When a command file holds more than one variable, each specified with a unique id, then different prediction (or simulation) modes are possible, depending on the complete model specification: predictions can be made independently (`multiple' prediction), dependently (`multivariable' prediction), or variables may correspond to certain portions (categories) in the mask map or prediction location data (`stratified' prediction). The modes hold equally for prediction and simulation. When only one variable is defined, predictions (or simulations) are made for this variable at every prediction location (`simple' mode).
The decision tree gstat uses to determine the mode is shown below. The choice for a certain mode is always implicit, and is made after gstat read the command file and examined the data.
Prediction mode decision tree:
