Friday, February 20, 2015

A new topic for a Ph.D. in Hydrology

GEOtop 2.0 (http://abouthydrology.blogspot.it/search/label/GEOtop%202.0) is a successful process-based model of the hydrological cycle. It integrates both the water and energy budget and it is supplied by the MeteoIO library for meteo data interpolations. GEOtop has a development history of more than fifteen years (http://abouthydrology.blogspot.it/2015/02/geotop-essentials.html). It can be used, and has been used, for soil moisture forecasting, eco-hydrology simulations, snow pack evolution forecasting, permafrost modelling, landslide triggering assessment. Its code is a mature C++ implementation of solid algorithms and physics. However it is conceived as a monolithic structure, in which improvements can be made with difficulty and after overcoming a huge learning curve.  At the same time, the user experience is far by being optimal, and must be structurally improved.

Therefore, during the same evolution of the model, it was envisioned to migrate it towards a more flexible informatics where improvements, maintenance and documentation could be pursued more easily. This refactoring of the code, is not a trivial operation, and would require to understand the present structure of GEOtop, and advanced concepts of software engineering. The first step would be obtaining a temporary implementation according to the guideline sketched in: http://abouthydrology.blogspot.it/2014/09/improve-geotop-informatics.html.

Subsequently a tied integration of the main modules/classes should be pursued inside the Object Modeling System Infrastructure (http://abouthydrology.blogspot.it/2013/10/the-summer-school-on-object-modelling.html), as suggested in http://abouthydrology.blogspot.it/2011/03/going-beyond-present-stato-of-art-in.html.  This would immediately open the road to use the intrinsic parallelism of OMS components, with a better treatment of river basins genmetry and topology, and  the integration of the various tools developed within the JGrass-NewAGE system with GEOtop own capabilities: but it will be a byproduct of the work, not the main objective of this Ph.D.

The main work in fact will be in implementing classes for the use of unstructured meshes, for the implementation of algorithms for solving partial differential equations in a matrix free formalism, and for a parallelisation of internal algorithms of GEOtop, by using standard matrix packages or enhancing them, either in Java or C++. The focus will be in the efficiency of the implementations inside standard-main stream techniques,  in supporting literate programming, and discovering appropriate design patterns in programming this science, more than on hydrology itself. 

Integration of some of the OpenDA (http://www.openda.org/joomla/index.php) classes in OMS and their prototypical use could also part of the Ph.D. work. 

Obviously, the candidate must have programming skills in Java and C++, or the willing to pursue them having outstanding knowledge of hydrological physical processes. All the code developed is intended to be free software, and must be produced with appropriate documentation, being the reflection about research reproducibility and replicability and the productions of tools for it being effective, part itself of the main work.

Who is interested can write to me. 
The call for the doctoral positions can be found here: http://www.unitn.it/ateneo/663/concorso-di-ammissione (in Italian) e  http://www.unitn.it/en/ateneo/1954/announcement-of-selection
(in English).
Who would like to finance such a research or know programs that can support it are also welcomed. 

18 comments:

  1. Great Job, Riccardo, especially for the possible phd student!!
    What is fulfilled after the the latest meeting of GEOtop? How 's going the project now? Now I'm working on GEOtop and I would like to know the actual state of the project and collaborate. How about the connection with OpenDA?

    ReplyDelete
  2. Dear Emanuele, thank you for your contribution. I do not know which is the status of things promised in the last GEOtop meeting. What I am proposing here is a possible evolution, and the one I think is feasible for me, which I will be able to develop just components, and the most appealing for third parties that will be able in the future to assemble components to produce their own modeling solutions.

    ReplyDelete
  3. I forgot OpenDA. Data assimilation is a necessity. OpenDA is there. I will see what we can do about, after saying that my first commitment, up to now is OMS and not OpenDA.

    ReplyDelete
  4. Dear professor Rigon,

    I would like to know more about the requirements for this research in particular... and if you have some time I would like to get some advice for early stage hydrologists "like me".

    I will be a bit honest here and say that I find it hard to find the right path in Hydrology, it is as you described once in your blog "very specialized" ... I learned the basics in my B.Sc and M.Sc and I even started teaching those basics but when it comes to research and finding a PhD topic it is like digging a hole with a match stick, even after learning the basics of GIS and R.

    I don't want to sound so negative especially that I believe in my ability learn anything, that's why I'm asking for advice in order to focus my efforts in the right direction.

    thank you for reading so far... and have a nice day!!

    ReplyDelete
    Replies
    1. Dear Wael,

      I deepen, but just a little, my thinking in this post: http://abouthydrology.blogspot.it/2015/03/early-stage-hydrologist-esh.html

      Delete
  5. Dear Wael, it will take time to answer to your request. I will dedicate to it my next post. Thank you for your contribution.

    ReplyDelete
  6. Dear Riccardo,
    even though I share the objectives of the project and personally I will try to support it, I think it requires further explanation on the hydrological-environmental point of view.
    You talk about unstructured grid, parallelisation, refactoring of code etc. but this sounds more like a task to a computer scientist (or a IT department) rather than a PhD topic to be given to a hydrologist.
    What is the research question? What is the background?
    This part is very weak. Will the candidate work (or apply) on snow modelling? Or on vegetation coupling? Or on shallow water equations to be linked to Richards equation?
    I think you should expand this part if you want your project to be accepted and to be successful in a PhD school of environmental engineering.

    ReplyDelete
    Replies
    1. Dear Matteo,

      thank you for your suggestion, and I will try to explain me a little better. I am reading a book about the history of research on photosynthesis, and there is a section where the researchers are trying to understand where it happens insides the vegetation cells. A first answer is inside mitochondria. However, the good researcjhers also deduced that photsythesis should also happen elsewhere. The candidate place in the cells where chloroplasts, and in fact they were able to show that this was the case. However, the demonstration needed that they could be able to separate chloroplasts from the rest of leaves, i.e. a lot of expertise in treating appropriately vegetation cells and separate their parts in lab.
      These science details "under the hood" where those that finally allowed to confirm the good ideas, and to make someone gain the Nobel Prize (I could remember incorrectly the last particular, but the meaning is clear). Now to do scientific discovers in hydrology we need, among other stuff, a way to treat appropriately all the nonlinearities present in the processes, and this can be done only with well written process-based-models which can allow inclusion of new algorithms, new process schemes, and so on, without altering the overall structure of model

      Take another example, that you, yourself suggested to me. Assume you have a music score with good music, but no violins, no musical instruments. Only some people will be able to understand the beauty contained in those rows and symbols. So it is a good hydrological theory without a numerical model that implement it.

      The work I am asking in this Ph. program is to build violins, trumpets, and all the missing instruments, and possibly (which is a little more demanding) envisioning a way to build even the undreamed ones.

      I strongly believe this is really a necessity at this stage of our research, without which we will continue to go around without doing real progress.

      Delete
  7. That's very impressive post.

    ReplyDelete
  8. Dear Prof Riccardo,
    I am enrolled in an Indian University for my PhD in water resources engineering and management. I am in second semester of my program and looking for a PhD topic. I have been planning to work on large scale high resolution hydrological model in one of the basins in India. I came across OMS3 just few days back and found quite interesting. However, I am planning to use Parflow or HydroGeosphere for my work. What could be the advantage of using OMS3? What research questions can be considered using OMS3?

    Thanking you.

    Regards,

    Surajit

    ReplyDelete
    Replies
    1. Dear Surajit, let me first say that Parflow, Hydrogeosphere and our JGrass-NewAGE (based on OMS3, please see http://abouthydrology.blogspot.it/2015/03/jgrass-newage-essentials.html) are different types of models. Parflow and Hydrogeosphere are more similar to GEOtop (http://www.geotop.org/wordpress/).
      I already wrote some consideration about these type of models in several posts. For instance:
      - here I make general considerations: http://abouthydrology.blogspot.it/2012/02/which-hydrological-model-is-better-q.html
      - here I talk of my research on process-based models: http://abouthydrology.blogspot.it/2012/09/my-past-research-on-process-based.html
      - here I talk on my research on physico-statistical models: http://abouthydrology.blogspot.it/2012/09/my-past-research-on-physico-statistical.html
      GEOtop, PARFLOW and Hydrogeosphere are models of the first type, JGrass-NewAGE of the second type.
      For large scale hydrology, I would frankly prefer to use JGrass-NewAGE. Recently we used it for the Upper Blue Nile. See Abera Ph.D. thesis: http://abouthydrology.blogspot.it/2016/04/wuletawus-abera-phd-defense.html

      In any case, if you want to stay in that category of models, both PARFLOW and Hydrogeosphere were written and are maintained by outstanding colleagues. However, GEOtop besides the stuff the other models simulates also the energy budget, snow and permafrost. The other model simply cannot do it, and I think it can be a great limitation for many of the applications. But you are the best judge by yourself.

      Maybe you are not looking at those models for doing hydrological simulation, but because what you really love is to build models. In this case, we can talk in more detail on what is the way to develop next generation models.
      I expressed my opinion here: http://abouthydrology.blogspot.it/2011/03/going-beyond-present-stato-of-art-in.html

      Hope this helps. In any case, next July we are organising a summer school on OMS3 and JGrass-NewAGE: http://abouthydrology.blogspot.it/2016/04/summer-school-on-modelling-catchment.html

      Delete
  9. Dear Prof Riccardo,
    Thanks for such a detailed post. I have gone through some of your posts earlier, but going through them again helped me to understand better the relevant topics of interest. Fully coupled models like parflow and HydroGeosphere are monolithic in nature though they represent entire terrestrial cycle. When you say that "Parflow, Hydrogeosphere and GEOtop are models of first type and JGRass-New-Age of second type", do you mean first type models have only physics based in-built, whereas second type models have statistical tools in addition to what first type have? Can I carry out model initialization studies using JGrass-New Age? How about feedbacks studies between surface water, groundwater and land surface, say for a Ganges-Brahmaputra basin using JGrass New Age? Thank you.

    Regards,

    Surajit

    ReplyDelete
    Replies
    1. Hi Surajit,

      1- Yes, for first type and second type I mean what you say
      2- If you want to study the entire Ganges-Brahmaputra basin you need, I guess to use JGrass-NewAGE (or some of the models used for "global hydrology" (http://abouthydrology.blogspot.it/2015/10/large-ccale-hydrology.html). It is not only a problem of computational time, but also of characterisation of the parameters and data. However, you have, first, to obtain a separation of the basin in parts, and this can be a setup that requires a lot of time. A feasible approach is the one used by Wuletawu Abera in his Ph.D. thesis (http://abouthydrology.blogspot.it/2016/04/wuletawus-abera-phd-defense.html).
      We are working hard to smooth the process, and we can help you if you chose our way.
      3 - Studying the surface water -groundwater interactions (and at such large scale) is another type of challenge. There is plenty of recent literature on this topic, but not real applications at the scale you mean: simply too complicate. Certainly, something can be done.

      My advise is to select carefully something that is up-to-date but that you also do not attack an overwhelming task. Your advisor can certainly help in this direction.

      Delete
    2. Dear Prof Riccardo,
      Thanks for the thoughtful advice. I have started to go through Wuletawu Abera's PhD thesis. Let me have a rough idea of his work before deciding to work on similar lines.

      Thanking you.

      Regards,

      Surajit

      Delete
  10. Hi Surajit,

    Do you have my PhD thesis already or do you want me to send you ?

    best regards,
    Wuletawu

    ReplyDelete
  11. Hi Dr Wuletawu,
    I did have your PhD thesis from Abouthydrology.blogspot.in of 246 pages. In case of any help I might require along the way I will surely contact you.

    Thanking you.

    Regards,

    Surajit

    ReplyDelete
  12. Dear Prof. Riccardo,
    I have gone through Dr Wuletawu's thesis few number of times to understand the contributions of his work. What I understand is that he performed water budget studies for the whole catchment for the upper blue river Nile basin given limited data. He also used different satellite data for a comparative study on rainfall evaluation.

    I started using parflow, but it was hard to learn as one needs huge learning curve even before one understands how model works. Being free model, it is quite difficult to get help from other users. The reason is that I had no prior exposure to such models earlier. Of course, it would take a while to learn and apply such models!
    I follow your posts regularly. As you mentioned in your previous post on Dr Wuletawu's work that you are working hard to smooth the process and can help me if I choose your way, how can I use JGrass-NewAge for Ganga-Brahmaputra basin studies? I would like to know what contribution I should be able to make for JGrass-NewAge for my PhD work.

    Cheers!
    Surajit

    ReplyDelete
    Replies
    1. Dear Surajit,

      PARFLOW is a very good model, but IMO non suitable for the application you have in mind.
      We can help you in applying our model to the Gange-Brahmaputra system and Wuletawu Abera can help you in doing this. Please write to us to our private mail:
      riccardo.rigon unitn.it and wuletawu979 gmail.com for more details.

      Delete