Monday, September 15, 2014

Opinions on the GEOtop roadmap

Just opinions indeed. For who landed here occasionally, information on GEOtop can be found here.

1 - Infrastructure

Mountain-eering and Exact-lab companies are doing a visible and positive effort in having a clean C++ code, after the huge work made by Stefano Endrizzi and Stephan Gruber to arrive to GEOtop 2.0. However, I believe the work on C++  should make some further steps, which I delineate in one of my previous posts here.
I deem necessary to do at least the first step described there, but the others would be also necessary in order that many people can work collaboratively on the project, and to maintain what we have while producing alternatives, and progress. My solution is, obviously just a proposal, and others (that I do not know) could exist. I plan to work myself part time on this from now to February, but if I can have help from someone this will be great. While changing the basic structure of the model's main( ), if  I will be really able to do it in this favourable astral conjunction,  I will take notes, and provide information to everybody. 

2 - Algorithms

On the side of the algorithms I believe we can work in several directions.  Using an unstructured grid will be a necessary intermediate step to pass through. 
Obviously this would imply a subsequent complete rewrite of the code, and in this rewriting we could take the occasion to: 1 - solve Richards equation with Casulli’s method, and 2 - better integrate surface waters with it. If resources would be available, even freezing soil could be rewritten, and the successful 2011 paper already contains how to do it.

3 - Processes

Among the processes, it would be important if the work by Stefano Dellachiesa on vegetation, and if the work of Florian Marshall on contaminant transport could be really embedded in the GEOtop main stream. My claim for more OO structure above, will go also in the direction to include these achievements, without cluttering the code and/or making more complicate the IO.
Personally, and in perspective, I am interested to include in GEOtop a modeling of the low atmosphere, integrating Navier-Stokes equation. I did some preliminary approach to the problem working conjointly with Dino Zardi and Michael Dumbser. A lot of resources are needed but I can work to grab them.

4 - OMS - Web services

My personal academic interest is also to push further  the evolution of GEOtop’s (beyond step 1) informatics for embedding GEOtop in OMS. This would allow splitting GEOtop  in components that can be joined at run-time through a scripting language, i.e. Groovy. This would enormously  increase the ability to run different modelling solutions, facilitate simulations reports, incapsulate even more the code. I am not claiming here, or pursuing, that this must be the main road followed by the GEOtop community. A certain “genetic variability” in the GEOtop versions and ecology would be indeed of benefit, and in favour of the model eventual survival. A passage necessary for OMS integration will be that task 1above will be pursued. That I think, should be really accomplished right away.

5 - Commitment and Executables

I already manifested my opinion to the developer group. Which is the following: in this moment using GEOtop require the total commitment of the guy/gal the want to embrace it. This is not acceptable if we want that our product would have some spreading. In this moment compiling GEOtop, still remains a challenge for a new users, and even for expert users, because they usually have a lot of other things to do, and they are pissed out when something that they consider trivial does not work as they expect. 
Downloading the source code directly from trunk should work for developers, but not for users. A tagged version, known to compile under the three major operating systems, should be used instead. More than that, a tarred zipped file, with the source code should be made available without passing through git. The best experience for a user would be, however, to download directly the executable for her/his machine and run it. I think that this move, together with improving the manual and documentation, will really boost GEOtop use. And, IMHO is the presence of a lot users that create a fertile soil for businesses.  Certainly a different pathway would be to implement GEOtop as a remote web-service. Recently at CSU,  I was impressed on what te eRams platform can offer, and how it helps in getting data and simulations together, and make easier collaborative work of groups.

6 -Testing

GEOtop comes with a series of tests, and, I verified, all of them compile, and all except one converge to the right solutions. This is useful for making users confident of what they are doing, and developers certain that they did not screwed up the system with their last minute modifications.  However, a special attention should be given to compare GEOtop with the tests provided by the increasing community of process-based model developers. They call their virtual experiment “the superslab” and you can find its definition here.  Being engaged with those guys would be really important, and would bring furhter recognition the our model. We could possibly be able also to challenge that community and propose tests that cope with the cryosphere, since none of its models does it. 

7 - Manual

The manual is not out-of -date but not up-to-date too. Efforts to get it improved would be necessary. I will try to put some students on it, if the occasion comes.

8 -License and Community

GEOtop is GPL and I believe it has to remain GPL. However, not in name of a single Author, but most probably in name of many. The creation of a GEOtop by components, could help in this direction, in the sense that single components, could be copyrighted then by single persons or group different from the original ones. The establishment of some GEOtop foundation with some organisation, and which determines some rules of behavior and/or an etiquette to follow for GEOtop use, will be a desired goal for future. I consider myself not eligible for any position in it though. 

9 - Courses

If a few of the conditions above would be met, it would not be difficult to start some schools either in Europe and in the Americas (or elsewhere). This also should be a goal to be pursued. Schools for Ph.D. students and Researchers, could not be very expensive, however some income can be generated.  Courses for professionals should be more expensive ... but maybe these could be directed to use web-services, and therefore the real income could arrive from subsequent use. 

10 - And then ?

Then it is matter to find resources and man/months to have all of it working. 

No comments:

Post a Comment