1

What books are available that address the fundamental principles all processed based modelers should be aware of?

Is there a guide, some holy grail of text books for process based simulation models?

Edit: Process based, or system models such as those used to simulate ecological/biogeochemical/landscape processes. As Naught101 mentioned: "process based modelling" refers to a computational-numerical mode built on a conceptual model that is mostly based on physical laws defined by empirical relationships. As opposed to a purely empirical/statistical model.

derelict
  • 315
  • 1
  • 7
  • 1
    I'm not sure what you mean by process based modeling.... perhaps edit the question to make it a bit more clear? – Neo Aug 06 '14 at 19:50
  • 1
    Process based models are used for every system that I can think of (e.g., ocean, atmosphere). The question needs more clarification – arkaia Aug 07 '14 at 01:37
  • I'm looking for do's and do nots, fundamental principles of process based modeling in general, not a specific system. – derelict Aug 07 '14 at 02:27
  • This might be more suitable for http://scicomp.stackexchange.com/. If not, you'd be better off removing the "This is a bit off topic, but " part at the start, and editing it to make it more on-topic, if possible (I think it should be). Also, you're generally better off not asking for book recommendations. Those questions don't usually work very well on these sites, especially when they're very general... – naught101 Aug 17 '14 at 02:18
  • 1
    -1 This question still needs improvement. What kind of processes are you interested in? Are they described by empirical relationships? Partial differential equations? There is absolutely no answer to modeling "really any process". – milancurcic Aug 19 '14 at 17:59
  • @Neo: I assume that "process based modelling" refers to a computational-numerical mode built on a conceptual model that is mostly based on physical laws. As opposed to an empirical model. – naught101 Sep 19 '14 at 03:31
  • The help center doesn't mention anything about asking questions that may have long answers. If you cannot answer the question, or are confused by what I'm asking, then don't try to answer the question. – derelict Sep 23 '14 at 14:04
  • This question is currently closed and thus no one can answer it. If you wish to have it re-opened, please edit your question to narrow its scope (this will automatically nominate it for re-opening). – casey Sep 23 '14 at 21:01
  • How is it too broad? – derelict Sep 23 '14 at 21:05

1 Answers1

1

I guess the only thing the modeler should know is either the mathematical equation describing the system or the statistical relationship between a predictor and a predictand. You might have "access" to the exact equations or you might need to fit a mathematical model onto empirical observations. Then you need to code these mathematical or statistical relationship into a software (i.e. numerical model).

Fairly general. The devil will be in the details.

Balinus
  • 439
  • 3
  • 9
  • 1
    Once you have an equation it helps to know how to discretize it and analyze it for numerical stability and computational noise. – casey Aug 12 '14 at 20:44
  • Indeed, thanks! Feel free to edit my answer. :) (or I might add that info later this week when I have some spare time). – Balinus Aug 13 '14 at 14:33
  • I'm afraid there is much more to know than simple relationships. Understanding the equations is just one small part of modeling, and most process based models contain hundreds of thousands of equations. Model calibration, model validation, model paramaterization, spinup, hind-casting, forecasting, the list goes on and on. I was hoping someone could direct me towards a modelers bible, if you will. Thank you all for your input. Pun intended. – derelict Aug 19 '14 at 14:54
  • 1
    the question should be divided into such parts (i.e. calbration, validation, etc...). Otherwise, the answer would need to be 40 page long. – Balinus Aug 20 '14 at 15:47