Fix PRAIN=precip_total_tend in P3?

[PeterCaldwell]

I think precip_total_tend (known as PRAIN in pbuf and output) is supposed to be rain + snow production (i.e. the things that create rain without things like evaporation that deplete rain). It is mistakenly(?) labeled "Total precipitation (rain + snow)" in several places in the code.

I think it is only used as a diagnostic output. I also think it is defined incorrectly in P3: https://github.com/E3SM-Project/scream/blob/4575d088af4d87cf44f4b7fbcf8eae5e965d5024/components/eam/src/physics/cam/micro_p3.F90#L862 . qccol is collection of qc by ice - I'm not sure why it has such an archaic name. This equation looks to me to be ignoring most sources of ice (like vapor deposition, etc). Is precip_total_tend supposed to just be conversion from qc to qr+qi? Why would anyone want that?

Because PRAIN/precip_total_tend seems to just be used for output and has a definition that I can't understand even when I look at its source code, I suggest just deleting it. Does anyone know why we shouldn't?

@AaronDonahue , @hassanbeydoun , @JS-WRF-SBM - do you have any opinion? Have we discussed this before?

[JS-WRF-SBM]

@PeterCaldwell

We discussed that briefly last week.

It is a diagnostic term which diagnose the amount of precipitate mass per time step [kg/kg/s] to be used in aerosols/gas wet scavenging. Since P3 doesn't explicitly differentiate between ice and snow, the question is what should we include as diagnostic for precipitate ice/snow?
In the case of the low-res P3 version for E3SMv2, I decided to introduce the total ice mass created in the current time step (i.e, d(qi)_dt), given it is a positive value (e.g., source, rather then sink). Since the low-res E3SMv2 is fully coupled with MAM4/Mozart, I need to keep this term and make sure it is realistic. Using this choice I can scale it down in case it is too strong, or alternatively refine it by the size of ice/snow (ongoing work).

Not sure what are the plans for SCREAMv1 and forward in terms of aerosol process explicit coupling.

[hassanbeydoun]

@JS-WRF-SBM, in your implementation did you include all positive rain AND ice tendencies?

[PeterCaldwell]

Ah yes, thanks Koby. Ok, if your version needs this quantity, we will definitely keep it to make switching to our version easier if/when you do it later. We should change our definition of PRAIN to match yours. But is it logically consistent to have the liquid terms be just the net production without any depletion tendencies but for the ice contribution to be the net of sources and sinks? Wouldn't it make more sense to explicitly sum up all the source terms for ice? I think wet scavenging should occur whenever you produce precip from cloud water = all sources of qr or qi arising solely from the qc reservoir. Is this correct?

[JS-WRF-SBM]

@hassanbeydoun @PeterCaldwell

I think wet scavenging can happen once the precipitate specie is large enough to facilitate a swept volume to collect the other inertia-like particle suspended in an air volume. If the collected particle is too small, there will be no collection. If this idea is accurate enough, there is little relevancy to where the swept particle (collector) originate from, and all it matters is it's size (or relative size to the collected particle) and terminal velocity, both of which are proportional to hydrometeor mass.

This is why I think we should consider the whole qr and qi balance at a particular time step, given that it results in positive terms since we can not introduce negative precip_total_tend.

Other approach would be: lets get the maximum amount of generated precipitate per time step by taking into account only source terms of qr and qi, so we can scale it down in case aerosol conc. becomes too high. This can provide larger degree of freedom for the aerosol-MP coupling.

[hassanbeydoun]

@JS-WRF-SBM, why not use the qi and qr sedimentation fluxes for scavenging? I'm still unclear on why scavenging wants information on process rates.

[JS-WRF-SBM]

@hassanbeydoun

I haven't dig into that, but I would have argue that describing a collection process between tracers (no-inertia) which are transported with the air velocity and a sedimentation mass flux is more complex, compared to just using the mass content of both species. Specifically, since aerosol/gaseous-phase are also a dynamic reservoir with sources and sinks, process rates make sense in prognoses tendencies, rather then absolute mass content.

[PeterCaldwell]

My main confusion is why you think the net tendency is appropriate. Suppose I have a cell with strong accretion in one area balanced by strong evaporation in another. The net tendency will be near zero, but wet deposition should be happening in the accretion area. That's why I think it would be more rational to sum the source terms rather than use the net tendency.

Also, I agree that small droplets/crystals will flow around rather precipitating drops rather than be collected, but I don't see how this is relevant to any of the parameterization strategies we've discussed so far. In particular, I'm not sure whether you're bringing up small droplets to argue against my assertion that we only care about mass starting in the qc category or that we should use net rather than just source terms in wet deposition... but I don't see how small drops is relevant to either argument. I was expecting you to argue that we should be using the rates of processes removing nc rather than qc, or maybe use the drop-size distribution...

[JS-WRF-SBM]

My main confusion is why you think the net tendency is appropriate. Suppose I have a cell with strong accretion in one area balanced by strong evaporation in another. The net tendency will be near zero, but wet deposition should be happening in the accretion area. That's why I think it would be more rational to sum the source terms rather than use the net tendency.

Ideally, if the aerosol/gas process were fully embedded in the MP code, during rain evaporation we should have release some dissolved aerosol/gaseous phases into the free atmosphere, so we would have wet-scavenging for the accretion and replenish of aerosol during evaporation (likely in different magnitudes), right? Let me ask you a comparable question -- why shouldn't we consider the melting of ice which is a sink of qi and source of qr?
When I tried to answer this question, I do not see any reason why we shouldn't take sink terms as well.

Having said that, for low-res, I agree (as mentioned in my previous reply) that using only the sources is possible as an approach of 'maximum-scavenging'. :-)

Also, I agree that small droplets/crystals will flow around rather precipitating drops rather than be collected, but I don't see how this is relevant to any of the parameterization strategies we've discussed so far. In particular, I'm not sure whether you're bringing up small droplets to argue against my assertion that we only care about mass starting in the qc category or that we should use net rather than just source terms in wet deposition... but I don't see how small drops is relevant to either argument. I was expecting you to argue that we should be using the rates of processes removing nc rather than qc, or maybe use the drop-size distribution...

Here I just meant that both particles (collector/collected) needs to be large enough for the scavenging to happen. We have the possibility to filter the size of the ice particles: from small spherical ice to dry/rimed snow, so we can limit the mass contribution for scavenging, but that was a comment served as a food for thought. It probably lead for more confusion.

[PeterCaldwell]

why shouldn't we consider the melting of ice which is a sink of qi and source of qr?

Isn't scavenging is about 2 particles fusing into one, causing a corresponding reduction of in-cloud aerosol? Thus an ice crystal melting doesn't change total particle number so shouldn't scavenge. That said, P3 has that nmltratio=0.3 scaling factor on the number of ice crystals becoming rain drops to account for small drops evaporating quickly... It seems like that should be a source of interstitial aerosol, but couldn't be described as scavenging...

We have the possibility to filter the size of the ice particles: from small spherical ice to dry/rimed snow, so we can limit the mass contribution for scavenging

Ok, that makes sense. Yes, using the size distribution to modify the wet deposition rate sounds like a future research project.

[JS-WRF-SBM]

Isn't scavenging is about 2 particles fusing into one, causing a corresponding reduction of in-cloud aerosol? Thus an ice crystal melting doesn't change total particle number so shouldn't scavenge. That said, P3 has that nmltratio=0.3 scaling factor on the number of ice crystals becoming rain drops to account for small drops evaporating quickly... It seems like that should be a source of interstitial aerosol, but couldn't be described as scavenging...

For deep convective precipitation, there are in-cloud scavenging resulted from aerosol nucleation, and in- / below- cloud scavenging based on collection process of un-activate aerosols. Both applicable for different modal size of the original aerosol. My understanding is that the PRAIN is used for the latter mechanism, since removing the aerosol number/mass after nucleation (in CLUBB) is straight forward. To put it simple, this is a cleaning mechanism of the atmosphere when there is precipitation. That is, we attempt to parameterize the interaction between dry aerosols of various sizes and precipitate products in two completely different schemes by diagnosing the precipitate mass per time step.

Ice particles primary nucleation in E3SM are coupled to the aerosol scheme (either through Lie & Penner or the CNT schemes), and thus when ice melts there should be also removal of the originated aerosol-IN.