Open jgomezdans opened 2 years ago
Need to update GUI and test "DA button" with big safe ensemble Tuesday.
I have tested the DA button, but my results are very different to yours @Ma-hy
Looking into your ensemble fitting code, @MarcYin noticed that you calculate the wrong LAI value:
np.nanmean(np.transpose(obs['mean_bios_all'][:,4,:])*obs['mean_bio_scales_all'][:,4], axis=1)
the LAI should be calculated as
f.f.mean_bios_all[:, 4, :].mean(axis=0)
We need to discuss this tomorrow morning 9am. I've sent an invite through.
That’s not good. Please keep me informed how this goes this morning. I hope it’s just something trivial.
I was thinking that if you are still having trouble, then a pragmatic solution would be to apply some tolerance to fengs empirical model for each sample as a first pass. That would ensure that all candidates broadly follow that pattern.
It would be better not having to impose that, as the danger is that you are ‘just’ applying the empirical model. Which isn’t really what we want. I think it would make it difficult to write it up.
If you do do that, check that the result isn’t just exactly the same as the empirical model then.
Feng - can you give them the coefficients for your empirical model, just in case.
Lewis
Professor Lewis, NCEO/UCL
From: José Gómez-Dans @.> Sent: Tuesday, March 1, 2022 7:17:09 PM To: UCL-EO/Workshop2022 @.> Cc: Subscribed @.***> Subject: Re: [UCL-EO/Workshop2022] DA notebook (Issue #15)
I have tested the DA button, but my results are very different to yours @Ma-hyhttps://github.com/Ma-hy [image]https://user-images.githubusercontent.com/139304/156233269-f330f347-8882-46c8-bca1-18f17b5eb442.png
Looking into your ensemble fitting code, @MarcYinhttps://github.com/MarcYin noticed that you calculate the wrong LAI value:
np.nanmean(np.transpose(obs['mean_bios_all'][:,4,:])*obs['mean_bio_scales_all'][:,4], axis=1)
the LAI should be calculated as
f.f.mean_bios_all[:, 4, :].mean(axis=0)
We need to discuss this tomorrow morning 9am. I've sent an invite through.
— Reply to this email directly, view it on GitHubhttps://github.com/UCL-EO/Workshop2022/issues/15#issuecomment-1055773890, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AAOC3GUQUN3RSALVBQ4YMQTU5ZULLANCNFSM5PKEBJSA. You are receiving this because you are subscribed to this thread.Message ID: @.***>
Also, can you do plots of the parameters? We surely need to see these to be able to interpret
Lewis
Professor Lewis, NCEO/UCL
From: Philip Lewis @.> Sent: Wednesday, March 2, 2022 7:41:06 AM To: UCL-EO/Workshop2022 @.>; UCL-EO/Workshop2022 @.> Cc: Subscribed @.>; Feng Yin @.***> Subject: Re: [UCL-EO/Workshop2022] DA notebook (Issue #15)
That’s not good. Please keep me informed how this goes this morning. I hope it’s just something trivial.
I was thinking that if you are still having trouble, then a pragmatic solution would be to apply some tolerance to fengs empirical model for each sample as a first pass. That would ensure that all candidates broadly follow that pattern.
It would be better not having to impose that, as the danger is that you are ‘just’ applying the empirical model. Which isn’t really what we want. I think it would make it difficult to write it up.
If you do do that, check that the result isn’t just exactly the same as the empirical model then.
Feng - can you give them the coefficients for your empirical model, just in case.
Lewis
Professor Lewis, NCEO/UCL
From: José Gómez-Dans @.> Sent: Tuesday, March 1, 2022 7:17:09 PM To: UCL-EO/Workshop2022 @.> Cc: Subscribed @.***> Subject: Re: [UCL-EO/Workshop2022] DA notebook (Issue #15)
I have tested the DA button, but my results are very different to yours @Ma-hyhttps://github.com/Ma-hy [image]https://user-images.githubusercontent.com/139304/156233269-f330f347-8882-46c8-bca1-18f17b5eb442.png
Looking into your ensemble fitting code, @MarcYinhttps://github.com/MarcYin noticed that you calculate the wrong LAI value:
np.nanmean(np.transpose(obs['mean_bios_all'][:,4,:])*obs['mean_bio_scales_all'][:,4], axis=1)
the LAI should be calculated as
f.f.mean_bios_all[:, 4, :].mean(axis=0)
We need to discuss this tomorrow morning 9am. I've sent an invite through.
— Reply to this email directly, view it on GitHubhttps://github.com/UCL-EO/Workshop2022/issues/15#issuecomment-1055773890, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AAOC3GUQUN3RSALVBQ4YMQTU5ZULLANCNFSM5PKEBJSA. You are receiving this because you are subscribed to this thread.Message ID: @.***>
Tried assimilating correct LAI, like below
Compared with the wrong LAI calculation
Your new thing is the same as my plot above: only a small variation between 2200-2500 kg/ha. Plot them with the same axes.
Yeah, the slope is quite flat, but keeps high R value, by the way, the pixel-by-pixel assimilation shows a bit larger variation.
So what is the way forward here? It sound like you don’t have a viable DA for the maize crop. Do we have a viable workshop then?
Did you try filtering from the empirical relationship?
Lewis
Professor Lewis, NCEO/UCL
From: Hy Ma @.> Sent: Wednesday, March 2, 2022 1:59:49 PM To: UCL-EO/Workshop2022 @.> Cc: Professor Philip Lewis @.>; Comment @.> Subject: Re: [UCL-EO/Workshop2022] DA notebook (Issue #15)
[image]https://user-images.githubusercontent.com/17977287/156375772-f2b64fbd-41a2-498e-a44e-8b3521efd702.png Yeah, the slope is quite flat, but keeps high R value, by the way, the pixel-by-pixel assimilation shows a bit larger variation.
— Reply to this email directly, view it on GitHubhttps://github.com/UCL-EO/Workshop2022/issues/15#issuecomment-1056961512, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AAOC3GSXUXSEHJLFDA23MG3U55X5LANCNFSM5PKEBJSA. You are receiving this because you commented.Message ID: @.***>
I'm currently filtering by the empirical relationship and will have some results in a bit. the ensembles from Hongyuan didn't cover empirical relationship line, so have changed things a bit to get to that point.
OK, so after fitting the empirical line, and changing a more sensible set of parameters, here's something more hopeful. Results first:
I'm doing this on a small ensemble, so I selected the 5 best solutions (for the correlation matrix, I selected the best 20). In general, this looks reasonable, and I think we can maybe expect better results with bigger ensembles that allow better fitting to LAI, and possibly reducing the input parameters space (not too bothered about that).
Assimilation code is working OK for fitting LAI etc, but using the DA parameters to forward model observations doesn't produce the same as the ensemble. Likely to be a discrepancy between ensemble generation code and notebook code.
s
Mostly there, bits missing are showing the per pixel DA results. Additionally, @Ma-hy has found a new smaller set of parameters for the model, that are maybe worthwhile exploring, but we'll see in discussion.
Can you show the growth stage on the graphs? To give some context to the amax parameters
Professor Lewis, NCEO/UCL
From: José Gómez-Dans @.> Sent: Thursday, March 3, 2022 8:11:37 PM To: UCL-EO/Workshop2022 @.> Cc: Professor Philip Lewis @.>; Comment @.> Subject: Re: [UCL-EO/Workshop2022] DA notebook (Issue #15)
Mostly there, bits missing are showing the per pixel DA results. Additionally, @Ma-hyhttps://github.com/Ma-hy has found a new smaller set of parameters for the model, that are maybe worthwhile exploring, but we'll see in discussion.
— Reply to this email directly, view it on GitHubhttps://github.com/UCL-EO/Workshop2022/issues/15#issuecomment-1058440490, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AAOC3GTCGWACY7ULU3JYEETU6EMHTANCNFSM5PKEBJSA. You are receiving this because you commented.Message ID: @.***>
It's one of the drop-down parameters, called DVS. Probably should use the full names and not the abbreviations
From: Professor Philip Lewis @.> Sent: Thursday, March 3, 2022 8:55:39 PM To: UCL-EO/Workshop2022 @.> Cc: Gomez-Dans, Jose @.>; Assign @.> Subject: Re: [UCL-EO/Workshop2022] DA notebook (Issue #15)
⚠ Caution: External sender
Can you show the growth stage on the graphs? To give some context to the amax parameters
Professor Lewis, NCEO/UCL
From: José Gómez-Dans @.> Sent: Thursday, March 3, 2022 8:11:37 PM To: UCL-EO/Workshop2022 @.> Cc: Professor Philip Lewis @.>; Comment @.> Subject: Re: [UCL-EO/Workshop2022] DA notebook (Issue #15)
Mostly there, bits missing are showing the per pixel DA results. Additionally, @Ma-hyhttps://github.com/Ma-hy has found a new smaller set of parameters for the model, that are maybe worthwhile exploring, but we'll see in discussion.
— Reply to this email directly, view it on GitHubhttps://github.com/UCL-EO/Workshop2022/issues/15#issuecomment-1058440490, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AAOC3GTCGWACY7ULU3JYEETU6EMHTANCNFSM5PKEBJSA. You are receiving this because you commented.Message ID: @.***>
— Reply to this email directly, view it on GitHubhttps://github.com/UCL-EO/Workshop2022/issues/15#issuecomment-1058487842, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AABCAKCC44KMGKUG567DNZDU6ERMXANCNFSM5PKEBJSA. Triage notifications on the go with GitHub Mobile for iOShttps://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fapps.apple.com%2Fapp%2Fapple-store%2Fid1477376905%3Fct%3Dnotification-email%26mt%3D8%26pt%3D524675&data=04%7C01%7Cj.gomez-dans%40ucl.ac.uk%7C5f72866b3aa04669a3d208d9fd582d76%7C1faf88fea9984c5b93c9210a11d9a5c2%7C0%7C0%7C637819377440728776%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=KWE5EAekGdfqPti9On8kjpwkOhiGsBjBxVvyviXgbMc%3D&reserved=0 or Androidhttps://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fplay.google.com%2Fstore%2Fapps%2Fdetails%3Fid%3Dcom.github.android%26referrer%3Dutm_campaign%253Dnotification-email%2526utm_medium%253Demail%2526utm_source%253Dgithub&data=04%7C01%7Cj.gomez-dans%40ucl.ac.uk%7C5f72866b3aa04669a3d208d9fd582d76%7C1faf88fea9984c5b93c9210a11d9a5c2%7C0%7C0%7C637819377440728776%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=T%2BLXCxPzblSF70Atyddvpy39rKIbaOB1b2pvH1auW%2Fs%3D&reserved=0. You are receiving this because you were assigned.Message ID: @.***>
Looking good people. Thanks.
I’d very much like to know more about what the empirical fn filtering did you the prior parameter space. As I said, I think this is pragmatic and seems to get us a solution for the workshop. But it might have other important implications for how to approach this.
Without that constraint, it seems to be suggesting that fitting lai alone can’t get us to yield using DA. I know there are some issues with much of the crop DA work out there, but that’s quite an important issue. Or maybe this is just a special case, since the drivers are so coarse and effectively the same for all fields, but the yield vary so much.
Does the empirical constraint filtering just put the prior in a more sensible place? Or … are we just really doing no more than mimicking the empirical relationship here?
Lewis
Professor Lewis, NCEO/UCL
From: José Gómez-Dans @.***> Sent: Thursday, March 3, 2022 9:11 pm To: UCL-EO/Workshop2022 Cc: Professor Philip Lewis; Comment Subject: Re: [UCL-EO/Workshop2022] DA notebook (Issue #15)
It's one of the drop-down parameters, called DVS. Probably should use the full names and not the abbreviations
From: Professor Philip Lewis @.> Sent: Thursday, March 3, 2022 8:55:39 PM To: UCL-EO/Workshop2022 @.> Cc: Gomez-Dans, Jose @.>; Assign @.> Subject: Re: [UCL-EO/Workshop2022] DA notebook (Issue #15)
⚠ Caution: External sender
Can you show the growth stage on the graphs? To give some context to the amax parameters
Professor Lewis, NCEO/UCL
From: José Gómez-Dans @.> Sent: Thursday, March 3, 2022 8:11:37 PM To: UCL-EO/Workshop2022 @.> Cc: Professor Philip Lewis @.>; Comment @.> Subject: Re: [UCL-EO/Workshop2022] DA notebook (Issue #15)
Mostly there, bits missing are showing the per pixel DA results. Additionally, @Ma-hyhttps://github.com/Ma-hy has found a new smaller set of parameters for the model, that are maybe worthwhile exploring, but we'll see in discussion.
— Reply to this email directly, view it on GitHubhttps://github.com/UCL-EO/Workshop2022/issues/15#issuecomment-1058440490, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AAOC3GTCGWACY7ULU3JYEETU6EMHTANCNFSM5PKEBJSA. You are receiving this because you commented.Message ID: @.***>
— Reply to this email directly, view it on GitHubhttps://github.com/UCL-EO/Workshop2022/issues/15#issuecomment-1058487842, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AABCAKCC44KMGKUG567DNZDU6ERMXANCNFSM5PKEBJSA. Triage notifications on the go with GitHub Mobile for iOShttps://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fapps.apple.com%2Fapp%2Fapple-store%2Fid1477376905%3Fct%3Dnotification-email%26mt%3D8%26pt%3D524675&data=04%7C01%7Cj.gomez-dans%40ucl.ac.uk%7C5f72866b3aa04669a3d208d9fd582d76%7C1faf88fea9984c5b93c9210a11d9a5c2%7C0%7C0%7C637819377440728776%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=KWE5EAekGdfqPti9On8kjpwkOhiGsBjBxVvyviXgbMc%3D&reserved=0 or Androidhttps://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fplay.google.com%2Fstore%2Fapps%2Fdetails%3Fid%3Dcom.github.android%26referrer%3Dutm_campaign%253Dnotification-email%2526utm_medium%253Demail%2526utm_source%253Dgithub&data=04%7C01%7Cj.gomez-dans%40ucl.ac.uk%7C5f72866b3aa04669a3d208d9fd582d76%7C1faf88fea9984c5b93c9210a11d9a5c2%7C0%7C0%7C637819377440728776%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=T%2BLXCxPzblSF70Atyddvpy39rKIbaOB1b2pvH1auW%2Fs%3D&reserved=0. You are receiving this because you were assigned.Message ID: @.***>
— Reply to this email directly, view it on GitHubhttps://github.com/UCL-EO/Workshop2022/issues/15#issuecomment-1058498826, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AAOC3GRPFPWWPXG4LZD2ORDU6ETG7ANCNFSM5PKEBJSA. You are receiving this because you commented.Message ID: @.***>