SALIENCE RANKINGS AND MAPS IN OBSERVATIONAL SCIENCE

We introduce the concept of salience to record the relevance of data points, or groups of data points, with respect to how well they support a specific model. As such, salience provides an explicit connection between data and its support for a model. We propose a 6-point scale to characterize the salience of data in supporting a specific model (from low to high): no relevance, negligible, peripheral, pertinent, important, and paramount. Although salience is applied to the data, it is connected to a specific model. For example, one dataset could be “paramount” to Model A and “pertinent” to Model B. Salience also does not provide any information about the quality of the data. There is a implicit relation, however, because low quality data cannot have high salience. Further, it is possible for data to have negative salience, indicating that the data is relevant but it does not support the model. We argue that scientists generally have an implicit and uncommunicated knowledge of the critical datapoints for a model.

For fields in which the spatial distribution of salience can be visually displayed, we introduce the concept of salience maps. Salience maps communicate results and mental classifications associated with geological studies that are often neglected (e.g., cryptic faults). Salience maps provide explicit links between data and models, as well as being communication tools between current and future practitioners. We provide an example of the use of salience rankings and construction of a salience map for the Sage Hen Flat pluton in the White Mountains in eastern California. The collection of this salience data was aided by the use of digital collection methods and digital data systems. The use of salience and salience maps is a way to provide increased reliability and trustworthiness of models, facilitate communication, and allow for scientists to more effectively build off earlier workers’ data in the observational sciences.