More about EM variables in ESML

ESML uses the term predictor to refer to independent variables (model inputs) and response for dependent variables (model outputs). Some models also have intermediary variables that are calculated from predictor variables in order to produce response variables.

Predictor Variables

Two kinds of predictor variables are defined:

Time- or Space-Varying Variable

Variable that tends to change with space and time and is causally related to (i.e., "drives") the response variable.  Typically must be measured or obtained from a dataset that is spatially and/or temporally specific.  (Excludes any computed intermediate variables.)

Constant or Parameter Variable

Typically, a factor that is relatively constant over space and time and modifies a driving variable, e.g. as a scalar multiplier or exponent applied to the driving variable.  Sometimes obtainable by lookup (i.e., from measurement or estimation from a similar setting), by statistical estimation or by calibration.  (Excludes any computed intermediate variables.)

Intermediate (Computed) Variable

Variable that is mathematically computed within the model yet not considered a response variable.  If model computation takes place in stages, an Intermediate is a variable computed at an early stage and then used in later stages.  For example, a water quality model may use an equation to estimate reach-specific stream discharge from mapped catchment data, and then use discharge in the computation of water quality.  In this context, discharge is a computed intermediate; it does not have to be obtained prior to the use of the model, nor would it be considered a response variable.  Computed intermediates are reported only if they are independently meaningful variables – i.e., they hold some relevance beyond the immediate context of the EM.

Response Variables

ESML uses the term predictor to refer to independent variables (model inputs) and response for dependent variables (model outputs).  Two kinds of response variables are defined:

Computed Variable

Response variable whose values are estimated by mathematical computation, using the model.

Observed Variable

Response variable whose values are obtained by observation, not by computation using the model. The primary situations in which this occurs are when a set of independent and dependent variable values are both obtained by measurement, then used either (a) to statistically estimate a model (e.g., a regression model) or (b) to calibrate or validate a model that has been constructed based on other means.

The Variable Relationship Diagram for EM-492 / EnviroAtlas - Restorable Wetlands demonstrates predictor, intermediate, and response variable roles. Learn more about Variable Relationship Diagrams.