trun-4.rst - pism - [fork] customized build of PISM, the parallel ice sheet model (tillflux branch)
 (HTM) git clone git://src.adamsgaard.dk/pism
 (DIR) Log
 (DIR) Files
 (DIR) Refs
 (DIR) LICENSE
       ---
       trun-4.rst (6856B)
       ---
            1 .. include:: ../../global.txt
            2 
            3 .. _sec-paleorun:
            4 
            5 Fourth run: paleo-climate model spin-up
            6 ---------------------------------------
            7 
            8 A this point we have barely mentioned one of the most important players in an ice sheet
            9 model: the surface mass balance (SMB) model. Specifically, an SMB model combines
           10 precipitation (e.g. :cite:`Balesetal2001` for present-day Greenland) and a model for melt.
           11 Melt models are always based on some approximation of the energy available at the ice
           12 surface :cite:`Hock05`. Previous runs in this section used a "constant-climate"
           13 assumption, which specifically meant using the modeled present-day SMB rates from the
           14 regional climate model RACMO :cite:`Ettemaetal2009`, as contained in the SeaRISE-Greenland
           15 data set ``Greenland_5km_v1.1.nc``.
           16 
           17 While a physical model of ice dynamics only describes the movement of the ice, the SMB
           18 (and the sub-shelf melt rate) are key inputs which directly determine changes in the
           19 boundary geometry. Boundary geometry changes then influence the stresses seen by the
           20 stress balance model and thus the motion.
           21 
           22 There are other methods for producing SMB than using present-day modeled values. We now
           23 try such a method, a "paleo-climate spin-up" for our Greenland ice sheet model. Of course,
           24 direct measurements of prior climates in Greenland are not available as data! There are,
           25 however, estimates of past surface temperatures at the locations of ice cores (see
           26 :cite:`JohnsenetalGRIP` for GRIP), along with estimates of past global sea level
           27 :cite:`Imbrieetal1984` which can be used to determine where the flotation criterion is
           28 applied---this is how PISM's ``mask`` variable is determined. Also, models have been
           29 constructed for how precipitation differs from the present-day values
           30 :cite:`Huybrechts02`. For demonstration purposes, these are all used in the next run. The
           31 relevant options are further documented in the :ref:`Climate Forcing Manual
           32 <sec-climate-forcing>`.
           33 
           34 As noted, one must compute melt in order to compute SMB. Here this is done using a
           35 temperature-index, :ref:`"positive degree-day" <sec-surface-pdd>` (PDD) model :cite:`Hock05`. Such a PDD model has
           36 parameters for how much snow and/or ice is melted when surface temperatures spend time
           37 near or above zero degrees. Again, see the :ref:`Climate Forcing Manual
           38 <sec-climate-forcing>` for relevant options.
           39 
           40 To summarize the paleo-climate model applied here, temperature offsets from the GRIP core
           41 record affect the snow energy balance, and thus the rates of melting and runoff calculated
           42 by the PDD model. In warm periods there is more marginal ablation, but precipitation may
           43 also increase (according to a temperature-offset model :cite:`Huybrechts02`). Additionally
           44 sea level undergoes changes in time and this affects which ice is floating. Finally we add
           45 an earth deformation model, which responds to changes in ice load by changing the bedrock
           46 elevation :cite:`BLKfastearth`.
           47 
           48 To see how all this translates into PISM options, run
           49 
           50 .. literalinclude:: scripts/run-4-echo.sh
           51    :language: bash
           52    :lines: 3-
           53 
           54 You will see an impressively-long command, which you can compare to the :ref:`first one
           55 <firstcommand>`. There are several key changes. First, we do not start from scratch but
           56 instead from a previously computed near-equilibrium result:
           57 
           58 .. code-block:: none
           59 
           60      -regrid_file g20km_10ka_hy.nc -regrid_vars litho_temp,thk,enthalpy,tillwat,bmelt
           61 
           62 For more on regridding see section :ref:`sec-regridding`. Then we turn on the earth
           63 deformation model with option ``-bed_def lc``; see section :ref:`sec-beddef`. After
           64 that the atmosphere and surface (PDD) models are turned on and the files they need are
           65 identified:
           66 
           67 .. code-block:: none
           68 
           69      -atmosphere searise_greenland,delta_T,precip_scaling -surface pdd \
           70      -atmosphere_precip_scaling_file pism_dT.nc -atmosphere_delta_T_file pism_dT.nc
           71 
           72 Then the sea level forcing module providing both a time-dependent sea level to the ice
           73 dynamics core, is turned on with ``-sea_level constant,delta_sl`` and the file it needs is
           74 identified with ``-ocean_delta_sl_file pism_dSL.nc``. For all of these "forcing" options,
           75 see the :ref:`Climate Forcing Manual <sec-climate-forcing>`. The remainder of the options
           76 are similar or identical to the run that created ``g20km_10ka_hy.nc``.
           77 
           78 To actually start the run, which we rather arbitrarily start at year `-25000`, essentially
           79 at the LGM, do:
           80 
           81 .. literalinclude:: scripts/run-4.sh
           82    :language: bash
           83    :lines: 3-
           84 
           85 This run should only take one or two hours, noting it is at a coarse 20 km resolution.
           86 
           87 The fields ``usurf``, ``velsurf_mag``, and ``velbase_mag`` from file
           88 ``g20km_25ka_paleo.nc`` are sufficiently similar to those shown in
           89 :numref:`fig-secondoutputcoarse` that they are not shown here. Close inspection reveals
           90 differences, but of course these runs only differ in the applied climate and run duration
           91 and not in resolution or ice dynamics parameters.
           92 
           93 To see the difference between runs more clearly, :numref:`fig-ivolconstpaleo` compares the
           94 time-series variable ``ice_volume_glacierized``. We see the effect of option ``-regrid_file
           95 g20km_10ka_hy.nc -regrid_vars ...,thk,...``, which implies that the paleo-climate run
           96 starts with the ice geometry from the end of the constant-climate run.
           97 
           98 .. figure:: figures/ivol-const-paleo.png
           99    :name: fig-ivolconstpaleo
          100 
          101    Time series of modeled ice sheet volume ``ice_volume_glacierized`` from constant-climate
          102    (blue; ``ts_g20km_10ka_hy.nc``) and paleo-climate (green; ``ts_g20km_25ka_paleo.nc``)
          103    spinup runs. Note that the paleo-climate run started with the ice geometry at the end
          104    of the constant-climate run.
          105 
          106 Another time-series comparison, of the variable ``ice_volume_glacierized_temperate``, the
          107 total volume of temperate (at `0^\circ C`) ice, appears in
          108 :numref:`fig-ivoltempconstpaleo`. The paleo-climate run shows the cold period from
          109 `\approx -25` ka to `\approx -12` ka. Both constant-climate and paleo-climate runs then
          110 come into rough equilibrium in the holocene. The bootstrapping artifact, seen at the start
          111 of the constant-climate run, which disappears in less than 1000 years, is avoided in the
          112 paleo-climate run by starting with the constant-climate end-state. The reader is
          113 encouraged to examine the diagnostic files ``ts_g20km_25ka_paleo.nc`` and
          114 ``ex_g20km_25ka_paleo.nc`` to find more evidence of the (modeled) climate impact on the
          115 ice dynamics.
          116 
          117 .. figure:: figures/ivoltemp-const-paleo.png
          118    :name: fig-ivoltempconstpaleo
          119 
          120    Time series of temperate ice volume ``ice_volume_glacierized_temperate`` from
          121    constant-climate (blue; ``ts_g20km_10ka_hy.nc``) and paleo-climate (green;
          122    ``ts_g20km_25ka_paleo.nc``) spinup runs. The cold of the last ice age affects the
          123    fraction of temperate ice. Note different volume scale compared to that in
          124    :numref:`fig-ivolconstpaleo`; only about 1\% of ice is temperate (by volume).