I am reflecting about what to do in the next Fall. Last year was heavy but the workflow was following a given track, and I have not to think about what to do . This incoming Fall will be a little different. Looking forward, I obviously have wishes, but reality are the people I am working with that keep me on more concrete goals. So I have four or five offsprings, since I am working directly with five students (let’s call them S1 to S5, they know who is who).
The first direction of work is the consolidation S1 work, but not limited to it. This work, during the last year, covered mainly the systematising of the theory of travel times, producing a (imho) beautiful paper which was not so much recognised yet. This happened because we could not access, so far, reliable-enough tracers' data to test the theory, improve details of our informatics accordingly, and continue the thread of publications in that research direction. One single publication on a topic is usually quite invisible if it is proposed by newcomers (as we are in this topic).
We also redirected part of S1 work towards the building of an operational hydrological model, which is a task with its own pain. It was based on a (relatively) new partition of (lumped) hillslope fluxes that was necessary to test by itself. This work is at a good stage but the initial goals, of fully testing and have it operational, has been only partially accomplished so far. Much work is required in the Fall to arrive to a full completion of the task which would allow also an easy (without pain) reproducibility of the results (it is a long story. To see models that work in papers and do not work so fairly in my implementation, is a constant of my research life). Getting a new cleaned version of the new model would mean also cleaning the Adige-Hymod component ("the old model"), to use it for comparison. Documentation of the new model has to be completed on the GEOframe blog too. This could involve distilling some of what already done for the 2017 hydrology class. Understanding the limitations of the new model would be also an achievement, especially regarding some details of the hydrographs that are reproduced systematically wrong. Hopefully this work would lead to a paper on modelling* (which could be under our - mine and student's- control) and others that the new arrangement of reservoirs allows in term of processes description (which is not fully under our control, since we need data produced by other researchers). S1 has to be pro-active in this.
Student 1 approach is very much based on a new engine to run modelling solutions described by graphs (to see what I mean, please see here). This task has been pretty much on the shoulders of S2 who did a good job in understanding how OMS works and in implementing a graph-based way to execute in parallel the models implemented in each node. This was constructed with the idea that graphs can describe many type of interaction, but the first implementations are to describe a river network hydrology, having hillslopes (or better HRU as, see the introduction of this paper). It cannot be considered a final work until the modifications he introduced will not flow into the OMS mainstream code. Besides, I am sure the idea is so fertile that several and several generalisation of the system will follow.
This informatics, behind the tree-graph engine, can produce a paper that could be written in the incoming Fall, especially if more than one exemplificative modelling solutions could be shown to work with it*. S2 tells me that the next main problem is to make the calibration machineries of OMS to work properly with various modeling solutions and spatial arrangements. But I am confident that, with the help of Olaf David (OD), this can be obtained reasonably soon. One further steep in informatics, would be to use the CSIP infrastructure to run our modelling solutions on the UNITN multiprocessor system.
Overall S2 has interaction with most of the other guys (see also below). His effort can be summarised in saying, that he is contributing practically and theoretically, in building an infrastructure for scientific productivity and science reproducibility. I think from his cooperative work, a new version of OMS will come out, and his work will also result in some new interesting papers.
The branch of the work in which S2 works is the one I know personally the least. I am conscious that this can cause frustration since he has to explore, mainly by himself, the unexplored. Hope that OD can fill the empty space, but I will try to do a little effort to reduce the gap myself, even if I cannot guarantee the results.
S3 is working on finding his way. The idea is to work with him on the irrigation demand. This has two non hydrological aspects, the climatological and the plant’s one. The climatological regards the expectation one has in certain places to have certain temperature, radiation (CO2 contents); the plants' response has to do with their response/adaptation to the climate forcing. However, I left out, to mention now, the third aspect, the hydrological one that has to do with water availability. This, in turn, has also two aspects, the natural one, which, in part, falls in the climatological area, and the human intervention. On the first, we dare to say, we are experts, and have models to treat with it, especially GEOtop and its evolutions, but also JGrass-NewAGE. However, we should not give for granted that the description of water movemente in “natural” settings, could not be improved. On the second, personally, I am a parvenù, and I have a lot to learn. Involving humans, accounting for anthropic behavior and conflicts, could not be out of the horizon, but also how to connect the hydraulics of human infrastructures with the “natural" environment cannot given for granted. So S3 has a perspective to build and a focus to create on just a few of the aspects of this complex question (he, we, will not able to deal with everything): with the second year of the Ph.D incoming S3 has to search this consistently and produce a couple of research questions to live along in the next two or three years. It could or could be not in one of the aspects I like the most (modelling, and thinking to the infrastructure -physics and informatics- to give answers) but he has (we have) to take decisions.
S4 is working on Richards equation (again!). The roadmap is well traced, into enriching the actual 1d code with, among possible other aspects: 1- the energy budget; 2 - the freezing: 3- the withdraw either a- natural and b-human; 4 - the interaction with surface water (either as source of pressure or element where runoff is produced). A few others: a - to make explicit the vapour flux inside soil, and therefore, add a further budget; b - to include an explicit treatment of the water (molecules) age, according to theories well defined around year 2000; c - be more explicit in describing water functioning by introducing plants thermodynamics and hydraulics (a topic, the latter where convergence can be found with S3). We focused our first attention on freezing soil aspects, but the whole CZO studies are waiting for tools.
This roadmap is to be broken into steps to get along with before to switch to a full 3D implementation. There are aspects of this research (also described from a different point o view here) that could represent possibilities with strong interactions with S2. S2 graph infrastructure, in fact, should be versatile enough to be able to run, at its node several of these codes, which exchanges, for instance, just water remaining at the surface. Besides, there remain the big problem on how to get the best computational parallel behavior in multiprocessor or multicores machines, under the hood, i.e. without the hydrologist has to take care too much about it (what we call implicit parallelism). But this is just one of the further possible directions to investigate on the informatics side.
There are a lot of technical mini-issues in all of these topics (which are not mini, at the end). For instance why we do not adopt a more reliable method than CSV for input and output files (I know at least two: one is netcdf, the second is sqlite files/databases) ?
S5 is concerned with Urban Hydrology. His committment is about absorbing SWMM (its site here) concepts and organisation into a new system which include a designing tool compatible with it. It's just a master thesis, but I have expectations on his work. Today SWMM relies on old physics and, above all, is not a designing tool. For who is interested, we did some experience and work around by using different tools (GISWATER, QGIS, Docker) in the experiment we made with students last semester. I think was hard but fun, and we will see the results with the next week finals. Let's what we will be able to produce in future to bring the infrastructure into OMS (OD told me that he already has a ported version of SWMM, let's see if we can find it in some drawer).
Seen from the bottom, and the real work of people, all our enterprise seems matter of ordinary work and, accordingly, of incremental work, while science, instead, is often said to be looking for paradigmatic changes. However, this is only an impression. There are only a few models (or models component) like ours and our model were conceived since long ago (here a brief compendium of ideas) to set the base for those paradigmatic changes just mentioned. Hydrology has been said to be a dilectantistic field, where physical issues are never really solved and people get accostumed with solutions that just work. My attention to informatics is for freeing people from the legacy of too much constraining tools, allowing third parties inspection of scientific work, and making easier the implementation of new modelling approaches and ideas. I think we are close to this, and we should perseverate on this road.
Before closing this, I just want to refocus on the two possible papers I mentioned above.
S1 paper cannot be justified itself by saying that it is "a new model". So the scientific question has to be a different one. For instance, 1 - which is the minimal set of reservoirs that allows to describe evapotranspiration correctly; 2 - How the introduction of these reservoirs alter the overall residence-travel time; 3 - Is the age of evapotranspirated water in this model different from the age of the runoff water ? 4 - can we say something abot the information that flows around ?
*S2 paper can be justified by saying that it make much more flexible the building of models, either because it is possible to include different types of "models" in each node and also because, we can add and delete nodes (and the relative connections) without altering the rest of the model structure at run time. For instance, in a spatially semi-distribute model, that works for giving the proper water budget of a catchment, we can add an intake or a reservoir, to see what does change. Or, in the same way, maintaining the same spatial organisation, we can change the runoff production mode in each node wihout excessive reprogramming burden. Looking also how the performance scales with the number of nodes, for instance in comparison, with ADIGE Hymod, could also be of interest.
In any of the two case, we have to keep in mind that a paper is a narrative of something that can be (or should ?) be different from simply describing the tool we have in our hand and it became interesting only when it is finalise to fill a gap in our current knowledge or overcome the current state of art. (What can we do that before was not possible ? Why it is good for research ? Why it is good for operational systems ? How it affects performances ? How it affects performances of researchers ?
No comments:
Post a Comment