Closed teixeirak closed 3 years ago
Note's on climate responses from the original papers :
Importantly, the sparse stands cover a different time period from intermediate and dense:
I suspect that this is what's behind the funny DBH response of PIPO, and could also be affecting climate.
@ValentineHerr , we should drop the sparse stands from the analysis.
Unfortunately, it looks like we have nothing indicating the stand density in our formatted data file, which means we'll have go back to the original, probably modifying @biancaglez 's scripts.
If that's complicated, another (probably easier) way to handle this would be to throw out trees starting after ~1960 or so (chronologies start ~1970, but presumably some trees go back farther).
Tala should be able to indicate which cores came from which stands for PIPO. Maybe then you can just have separate DBH model fits for stands at different densities. I can try to get on a zoom call with Tala, if you want me to help figure that out. I think a zoom call will be easier than trading many emails, as that seemed to lead to mis-communications last time. I'd probably need a bit of orientation to the data file that Bianca/Valentine are using, or one of them can be on the call. Let me know how I can help!
Thanks, @srusso2. As far as I know this info is clear in Tala's data sheets (looks clear, although its possible that questions could arise). What's needed is modifying Bianca's script to ignore the cores from the sparse treatment. Bianca's no longer with us, so I'll let @ValentineHerr determine the easiest way to deal with this.
How about just deleting those lines that you don’t want to use from the data itself?
That way, you don’t have to modify the code.
If you want an easy solution...I assume that’s what you meant by “ignore”?
Sabrina E. Russo Professor, School of Biological Sciences & Center for Plant Science Innovation University of Nebraska-Lincoln, USA
From: Kristina Anderson-Teixeira notifications@github.com Sent: Monday, November 9, 2020 6:36:04 AM To: EcoClimLab/ForestGEO-climate-sensitivity ForestGEO-climate-sensitivity@noreply.github.com Cc: Sabrina Russo srusso2@unl.edu; Mention mention@noreply.github.com Subject: Re: [EcoClimLab/ForestGEO-climate-sensitivity] what's going on with NE? (#89)
Thanks, @srusso2https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_srusso2&d=DwMCaQ&c=Cu5g146wZdoqVuKpTNsYHeFX_rg6kWhlkLF8Eft-wwo&r=ozUfXq8GhmyNFrTdmFFL6Q&m=_3XlProqZjdqbCG2HueUeOM_4XV05RIzKy3m7Aw2AcM&s=pQrBD9mcMHx3M5AMmV1BW-R1M1TchMYoYzi6Y7vKkkM&e=. As far as I know this info is clear in Tala's data sheets (looks clear, although its possible that questions could arise). What's needed is modifying Bianca's script to ignore the cores from the sparse treatment. Bianca's no longer with us, so I'll let @ValentineHerrhttps://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_ValentineHerr&d=DwMCaQ&c=Cu5g146wZdoqVuKpTNsYHeFX_rg6kWhlkLF8Eft-wwo&r=ozUfXq8GhmyNFrTdmFFL6Q&m=_3XlProqZjdqbCG2HueUeOM_4XV05RIzKy3m7Aw2AcM&s=aFKF2CTXQy3WbxwJk3WosZPyiknPrnGdirywi5jCiTg&e= determine the easiest way to deal with this.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_EcoClimLab_ForestGEO-2Dclimate-2Dsensitivity_issues_89-23issuecomment-2D723986319&d=DwMCaQ&c=Cu5g146wZdoqVuKpTNsYHeFX_rg6kWhlkLF8Eft-wwo&r=ozUfXq8GhmyNFrTdmFFL6Q&m=_3XlProqZjdqbCG2HueUeOM_4XV05RIzKy3m7Aw2AcM&s=36BiCcKFLM_yZDdIUJCBzmrREFV-q9QP_eADoeiOGAw&e=, or unsubscribehttps://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_notifications_unsubscribe-2Dauth_AHHXIUEKJBWSCIGO2YTMFVLSO7O3JANCNFSM4TNUREOA&d=DwMCaQ&c=Cu5g146wZdoqVuKpTNsYHeFX_rg6kWhlkLF8Eft-wwo&r=ozUfXq8GhmyNFrTdmFFL6Q&m=_3XlProqZjdqbCG2HueUeOM_4XV05RIzKy3m7Aw2AcM&s=N879n-7U8Mc5M-m8xHU8Ab0bYa45Q7YCwWfCEskOsGg&e=.
Sorry for the delayed response. I think I can easily change the code to ignore cores that are in the sparse stand.
So just to make sure, I need to ignore the "Treatment" = "Sparse Pine" in this file ?
Correct. Thanks!
@ValentineHerr , looking at the figure again, I think we also need to drop "Treatment" = "Mid Pine" for JUNI only.
There's an un-resolved question on the units here. (@srusso2, you can't see that link, but I'm about to send an email.) I think we're okay, but it's possible there will be a change.
The measurement units are correct. Now we just need to remove the young trees that are biasing the analysis.
"Treatment" = "Sparse Pine" includes a couple JUNI, is that okay?
Will also remove "Treatment" = "Mid Pine" for JUNI only.
Please remove the "sparse pine" JUNI as well.
Nebraska results make a bit more sense after removing sparse pine for JUNI and PIPO and mid pine for JUNI :
However, the selection of Dec PET is strange, and they still don't match previous observations:
Climwin plots show little correlation, which is to be expected for BEPA (based on table above) but not JUNI and PIPO:
I'm not sure what to try next... Remove mid- PIPO? Or maybe the two sites need to be split, and just one included? (I'd hate to do that, but perhaps our assumption that they could be reasonably combined was wrong.)
@srusso2, it would be great to hear any thoughts you may have on this.
I'm noting that the analysis time frame for BEPA at Niobrara is 3-4 decades longer than that of the 2 species at Halsey. Perhaps we need to standardize the analysis time frame? (We don't do this at other sites, but differences are smaller.)
Indeed, the near-zero Dec PET that's showing up as an outlier in the climwin PET plot occurred prior to the start of the JUVI and PIPO records:
@ValentineHerr, let's try limiting the analysis timeframe. To make this general, what if we say that the analysis timeframe doesn't start until there are records for at least half the trees at a site?
Hi Krista, can you zoom this morning about this, in about 30 min? 8:30 my time? just a short meeting
Sabrina E. Russo Professor, School of Biological Sciences & Center for Plant Science Innovation University of Nebraska-Lincoln, USA
From: Kristina Anderson-Teixeira notifications@github.com Sent: Monday, November 16, 2020 7:25:17 AM To: EcoClimLab/ForestGEO-climate-sensitivity ForestGEO-climate-sensitivity@noreply.github.com Cc: Sabrina Russo srusso2@unl.edu; Mention mention@noreply.github.com Subject: Re: [EcoClimLab/ForestGEO-climate-sensitivity] what's going on with NE? (#89)
Nebraska results make a bit more sense after removing sparse pine for JUNI and PIPO and mid pine for JUNI :
However, the selection of Dec PET is strange, and they still don't match previous observations: [image]https://urldefense.proofpoint.com/v2/url?u=https-3A__user-2Dimages.githubusercontent.com_6355854_99256096-2Ddb48b700-2D27e2-2D11eb-2D8a74-2D665b869a748c.png&d=DwMCaQ&c=Cu5g146wZdoqVuKpTNsYHeFX_rg6kWhlkLF8Eft-wwo&r=ozUfXq8GhmyNFrTdmFFL6Q&m=ygwKpdM5j2hXyMFv3E4dOpEivd-ufm-TuRyj8a-utuM&s=QCQ5Bu4YIwWnuDKCV3RloA7ZqejnsrU2jtEk7Ma2oZE&e=
Climwin plots show little correlation, which is to be expected for BEPA (based on table above) but not JUNI and PIPO: [image]https://urldefense.proofpoint.com/v2/url?u=https-3A__user-2Dimages.githubusercontent.com_6355854_99256180-2D00d5c080-2D27e3-2D11eb-2D8ee2-2D5e5cf68db554.png&d=DwMCaQ&c=Cu5g146wZdoqVuKpTNsYHeFX_rg6kWhlkLF8Eft-wwo&r=ozUfXq8GhmyNFrTdmFFL6Q&m=ygwKpdM5j2hXyMFv3E4dOpEivd-ufm-TuRyj8a-utuM&s=29N47OvqjzlaqXrhzrnR0_LkTzhcY3skvFJspES6uDk&e=
I'm not sure what to try next... Remove mid- PIPO? Or maybe the two sites need to be split, and just one included? (I'd hate to do that, but perhaps our assumption that they could be reasonably combined was wrong.)
@srusso2https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_srusso2&d=DwMCaQ&c=Cu5g146wZdoqVuKpTNsYHeFX_rg6kWhlkLF8Eft-wwo&r=ozUfXq8GhmyNFrTdmFFL6Q&m=ygwKpdM5j2hXyMFv3E4dOpEivd-ufm-TuRyj8a-utuM&s=8pUWoz8wZRltBarZgHkSJriKxljU1jKNZviXQl6MnFQ&e=, it would be great to hear any thoughts you may have on this.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_EcoClimLab_ForestGEO-2Dclimate-2Dsensitivity_issues_89-23issuecomment-2D728007616&d=DwMCaQ&c=Cu5g146wZdoqVuKpTNsYHeFX_rg6kWhlkLF8Eft-wwo&r=ozUfXq8GhmyNFrTdmFFL6Q&m=ygwKpdM5j2hXyMFv3E4dOpEivd-ufm-TuRyj8a-utuM&s=iXzNHnax2JiNz85AXW1f-X2pPnTZAYEYebEbC_Cm39k&e=, or unsubscribehttps://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_notifications_unsubscribe-2Dauth_AHHXIUFBLGXOEJHRZTPJ3PTSQER33ANCNFSM4TNUREOA&d=DwMCaQ&c=Cu5g146wZdoqVuKpTNsYHeFX_rg6kWhlkLF8Eft-wwo&r=ozUfXq8GhmyNFrTdmFFL6Q&m=ygwKpdM5j2hXyMFv3E4dOpEivd-ufm-TuRyj8a-utuM&s=ukJsTtbJVnzC70TXEYuaR8X-Es4EvEr7dk4uE6wYsPw&e=.
From conversation with @srusso2 , we wouldn't expect BEPA to be responding to the same variables as PIPO and JUVI, so it may make sense to just drop BEPA altogether.
I think it still makes sense to limit the analysis time frame such that it doesn't start until there are records for at least half the trees at a site. If that gives reasonable results, I personally think it's okay to keep both, but if not we should drop BEPA. (@ValentineHerr , it it's easy to code maybe try it both ways?)
I looked at the PRISM climate data, and it's fairly similar-- Niobrara tends to be slightly (~1C) warmer, but the two are highly correlated. Thus, I don't think that combining climate data is a problem. Obviously, the hydraulic setting is very different at the two sites, but our results won't be "wrong".
Should I do this only for the early years or also for the late years?
And to make sure, for each site, I remove any year that has n_trees_that_year < max_n_trees /2
, n_trees_that_year
and max_n_trees
beeing calculated regardless of species, correct?
Should I do this only for the early years or also for the late years?
We won't differentiate early and late years at this site. In cases where we do, this will only affect the early years.
And to make sure, for each site, I remove any year that has
n_trees_that_year < max_n_trees /2
,n_trees_that_year
andmax_n_trees
beeing calculated regardless of species, correct?
correct.
Well, that change didn't do much to help NE, and made some undesirable changes at a couple other sites, so we'll want to revert that.
That changed the start date for both sites to 1946, and resulted in these results:
I'll have to come back to this later, but two thoughts:
1- We need to confirm that we're really removing the sparse PIPO. I don't know where those high growth rates are coming from... This might include needing to check our data script.
2- Assuming the data are all correct, we'll need to split the sites and include only one.
1- We need to confirm that we're really removing the sparse PIPO. I don't know where those high growth rates are coming from... This might include needing to check our data script.
I don't see any reason to suspect an error in the script. I've traced back the PIPOs with the highest growth rates to the original data sheet, and they are mid pine.
Besides splitting the sites, we might also try dropping PIPO mid-pine.
I would split the sites first and see what happens! Sabrina
From: Kristina Anderson-Teixeira notifications@github.com Sent: Tuesday, November 24, 2020 4:28 PM To: EcoClimLab/ForestGEO-climate-sensitivity ForestGEO-climate-sensitivity@noreply.github.com Cc: Sabrina Russo srusso2@unl.edu; Mention mention@noreply.github.com Subject: Re: [EcoClimLab/ForestGEO-climate-sensitivity] what's going on with NE? (#89)
Besides splitting the sites, we might also try dropping PIPO mid-pine.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_EcoClimLab_ForestGEO-2Dclimate-2Dsensitivity_issues_89-23issuecomment-2D733269223&d=DwMCaQ&c=Cu5g146wZdoqVuKpTNsYHeFX_rg6kWhlkLF8Eft-wwo&r=ozUfXq8GhmyNFrTdmFFL6Q&m=lEcwWc-dh1BY4dJauWDJSFk-_ZQu5D8Ol89ziIVJ2mU&s=vzBbSKY4wWaiOrCcWJcxd3TNnM83flUlWmpGgq-n0KE&e=, or unsubscribehttps://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_notifications_unsubscribe-2Dauth_AHHXIUFGOAZEWSPIYZQ6SJTSRQXNLANCNFSM4TNUREOA&d=DwMCaQ&c=Cu5g146wZdoqVuKpTNsYHeFX_rg6kWhlkLF8Eft-wwo&r=ozUfXq8GhmyNFrTdmFFL6Q&m=lEcwWc-dh1BY4dJauWDJSFk-_ZQu5D8Ol89ziIVJ2mU&s=VOKxG941YmdYqvTiw1CVvNLYJ0OCCCCcKgNz3iuGpe8&e=.
Arggg... NE is being tricky!
with CO2 (year would be similar):
and the climwin plots:
this doesn't really conflict with the original pub, but...
(I'll need to return to this later)
with CO2 (as proxy for year):
These results are still weird. I think we need to see how it looks without BEPA.
Where this stands:
My current inclination would be to add streamflow to Niobrara, as that's the ForestGEO site. (In this case, @ValentineHerr , no need to do the run without BEPA.)
I loaded the Niobrar streamflow data here.
@ValentineHerr , let's make this the next (and hopefully last!) thing we try.
Specifically, we'll want to add streamflow ("SF") as a variable in the precipitation category (now "water"). Obviously we'll need to cut the analysis start date to the start of this record (1946). And, to make sure it's clear, this will be for just BEPA.
working on this now
Stream Flow was never picked up as best water candidate. (at least not in the regular analysis, have not run Year analysis)
at least not in the regular analysis, have not run Year analysis)
It shouldn't change, should it? Climwin variable selection should be the same for both, right?
oh yes, correct.
This site continues to be frustrating, and seems to be the one site that is not aligning all that well with the original pubs. Not sure what's going on....
Just some sanity checks:
climate data:
cores data:
all_sites_cores.csv
match Paper birch data_original.xlsx
.all_sites_cores.csv
match nebraska_data_clean.xlsx
@ValentineHerr , one potentially important observation here: for WET, the lowest AIC is current May (open=4, close=4), but that's more of a spurious correlation: the 95% CI window is open=11, close=2. For the latter, there would be some sort of (probably weak) concave-up fit. I'm not sure what happens if that's selected-- maybe streamflow then becomes the best variable.
We've determined that we want to continue using the lowest AIC (#95), and that even if we were using windows based on the 95% confidence set, streamflow would not come out for Niobrara.
This post shows that if precipitation day frequency were excluded, streamflow would come out as the top water variable, but would remain a poor predictor and not come out significant in the GLS model.
(The results shown above for climate only are stable when DBH is included, and for any metric of growth.
(when year shows up in best model, climate response remains the same.)
My take-home on Niobrara:
@srusso2 , what do you think? (I'm going to post next about Hansley, but I'm leaning towards including Niobrara. Hansley still isn't behaving all that well...)
(I don't like the shape of the PIPO DBH response)
There's a lot going on there, but basically a lot of these climate relationships look suspect to me. Given that we don't have bandwidth to dig into what's going on, and that Niobrara is the ForestGEO site, I'm inclined to just go with Niobrara.
@srusso2 , does this sound okay to you?
Krista, I am sorry but I am totally under a deadline to turn in my grades by Dec 11 and I am barely going to make it if I do make it at all. I literally have no time to look at this thoughtfully. Can it wait until Monday 14th? I am so sorry because I know you are trying to advance work on this… Sabrina
From: Kristina Anderson-Teixeira notifications@github.com Sent: Tuesday, December 8, 2020 2:54 PM To: EcoClimLab/ForestGEO-climate-sensitivity ForestGEO-climate-sensitivity@noreply.github.com Cc: Sabrina Russo srusso2@unl.edu; Mention mention@noreply.github.com Subject: Re: [EcoClimLab/ForestGEO-climate-sensitivity] what's going on with NE? (#89)
@srusso2https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_srusso2&d=DwMCaQ&c=Cu5g146wZdoqVuKpTNsYHeFX_rg6kWhlkLF8Eft-wwo&r=ozUfXq8GhmyNFrTdmFFL6Q&m=u2Dm3ya1nUhPOYat9I4a5WaIzauGbuK1Vjt3mh9Gmig&s=4L_3_ZoOcqswlDgSy5yWMQKgca-MtKm79a9zxSQQtZ8&e= , does this sound okay to you?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_EcoClimLab_ForestGEO-2Dclimate-2Dsensitivity_issues_89-23issuecomment-2D741012485&d=DwMCaQ&c=Cu5g146wZdoqVuKpTNsYHeFX_rg6kWhlkLF8Eft-wwo&r=ozUfXq8GhmyNFrTdmFFL6Q&m=u2Dm3ya1nUhPOYat9I4a5WaIzauGbuK1Vjt3mh9Gmig&s=HrOsBOCsauqnT4b3ukOGaoVExsf9Hl90rswi09lEX_w&e=, or unsubscribehttps://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_notifications_unsubscribe-2Dauth_AHHXIUAFJJRQWWQXULM6TBDST2G47ANCNFSM4TNUREOA&d=DwMCaQ&c=Cu5g146wZdoqVuKpTNsYHeFX_rg6kWhlkLF8Eft-wwo&r=ozUfXq8GhmyNFrTdmFFL6Q&m=u2Dm3ya1nUhPOYat9I4a5WaIzauGbuK1Vjt3mh9Gmig&s=xrCH9N6GnIoYIm2yfIdlCHMy2oDHUnoqpCPStYbqCQQ&e=.
No worries, I completely understand. Good luck with your grading!
@srusso2 , I'm about 95% sure I want to go with the Niobrara route and am assuming that as I write, but let me know in case you object.
I'm calling this decision final. We are removing Hansley (#103).
Results are strange: