add different sizes for dust and sea salt

[szy21]

Purpose

co-authored-by: @charleskawczynski

To-do

Content

---

• I have read and checked the items on the review checklist.

[szy21]

This works fine in amip: https://buildkite.com/clima/climacoupler-coarse-nightly-amip/builds/237. It doesn't change the result right now. I need to update ClimaAtmos to see if this new capability improves the simulation. I would like to make a patch release after this is merged.

[szy21]

Actually this is going to be a breaking change as we add more aerosols in RRTMGP, and we need to initialize them correctly in ClimaAtmos. I'll make a patch release with the current main.

[sriharshakandala]

Does this not work for our purpose to use (compute) one statistically representative particle size for each particle type, statistically computed based on available particle size distribution for that particle type, and use that instead?

[szy21]

We talked about it offline. This gives us the most accurate optics with information about different size bins, but is slow. I can think about if we can come up with effective radius for dust and sea salt when we have an aerosol model.

[sriharshakandala]

I started a benchmarking build here: https://buildkite.com/clima/rrtmgp-clima-a100-pipeline/builds/45

[szy21]

I started a benchmarking build here: https://buildkite.com/clima/rrtmgp-clima-a100-pipeline/builds/45

I don't think the benchmark is accurate as it reads the aerosol state from the rrtmgp example. So no matter how many aerosol categories are added, only 2 type (1 and 3) are used.

[sriharshakandala]

We talked about it offline. This gives us the most accurate optics with information about different size bins, but is slow. I can think about if we can come up with effective radius for dust and sea salt when we have an aerosol model.

I am ok with this if this is a

I started a benchmarking build here: https://buildkite.com/clima/rrtmgp-clima-a100-pipeline/builds/45

I don't think the benchmark is accurate as it reads the aerosol state from the rrtmgp example. So no matter how many aerosol categories are added, only 2 type (1 and 3) are used.

Yes. This only checks if there are no performance regressions from this PR!

[sriharshakandala]

We talked about it offline. This gives us the most accurate optics with information about different size bins, but is slow. I can think about if we can come up with effective radius for dust and sea salt when we have an aerosol model.

As discussed, this use case deviates from the design intent, which expects one representative particle size for each particle type.

As discussed, we are ok with this as a temporary fix, if immediately needed.

Eventually, we should either try to get one statistically representative particle size for each particle type, roll this back, or we can redesign this so that users can choose number of particle types they would like to use and number of bins for each particle type!

[szy21]

Sounds good, thanks @sriharshakandala.