Questions about SR blog post (2019-04-20)

[Karl Dunkle Werner]

Thanks for writing the source-receptor blog post! I found it really helpful in fixing ideas. I have a couple of follow-up questions -- thanks for your understanding as I get up to speed with this model.

1. Is it correct to think of the SR ISRM analysis as equivalent to running the standard InMAP "run steady" myself with the same inputs? If not, how do they differ?

2. You mentioned the InMAP cloud commands run on your kubernetes server. Are there resource limits I should be aware of?

3. It looks like inmap sr start creates a new SR matrix, correct? (I was confused about the distinction between "sr" and "srpredict' and the use of SR.OutputFile, but I think I've sorted it out.) It would be great to have example code of how you created the ISRM matrix referenced in the blog.

4. I would rather estimate lost life years than deaths. Do you have any examples of people using InMAP to estimate lost years?

Best,

Karl

[Chris Tessum]

Hi Karl,

Thanks for your questions! In response:

1. Is it correct to think of the SR ISRM analysis as equivalent to running the standard InMAP "run steady" myself with the same inputs? If not, how do they differ?

They should be basically the same. The results may differ some because the different simulations may use different grid cell sizes.

 

2. You mentioned the InMAP cloud commands run on your kubernetes server. Are there resource limits I should be aware of?

Currently there are no official resource limits but the size of the cluster is limited to control costs. So there is a practical limit to the number of simulations that can be run without overwhelming the cluster (which is something like 1 at a time.) As use of these tools increases, we will work on putting policies into place, but in the meantime if anyone wants to use the cloud cluster to do a lot of simulations, contact me and we'll work something out.

 

3. It looks like inmap sr start creates a new SR matrix, correct? (I was confused about the distinction between "sr" and "srpredict' and the use of SR.OutputFile, but I think I've sorted it out.) It would be great to have example code of how you created the ISRM matrix referenced in the blog.

Yes, the 'inmap sr' group of commands is for creating a new SR matrix, and the 'inmap srpredict' command is for making predictions using an already created SR matrix. The use of 'SR.OutputFile' as an input file in some cases is an admittedly confusing legacy of a previous version of the program, which we will change at some point. We haven't worked on making it clear how to generate new SR matrices because we don't yet have the infrastructure in place to allow users to pay for the computational resources required to make one, but as above, anyone interested in doing this can contact me and we'll work something out.

 

4. I would rather estimate lost life years than deaths. Do you have any examples of people using InMAP to estimate lost years?

I haven't done that, but we do try to make it as easy as possible to use different relationships between concentrations and health impacts. After settling on a relationship that you want to use, all you need to do is change the "OutputVariable"  formulas to meet your needs (more information is here: https://inmap.run/docs/results/). Of course, if your relationship depends on information that is not already available in the model (as listed at https://inmap.run/blog/2019/04/20/sr/), then you will have more work cut out for you.

 

Chris

Best,

Karl

[Karl Dunkle Werner]

Hi Chris,

Thank you so much!

Best,

Karl