Developing Conceptual Models of Relevant
Ecosystem Components

A conceptual model is a visual or narrative summary that describes the important components of the ecosystem and the interactions among them. Development of a conceptual model helps in understanding how the diverse components of a monitoring program interact, and promotes integration and communication among scientists and managers from different disciplines. Conceptual model diagrams often take the form of a "boxes and arrows" diagram, whereby mutually exclusive components are shown in boxes and interactions among the components are shown with arrows, but many conceptual models include tables, matrices, sentences or paragraphs to summarize and communicate our understanding of the system.

The next section is condensed from a more detailed paper (with appendices) on guidelines for developing and preparing conceptual models.

Conceptual models are important throughout all phases of development of a monitoring program. Early in the process, simple conceptual models provide a framework that relates information in discussions and literature reviews to a broader context - it is a structure to organize information. Learning that accompanies the design, construction, and revision of the models contributes to a shared understanding of system dynamics and appreciation of the diversity of information needed to identify an appropriate suite of ecosystem indicators.

Well designed conceptual models will:

• Formalize current understanding of system processes and dynamics
• Identify linkages of processes across disciplinary boundaries
• Identify the bounds and scope of the system of interest
• Contribute to communication:
  • Among scientists and program staff
  • Between scientists and managers
  • With the general public.

These roles are important throughout the life of a monitoring program. Once the program is underway, proper interpretation of indicators is greatly facilitated by sound and defensible linkages between the indicator and the ecological function or critical resource it is intended to represent (Kurtz et al. 2001). These key linkages should be explicit in conceptual models and their articulation is essential to justifying and interpreting ecological measurements.

Conceptual models can take the form of any combination of narratives, tables, matrices of factors, or box-and-arrow diagrams. Jorgensen (1988) discusses 10 kinds of models and evaluates their advantages and disadvantages. Most monitoring programs will use a combination of these forms, and it may occasionally be useful to combine several forms in the same figure.

Tables and matrices provide a convenient means to summarize large quantities of information, including interactions between components. However, many people find it difficult to comprehend how a system works from tabulated data, especially where the spatial context is significant.

Diagrams are usually necessary to clearly communicate linkages between systems or system components. Most monitoring programs develop a set of conceptual models that consist of diagrams and accompanying narratives. Narratives describe the diagrams, justify functional relationships in figures, and cite sources of information and data on which the models are based.

The process of constructing system diagrams almost always identifies inadequately understood or controversial model components. There isn't a single correct conceptual model, and it can be insightful to explore alternative ways to represent the system. These different representations of the system can help articulate important, and often exclusive, hypotheses about drivers, stressors, or interactions that are central to understanding how the system operates. These alternative hypotheses can form the basis of an effective adaptive management program, and it will likely be worthwhile to make the extra effort to clearly document and archive alternatives that arise during the process of model construction. Workshops to construct conceptual models are brainstorming sessions, and they provide an important opportunity to explore alternative ways to compress a complex system into a small set of variables and functions.

Most ecological systems are complex and management decisions are based on ecological, social, political, and economic considerations. To accommodate the full range of considerations, a set of models with different spatial domains and relevant subsystems will be necessary. Thus you can anticipate the need to construct different models that vary in scope, detail, spatial extent, relevant time frame, and focus. For realistic systems, it probably will not be particularly insightful or rewarding to attempt to construct a single model with all important components and interactions. An all-encompassing model will be too complex for most people to understand.

While the monitoring program does not intend to develop quantitative ecosystem models or dictate management policy, constructing a set of realistic, focused conceptual models is an important starting point for designing effective monitoring programs and for evaluating effective management policies. Monitoring programs founded on a solid conceptual model are more likely to identify key processes and indicators, and thereby contribute significantly to Parks management. The central role of models (both conceptual and quantitative) is well illustrated in the Applied Science Strategy adopted by the South Florida Ecosystem Restoration Working Group.

Developing Conceptual Models

In many cases it will be difficult to create even a single conceptual model, and the more complex the system is, the more difficult it will be to reach consensus on the elements to be included, the key interactions between elements, and the response of the system to drivers and stressors. It may require a multiple meetings to obtain general agreement on model structure and content. Keep the end in mind - you want to develop a suite of models that address the time and spatial scales of interest, at an appropriate level of detail.

Control and stressor models

Depending on the intended use of the conceptual model, two fundamentally different model structures have been used by I & M Networks and other agencies. A control model is a conceptualism of the actual controls, feedback, and interactions responsible for system dynamics. A control model therefore needs to represent, in a mechanistic way, the key processes, interactions, and feedbacks. Quantitative ecosystem simulation models are control models, and they vary in complexity from relatively simple to highly complex. Most groups begin by constructing a set of control models since this is the way we typically think about how systems operate. For a particular system (e.g., Park or other land) control models are typically hierarchical, with a top level, highly aggregated model and more detailed models of subsystems. In quantitative simulation models, the subsystems are usually functional units (e.g., soils, plant, fire, etc.) that overlap in space, whereas conceptual models often first decompose a larger system into more-or-less spatially distinct vegetation or habitat types. Jackson et al. (2000) describe the process of creating simple simulation models.

Stressor models are designed to articulate the relationships between stressors, ecosystem components, effects, and (sometimes) indicators. Stressor models normally do not represent feedbacks and they include only those system components that are most pertinent to the monitoring program. The intent of a stressor model is to illustrate sources of stress, ecological responses, and system attributes of most interest. These models are founded on known or hypothesized ecological relationships, frequently derived from control models, but they do not attempt a mechanistic representation of the system . The Everglades restoration program produced a comprehensive set of stressor models, and they have excellent documentation on how the models contribute to their overall management strategy (e.g., Gentile et al. 2001). The Greater Yellowstone and Northeast Coastal and Barrier Networks have developed sets of stressor models to guide their monitoring programs.

It may be necessary to develop both kinds of models, at least for some subsystems or habitats. Control models present a more complete and accurate picture of system components and their interactions. Stressor models are likely to more clearly communicate the direct linkages between stressors, ecological responses, and indicators. The appendices to Everglades Restoration Plan include a set of well constructed and documented stressor models; some of these are reproduced in Appendix IV of the full document.

Steps in Constructing Conceptual Models

A systematic program that leads to a set of conceptual models will include the following tasks. These tasks are described in more detail in the documents listed near the bottom of this website.

1. Clearly state the goals of the conceptual models.
2. Identify bounds of the system of interest.
3. Identify key model components, subsystems, and interactions.
4. Develop control models of key systems and subsystems.
5. Identify natural and anthropogenic stressors
6. Describe relationships of stressors, ecological factors, and responses.
7. Articulate key questions or alternative approaches.
8. Identify inclusive list of indicators.
9. (Prioritize indicators - a separate process)
10. Review, revise, refine models.

These steps appear in a sequential list, but it will be necessary to at least partially address the goals of some tasks simultaneously. For example, the construction of control models (steps 3 & 4) must include substantial discussion and consideration of stressors and relationships between stressors and ecological functions (steps 5 & 6).

Execution and Network Experiences

Networks and prototypes have employed a wide variety of processes to develop conceptual models and the resulting models reflect this diversity. Here are some general observations from Network’s experiences:

• It is very useful to have a general (high level) conceptual model to focus groups on linkages between submodels and to encourage model builders to conform to a common model structure.
• Hierarchical sets of models work well. At intermediate levels, submodels most commonly focus on vegetation types. The lowest-level models may focus on species, soils, or nutrients.
• It can be difficult to include animal species or animal communities in ecosystem models. Separate models may be required for a particular species or community.
• Models that address different scales are insightful, even when they focus on the same process or variables, but at different scales.
• It is very time-consuming to build useful conceptual models. Engage collaborators with appropriate disciplinary expertise as early as possible and allow time for repeated revision.
• There is a large return on investment in documenting the ecological theory that underpins a modeling approach. The underlying theory supports use of a common approach and shared vision of system processes and linkages. The NCPN report (currently being revised) is an excellent example.
• At the lowest levels, models must include sufficient detail to link indicators to ecological processes and, where possible, to management actions. Insufficiently detailed models have limited utility. It is a substantial challenge to construct a model with just the right amount of detail, and to decide when to split a model into separate submodels to avoid an overly-complicated model.
• Provide definitions of key terms and phrases. Syntax is important.

Greater Yellowstone Network - is using the I&M program as an opportunity to review and integrate a variety of NR programs. Up to July 2003, they have developed a comprehensive set of control and stressor models, and a few hybrids. The models operate on a variety of scales (e.g., they include a dry timberland model as well as a Lake Bob model).

Northern Colorado Plateau Network - report has an excellent discussion of underlying ecosystem theory. They have adopted state and transition models as a structural framework for representing dynamics of many systems. In conversation, they noted that insufficient detail in early models limited their usefulness.

Mediterranean Coast Network - Developed an initial set of Everglades-type stressor models, but had difficulties adequately incorporating animal communities. The Network is currently developing energy flow models to better represent trophic relationships.

Cape Cod - Implementation of stressor models and tables. Excellent early work on conceptual foundation of these models (Roman and Barrett 1997).

Developing Conceptual Models for Ecological Monitoring Programs

Documents on conceptual models and the role of conceptual models in monitoring programs. Also see the example below, especially the special issue of Wetlands.
Gross, J.E. 2003.  Developing conceptual models for monitoring programs. (pdf, 650 KB, DRAFT)
Gross, J.E. 2003.  Developing conceptual models.  Appendix IV. Figures. (pdf, 880 KB, DRAFT)

IAN Newsletters.  Excellent newsletters from U. MD Center for Environmental Studies  (link to UMCES site)
Lookingbill, T., et al. 2007.  Conceptual models as hypotheses in monitoring urban landscapes.  (pdf, 240 KB)

Roman and Barrett 1999. Conceptual framework for the development of long-term monitoring protocols at Cape Cod National Seashore.  (pdf, 900 KB)
Plumb, G. 2003.  Really useful conceptual models.  Paper presented at the Greater Yellowstone workshop. (pdf, 820 KB)

Haefner, J.W. 1996.   Chapter 3. Qualitative model formulation.  Excellent; see section 3.7 for strategies to simplify models. (pdf, 1.4 MB)
Maddox, D. et al. 1999.  Evaluating management success: Using ecological models to ask the right monitoring questions.   Great consideration of conceptual models in the context of monitoring. (pdf, 1.7 MB)

Grant, W.E., et al. 1997.  Chapter 3 - Conceptual model formulation.  (pdf, 885 KB)
Jorgensen, S.E. 1988.  Conceptual models.  (pdf, 1.7 MB)


Really good I&M Network reports
The state-of-the art in conceptual models is rapidly evolving, particulary the use of the IAN toolkit and development of park-specific models. This list does not include some new, outstanding examples and reports. I suggest you call me (John Gross, (970) 267-2111) to discuss what you´re looking for. If you need to obtain access to a report, contact us or the network coordinator and we´ll work something out.

• Greater Yellowstone Network Phase 2 Very complete set of models across scales and levels of aggregation; use of different model structures
  
• Southwest  Alaska Network Phase 1 Simple, clear, and attractive models. Strong resemblance to the Gulf of Alaska Ecosystem Monitoring approach
  
• Northern Colorado Plateau Network Phase 2 See for Chapin top-level model. Adopted state-and-transition models as basic structure. Very good consideration of scientific underpinning
  
• Northeast Coastal & Barrier Network Phase 2 Excellent examples of effective driver-stressor type models.

Presentations on Conceptual Modeling
Developing conceptual models for I&M Networks. John Gross, September 2005. Most up-to-date and complete overview (5.9 MB, .ppt)
Conceptual models for I&M Networks. John Gross, January 2005. PowerPoint format (3.9 MB)
Developing conceptual ecosystem models for long-term monitoring. Lisa Thomas, 2002. PowerPoint file (550 KB)

Conceptual models: what are they and how do we use them to design monitoring programs? Lisa Thomas, 2001. PowerPoint file (311 KB)
The use of conceptual models in designing and implementing long-term ecological monitoring. Lisa Thomas, 2001. MS-Word (160 KB)

Good examples - models, and/or programs using conceptual models (in no particular order)
Univ. of Maryland Center for Environmental Studies, Integration and Application Network.  Comprehensive, state-of-the-art site for science communication, including use of conceptual diagrams and how to use them to reach different audiences. Developed and provide a free library to greatly facilitate production of professional-quality diagrams. http://ian.umces.edu/

Strategic Plan for the U.S. Climate Change Science Program..  This well-written document makes extensive use of very thoughtful and informative conceptual models. Many models in the report are broad-scale and they will likely be useful to illustrate drivers and impacts of climate change that are relevant to ecological monitoring programs. The IPCC reports also include many climate-specific models that can be useful to monitoring programs. http://www.climatescience.gov/Library/stratplan2003/default.htm

Wetlands, volume 25, issue 4 (December 2005) is coprised of a set of 14 papers on conceptual models for the Everglades and other subtropical (mostly wetland) systems. This is a very good general reference on conceptual models.

Mark E. Miller. 2005. The structure and functioning of dryland ecosystems. Conceptual models to inform the vital-sign selection process. Link to USGS site, from which you can download the very large file.

Mike L. Scott et al. 2005. The structure and functioning of riparian and aquatic ecosystems of the Colorado Plateau. Conceptual models to inform monitoring. (pdf, 3.3 MB).

Gulf of Alaska Ecosystem Monitoring (GEM) Program. Very nice treatment of conceptual models in chapters 2 & 8. Particularly good for coastal parks/networks. Update Feb. 2008: Could not locate the document on the web. Contact John Gross at NPS for a copy.

The Western Port Project (Victoria, Australia) This reports on a study commissioned specifically to produce conceptual models of a large marine bay. An excellent report with professional "picture" models - released October 31, 2003. http://www.coastal.crc.org.au/Publications/WesternPort.html

Healthy Waterways Conceptual Models. About 50 aesthetic models of Australian rivers, creeks, bay, and estuaries. Update February 2008: The models seem to have mostly disappeared from the site. See the Background documents for an example that remains on the site. http://www.ehmp.org/index.html

USDA NRCS state and transition models. ** see areas SD2, SD3, WP3 - only some sites have embedded models. http://www.nm.nrcs.usda.gov/technical/fotg/section-2/ESD.html


Software U. MD Adobe Illustrator diagram library.   *Very useful* library of symbols, landscapes, backdrops for building conceptual diagrams. These are great. Follow the link to Symbols Library to download symbols, and check out the very extensive set of images that are also available. This site also has an excellent examples of science communication. http://ian.umces.edu/

Most people construct diagrams in PowerPoint, which isn’t all that well suited to the task (but most of us already know how to use the program). If you use PowerPoint, it’s generally easier to use "connectors" rather than lines to connect the parts of your model. With connectors, the lines stay attached to the objects as you move them around. Other programs are better suited to constructing models, but there’s likely to be a learning curve and, more importantly, you may be the only one with the software and the knowledge necessary to modify the diagrams.

Vensim.   Downloadable software for building dynamic models or creating flow diagrams. Excellent software for constructing simple dynamic models, but a steep learning curve if all you want are diagrams. http://www.vensim.com