- TWS-NEE discussion with Sujan, Martin, Nuno, and Markus
- 밥 & 술
2020. 11. 26. 목요일
TWS-NEE discussion with Sujan, Martin, Nuno, and Markus
9월부터 11월까지 내 연구를 어떻게 진행해 왔고 어떻게 해나가면 좋을지 토의하는 자리다.
내 발표는 Sujan이 만들어준 C-W coupled SINDBAD configuration을 테스트하는 것 위주로 진행됐다. Global run 시행 전에 site-level에서 잘 맞는지 확인해보는 단계. 그리고 몇몇 논문 요약.
Martin과 Markus는 장기 전략에 대해 조언을 많이 해줬다. 내 프로젝트의 주요 질문인 TWS-NEE IAV relationship은 바로 해결하기 어렵다. 때문에 잘게 쪼개서 쉬운 것 먼저 진행하는 것이다(divide&conquer). TWS IAV 혹은 NEE IAV를 먼저 살펴봐서 각각 component가 다른 것에 어떻게 반응하는지, 어떤 게 가장 영향력 있는지 등을 먼저 살피면 좀 더 내 질문을 명확하게 느낄 수 있을 것 같다. 미팅 후 Martin이 코멘트를 정리해서 메일로 보내줬다.
thanks again for the talk and all for the discussion. I briefly write again my main comments, for clarity, and transparency.
It’s really great and important that you started diving into SINDBAD! For this stage of the PhD and esp for the PAC it’s important to get also the problem and concept clear, and a tentative plan of attack straight. Markus suggestion on how different TWS components may be interacting with different C cycle processes on scales relevant to this project would also help to get things clearer conceptually.
With divide and conquer I was referring to trying to identify relevant chunks of lower complexity which could form papers or work packages, rather than starting with full complexity. In my opinion you can divide the problem in two bigger pieces: a) getting TWS IAV ‘right’, b) getting NEE IAV (and trend) right. Of course we expect some interactions, so we need to put them together, but maybe not from the beginning.
Starting from TWS IAV would be conceptually nicer as the water cycle constraints are likely more important for the carbon cycle IAV than carbon cycle constraints on water cycle IAV. But getting TWS IAV right might also be a ‘trap’ for a first project. To understand this better it would be good to analyse Tina’s and Basil’s results with respect to TWS IAV. In particular we’d like to infer which processes and data constraints might be missing or how difficult it may be (or not). My gut feeling is that we have a problem with surface water storage variations. To look at this you can look at regions and times where this should matter most, compared to other regions. For example surface water storage variations should matter most in the wet period of the wet tropics (esp amazon), and during/after snow melt in the high-latitude snow regions … or other regions with high inland water body area. As far as I remember the global TWS IAV is controlled by the tropics where e.g. Tina’s model doesn’t get it while it wasn’t too bad in other regions. But maybe you find other things!
Starting with global NEE IAV and trend is in my opinion also a viable strategy. Keenan (https://www.nature.com/articles/ncomms13428) showed that you can capture the big picture decadal variability / trend with a simple light use efficiency model with CO2 fertilization and Mirco’s no-pool TER model. Constraining CO2 fertilization with a global long-term inversion should really not be difficult, even when starting with a very simple (and fast) respiration model. Getting the NEE IAV on ‘eye-level’ with TRENDY models should also not be difficult - also FLUXCOM get’s the patterns (not the variance correct though) without having any carbon pool. I’m certainly not arguing that we shouldn’t have C pools in the model but i would start with something simple. Land use change aspects had basically no relevant effect on global NEE IAV in trendy models. Let’s assume that with a relatively simple setup we can get global NEE IAV and trend reasonable one could look at the effect of co2 fertilization on (spatial) NEE IAV variance changes with factorial experiments (to have an interesting research question :)).
When trying to make some progress on global NEE IAV beyond state-of-the art (which is maybe not necessary but would be nice) you’ll encounter probably three things: a) the vulcano eruptions, b) the change in variance, and c) something going on in the wet tropics. Wrt a) most people think it’s related to diffuse radiation which is not or not well represented in global meteo forcing data and/or models. I know that Stephen Sitch and Lina Mercado were after generating a better radiation product for that. We can ask … ; for b) it’s still unclear as far as i know; for c) it could be a ‘wet’ stress signal (either suppression of respiration, or wet stress on GPP … GPP of tropical forests seem quite sensitive to VPD otherwise.
Starting with the NEE would have also some advantage in terms of learning from Tina - in her final analysis over the next months she’ll look at the effect of lateral fluxes/river routing on TWS variations… so there are pros and cons for starting with TWS or NEE. I would try to start with whatever seems easier and more straightforward to you. But it does matter of course for your sindbad setup as you need different model structures, data constraints, and forcing data (e.g. long-term for nee).
Wrt attribution problem and the different methods you read about: Most of them were ‘empirical’ and for using with ‘observations’ only. If we have a model explaining the observations we should focus on using the model to do attribution, e.g. by designing factorial experiments, and by understanding equifinality due to parameter uncertainties.
Sorry for the long email. Trying to compensate a bit for my lack of availability over the next weeks:) Looking very much forward to this journey and more in-depth discussion!
P.S.: Sooner or later you’ll likely encounter different opinions and suggested directions. Sujan is your main adviser here and Matthias in Dresden so discuss and decide together with them in case of conflicts. It’s important to prioritize to make progress, not always you can make everyone happy simultaneously.
미팅 후 Sujan과 추가 미팅을 했고, TWS IAV 먼저 시작해보기로 했다. Tina와 Basil이 한 것을 토대로. 그리고 Cluster 작업에 더 익숙해지기 위해 Linux를 사용해보기로 했다.
2020. 11. 28. 토요일
중국 애들과 저녁 & 한 잔 했다. 이번에 내가 좀 만들어 줬다. 메뉴는 (최근에 배운) 떡볶이, (그날 아침 엄마가 레시피를 보내준) 소세지 야채 볶음, 김치찌개다. 이를 위해 판다판다에서 (드디어) 김치를 사왔다. 음… 대체로 만족이다. 떡볶이는 어묵 맛이 좀 에러였지만 다음부터 부산어묵 사 쓰면 될 것 같다. 고춧가루를 1T만 쓴 것도 잘 먹혔다. 쏘야볶은 소세지가 에러. 독일 소세지는 대체로 짜다. 김치찌개는 김치가 너무 설익어서 에러였지만 그래도 먹을만했다.
술은 따뜻한 레드와인 –> 보드카+레모네이드 –> 고량주. 유일하게 술 잘하고 좋아하는 중국 친구가 고량주를 가지고 있었다. 내가 좀 마시는 듯 해보인다면서 오피스에서(?!) 고량주를 가져와줬다. 보드카보다는 맛있었는데 좀 쎄다 (52%).
2020. 11. 29. 일요일
김치와 함께 산 된장, 두부. 드디어 된장찌개를 해먹는다. 역시 된장찌개는 밥에다 슥삭이다.
- 올 가을 처음으로 서리가 내렸다. 눈 내린 풍경이 기대된다. 이사 가기 전에 눈이 올까.
- 완성도 있던 접시와 성공적이었던 파스타
- 크리스마스 단장한 괴테갤러리. 블럭마다 동화를 묘사한 인형들이 있었다. 백설공주, 빨간 망토…. 저 늑대는 심지어 들숨/날숨에 맞춰 움직이기까지 한다. 독일은 뜬금 없는 곳에서 디테일하다.