Version 34 (modified by aducharne, 3 months ago) (diff) |
---|
ORCHIDEE development for the ROUTING SCHEME
Meeting 29 Mars 2021
Participants : Jan P., Anthony Schrapffer, Agnes D., Yann Meurdesoif, Philippe P., Lucia Rinchiuso
1) Présentation par Agnès des bases physiques du routage], pour éclairer la définition de l'indice topographique utilisé par ORCHIDEE, et les difficultés pour l'agréger et obtenir une valeur effective par HTU.
2) Présentation par Anthony des nouveautés dans le routage haute résolution, avec développements en cours pour déduire les HTUs de la structure des rivières principales :
Meeting 15 Janvier 2021
Participant: Jan P., Anthony Schrapffer, Agnes D., Catherine O., Olivier Boucher, Yann Meurdesoif, Thomas Dubos, Emmanuel Mouche, Olivier Marti, Philippe P., Pascal Maugis, Lucia Rinchiuso, Elliot.
Jan et Anthony ont fait une présentation des récents progrès du schéma de routage haute résolution RoutageJanvier2021
Points discutés :
- Numerical stability (link to the respect of the CFL condition and thus the choice of the maximum afordable time step):
- a semi-implicit would be possible with Q{t+1}_{stream,i} and we agreed after some thinking that it can be made to conserve Water.
- A local full implicit scheme would also be possible. That would be local to the atmospheric grid.
- Numerical tests
- \tau = \lambda . g_stream is the key parameter.
- \lambda (topoindex) and g_stream are probably not independent and thus it is only on the adjusted model that the maximal time step can be selected.
- Current tests seem to indicate that the order of magnitude would be hours.
- At high flow the Amazon flow at 6km/h. If an HTU has a dominant river segment of length 12km it would be traversed in two hours.
- Topoindex or \lambda
- This will be the key parameter. It describes the geometry of the HTU and needs to be aggregated from the high resolution data available. One also sees by the calculation above that the result of this integration will be key for the numerical stability.
- Looking at the distribution of the topoindex over the domain simulated indicate that a reasonable maximum time step can be selected.
- A lot of discussion was on how to "aggregate" topoindexes: Pascal M. suggested some approaches; Agnes suggested a test of physical consistency (linked to the length of the river, To be precised ?)
- Floodplains
- The fraction of flooded area in low flow conditions needs to be verified; given the choice to put firt the incoming water in a grid cell into the floodplains first!
- The explanations lacked the hypothesis on the transect perpendicular to the dominant river of the HTU.
- The results of Ronny Lauerwald need to be taken into account.
Conclusions:
- In the coming weeks Jan will advance on the topoindex aggregation and discuss the results with Agnès.
- Once there is a satisfactory solution it will be presented to the group.
- Then we will move to adjust the parameters (i.e. the three time constants (g)) to produce a usable model.
- Then we can move to improve the numerical scheme.
- If possible, there will be another meeting before Anthony travels back to Buenos Aires, else it will be later.
Meeting 08 October 2020
Participant: Jan P., Anthony Schrapffer (from Argentina), Agnes D., Catherine O., Olivier Boucher, Jerome Servonnat, Yann Meurdesoif, Thomas Dubos, Emmanuel Mouche, Olivier Marti, Philippe P.
Jan and Anthony gave a presentation on the recent progresses about the developement of their new routing scheme and the overall procedure to create the necessary input files for the routing scheme in ORCHIDEE: Routage_progress_Oct2020
Note that the developments for this new routing branch are stored under the ipsl gitlab: https://gitlab.in2p3.fr/ipsl/lmd/intro/routingpp
Main actions / points discussed are:
- Construction of the Hydrological Transfer Units (HTUs) done offline with the pre-processing routing: Routing_PP
- Need to test/update routing_PP in order to work with the icosaedric grid (especially for the number of neighbours)
- Potentially test the "efficiency" of routing_PP with a very large number of processor to gain speed (currently several hours in one node for a regional configuration)
- Test the impact of the maximum number of HTUs per grid cell on the overall routing scheme
- The Astropy http://spacetelescope.github.io/spherical_geometry/spherical_geometry/user.html spherical geometry package used is probably not the most efficient one.
- Routing scheme in ORCHIDEE
- NEED to test the intrinsic/numerical properties of the scheme in order to better understand why the results (river discharge for the Amazon river at Obidos) seems to critically depend on the time step in routing and possibly on the number of HTUs per grid box, etc...
- For that a specific "unitary" test with a pulse of rain (or runoff) should be done. To ease this approach Jan will update the "force_routing" routine in order to perform some tests with allways the same inputs of water.
- Agnes also suggested to test the properties of the routing scheme with only the "river reservoir"
- She reminds her note on the routing scheme and its application at high and coarse resolutions, with discussion elements on:
- the CFL condition and implication for the HTU size,
- the way to calculate an effective topographic index at the HTU scale from high resolution topographic information,
- the resulting effective stream length and its influence on timescale of the slow and fast reservoirs.
==> Different tests on the choice of time step, the choise of the HTUs max per grid, the calculation of the topoindex, etc need to be done and discussed together at the next meeting
- Contributions to this work
- Agnes wants to apply the scheme over France to test its properties
- Thomas will help with the icosaedric grid to test routing_pp
- Emmanuel will advise on the numerical problems and the physics of the hydrological transfers
- Yan M. will help with respect to reproducing / understanding the difference between routing the water at the highest resolution (using the original basin description from Hydroshed for instance) and routing the water at a coarser ORCH model resolution.
- Next big step will be to parallelise "routing.f90"
- Yan will be able to help with XIOS to compute the Halo for each domain, given that the conectivities are pre-calculated by "routing_PP"
==> Next meeting should be in December 2020 or January 2021 with the potential visit of Anthony in France.
Meeting 18 Fevrier 2019
Participant: Jan P., Anthony S., Agnes D., Catherine O., Yan M., Thomas D., Olivier DM, Philippe P.
First meeting to discuss the strategy of the developments of the routing scheme in ORCHIDEE.
- Jan P. presented the approach he is currently following: Routage_JanP_Fev2019
- Yan M. presented a parallele approach which is to perform the routing on a regular lat/lon grid at high resolution, which is easier to parallelise; the scheme then involve to interpolate of the fluxes to the regular grid Routage_YanM_Fev2019
Attachments (6)
- Routage_Jan_Presentation_Oct2020_VAS.pdf (1.3 MB) - added by peylin 10 months ago.
- routage_orchidee_YanM.pdf (2.1 MB) - added by peylin 9 months ago.
- routing_k_upscaling_etc.pdf (819.4 KB) - added by aducharne 9 months ago.
- Routing_Meeting_Jan21.pdf (1.5 MB) - added by peylin 6 months ago.
- Les bases physiques du routage_v2.pptx (13.6 MB) - added by aducharne 3 months ago.
- Status_of_the_ORCHIDEE_routing_scheme_development_V290321.pdf (3.3 MB) - added by aducharne 3 months ago.