to make qualitative scenarios more rigorous and to elicit models from diverse groups of people. They have similar strengths and weaknesses to Bayesian belief networks.

 
Matrix models

Matrix models are a common way of integrating expert opinion, land-cover data and other empirical data. These models estimate the capacity (i.e. ability based on ecological condition and integrity) of a landscape to supply ecosystem services. Combining maps of land cover and land cover’s contribution to ecosystem services using Geographic Information Systems (GIS) and matrices allows simple and rapid exploratory ecosystem service assessment that does not require access to or training in other ecosystem service assessment models.

Matrix models have gained popularity as a pragmatic way of quantifying spatio-temporal changes in the supply of multiple ecosystem services under scenarios and drivers of environmental change (especially in data-sparse locations), and of meeting the co-design, participatory and transdisciplinary needs inherent in ecosystem service assessments.

This approach:

  • allows the relatively cheap and rapid identification of key areas and issues,
  • enables useful discussions between ecosystem service experts and urban managers, and
  • facilitates an analysis of changes in land management or differences among particular sites.