|A Song Sparrow in marsh vegetation. A model might describe the relationship between bird abundance and shrub cover. Photo by Leonard Liu/PRBO.|
PRBO's foundation is data collected from carefully designed field studies over long time periods and organized to be easily and widely accessible. But data alone are not enough to enhance our decisions about conservation and management. To improve these decisions, we need to see patterns, deal with complexity and uncertainty, and peer into the future. Models are one of the tools that we use to do this.
But what is a model, and how do models help us conserve biodiversity in a rapidly changing world?
Dictionaries contain many definitions of "model." In the sciences, the term generally refers to a simplified representation of reality—a metaphor or abstraction of the real world. Models describe relationships among components of an ecological system. An example might be, "Bird abundance increases as shrub cover increases." By clarifying such relationships, we may gain a better understanding of how nature works. Sometimes, even relatively simple models reveal unanticipated relationships, and these prompt new questions that we can answer with new research (for example, "Why is increased bird abundance not associated with increased nesting success?").
Models also help us communicate ideas about complex ecosystems. They require us to simplify our descriptions of these systems to only the most important characteristics. They demand that vague terms be defined in clear and consistent ways. And in the case of mathematical models, they lead us to adopt quantitative definitions for each of these characteristics (for example, "What are the measurements of ‘abundance' and ‘shrub cover'?").
|Roles of Models in the Scientific Process. To see a larger version of this diagram, click here.|
Often, we use models to describe patterns that we cannot measure directly. For example, we may want to know where birds occur in a large study area, even though it's not feasible to survey the entire area. By using the data from the places we do visit to build a model, we can extrapolate to the places that we did not visit (see "Predicting in Space" on page 5). Similarly, we may want to know how long individual birds live, even though it's rarely possible to measure a bird's actual lifespan. By using data about banded birds—how frequently they are recaptured—we can develop models that estimate how long they live.
Models may be especially useful for predicting the future: given a plate of data about past relationships, sprinkled with a dash of assumptions about future conditions, what might we expect? Sometimes these models may be used to predict conditions over relatively short time periods, for example, the number of salmon that will return from the sea next year (see "Predicting Salmon Abundance," page 7). In other cases, we may use models to make projections about conditions many decades or even a century into the future. Model predictions can provide early warnings that certain management practices may be going astray, or confirm that others are likely to have the desired effects.
Models in science are of different sorts: they enter into the process of understanding nature at different points (see diagram, above). When scientists analyze data, they often use correlative models (see Glossary, page 4) that establish a relationship among variables of interest, usually by means of statistical tests. The relationship between bird abundance and shrub cover is an example of such a correlative model. We can then interpret these relationships or correlations using a mechanistic model, usually an explanation of what we think caused the relationship.
To determine if a model is "right" requires that it also be validated—the model must be tested with data. For example, we might propose that birds are more abundant in areas with shrubs because shrubs harbor concentrations of insects that attract feeding birds. We could design an experiment to validate this model, such as: "What happens to bird abundance if shrub cover is reduced by 20%, or if insect concentrations on shrubs are experimentally reduced?" This process reveals whether the causal interpretation holds up. It is through this iterative process—building models, validating them with new data, and then improving the models—that we learn how nature works.
The interpretation developed in a mechanistic model can also lead to scenario models that enable us to consider hypothetical, "what if" questions. For example, we might use climate models to make predictions about changes in shrub cover over the next century, which we could then use to make projections about how birds might respond over the same time period. This type of modeling is growing in importance as we move into an uncertain future increasingly influenced by climate change (the subject of the spring 2009 Observer).
|PRBO biologists (here, Leonard Liu) collect data on birds using diverse habitats. Photo by Jenny Erbes/PRBO.|
Models help us explore possibilities or establish probabilities, but they don't have the final say. Data provide the final say. The examples that follow illustrate the ways in which PRBO scientists use models to understand ecosystems, predict what may happen in the future, and inform conservation and management. Modeling will be increasingly important as we navigate the uncharted waters of the future. But data, and the field biologists who collect the data, will remain our foundation.