search

BAAQMD / InMAP-SFAB

InMAP analyses scoped to SF air basin
1 stars 0 forks source link

Bay Area ISRM grid cells #12

Open bujinb opened 1 year ago

bujinb commented 1 year ago

This issue is opened to discuss the grid cell number and configuration for ISRM grid cells. The current InMAP configuration grid is

xorig,yorig,xcell,ycell,ncols,nrows,nthik -220000.0, -16000.0, 1000.0, 1000.0, 164, 224, 1

If InMAP is run under static condition, it produces 164x224=36736 grid cells (1km by 1km, attached) at the maximum. We could do dynamic run, and get larger grid cells in some area and smaller grid cells in the cities, but that will result in lower grid cell number. How should we proceed?

Screen Shot 2023-03-29 at 9 16 24 PM
bujinb commented 1 year ago

Kubernetes is ready to go. We are working on estimating cost of single run as accurately as possible.

bujinb commented 1 year ago

@bkoo-git Can I send you the configuration file, so you can check if the grid cells are good?

bkoo-git commented 1 year ago

We could do dynamic run, and get larger grid cells in some area and smaller grid cells in the cities

@bujinb Can you tell me how you define the city areas? Do you have a shapefile that shows the proposed grid configuration?

I'm also adding @pmartien and @dholstius to this discussion.

bujinb commented 1 year ago

Hi @bkoo-git Here is an example shape file of the very first inmap run of the isrm. Configuration: -220000.0, -16000.0, 1000.0, 1000.0, 160, 220, 1

test_isrm.zip

bkoo-git commented 1 year ago

@bujinb I thought you were proposing a dynamic grid (larger grid cells in some area and smaller grid cells in the cities) and I was wondering how you would define the city areas. But the shapefile you attached is of a static grid, slightly smaller than the CMAQ 1km grid (the same origin coordinates, but you cut off 4 columns at the eastern edge and 4 rows at the northern edge)?

bujinb commented 1 year ago

@bkoo-git Hi, yes! This takes 1.3-2 hours per run, and could even be longer depending on the computing power. And one hour of isrm run is about 70cents for us right now. For dynamic grid cell, I can only tell Inmap to nest whenever the concentration or the population is more than certain threshold, and it will do the nesting itself. I can send you some configurations, but if you do the configuration yourself, and happy with the resulting grids and send it to me to run, I think that would be best for me.

Thanks Bujin

dholstius commented 1 year ago

Just some housekeeping: Should we close #7 in favor of this thread? (My vote is yes; 👍 if you agree)