gcpy
gcpy copied to clipboard
geoschem
Update mass conservation tables to include Ref & Dev in the same file
Name and Institution (Required)
Name: Bob Yantosca Institution: Harvard + GCST
Confirm you have reviewed the following documentation
Describe the update
This PR updates the mass conservation table benchmark code as follows:
-
A new module
gcpy/benchmark/modules/benchmark_mass_conc_table.pyhas been created. This contains the code from themake_benchmark_mass_conservation_tablesin thebenchmark_funcs.pymodule. The code has been refactored for clarity. It now accepts arguments from both Ref and Dev versions. -
Total masses and statistics from Ref and Dev model versions are now printed in the same file.
-
TransportTracers benchmark script
run_1yr_tt_benchmark.pynow imports routinemake_benchmark_mass_conservation_tablefrombenchmark_mass_conc_table.py. -
The old make
make_benchmark_mass_conservation_tablehas now been removed
Expected changes
Prior to this change, the mass conservation table looked like this:
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Global Mass of Passive Tracer in 14.2.0-rc.1/GCClassic/TransportTracers
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Date Mass [Tg]
---------- ----------------
2019-01-01 17.6561658812290
2019-02-01 17.6561659006875
2019-03-01 17.6561658888524
2019-04-01 17.6561659034459
2019-05-01 17.6561658979651
2019-06-01 17.6561658832719
2019-07-01 17.6561658830644
2019-08-01 17.6561658772005
2019-09-01 17.6561658896849
2019-10-01 17.6561658980134
2019-11-01 17.6561658915458
2019-12-01 17.6561658810218
Summary
------------------------------
Max mass = 17.6561659034459 Tg
Min mass = 17.6561658772005 Tg
Abs diff = 26245.367 g
Pct diff = 0.0000001486 %
With the updates in this PR, it now looks like this:
===============================================================================
Global mass of PassiveTracer
Ref = gcc-14.2.0-rc.0
Dev = gcc-14.2.0-rc.2
===============================================================================
Date & Time Ref mass [Tg] Dev mass [Tg]
----------------- -------------------- --------------------
2019-01-01 00:00 17.6561658812290 17.6561658812290
2019-02-01 00:00 17.6561659006875 17.6561659006875
2019-03-01 00:00 17.6561658888524 17.6561658888524
2019-04-01 00:00 17.6561659034459 17.6561659034459
2019-05-01 00:00 17.6561658979651 17.6561658979651
2019-06-01 00:00 17.6561658832719 17.6561658832719
2019-07-01 00:00 17.6561658830644 17.6561658830644
2019-08-01 00:00 17.6561658772005 17.6561658772005
2019-09-01 00:00 17.6561658896849 17.6561658896849
2019-10-01 00:00 17.6561658980134 17.6561658980134
2019-11-01 00:00 17.6561658915458 17.6561658915458
2019-12-01 00:00 17.6561658810218 17.6561658810218
Summary Ref Dev
----------------- -------------------- --------------------
Maximum mass [Tg] 17.6561659034459 17.6561659034459
Minimum mass [Tg] 17.6561658772005 17.6561658772005
Abs diff [g] 26245.3667687623238 26245.3667687623238
% difference 0.0000001486470 0.0000001486470
Reference(s)
If this is a science update, please provide a literature citation.
Related Github Issue(s)
- This PR partially addresses #308.
@lizziel @msulprizio: Also note I expanded the date tag to date & time so that you can do mass conservation tests for shorter intervals (1 month, 1 week) with archived estarts for every 6 or 12 hours if you wish.
This is a very useful update! Is it possible to have a column for difference? This will help see if the difference goes up and down, or if there is a steady bias. We also sometimes see a jump between the first and second entries due to the restart file.
Along these lines, if it is easy to do I think we should have metrics at the bottom for mean, variance, and difference between start and stop. Something like this (I added columns for difference and % difference too).
Summary Ref Dev Difference % Difference
----------------- -------------------- -------------------- -------------------- --------------------
Maximum mass [Tg] 17.6561659034459 17.6561659034459
Minimum mass [Tg] 17.6561658772005 17.6561658772005
Abs diff [g] 26245.3667687623238 26245.3667687623238
% difference 0.0000001486470 0.0000001486470
Start mass [Tg] value here value here
End mass [Tg] value here value here
Abs diff [g] value here value here
% difference value here value here
Mean mass [Tg] value here value here
Variance [Tg] value here value here
@lizziel: I will try to implement your suggestions. It should not be that difficult to do.
@lizziel: In commit 5c8df1325 I added the extra statistics output you requested. The table now looks like:
===============================================================================
Global mass of PassiveTracer
Ref = alpha.9
Dev = PR_2072
===============================================================================
Date & Time Ref mass [Tg] Dev mass [Tg]
----------------- -------------------- --------------------
2019-01-01 00:00 17.6563096646018 17.6563096646018
2019-01-02 00:00 17.6563097334271 17.6563097334271
2019-01-03 00:00 17.6563096472464 17.6563096472464
2019-01-04 00:00 17.6563097381031 17.6563097381031
2019-01-05 00:00 17.6563096420086 17.6563096420086
2019-01-06 00:00 17.6563096769969 17.6563096769969
2019-01-07 00:00 17.6563097517587 17.6563097517587
2019-01-08 00:00 17.6563096589453 17.6563096589453
2019-01-09 00:00 17.6563096255438 17.6563096255438
2019-01-10 00:00 17.6563096669641 17.6563096669641
2019-01-11 00:00 17.6563095824931 17.6563095824931
2019-01-12 00:00 17.6563096810990 17.6563096810990
2019-01-13 00:00 17.6563097035783 17.6563097035783
2019-01-14 00:00 17.6563096224065 17.6563096224065
2019-01-15 00:00 17.6563097281873 17.6563097281873
2019-01-16 00:00 17.6563096448919 17.6563096448919
2019-01-17 00:00 17.6563096999110 17.6563096999110
2019-01-18 00:00 17.6563096911278 17.6563096911278
2019-01-19 00:00 17.6563096524387 17.6563096524387
2019-01-20 00:00 17.6563096730235 17.6563096730235
2019-01-21 00:00 17.6563096880015 17.6563096880015
2019-01-22 00:00 17.6563096931819 17.6563096931819
2019-01-23 00:00 17.6563096625323 17.6563096625323
2019-01-24 00:00 17.6563096789355 17.6563096789355
2019-01-25 00:00 17.6563096582845 17.6563096582845
2019-01-26 00:00 17.6563096609817 17.6563096609817
2019-01-27 00:00 17.6563096687468 17.6563096687468
2019-01-28 00:00 17.6563096712458 17.6563096712458
2019-01-29 00:00 17.6563096699832 17.6563096699832
2019-01-30 00:00 17.6563096648479 17.6563096648479
2019-01-31 00:00 17.6563096684741 17.6563096684741
2019-02-01 00:00 17.6563096665982 17.6563096665982
Summary Ref Dev Abs Diff % Diff
----------------- -------------------- -------------------- ------------- -------
Maximum mass [Tg] 17.6563097517587 17.6563097517587 0.000000 0.000
Minimum mass [Tg] 17.6563095824931 17.6563095824931 0.000000 0.000
Abs diff [g] 169265.6539376002911 169265.6539376002911 0.000000 0.000
% difference 0.0000009586695 0.0000009586695 0.000000 0.000
Start mass [Tg] 17.6563096646018 17.6563096646018 0.000000 0.000
End mass [Tg] 17.6563096665982 17.6563096665982 0.000000 0.000
Abs diff [g] 1996.3231068231835 1996.3231068231835 0.000000 0.000
% difference 0.0000000113066 0.0000000113066 0.000000 0.000
Mean mass [Tg] 17.6563096730177 17.6563096730177 0.000000 0.000
Variance [Tg] 0.0000000000000 0.0000000000000 0.000000 0.000
@lizziel: I added abs diff & % diff columns as such in 0c3bca7:
===============================================================================
Global mass of PassiveTracer
Ref = alpha.9
Dev = PR_2072
===============================================================================
Date & Time Ref mass [Tg] Dev mass [Tg] Abs Diff % Diff
----------------- -------------------- -------------------- ------------- -------
2019-01-01 00:00 17.6563096646018 17.6563096646018 0.000000 0.000
2019-01-02 00:00 17.6563097334271 17.6563097334271 0.000000 0.000
2019-01-03 00:00 17.6563096472464 17.6563096472464 0.000000 0.000
2019-01-04 00:00 17.6563097381031 17.6563097381031 0.000000 0.000
2019-01-05 00:00 17.6563096420086 17.6563096420086 0.000000 0.000
2019-01-06 00:00 17.6563096769969 17.6563096769969 0.000000 0.000
2019-01-07 00:00 17.6563097517587 17.6563097517587 0.000000 0.000
2019-01-08 00:00 17.6563096589453 17.6563096589453 0.000000 0.000
2019-01-09 00:00 17.6563096255438 17.6563096255438 0.000000 0.000
2019-01-10 00:00 17.6563096669641 17.6563096669641 0.000000 0.000
2019-01-11 00:00 17.6563095824931 17.6563095824931 0.000000 0.000
2019-01-12 00:00 17.6563096810990 17.6563096810990 0.000000 0.000
2019-01-13 00:00 17.6563097035783 17.6563097035783 0.000000 0.000
2019-01-14 00:00 17.6563096224065 17.6563096224065 0.000000 0.000
2019-01-15 00:00 17.6563097281873 17.6563097281873 0.000000 0.000
2019-01-16 00:00 17.6563096448919 17.6563096448919 0.000000 0.000
2019-01-17 00:00 17.6563096999110 17.6563096999110 0.000000 0.000
2019-01-18 00:00 17.6563096911278 17.6563096911278 0.000000 0.000
2019-01-19 00:00 17.6563096524387 17.6563096524387 0.000000 0.000
2019-01-20 00:00 17.6563096730235 17.6563096730235 0.000000 0.000
2019-01-21 00:00 17.6563096880015 17.6563096880015 0.000000 0.000
2019-01-22 00:00 17.6563096931819 17.6563096931819 0.000000 0.000
2019-01-23 00:00 17.6563096625323 17.6563096625323 0.000000 0.000
2019-01-24 00:00 17.6563096789355 17.6563096789355 0.000000 0.000
2019-01-25 00:00 17.6563096582845 17.6563096582845 0.000000 0.000
2019-01-26 00:00 17.6563096609817 17.6563096609817 0.000000 0.000
2019-01-27 00:00 17.6563096687468 17.6563096687468 0.000000 0.000
2019-01-28 00:00 17.6563096712458 17.6563096712458 0.000000 0.000
2019-01-29 00:00 17.6563096699832 17.6563096699832 0.000000 0.000
2019-01-30 00:00 17.6563096648479 17.6563096648479 0.000000 0.000
2019-01-31 00:00 17.6563096684741 17.6563096684741 0.000000 0.000
2019-02-01 00:00 17.6563096665982 17.6563096665982 0.000000 0.000
Summary Ref Dev Abs Diff % Diff
----------------- -------------------- -------------------- ------------- -------
Maximum mass [Tg] 17.6563097517587 17.6563097517587 0.000000 0.000
Minimum mass [Tg] 17.6563095824931 17.6563095824931 0.000000 0.000
Abs diff [g] 169265.6539376002911 169265.6539376002911 0.000000 0.000
% difference 0.0000009586695 0.0000009586695 0.000000 0.000
Start mass [Tg] 17.6563096646018 17.6563096646018 0.000000 0.000
End mass [Tg] 17.6563096665982 17.6563096665982 0.000000 0.000
Abs diff [g] 1996.3231068231835 1996.3231068231835 0.000000 0.000
% difference 0.0000000113066 0.0000000113066 0.000000 0.000
Mean mass [Tg] 17.6563096730177 17.6563096730177 0.000000 0.000
Variance [Tg] 0.0000000000000 0.0000000000000 0.000000 0.000
Thanks @yantosca. We might need to increase precision on the diff column to capture the actual differences. Does it expand automatically if diffs are non-zero?
@lizziel: I used np.var to compute the variance. Maybe I used the wrong argument to that. Let me check.
@lizziel: I've now made the abs diff in scientific notation, since those numbers may be very small. Also the variance was very small so that's why it was showing up as zero. Changing it to scientific notation fixes that. See commit 486185b.
===============================================================================
Global mass of PassiveTracer
Ref = alpha.9
Dev = PR_2072
===============================================================================
Date & Time Ref mass [Tg] Dev mass [Tg] Abs Diff % Diff
----------------- -------------------- -------------------- ------------- --------
2019-01-01 00:00 17.6563096646018 17.6563096646018 0.0000e+00 0.000
2019-01-02 00:00 17.6563097334271 17.6563097334271 0.0000e+00 0.000
2019-01-03 00:00 17.6563096472464 17.6563096472464 0.0000e+00 0.000
2019-01-04 00:00 17.6563097381031 17.6563097381031 0.0000e+00 0.000
2019-01-05 00:00 17.6563096420086 17.6563096420086 0.0000e+00 0.000
2019-01-06 00:00 17.6563096769969 17.6563096769969 0.0000e+00 0.000
2019-01-07 00:00 17.6563097517587 17.6563097517587 0.0000e+00 0.000
2019-01-08 00:00 17.6563096589453 17.6563096589453 0.0000e+00 0.000
2019-01-09 00:00 17.6563096255438 17.6563096255438 0.0000e+00 0.000
2019-01-10 00:00 17.6563096669641 17.6563096669641 0.0000e+00 0.000
2019-01-11 00:00 17.6563095824931 17.6563095824931 0.0000e+00 0.000
2019-01-12 00:00 17.6563096810990 17.6563096810990 0.0000e+00 0.000
2019-01-13 00:00 17.6563097035783 17.6563097035783 0.0000e+00 0.000
2019-01-14 00:00 17.6563096224065 17.6563096224065 0.0000e+00 0.000
2019-01-15 00:00 17.6563097281873 17.6563097281873 0.0000e+00 0.000
2019-01-16 00:00 17.6563096448919 17.6563096448919 0.0000e+00 0.000
2019-01-17 00:00 17.6563096999110 17.6563096999110 0.0000e+00 0.000
2019-01-18 00:00 17.6563096911278 17.6563096911278 0.0000e+00 0.000
2019-01-19 00:00 17.6563096524387 17.6563096524387 0.0000e+00 0.000
2019-01-20 00:00 17.6563096730235 17.6563096730235 0.0000e+00 0.000
2019-01-21 00:00 17.6563096880015 17.6563096880015 0.0000e+00 0.000
2019-01-22 00:00 17.6563096931819 17.6563096931819 0.0000e+00 0.000
2019-01-23 00:00 17.6563096625323 17.6563096625323 0.0000e+00 0.000
2019-01-24 00:00 17.6563096789355 17.6563096789355 0.0000e+00 0.000
2019-01-25 00:00 17.6563096582845 17.6563096582845 0.0000e+00 0.000
2019-01-26 00:00 17.6563096609817 17.6563096609817 0.0000e+00 0.000
2019-01-27 00:00 17.6563096687468 17.6563096687468 0.0000e+00 0.000
2019-01-28 00:00 17.6563096712458 17.6563096712458 0.0000e+00 0.000
2019-01-29 00:00 17.6563096699832 17.6563096699832 0.0000e+00 0.000
2019-01-30 00:00 17.6563096648479 17.6563096648479 0.0000e+00 0.000
2019-01-31 00:00 17.6563096684741 17.6563096684741 0.0000e+00 0.000
2019-02-01 00:00 17.6563096665982 17.6563096665982 0.0000e+00 0.000
Summary Ref Dev Abs Diff % Diff
----------------- -------------------- -------------------- ------------- --------
Maximum mass [Tg] 17.6563097517587 17.6563097517587 0.0000e+00 0.000
Minimum mass [Tg] 17.6563095824931 17.6563095824931 0.0000e+00 0.000
Abs diff [g] 169265.6539376002911 169265.6539376002911 0.0000e+00 0.000
% difference 0.0000009586695 0.0000009586695 0.0000e+00 0.000
Start mass [Tg] 17.6563096646018 17.6563096646018 0.0000e+00 0.000
End mass [Tg] 17.6563096665982 17.6563096665982 0.0000e+00 0.000
Abs diff [g] 1996.3231068231835 1996.3231068231835 0.0000e+00 0.000
% difference 0.0000000113066 0.0000000113066 0.0000e+00 0.000
Mean mass [Tg] 17.6563096730177 17.6563096730177 0.0000e+00 0.000
Variance [Tg] 1.1452659234774e-15 1.1452659234774e-15 0.0000e+00 0.000
I just tested this out with some recent 14.3.1 GCHP data courtesy of @yuanjianz. This is what I see.
Here are my thoughts on the format:
- I think the mass in Tg should have 12 digits beyond the decimal, so equivalent to whole numbers of grams
- Abs difference could be in whole grams (no decimal) and aligned right
- Both % difference and variance in all places could be scientific notation with 4 digits past the decimal for consistency
Aside: if you are wondering why there is such a huge jump after the start.... I think it's because the run uses a 2019 restart file to start in 2022.
Also note that the times printed to the file are the file times and not the times in the filename. This is a limitation that could be addressed later on.
Thanks @lizziel for the great suggestions. I think the table is looking really nice now!
===============================================================================
Global mass of PassiveTracer
Ref = gcc-14.2.0-rc.2
Dev = gchp-14.2.0-rc.2
===============================================================================
Date & Time Ref mass [Tg] Dev mass [Tg] Abs Diff % Diff
----------------- ----------------- ----------------- ------------- -------------
2019-01-01 00:00 17.656165881229 17.656280192674 1.1431e-04 6.4743e-04
2019-02-01 00:00 17.656165900687 17.656119120463 -4.6780e-05 -2.6495e-04
2019-03-01 00:00 17.656165888852 17.656119334500 -4.6554e-05 -2.6367e-04
2019-04-01 00:00 17.656165903446 17.656119237693 -4.6666e-05 -2.6430e-04
2019-05-01 00:00 17.656165897965 17.656119239163 -4.6659e-05 -2.6426e-04
2019-06-01 00:00 17.656165883272 17.656119120917 -4.6762e-05 -2.6485e-04
2019-07-01 00:00 17.656165883064 17.656119294423 -4.6589e-05 -2.6387e-04
2019-08-01 00:00 17.656165877200 17.656119153148 -4.6724e-05 -2.6463e-04
2019-09-01 00:00 17.656165889685 17.656119076444 -4.6813e-05 -2.6514e-04
2019-10-01 00:00 17.656165898013 17.656119174781 -4.6723e-05 -2.6463e-04
2019-11-01 00:00 17.656165891546 17.656119148337 -4.6743e-05 -2.6474e-04
2019-12-01 00:00 17.656165881022 17.656119245409 -4.6636e-05 -2.6413e-04
Summary Ref Dev Abs Diff % Diff
----------------- ----------------- ----------------- ------------- -------------
Maximum mass [Tg] 17.656165903446 17.656280192674 1.1429e-04 6.4730e-04
Minimum mass [Tg] 17.656165877200 17.656119076444 -4.6801e-05 -2.6507e-04
Abs diff [g] 26245 161116229 1.6109e+08 6.1378e+05
% difference 0.000000148647 0.000912523463 9.1237e-04 6.1379e+05
Start mass [Tg] 17.656165881229 17.656280192674 1.1431e-04 6.4743e-04
End mass [Tg] 17.656165881022 17.656119245409 1.1431e-04 6.4743e-04
Abs diff [g] -207 -160947264 -1.6095e+08 7.7677e+07
% difference -0.000000001174 -0.000911558172 -9.1156e-04 7.7676e+07
Mean mass [Tg] 17.656165889665 17.656132611496 -3.3278e-05 -1.8848e-04
Variance [Tg] 6.998878437967e-17 1.980023970141e-09 1.9800e-09 2.8291e+09
@lizziel: We now get the datetimes from the restart file names in commit b889d2e4348883f9a52c888c01628c9e387b8845,
@lizziel: I added a function replace_whitespace to util.py that will replace any whitespace character with a supplied character (_ is default). Then I added spaces to the version labels:
data:
ref:
gcc:
version: gcc 14.2.0 rc 0
...
gchp:
version: gchp 14.2.0 rc 0
...
dev:
gcc:
version: gcc 14.2.0 rc 2
...
gchp:
version: gchp 14.2.0 rc 2
...
The code generated a file Passive_mass.gcc_14.2.0_rc_0_vs_gcc_14.2.0_rc_2.txt that looks like this:
===============================================================================
Global mass of PassiveTracer
Ref = gcc_14.2.0_rc_0
Dev = gcc_14.2.0_rc_2
===============================================================================
Date & Time Ref mass [Tg] Dev mass [Tg] Abs Diff % Diff
----------------- ----------------- ----------------- ------------- -------------
2019-01-01 00:00 17.656165881229 17.656165881229 0.0000e+00 0.0000e+00
2019-02-01 00:00 17.656165900687 17.656165900687 0.0000e+00 0.0000e+00
2019-03-01 00:00 17.656165888852 17.656165888852 0.0000e+00 0.0000e+00
2019-04-01 00:00 17.656165903446 17.656165903446 0.0000e+00 0.0000e+00
2019-05-01 00:00 17.656165897965 17.656165897965 0.0000e+00 0.0000e+00
2019-06-01 00:00 17.656165883272 17.656165883272 0.0000e+00 0.0000e+00
2019-07-01 00:00 17.656165883064 17.656165883064 0.0000e+00 0.0000e+00
2019-08-01 00:00 17.656165877200 17.656165877200 0.0000e+00 0.0000e+00
2019-09-01 00:00 17.656165889685 17.656165889685 0.0000e+00 0.0000e+00
2019-10-01 00:00 17.656165898013 17.656165898013 0.0000e+00 0.0000e+00
2019-11-01 00:00 17.656165891546 17.656165891546 0.0000e+00 0.0000e+00
2019-12-01 00:00 17.656165881022 17.656165881022 0.0000e+00 0.0000e+00
Summary Ref Dev Abs Diff % Diff
----------------- ----------------- ----------------- ------------- -------------
Maximum mass [Tg] 17.656165903446 17.656165903446 0.0000e+00 0.0000e+00
Minimum mass [Tg] 17.656165877200 17.656165877200 0.0000e+00 0.0000e+00
Abs diff [g] 26245 26245 0.0000e+00 0.0000e+00
% difference 0.000000148647 0.000000148647 0.0000e+00 0.0000e+00
Start mass [Tg] 17.656165881229 17.656165881229 0.0000e+00 0.0000e+00
End mass [Tg] 17.656165881022 17.656165881022 0.0000e+00 0.0000e+00
Abs diff [g] -207 -207 0.0000e+00 -0.0000e+00
% difference -0.000000001174 -0.000000001174 0.0000e+00 -0.0000e+00
Mean mass [Tg] 17.656165889665 17.656165889665 0.0000e+00 0.0000e+00
Variance [Tg] 6.998878437967e-17 6.998878437967e-17 0.0000e+00 0.0000e+00
So this will make sure the generated file doesn't have spaces in it.
This looks great. I think it is good to go except that I couldn't get it to work for GCHP vs GCC tables. I'll troubleshoot that with my data and let you know what I find.
Thanks @lizziel. For what it's worth, I was able to get GCHP vs GCC output with my sample data:
===============================================================================
Global mass of PassiveTracer
Ref = gcc_14.2.0_rc_2
Dev = gchp_14.2.0_rc_2
===============================================================================
Date & Time Ref mass [Tg] Dev mass [Tg] Abs Diff % Diff
----------------- ----------------- ----------------- ------------- -------------
2019-01-01 00:00 17.656165881229 17.656280192674 1.1431e-04 6.4743e-04
2019-02-01 00:00 17.656165900687 17.656119120463 -4.6780e-05 -2.6495e-04
2019-03-01 00:00 17.656165888852 17.656119334500 -4.6554e-05 -2.6367e-04
2019-04-01 00:00 17.656165903446 17.656119237693 -4.6666e-05 -2.6430e-04
2019-05-01 00:00 17.656165897965 17.656119239163 -4.6659e-05 -2.6426e-04
2019-06-01 00:00 17.656165883272 17.656119120917 -4.6762e-05 -2.6485e-04
2019-07-01 00:00 17.656165883064 17.656119294423 -4.6589e-05 -2.6387e-04
2019-08-01 00:00 17.656165877200 17.656119153148 -4.6724e-05 -2.6463e-04
2019-09-01 00:00 17.656165889685 17.656119076444 -4.6813e-05 -2.6514e-04
2019-10-01 00:00 17.656165898013 17.656119174781 -4.6723e-05 -2.6463e-04
2019-11-01 00:00 17.656165891546 17.656119148337 -4.6743e-05 -2.6474e-04
2019-12-01 00:00 17.656165881022 17.656119245409 -4.6636e-05 -2.6413e-04
Summary Ref Dev Abs Diff % Diff
----------------- ----------------- ----------------- ------------- -------------
Maximum mass [Tg] 17.656165903446 17.656280192674 1.1429e-04 6.4730e-04
Minimum mass [Tg] 17.656165877200 17.656119076444 -4.6801e-05 -2.6507e-04
Abs diff [g] 26245 161116229 1.6109e+08 6.1378e+05
% difference 0.000000148647 0.000912523463 9.1237e-04 6.1379e+05
Start mass [Tg] 17.656165881229 17.656280192674 1.1431e-04 6.4743e-04
End mass [Tg] 17.656165881022 17.656119245409 1.1431e-04 6.4743e-04
Abs diff [g] -207 -160947264 -1.6095e+08 7.7677e+07
% difference -0.000000001174 -0.000911558172 -9.1156e-04 7.7676e+07
Mean mass [Tg] 17.656165889665 17.656132611496 -3.3278e-05 -1.8848e-04
Variance [Tg] 6.998878437967e-17 1.980023970141e-09 1.9800e-09 2.8291e+09
I tried again to create the table comparing GC-Classic 4x5 and GCHP C180 output. I get this warning and no table is produced:
PerformanceWarning: \
Slicing is producing a large chunk. To accept the large
chunk and silence this warning, set the option
>>> with dask.config.set(**{'array.slicing.split_large_chunks': False}):
... array[indexer]
To avoid creating the large chunks, set the option
>>> with dask.config.set(**{'array.slicing.split_large_chunks': True}):
... array[indexer]
value = value[(slice(None),) * axis + (subkey,)]
I am able to create the separate ref and dev tables using the GCPy dev branch so this must be something new in this branch.
@lizziel: This must be an under-the-hood xarray/dask issue due to the large size of the data. I only tested with c24 and that works OK. I"ll see if there is a published workaround.
@lizziel: I pushed a fix in commit df1e984. Please grab this update and let me know if it works. Thanks!
Using the updated code results in the program being killed, even when I up the memory 160G.
@lizziel found that the problem may be due to reading all files up-front, which takes too much memory. We can try looping over each file and then purging the memory afterwards.
@lizziel: I modified the code so that we now loop over individual files (and delete objects afterwards to force garbage collection) in commit af8ef64. Try it again and let me know if it works.