When GPEP is run multiple times for different periods, the random fields between time periods will be inconsistent, causing potential inhomgeniety in time series. This problem can happen if
The entire period is divided into several parts to balance computation load
A dataset needs real-time or near-real-time updates
This problem could be more important for variables like snow with a lot of persistence than prcp and temperature.
This task can be done by outputing the random field at the last time step, and the next model run can read those to calculate the initial random field.
When GPEP is run multiple times for different periods, the random fields between time periods will be inconsistent, causing potential inhomgeniety in time series. This problem can happen if
This problem could be more important for variables like snow with a lot of persistence than prcp and temperature.
This task can be done by outputing the random field at the last time step, and the next model run can read those to calculate the initial random field.