Documentation for PISM, the Parallel Ice Sheet Model

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision
Previous revision
history [2015/12/16 18:51]
Ed Bueler add subheading in 2008
history [2016/08/18 19:27]
Ed Bueler fiddles
Line 20: Line 20:
  
 Bueler was interested in understanding numerical models by checking them against exact predictions (solutions) of the differential equations. ​ This became a paper [E. Bueler, C. Lingle, J. Kallen-Brown,​ D. Covey, L. Bowman, 2005.  //Exact solutions and verification of numerical models for isothermal ice sheets//, J. Glaciol. 51 (173), 291--306], but the first submission in 2003 did not get accepted. ​ This work used Matlab scripts, instead of the Fortran code, but it provided tests which, around the time of the re-submission,​ were used to check the code that would become PISM.  In this period, Latrice was the first of Bueler'​s graduate students to work on ice flow, with MS Math based on this work in 2002. Bueler was interested in understanding numerical models by checking them against exact predictions (solutions) of the differential equations. ​ This became a paper [E. Bueler, C. Lingle, J. Kallen-Brown,​ D. Covey, L. Bowman, 2005.  //Exact solutions and verification of numerical models for isothermal ice sheets//, J. Glaciol. 51 (173), 291--306], but the first submission in 2003 did not get accepted. ​ This work used Matlab scripts, instead of the Fortran code, but it provided tests which, around the time of the re-submission,​ were used to check the code that would become PISM.  In this period, Latrice was the first of Bueler'​s graduate students to work on ice flow, with MS Math based on this work in 2002.
 +
 +=== 2003: PETSc and C ... and PISM ===
  
 In 2002 Jed Brown became involved with PISM as an undergraduate,​ working for a while with the Fortran code from Craig and Elena. ​ Around 2003/04 Brown came into Bueler'​s office and said, essentially,​ that there was this nice library that would allow us to work in parallel at a higher conceptual level, namely [[http://​www.mcs.anl.gov/​petsc/​|PETSc]]. ​ And that we should switch to C-plus-plus so that the code could be more modular. ​ This suggestion was not fully appreciated by Bueler, but it was fully accepted. ​ The Fortran code was dropped, object classes were defined, and the isothermal SIA model, with under-development thermomechanical-coupling code, rebuilt based on collaboration between Brown and Bueler. In 2002 Jed Brown became involved with PISM as an undergraduate,​ working for a while with the Fortran code from Craig and Elena. ​ Around 2003/04 Brown came into Bueler'​s office and said, essentially,​ that there was this nice library that would allow us to work in parallel at a higher conceptual level, namely [[http://​www.mcs.anl.gov/​petsc/​|PETSc]]. ​ And that we should switch to C-plus-plus so that the code could be more modular. ​ This suggestion was not fully appreciated by Bueler, but it was fully accepted. ​ The Fortran code was dropped, object classes were defined, and the isothermal SIA model, with under-development thermomechanical-coupling code, rebuilt based on collaboration between Brown and Bueler.
  
-The two goals for major additions ​were then split:  ​First, ​Bueler led the effort to add thermocoupling to the SIA, with Brown and Lingle assistance, and emphasizing exact solutions to check. ​ This became E. Bueler, J. Brown, and C. Lingle, (2007). //Exact solutions to the thermomechanically coupled shallow-ice approximation:​ effective tools for verification//,​ J. Glaciol. 53 (182), 499--516.+Three goals for major additions then followed, in a period when Brown became very familiar with PETSc and Bueler finally learned C: 
 +  Bueler led the effort to add thermocoupling to the SIA, with Brown and Lingle assistance, and emphasizing exact solutions to check. ​ This became E. Bueler, J. Brown, and C. Lingle, (2007). //Exact solutions to the thermomechanically coupled shallow-ice approximation:​ effective tools for verification//,​ J. Glaciol. 53 (182), 499--516. 
 +  - Because Brown was now an MS student in math, Bueler suggested that Brown'​s MS project be the addition of, and testing of, a SSA solver in PISM.  This led to a successful [[http://​pism.github.io/​uaf-iceflow/​slidesJBrown.pdf|August 2006 MS project defense]]. ​ At that time the model had this (Bueler'​s suggestion) name: the C-plus-plus Object-oriented Multi-Modal,​ Verifiable Numerical Ice Sheet Model, a.k.a. COMMVNISM. 
 +  - [[http://​www.unidata.ucar.edu/​software/​netcdf/​|NetCDF]] was adopted as the input/​output format. ​ Before this, PETSc binary files were used.  (This fast format lacks included and standardized metadata.)
  
-Second, because Brown was now an MS student in math, Bueler suggested that Brown'​s MS project be the addition of, and testing of, a SSA solver in PISM.  This led to a successful [[http://​pism.github.io/​uaf-iceflow/​slidesJBrown.pdf|August 2006 MS project defense by Brown]], describing the model, at that time, by Bueler'​s suggested name: the C-plus-plus Object-oriented Multi-Modal,​ Verifiable Numerical Ice Sheet Model, a.k.a. COMMVNISM. +As this work was finishing, three things became clear: the multi-modal aspect was not actually working, ​Brown would be graduating ​and leaving for a PhD in Zurich, and another proposal would soon need to be written.  ​Brown renamed the model one day around this time---with no opposition---to the less cumbersome and better-suited-to-a-proposal name of "​Parallel Ice Sheet Model",​ PISM.
- +
-As this work was finishing, three things became clear: the multi-modal aspect was not actually working, ​Jed would be graduating, and another proposal would soon need to be written.  ​Jed renamed the model one day around this time---with no opposition---to the less cumbersome and better-suited-to-a-proposal name of "​Parallel Ice Sheet Model",​ PISM.  ​Bueler learned C as this went on, which was critical because Jed was graduating and leaving for a PhD in Zurich.  ​+
  
 === 2006: PISM goes public === === 2006: PISM goes public ===
  
-In September 2006 PISM was for the first time hosted publicly, [[http://​gna.org/​projects/​pism/​|on GNA]] with a [[http://​svn.gna.org/​viewcvs/​pism/​trunk/​COPYING?​view=log|GNU General Public License]].  ​(//We benefited greatly from using SVN and having free GNA hosting, even though we eventually moved happily to git and github.//)+In September 2006 PISM was for the first time hosted publicly, [[http://​gna.org/​projects/​pism/​|on GNA]] with a [[http://​svn.gna.org/​viewcvs/​pism/​trunk/​COPYING?​view=log|GNU General Public License]]. ​ We benefited greatly from using [[https://​subversion.apache.org/​|SVN]] and having free GNA hosting, even though we eventually moved happily to [[https://git-scm.com/​|git]] ​and [[https://github.com/|github]].
  
-The next idea, circa mid-2007, that went into PISM was that the SSA should be solved //​everywhere//​. ​ This is because, in a Coulomb or near-Coulomb basal drag regime, the solution simply returns zero sliding where the base is sufficiently strong. ​ Solving everywhere thus defines the ice stream regions organically. ​ This idea arose because Bueler actually read C. Schoof'​s isothermal paper [C. Schoof (2006). //A variational approach to ice stream flow//, J. Fluid Mech. 556, 227--251], and realized this made just as much sense in a thermocoupled context. ​ Additionally,​ solving the SIA everywhere would do no harm because in low-angle ​streams ​and shelves it produced low predicted velocities. ​ Furthermore,​ a convex combination of two reasonable stress balance solutions (i.e. SIA+SSA) was reasonable.+=== 2007PISM gets ice streams ​===
  
-Almost as importantin driving the solve-SSA-everywhere model, was the failure of PISM, and every other model, to produce anything sensible from the [[http://www.ingentaconnect.com/content/igsoc/​jog/​2010/​00000056/​00000197/​art00001|ISMIP-HEINO modeling assumptions/​requirements and boundary conditions]].  ​The issue seen in that project ​is that, in essence, there is no way to switch sliding on or off, in a physically-based thermomechanically-coupled mannerentirely within ​the SIA paradigm.  ​One needs to balance ​the transitions in boundary shear stress with //membrane stresses// within the ice The SIA is a good modelbut not of sliding (or ice shelvesfor that matter).+The next ideacirca mid-2007that went into PISM was that the SSA should be solved ​//everywhere//​.  ​This is because, in a Coulomb ​or near-Coulomb basal drag regime, the solution simply returns zero sliding where the base is sufficiently strong.  ​Solving everywhere thus defines ​the ice stream regions organically. ​ This idea arose because Bueler actually read C. Schoof'​s isothermal paper [C. Schoof (2006). ​//A variational approach to ice stream flow//, JFluid Mech. 556227--251]and realized this made just as much sense in a thermocoupled context.
  
-These ideas, and Jed's work on the PISM SSA implementation,​ led to actually having [[http://​pism.github.io/​uaf-iceflow/​talkagu.pdf|ice ​streams ​in the model for the right reasons by 2007]].  ​The paper E. Bueler and J. Brown, (2009)//Shallow shelf approximation as a “sliding law” in a thermomechanically coupled ice sheet model//, JGeophys. Res. 114 (F3) was the result. ​ This paper turned out to be the core of PISM, and it is the most-cited of the PISM-related papers.+Solving ​the SIA everywhere would do no harm because in low-angle streams ​and shelves ​the SIA produces low velocities.  ​Furthermorea convex combination of two reasonable stress balance solutions ​(i.eSIA+SSA) was reasonable.
  
 +Almost as important, in driving the creation of the solve-SSA-everywhere model, was the failure of PISM, and every other model, to produce anything sensible from the [[http://​www.ingentaconnect.com/​content/​igsoc/​jog/​2010/​00000056/​00000197/​art00001|ISMIP-HEINO modeling assumptions/​requirements and boundary conditions]]. ​ The issue seen in that project is that, in essence, there is no way to switch sliding on or off, in a physically-based thermomechanically-coupled manner, entirely within the SIA paradigm. ​ One needs to balance transitions in boundary shear stress with //membrane stresses// within the ice.
 +
 +In other words, the SIA is a good model, but not of sliding (//or ice shelves, for that matter//).
 +
 +These ideas, and Jed's work on the PISM SSA implementation,​ led to actually having [[http://​pism.github.io/​uaf-iceflow/​talkagu.pdf|ice streams in the model for the right reasons by 2007]]. ​ The paper E. Bueler and J. Brown, (2009). //Shallow shelf approximation as a “sliding law” in a thermomechanically coupled ice sheet model//, J. Geophys. Res. 114 (F3) was the result. ​ This paper turned out to be the core of PISM, and it is the most-cited of the PISM-related papers.
  
 === 2008: new team === === 2008: new team ===
Line 47: Line 55:
  
 Without needing excess tuning based on generally-unavailable data (i.e. inversion of measured surface velocities) the results for surface velocity have the right look.  Indeed, by early 2009 we saw that nearly-untuned results for Greenland [[http://​pism.github.io/​uaf-iceflow/​BKAJS_submit2_twocolumn.pdf|matched observations reasonably well]]. Without needing excess tuning based on generally-unavailable data (i.e. inversion of measured surface velocities) the results for surface velocity have the right look.  Indeed, by early 2009 we saw that nearly-untuned results for Greenland [[http://​pism.github.io/​uaf-iceflow/​BKAJS_submit2_twocolumn.pdf|matched observations reasonably well]].
 +
 +=== 2008: PIK collaboration ===
  
 Fall 2008 involved another big change: Anders Levermann and students (Maria Martin and Ricarda Winkelmann) came to Fairbanks to propose a collaboration in which they would add what PISM needed. ​ To be continued ... Fall 2008 involved another big change: Anders Levermann and students (Maria Martin and Ricarda Winkelmann) came to Fairbanks to propose a collaboration in which they would add what PISM needed. ​ To be continued ...
history.txt · Last modified: 2016/08/18 19:27 by Ed Bueler
© 2020 by PISM | webmaster