This paper describes a software tool for: (1) assessing uncertainties in environmental data; and (2) generating realisations of uncertain data for use in uncertainty propagation analyses: the ¿Data Uncertainty Engine (DUE)¿. Data may be imported into DUE from file or from a database, and are represented in DUE as objects whose positions and attribute values may be uncertain. Objects supported by DUE include spatial vectors, spatial rasters, time-series of spatial data, simple time-series and objects that are constant in space and time. Attributes supported by DUE include continuous numerical variables (e.g. rainfall), discrete numerical variables (e.g. bird counts) and categorical variables (e.g. land-cover). Once data are imported, an uncertainty model can be developed for the positional and attribute uncertainties of environmental objects. This is currently limited to probability models, but confidence intervals and scenarios will be provided in the future. Using DUE, the spatial and temporal patterns of uncertainty (autocorrelation), as well as cross-correlations between related inputs, can be incorporated into an uncertainty analysis. Alongside expert judgement, sample data may be used to help estimate uncertainties, and to reduce the uncertainty of the simulated output by ensuring each realisation reproduces the sample data. Most importantly, DUE provides a conceptual framework for structuring an uncertainty analysis, allowing users without direct experience of uncertainty methods to develop realistic uncertainty models for their data.
- modeling error