{
"cells": [
{
"cell_type": "markdown",
"id": "cb1f8fb6-5210-415f-9766-9ef92bd3f085",
"metadata": {},
"source": [
"# AWS Access"
]
},
{
"cell_type": "markdown",
"id": "6a78c345-4330-4638-aa10-811e62f9cca7",
"metadata": {},
"source": [
"A growing number of CORDEX datasets is now available from AWS S3. In this examples, we will show how to access the datasets using an ESM collection and derive an ensemble mean."
]
},
{
"cell_type": "code",
"execution_count": 1,
"id": "518e0530-6acc-4752-8d6e-e2b356064549",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"
This data has been produced in the context of the C3S_34b_Lot1 and Lot 2 projects (PRINCIPLES/CORDEX4CDS) as a data provider for the Climate Data Store within the Copernicus Climate Change Service (C3S - https://climate.copernicus.eu/). While abiding by the highest scientific and technical standards, ECMWF cannot warrant that any information provided by the C3S will be entirely free from errors or omissions or that such errors or omissions can or will be rectified entirely. This applies to data from projects that continue to be developed, but are made publicly available for the purpose of feedback and testing. Some data or metadata may have been created or structured in files or formats that are not error-free. ECMWF accepts no responsibility with regard to such problems incurred as a result of using this data (see http://climate.copernicus.eu/disclaimer-privacy for the full disclaimer).
comment :
Please use the following reference for this climate data: CORDEX climate projection performed by ETH Zurich in collaboration with the CLM-Community
config_cclm :
EUR-11_CLMcom-ETH-COSMO-crCLIM-v1-1_01_config
config_int2lm :
EUR-11_int2lm2.5.1_01_config
contact :
cordex-eth@env.ethz.ch
conventionsURL :
http://www.cfconventions.org
creation_date :
2019-10-02 04:55:25
driving_experiment :
MPI-M-MPI-ESM-LR, historical, r1i1p1
driving_experiment_name :
historical
driving_model_ensemble_member :
r1i1p1
driving_model_id :
MPI-M-MPI-ESM-LR
experiment :
Historical run with MPI-M-MPI-ESM-LR GCM forcing
experiment_id :
historical
frequency :
mon
institute_id :
CLMcom-ETH
institution :
ETH Zurich, Zurich, Switzerland in collaboration with the CLM-Community
model_id :
CLMcom-ETH-COSMO-crCLIM-v1-1
product :
output
project_id :
CORDEX
rcm_version_id :
v1
references :
http://cordex.clm-community.eu/
source :
Climate Limited-area Modelling Community (CLM-Community)
title :
CLMcom-ETH-COSMO-crCLIM-v1-1 model output prepared for CORDEX
Processing for CORDEX archive at DKRZ (SVN revision 4244 http://svn-mad.zmaw.de/svn/mad/Model/IMDI/tags/cclm/cosmo_090213_4.8_clm17_cordex/util/running) 2014-03-19T07:56:10Z CMOR rewrote data to comply with CF standards and CORDEX requirements.
initialization_method :
1
institute_id :
CLMcom
institution :
Climate Limited-area Modelling Community (CLM-Community)
"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"import seaborn as sns\n",
"\n",
"palette = {\"historical\": \"C0\", \"rcp26\": \"C2\", \"rcp45\": \"C1\", \"rcp85\": \"C3\"}\n",
"g = sns.relplot(\n",
" data=df[(df[\"year\"] >= 1950) & (df[\"year\"] <= 2098)],\n",
" x=\"year\",\n",
" y=\"tas\",\n",
" hue=\"experiment_id\",\n",
" kind=\"line\",\n",
" errorbar=(\"ci\", 95),\n",
" aspect=2,\n",
" palette=palette,\n",
")\n",
"g.set(title=\"Yearly model mean with 95th percentile confidence interval\")"
]
},
{
"cell_type": "markdown",
"id": "94747571-7bba-4196-9344-1d6cb476c719",
"metadata": {
"tags": []
},
"source": [
"## Compute the change in the rolling yearly mean"
]
},
{
"cell_type": "markdown",
"id": "ccf24fe4-304c-40f2-847a-6a742bcbf052",
"metadata": {},
"source": [
"Ususally, it's easier to look at at rolling mean with a window size of a climate timescale, e.g., a decade. We can achieve this but we have to take care to fill the scenario datasets with historical data so that we don't get a 'gap' in the data when computing the rolling mean. We can achive this by using xarrays index notation similar to pandas:"
]
},
{
"cell_type": "code",
"execution_count": 29,
"id": "a34f91ee-97f7-4f82-945d-c5e94cf40037",
"metadata": {},
"outputs": [],
"source": [
"def rolling_mean(ensemble, n):\n",
" \"\"\"compute a rolling mean\n",
"\n",
" Take care to fill scenario timelines with historical data.\n",
"\n",
" \"\"\"\n",
" year = slice(2005 - n + 1, 2005)\n",
" ensemble.loc[dict(experiment_id=[\"rcp26\", \"rcp45\", \"rcp85\"], year=year)] = (\n",
" ensemble.sel(experiment_id=\"historical\").sel(year=year)\n",
" )\n",
"\n",
" return ensemble.rolling(year=n, center=False).mean()"
]
},
{
"cell_type": "code",
"execution_count": 30,
"id": "9c35d851-a831-43f5-93e6-6ff8f5f0fe95",
"metadata": {},
"outputs": [],
"source": [
"mean = rolling_mean(ensemble, 10)"
]
},
{
"cell_type": "markdown",
"id": "e5408b2d-2feb-4939-8991-6144d9df84c8",
"metadata": {},
"source": [
"Instead of looking at absolute values, it's more interesting to look at the change in temperature with respect to a reference period. Here we choose a reference period from 1961-1990 from the historical runs."
]
},
{
"cell_type": "code",
"execution_count": 31,
"id": "62f2b8fc-e29b-48dd-b1be-da01dab78d94",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"