Multi-day modelling

[Soumyajeet-Patra]

Hello episim developer,

The project is really great work and it is really helpful to us. while reading the mechanism of the model we understood that multi-day modeling is not supported at the moment and the data ends at the end of the day ie the simulations run the same person trajectories. Kindly could you help with the information that whether this model can be updated to support multiday modeling and if yes could you give me a lead as to what kind of changes could be done in order to support a basic version of it? I have tried to go through the whole code but I am having difficulty understanding how to incorporate this logic at a basic level.

[rakow]

It is possible to model multiple days and run trajectories based on the current weekday. This works out-of-the-box by defining different events files for each day: https://github.com/matsim-org/matsim-episim/blob/ab0d1664a2badfe734fdaaef65b5c95e59d2e905/src/main/java/org/matsim/run/modules/SnzBerlinProductionScenario.java#L180-L187

Modelling each day uniquely is not possible at the moment. It might be possible by using the current approach as starting point, but would require a separate event file for every day.

[Soumyajeet-Patra]

Hello,

Thank you for your suggestions. I tried to run the SnzBerlinProductionScenario.java but it said that a file named BerlinSnzData_daily_until20201107.csv is missing. When I am trying to run the code excluding the file I am getting an error

Error in custom provider, java.lang.IllegalArgumentException: org.opengis.referencing.NoSuchAuthorityCodeException: No authority was defined for code "GK4". Did you forget "AUTHORITY:NUMBER"?
at org.matsim.run.modules.SnzBerlinProductionScenario.scenario(SnzBerlinProductionScenario.java:123) (via modules: com.google.inject.util.Modules$OverrideModule -> org.matsim.run.modules.SnzBerlinProductionScenario)
while locating org.matsim.api.core.v01.Scenario

Kindly could tell me whether the file BerlinSnzData_daily_until20201107.csv is available anywhere publicly. Also could you tell me if I can generate seperate events file for each day does the current episim model support the logic to spreading of imfections with repect to multi day modelling or the logic is yet to be incorporated in the episim model.

[rakow]

This file and also some other you will need are currently not publicly available. However, we are currently looking into ways to make them available, but I can't promise anything at the moment.

As you see in the example you can add at most seven separate event files for:
MONDAY, TUESDAY, ..., SUNDAY

Person trajectories will be repeated every week. Anything beyond that is not yet incorporated.

[Soumyajeet-Patra]

Hi ,
could you tell me in the base parameter calibration how do you get the target_rate, target_interval, and days in the function infection_rate(). I have the infection data of a particular country. Kindly could you let me know what method you used to calculate these?

[rakow]

For this you need infection data from unrestricted spread, meaning without any interventions such as lockdown. Probably there is not too much data for this, but anyway you can try to extract it.

For example target_rate=2 and target_interval=3 means that the infection numbers double every 3 days. This could be a good starting point. I would check if this rate roughly fits your data, and if not adjust it accordingly.

[Soumyajeet-Patra]

Thank you for your reply. Please can you tell me how do you fit the curve for the lockdown data after restrictions are taking place? Also for the part without any intervention, the infection rate you have assumed is a constant but it is not a constant if I am not mistaken. So did you just assume it or have you used some kind of linear regression or time series modeling to find out the rate.

[rakow]

The assumption that the infection rate is constant is only made for a certain period of time, e.g. 20 days during early phase, without interventions and other effects such as herd immunity,

Calibrating theta corresponds to calibrating this base infection rate. Of course the actual infection rate will depend on a lot of other factors, such as:

• Lockdowns, i.e. reduced activity participation, or curfews
• Disease import from persons traveling abroad
• Contact tracing and quarantine measures
• Vaccinations

With episim you can model all of these explicitly and have an effect on the infection rate. In short, we don't define the infection rate, but it will be a result of all these interventions.
So you would have to define the interventions in your scenario to match the reality as closely as possible and calibrate these with actual infection numbers. There is unfortunately no automated way to do this.

You may look into our berlin scenario as an example:
https://github.com/matsim-org/matsim-episim/blob/ae929d29e76d3c2dc7136f65d29b1622271e0035/src/main/java/org/matsim/run/modules/SnzBerlinProductionScenario.java