## Model skills notebook with ERA5

It was a nice idea, but I will not champion it.

Research question (thanks Iuliia Polkova!) **"Evaluation of prediction skill for surface air temperature and precipitation from the CMIP6 models using ERA5"**

**STEP 1**. DONE find ERA in dkrz data pool --> /pool/data/ERA5/ (thanks Fabi! also for the docs: https://confluence.ecmwf.int/display/CKB/ERA5%3A+data+documentation#ERA5:datadocumentation-Introduction)

**STEP 2**. DONE how to load ERA5 --> no problem with GRIB: http://xarray.pydata.org/en/stable/examples/ERA5-GRIB-example.html (thanks Marco!)

**STEP 3**. DONE load model data --Y intake and xarray, just copy-paste from the summer days nb's

**STEP 4**. TODO how to compare ERA-model: Iuliia's answer:
"load the data, calculate ensemble mean, calculate anomalies with respect to the long-term mean, calculate RMSE or CORR for different lead times et voila. If you have problems just let me know, I will help. The skill is usually calculated for anomalies and for lead times, so for prediction year 1, and multi-year averages such as years 2-5. Depending on the research problem, we calculate skill for each grid-point or for the anomalies averaged over a certain region. Here is one of my papers with the examples of skill calculation https://doi.org/10.1029/2018MS001439"