![binder badge](\"https://mybinder.org/badge_logo.svg\"){kind=icon}
](https://binder.pangeo.io/v2/gh/pangeo-data/climpred/main?urlpath=lab/tree/docs/source/examples/decadal/perfect-model-predictability-demo.ipynb) or view it [on Github](https://github.com/pangeo-data/climpred/blob/main/docs/source/examples/decadal/perfect-model-predictability-demo.ipynb)."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Demo of Perfect Model Predictability Functions\n",
"\n",
"This demo demonstrates `climpred`'s capabilities for a perfect-model framework ensemble simulation with {py:class}`.PerfectModelEnsemble`.\n",
"\n",
"**What's a perfect-model framework simulation?**\n",
"\n",
"A perfect-model framework uses a set of ensemble simulations that are based on a General Circulation Model (GCM) or Earth System Model (ESM) alone. There is *no* use of any reanalysis, reconstruction, or data product to initialize the decadal prediction ensemble. An arbitrary number of `members` are initialized from perturbed initial conditions, and the control simulation can be viewed as just another member, in `climpred`'s view as member 0.\n",
"\n",
"\n",
"**How to compare predictability skill score:**\n",
"As no observational data interferes with the random climate evolution of the model, we cannot use an observation-based reference for computing skill scores. Therefore, we can compare the members with one another (`\"m2m\"`) {cite:p}`Seferian2018`, against the ensemble mean (`\"m2e\"`) {cite:p}`Griffies1997`, or against the control (`\"m2c\"`) {cite:p}`Seferian2018`. We can also compare the ensemble mean to the control member (`\"e2c\"`) {cite:p}`Griffies1997`. See [comparisons](../../comparisons.html#comparisons) for more information.\n",
"\n",
"**When to use perfect-model frameworks:**\n",
"\n",
"- You don't have a sufficiently long observational record to use as a reference.\n",
"- You want to avoid biases between model climatology and reanalysis climatology.\n",
"- You want to avoid sensitive reactions of biogeochemical cycles to disruptive changes in ocean physics due to assimilation.\n",
"- You want to delve into process understanding of predictability in a model without outside artifacts."
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [
{
"data": {
"application/javascript": [
"\n",
" setTimeout(function() {\n",
" var nbb_cell_id = 1;\n",
" var nbb_unformatted_code = \"# linting\\n%load_ext nb_black\\n%load_ext lab_black\";\n",
" var nbb_formatted_code = \"# linting\\n%load_ext nb_black\\n%load_ext lab_black\";\n",
" var nbb_cells = Jupyter.notebook.get_cells();\n",
" for (var i = 0; i < nbb_cells.length; ++i) {\n",
" if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
" if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
" nbb_cells[i].set_text(nbb_formatted_code);\n",
" }\n",
" break;\n",
" }\n",
" }\n",
" }, 500);\n",
" "
],
"text/plain": [
"<Initialized Ensemble>\n",
"Dimensions: (lead: 20, init: 12, member: 10)\n",
"Coordinates:\n",
" * lead (lead) int64 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20\n",
" * init (init) object 3014-01-01 00:00:00 ... 3257-01-01 00:00:00\n",
" * member (member) int64 0 1 2 3 4 5 6 7 8 9\n",
" valid_time (lead, init) object 3015-01-01 00:00:00 ... 3277-01-01 00:00:00\n",
"Data variables:\n",
" tos (lead, init, member) float32 10.78 10.92 11.11 ... 10.83 10.89\n",
" sos (lead, init, member) float32 33.38 33.37 33.28 ... 33.4 33.48\n",
" AMO (lead, init, member) float32 0.07232 0.1894 ... -0.01757 0.06075array([ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18,\n",
" 19, 20])array([cftime.DatetimeProlepticGregorian(3014, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3023, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3045, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3061, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3124, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3139, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3144, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3175, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3178, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3228, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3237, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3257, 1, 1, 0, 0, 0, 0, has_year_zero=True)],\n",
" dtype=object)array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
array([[cftime.DatetimeProlepticGregorian(3015, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3024, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3046, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3062, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3125, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3140, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3145, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3176, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3179, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3229, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3238, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3258, 1, 1, 0, 0, 0, 0, has_year_zero=True)],\n",
" [cftime.DatetimeProlepticGregorian(3016, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3025, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3047, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3063, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3126, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3141, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3146, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3177, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
"...\n",
" cftime.DatetimeProlepticGregorian(3158, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3163, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3194, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3197, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3247, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3256, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3276, 1, 1, 0, 0, 0, 0, has_year_zero=True)],\n",
" [cftime.DatetimeProlepticGregorian(3034, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3043, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3065, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3081, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3144, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3159, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3164, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3195, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3198, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3248, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3257, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3277, 1, 1, 0, 0, 0, 0, has_year_zero=True)]],\n",
" dtype=object)array([[[10.781249, 10.917702, ..., 10.897053, 10.843189],\n",
" [10.727219, 10.813577, ..., 10.803096, 10.651206],\n",
" ...,\n",
" [10.294819, 10.539777, ..., 10.442225, 10.412181],\n",
" [10.385962, 10.594727, ..., 10.409877, 10.403644]],\n",
"\n",
" [[11.026904, 10.881209, ..., 10.903457, 10.85706 ],\n",
" [10.936329, 11.062461, ..., 10.968625, 11.082683],\n",
" ...,\n",
" [10.113186, 10.293538, ..., 10.450742, 10.182688],\n",
" [10.549644, 10.482872, ..., 10.614776, 10.51542 ]],\n",
"\n",
" ...,\n",
"\n",
" [[10.877393, 10.962437, ..., 10.858264, 10.441303],\n",
" [10.503685, 10.489896, ..., 11.052794, 10.790496],\n",
" ...,\n",
" [10.694736, 10.678864, ..., 10.324147, 10.555562],\n",
" [10.533122, 10.690272, ..., 10.522203, 11.049645]],\n",
"\n",
" [[11.197919, 10.93896 , ..., 10.838304, 10.451338],\n",
" [10.67775 , 10.550334, ..., 10.607387, 10.692555],\n",
" ...,\n",
" [10.642373, 10.502099, ..., 10.11778 , 10.391133],\n",
" [10.826764, 10.684681, ..., 10.829444, 10.894786]]], dtype=float32)array([[[33.379833, 33.366467, ..., 33.36124 , 33.32225 ],\n",
" [33.377155, 33.373028, ..., 33.346344, 33.38318 ],\n",
" ...,\n",
" [33.217045, 33.2169 , ..., 33.24268 , 33.241886],\n",
" [33.27586 , 33.298878, ..., 33.232338, 33.308716]],\n",
"\n",
" [[33.408016, 33.325253, ..., 33.42898 , 33.345768],\n",
" [33.3932 , 33.400852, ..., 33.3288 , 33.398758],\n",
" ...,\n",
" [33.277046, 33.27699 , ..., 33.328503, 33.218845],\n",
" [33.33428 , 33.243446, ..., 33.216988, 33.290936]],\n",
"\n",
" ...,\n",
"\n",
" [[33.39525 , 33.51696 , ..., 33.341 , 33.238438],\n",
" [33.367573, 33.445225, ..., 33.315975, 33.344593],\n",
" ...,\n",
" [33.387165, 33.345234, ..., 33.189358, 33.299137],\n",
" [33.420033, 33.389782, ..., 33.43263 , 33.552074]],\n",
"\n",
" [[33.36968 , 33.55584 , ..., 33.390663, 33.31099 ],\n",
" [33.402542, 33.388348, ..., 33.343822, 33.431545],\n",
" ...,\n",
" [33.246838, 33.30958 , ..., 33.144096, 33.204456],\n",
" [33.4219 , 33.406277, ..., 33.40383 , 33.48065 ]]], dtype=float32)array([[[ 0.072317, 0.189433, ..., 0.183764, 0.139705],\n",
" [-0.05801 , -0.092494, ..., 0.131268, -0.141136],\n",
" ...,\n",
" [-0.250911, -0.144247, ..., -0.144506, -0.102277],\n",
" [-0.208696, -0.057604, ..., -0.150325, -0.327045]],\n",
"\n",
" [[ 0.222662, 0.05584 , ..., 0.081747, -0.005777],\n",
" [ 0.085213, -0.02169 , ..., 0.237478, 0.120696],\n",
" ...,\n",
" [-0.239757, -0.290068, ..., -0.190722, -0.152883],\n",
" [-0.283574, -0.091263, ..., -0.003737, -0.199744]],\n",
"\n",
" ...,\n",
"\n",
" [[ 0.034137, 0.121907, ..., 0.014317, -0.101567],\n",
" [-0.160837, -0.123508, ..., 0.303377, 0.276406],\n",
" ...,\n",
" [ 0.005184, 0.020541, ..., -0.080405, -0.202475],\n",
" [-0.008904, -0.089735, ..., -0.026045, 0.194669]],\n",
"\n",
" [[ 0.325681, 0.06963 , ..., -0.17881 , -0.209799],\n",
" [-0.122161, 0.146038, ..., -0.013272, 0.045822],\n",
" ...,\n",
" [ 0.105318, -0.048179, ..., -0.166912, -0.266466],\n",
" [-0.01501 , -0.08603 , ..., -0.017568, 0.060747]]], dtype=float32)<Control Simulation>\n",
"Dimensions: (time: 300)\n",
"Coordinates:\n",
" * time (time) object 3000-01-01 00:00:00 ... 3299-01-01 00:00:00\n",
"Data variables:\n",
" tos (time) float32 10.91 10.96 10.93 11.12 ... 10.54 10.52 10.59 10.84\n",
" sos (time) float32 33.45 33.37 33.42 33.44 ... 33.4 33.46 33.39 33.38\n",
" AMO (time) float32 0.1678 0.1614 -0.1101 ... 0.01479 -0.02503 0.07905array([cftime.DatetimeProlepticGregorian(3000, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3001, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3002, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" ...,\n",
" cftime.DatetimeProlepticGregorian(3297, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3298, 1, 1, 0, 0, 0, 0, has_year_zero=True),\n",
" cftime.DatetimeProlepticGregorian(3299, 1, 1, 0, 0, 0, 0, has_year_zero=True)],\n",
" dtype=object)array([10.906495, 10.958404, 10.93251 , ..., 10.516892, 10.586926, 10.836343],\n",
" dtype=float32)array([33.449768, 33.3701 , 33.424828, ..., 33.455902, 33.391457, 33.38208 ],\n",
" dtype=float32)array([ 0.167775, 0.161447, -0.110121, ..., 0.01479 , -0.02503 , 0.079047],\n",
" dtype=float32)