-
Notifications
You must be signed in to change notification settings - Fork 6
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Allow FastSlowModel to take initial realised values #82
Comments
Current Realized value of the slow parameters is calculated as:
R_{t-1} = Realized value at previous timestep Optimal value is the ideal value at a time step whereas realised value is the true behaviour of the plant. In current setup, for t=0, the realized value is equal to Refer to https://pyrealm.readthedocs.io/en/latest/users/pmodel/subdaily_details/subdaily_overview.html for more details |
We just need to be able to pass those initial values in as arguments somehow. They could be three arguments ( The obvious test would be to run a model in two time chunks and it should then be identical to running it one go. |
Self Note:
Ques: |
Not quite. The acclimation window is simply used to take a sample from the fast data and the averages within that window are used to calculate So, in the scenario we want to support, someone has run a previous time series and ended up with realised daily values. We want to be able to take the last set of daily realised values |
@davidorme do we want the |
It would be neat if a user could feed a previous |
Is your feature request related to a problem? Please describe.
When a FastSlowModel is fitted, the first 'realised' values are equal to the optimal value at time$t=0$ and thereafter the memory effect is used to impose a lag. If you break up a time series into chunks, then each chunk will 'reset' to the optimal value at the start of the chunk, so there is a brief period where a memory reset happens.
Describe the solution you'd like
The FastSlowPModel could accept a dictionary of realised$V_{cmax}$ , $J_{max}$ and $\xi$ values, which are used instead of the optimal value at $t=0$ . This would allow a time series to maintain a continuous memory effect, even if the models are fitted in chunks. The class would also provide a method to optionally extract that dictionary from a fitted FastSlowPModel.
You would still need to run the time chunks in series (not in parallel), to be able to feed those outputs into the next time chunk.
The text was updated successfully, but these errors were encountered: