WeatherBench 2

https://weatherbench2.readthedocs.io/en/latest/
Admin
WeatherBench 2 is a framework for evaluating and comparing data-driven and traditional numerical...
Tokens:126,368
Snippets:480
Trust Score:8.3
Update:1 year ago
Show doc for...
Tokens:
Raw
### Install WeatherBench 2 and Dependencies

Source: https://weatherbench2.readthedocs.io/en/latest/_sources/evaluation

Installs the WeatherBench 2 package and essential dependencies using pip. Ensure you have Python and pip installed before running this command.

```bash
pip install weatherbench2
```

--------------------------------

### Example of Running Evaluation with Apache Beam in Python

Source: https://weatherbench2.readthedocs.io/en/latest/evaluation

Provides a commented-out example of how to run the evaluation using Apache Beam, which is recommended for larger datasets. It shows how to specify the Beam runner and input chunk sizes for distributed processing.

```python
# evaluation.evaluate_with_beam(
#     data_config,
#     eval_configs,
#     runner='DirectRunner',
#     input_chunks={'time': 20},
# )
```

--------------------------------

### Install WeatherBench 2 using pip

Source: https://weatherbench2.readthedocs.io/en/latest/evaluation

Installs the WeatherBench 2 framework from its GitHub repository using pip. This command may sometimes trigger warnings regarding the Pandas version, but the installation should proceed correctly.

```shell
# Pip might complain about the Pandas version. The notebook should still work as expected.
!pip install git+https://github.com/google-research/weatherbench2.git
```

--------------------------------

### Basic Data Loading Example

Source: https://weatherbench2.readthedocs.io/en/latest/_sources/evaluation

Demonstrates how to load weather data using WeatherBench 2. This involves specifying the dataset and time range for data retrieval.

```python
import weatherbench2 as wb2

data = wb2.load_data(
    dataset="era5",
    times=('2000-01-01', '2000-12-31')
)
```

--------------------------------

### Setup WeatherBench Configuration in Python

Source: https://weatherbench2.readthedocs.io/en/latest/evaluation

Imports the necessary configuration module from the WeatherBench2 library to begin setting up evaluation parameters. This is the initial step before defining data or evaluation specifics.

```python
from weatherbench2 import config
```

--------------------------------

### Clone WeatherBench 2 Repository

Source: https://weatherbench2.readthedocs.io/en/latest/index

This command clones the WeatherBench 2 GitHub repository to your local machine. It's the first step to get started with the project.

```bash
git clone git@github.com:google-research/weatherbench2.git
```

--------------------------------

### Simple Linear Regression Forecasting Model

Source: https://weatherbench2.readthedocs.io/en/latest/_sources/evaluation

A basic example of a forecasting model using linear regression. This function takes historical data and predicts future values.

```python
import weatherbench2 as wb2
import numpy as np

def simple_linear_regression(data):
    # Placeholder for actual linear regression implementation
    # This function should take historical data and return predictions
    print("Running simple linear regression...")
    # Example: return a copy of the input data as a prediction
    return data.copy()
```

--------------------------------

### Running a Forecasting Experiment

Source: https://weatherbench2.readthedocs.io/en/latest/_sources/evaluation

Shows how to set up and run a forecasting experiment using a specified model and configuration. This involves defining parameters and executing the forecast.

```python
import weatherbench2 as wb2

# Load data (assuming data is already loaded as shown in previous examples)
data = wb2.load_data(dataset="era5", times=('2000-01-01', '2000-12-31'))

# Define the forecasting model (e.g., using a predefined model or a custom one)
model = wb2.models.LinearRegression(
    resolution=wb2.Resolution.T63,
    time_range=(0, 360)  # Example time range in days
)

# Run the forecasting experiment
predictions = wb2.run_experiment(
    model=model,
    data=data,
    output_dir="./forecast_output"
)
```

--------------------------------

### WeatherBench 2 Evaluation Script Example

Source: https://weatherbench2.readthedocs.io/en/latest/command-line-scripts

This example demonstrates how to execute the `evaluate.py` script with specific command-line arguments. It showcases a practical application of the script for running evaluations, likely involving setting paths to forecast and observation data, and potentially other configuration options.

```bash
python evaluate.py \

```

--------------------------------

### Install WeatherBench 2 with GCP Requirements

Source: https://weatherbench2.readthedocs.io/en/latest/index

This command installs WeatherBench 2 along with optional requirements for launching jobs on Google Cloud Platform (GCP).

```bash
pip install .[gcp]
```

--------------------------------

### Data Handling and Preprocessing

Source: https://weatherbench2.readthedocs.io/en/latest/_sources/evaluation

Details on loading and preprocessing weather data for use with forecasting models. This includes selecting datasets and specifying time periods.

```python
import weatherbench2 as wb2

# Load data from a specific dataset (e.g., 'era5') for a given time range
data = wb2.load_data(
    dataset="era5",
    times=("1980-01-01", "2014-12-31")
)

# Data preprocessing steps would typically follow here, such as:
# - Normalization
# - Feature selection
# - Splitting into training and validation sets
```

--------------------------------

### Install WeatherBench 2

Source: https://weatherbench2.readthedocs.io/en/latest/index

These commands install the WeatherBench 2 package using pip. The first command changes the directory to the cloned repository, and the second installs the package. An editable install is available for development.

```bash
cd weatherbench2
pip install .
```

```bash
pip install -e .
```

--------------------------------

### Configuration for a Forecasting Model

Source: https://weatherbench2.readthedocs.io/en/latest/_sources/evaluation

Illustrates how to configure a forecasting model, specifying resolution, time range, and other relevant parameters. This configuration is used when initializing a model object.

```python
from weatherbench2.models import Resolution

# Example configuration for a model
resolution = Resolution.T63
time_range = (0, 360)  # Forecast for 360 days

# This configuration would be passed to a model constructor, e.g.:
# model = SomeModel(resolution=resolution, time_range=time_range)
```

--------------------------------

### Open Zarr Dataset with xarray

Source: https://weatherbench2.readthedocs.io/en/latest/_sources/data-guide

Opens a Zarr dataset from a Google Cloud Storage (GCS) path using the xarray library. This function requires the xarray library to be installed and assumes access to the specified GCS bucket.

```python
import xarray as xr

xr.open_zarr('gs://weatherbench2/datasets/neuralgcm_deterministic/2020-240x121_equiangular_with_poles_conservative.zarr')
```

--------------------------------

### Load Keisler Dataset with xarray

Source: https://weatherbench2.readthedocs.io/en/latest/_sources/data-guide

This snippet demonstrates how to open and load the Keisler dataset from a Zarr store located in Google Cloud Storage using the xarray library. It assumes xarray and dask are installed. The output will be a dask-backed xarray DataArray representing weather data.

```python
xr.open_zarr('gs://weatherbench2/datasets/keisler/2020-360x181.zarr')
```

--------------------------------

### Open Zarr Dataset with Xarray

Source: https://weatherbench2.readthedocs.io/en/latest/data-guide

This snippet demonstrates how to open a Zarr dataset from a Google Cloud Storage (GCS) path using the xarray library in Python. It assumes the necessary libraries are installed and credentials are configured.

```python
import xarray as xr

xr.open_zarr('gs://weatherbench2/datasets/ifs_ens/2018-2022-1440x721.zarr')
```

--------------------------------

### Open Zarr Dataset with Xarray

Source: https://weatherbench2.readthedocs.io/en/latest/data-guide

Opens a dataset stored in Zarr format from a Google Cloud Storage path using the xarray library. This is a common method for accessing large, chunked datasets. It requires the xarray library to be installed.

```python
import xarray as xr

xr.open_zarr('gs://weatherbench2/datasets/hres/2016-2022-0012-1440x721.zarr')
```

--------------------------------

### Evaluate ERA5 Forecasts vs ERA5 using Bash

Source: https://weatherbench2.readthedocs.io/en/latest/_sources/official-evaluation

This script demonstrates the setup for evaluating ERA5 forecasts against ERA5 observational data. It specifies the necessary paths for forecast data, observations, and climatology, along with output configurations. This example is incomplete as it lacks the full set of arguments.

```bash
python evaluate.py -- \
  --forecast_path=gs://weatherbench2/datasets/era5-forecasts/2020-64x32_equiangular_conservative.zarr \
  --obs_path=gs://weatherbench2/datasets/era5/1959-2022-6h-64x32_equiangular_conservative.zarr \
  --climatology_path=gs://weatherbench2/datasets/era5-hourly-climatology/1990-2019_6h_64x32_equiangular_conservative.zarr \

```

--------------------------------

### Open ERA5 Zarr Dataset with xarray

Source: https://weatherbench2.readthedocs.io/en/latest/_sources/data-guide

Opens the ERA5 dataset stored in Zarr format from a Google Cloud Storage bucket using the xarray library. This is a common starting point for accessing and analyzing the WeatherBench 2 data.

```python
xr.open_zarr('gs://weatherbench2/datasets/era5/1959-2023_01_10-wb13-6h-1440x721_with_derived_variables.zarr')
```

--------------------------------

### Open ERA5 Forecasts with Xarray

Source: https://weatherbench2.readthedocs.io/en/latest/data-guide

Opens a specific ERA5 forecast dataset (2020-1440x721.zarr) from Google Cloud Storage using the xarray library. This function is used to load the forecast data for analysis. It requires the xarray library to be installed.

```python
xr.open_zarr('gs://weatherbench2/datasets/era5-forecasts/2020-1440x721.zarr')
```

--------------------------------

### Evaluate FUXI Forecasts (Python)

Source: https://weatherbench2.readthedocs.io/en/latest/official-evaluation

This Python script evaluates the FUXI forecast model. It configures paths for forecast, observation, and climatology data, along with output settings, input chunking, and evaluation criteria. This example also demonstrates specifying specific pressure levels for evaluation and including derived variables.

```python
python evaluate.py -- \
 --forecast_path=gs://weatherbench2/datasets/fuxi/2020-64x32_equiangular_conservative.zarr \
 --obs_path=gs://weatherbench2/datasets/era5/1959-2022-6h-64x32_equiangular_conservative.zarr \
 --climatology_path=gs://weatherbench2/datasets/era5-hourly-climatology/1990-2019_6h_64x32_equiangular_conservative.zarr \
 --output_dir=$OUTDIR/64x32/deterministic/ \
 --output_file_prefix=fuxi_vs_era_2020_ \
 --input_chunks=init_time=1 \
 --fanout=27 \
 --regions=all \
 --eval_configs=deterministic,deterministic_temporal \
 --evaluate_climatology=False \
 --evaluate_persistence=False \
 --time_start=2020-01-01 \
 --time_stop=2020-12-31 \
 --levels=500,850 \
 --variables=geopotential,temperature,u_component_of_wind,v_component_of_wind,2m_temperature,10m_u_component_of_wind,10m_v_component_of_wind,mean_sea_level_pressure,10m_wind_speed,wind_speed \
 --add_land_region=False \
 --use_beam=True
```

--------------------------------

### Open Zarr Dataset with Xarray in Python

Source: https://weatherbench2.readthedocs.io/en/latest/data-guide

Opens a Zarr dataset from a Google Cloud Storage (GCS) path using the xarray library. This is useful for loading large, chunked datasets for analysis. Ensure you have xarray and the necessary GCS connectors installed.

```python
import xarray as xr

dataset = xr.open_zarr('gs://weatherbench2/datasets/hres_t0/2016-2022-6h-1440x721.zarr')
```

--------------------------------

### Displaying Coordinate Data

Source: https://weatherbench2.readthedocs.io/en/latest/_sources/evaluation

This snippet shows how to display the data associated with coordinates like 'latitude', 'level', and 'longitude'. It extracts preview data and associated metadata.

```html
<li class='xr-var-item'><div class='xr-var-name'><span class='xr-has-index'>level</span></div><div class='xr-var-dims'>(level)</div><div class='xr-var-dtype'>int64</div><div class='xr-var-preview xr-preview'>50 100 150 200 ... 700 850 925 1000</div><input id='attrs-afedcd48-de82-453a-a658-7bb58e7ffbac' class='xr-var-attrs-in' type='checkbox' disabled><label for='attrs-afedcd48-de82-453a-a658-7bb58e7ffbac' title='Show/Hide attributes'><svg class='icon xr-icon-file-text2'><use xlink:href='#icon-file-text2'></use></svg></label><input id='data-d5260f83-aae9-45a2-a3f6-817df555e95e' class='xr-var-data-in' type='checkbox'><label for='data-d5260f83-aae9-45a2-a3f6-817df555e95e' title='Show/Hide data repr'><svg class='icon xr-icon-database'><use xlink:href='#icon-database'></use></svg></label><div class='xr-var-attrs'><dl class='xr-attrs'></dl></div><div class='xr-var-data'><pre>array([  50,  100,  150,  200,  250,  300,  400,  500,  600,  700,  850,  925,
       1000])</pre></div></li>
```

```html
<li class='xr-var-item'><div class='xr-var-name'><span class='xr-has-index'>longitude</span></div><div class='xr-var-dims'>(longitude)</div><div class='xr-var-dtype'>float64</div><div class='xr-var-preview xr-preview'>0.0 5.625 11.25 ... 348.8 354.4</div><input id='attrs-cf6c1df6-07e7-4506-b1b1-648774f5ab6d' class='xr-var-attrs-in' type='checkbox' disabled><label for='attrs-cf6c1df6-07e7-4506-b1b1-648774f5ab6d' title='Show/Hide attributes'><svg class='icon xr-icon-file-text2'><use xlink:href='#icon-file-text2'></use></svg></label><input id='data-9b4e6ea1-bbe4-467a-99d1-0d0bc1c04804' class='xr-var-data-in' type='checkbox'><label for='data-9b4e6ea1-bbe4-467a-99d1-0d0bc1c04804' title='Show/Hide data repr'><svg class='icon xr-icon-database'><use xlink:href='#icon-database'></use></svg></label><div class='xr-var-attrs'><dl class='xr-attrs'></dl></div><div class='xr-var-data'><pre>array([  0.   ,   5.625,  11.25 ,  16.875,  22.5  ,  28.125,  33.75 ,  39.375,
        45.   ,  50.625,  56.25 ,  61.875,  67.5  ,  73.125,  78.75 ,  84.375,
        90.   ,  95.625, 101.25 , 106.875, 112.5  , 118.125, 123.75 , 129.375,
       135.   , 140.625, 146.25 , 151.875, 157.5  , 163.125, 168.75 , 174.375,
       180.   , 185.625, 191.25 , 196.875, 202.5  , 208.125, 213.75 , 219.375,
       225.   , 230.625, 236.25 , 241.875, 247.5  , 253.125, 258.75 , 264.375,
       270.   , 275.625, 281.25 , 286.875, 292.5  , 298.125, 303.75 , 309.375,
       315.   , 320.625, 326.25 , 331.875, 337.5  , 343.125, 348.75 , 354.375])</pre></div></li>
```

--------------------------------

### Open Zarr Archive with xarray

Source: https://weatherbench2.readthedocs.io/en/latest/_sources/data-guide

This code snippet demonstrates how to open a Zarr archive from a Google Cloud Storage bucket using the xarray library in Python. It assumes the 'xarray' library is installed. The function returns an xarray.Dataset object representing the data.

```python
import xarray as xr


# Open the Zarr archive from Google Cloud Storage
dataset = xr.open_zarr('gs://weatherbench2/datasets/ifs_ens/2018-2022-1440x721.zarr')
```

--------------------------------

### Pandas Index: longitude

Source: https://weatherbench2.readthedocs.io/en/latest/_sources/evaluation

Provides the Pandas Index for the 'longitude' variable. This index includes floating-point values representing longitude coordinates, starting from 0.0 and increasing in increments.

```python
PandasIndex(Index([               0.0,              5.625,              11.25,
                   16.875,               22.5,             28.125,
                    33.75,             39.375,               45.0,
                   50.625,              56.25,  61.87499999999999,
                     67.5,             73.125,              78.75,
                   84.375,               90.0,             95.625,
                   101.25,            106.875,              112.5,
                  118.125, 123.74999999999999,            129.375,
                    135.0,            140.625,             146.25,
                  151.875,              157.5,            163.125,
                   168.75,            174.375,              180.0,

```

--------------------------------

### Time Data Representation

Source: https://weatherbench2.readthedocs.io/en/latest/_sources/data-guide

Details the data representation for the 'time' dimension. This array contains datetime objects, showing timestamps for the data, including start and end dates for the year 2020.

```python
array(['2020-01-01T00:00:00.000000000', '2020-01-01T12:00:00.000000000',
       ...])
```

--------------------------------

### Pandas TimedeltaIndex for Prediction Intervals in WeatherBench2

Source: https://weatherbench2.readthedocs.io/en/latest/data-guide

This snippet illustrates the Pandas TimedeltaIndex for the 'prediction_timedelta' dimension in WeatherBench2. It defines various time intervals used for predictions, starting from 0 days.

```python
PandasIndex(TimedeltaIndex([ '0 days 00:00:00',  '0 days 06:00:00',  '0 days 12:00:00',
                 '0 days 18:00:00',  '1 days 00:00:00',  '1 days 06:00:00',
                 '1 days 12:00:00',  '1 days 18:00:00',  '2 days 00:00:00',
                 '2 days 06:00:00',  '2 days 12:00:00',  '2 days 18:00:00',
                 '3 days 00:00:00',  '3 days 06:00:00',  '3 days 12:00:00',
                 '3 days 18:00:00',  '4 days 00:00:00',  '4 days 06:00:00',
                 '4 days 12:00:00',  '4 days 18:00:00',  '5 days 00:00:00',
                 '5 days 12:00:00',  '6 days 00:00:00',  '6 days 12:00:00',
                 '7 days 00:00:00',  '7 days 12:00:00',  '8 days 00:00:00',
                 '8 days 12:00:00',  '9 days 00:00:00',  '9 days 12:00:00',
                '10 days 00:00:00'],
               dtype='timedelta64[ns]', name='prediction_timedelta', freq=None))

```

--------------------------------

### Pandas TimedeltaIndex Example

Source: https://weatherbench2.readthedocs.io/en/latest/data-guide

This code snippet displays a Pandas TimedeltaIndex, which is used to represent time differences. It's commonly employed in datasets to denote prediction lead times or intervals between forecasts.

```python
from pandas import TimedeltaIndex

prediction_timedelta = TimedeltaIndex([
 '0 days 00:00:00',
 '0 days 12:00:00',
 '1 days 00:00:00',
 '1 days 12:00:00',
 '2 days 00:00:00',
 '2 days 12:00:00',
 '3 days 00:00:00',
 '3 days 12:00:00',
 '4 days 00:00:00',
 '4 days 12:00:00',
 '5 days 00:00:00',
 '5 days 12:00:00',
 '6 days 00:00:00',
 '6 days 12:00:00',
 '7 days 00:00:00',
 '7 days 12:00:00',
 '8 days 00:00:00',
 '8 days 12:00:00',
 '9 days 00:00:00',
 '9 days 12:00:00',
 '10 days 00:00:00',
 '10 days 12:00:00',
 '11 days 00:00:00',
 '11 days 12:00:00',
 '12 days 00:00:00',
 '12 days 12:00:00',
 '13 days 00:00:00',
 '13 days 12:00:00',
 '14 days 00:00:00',
 '14 days 12:00:00',
 '15 days 00:00:00'
], dtype='timedelta64[ns]', name='prediction_timedelta', freq=None))
```

--------------------------------

### Pandas DatetimeIndex Example

Source: https://weatherbench2.readthedocs.io/en/latest/data-guide

This code snippet shows a Pandas DatetimeIndex, which represents a sequence of dates and times. It's crucial for time-series analysis, marking specific points in time for weather observations or forecasts.

```python
from pandas import DatetimeIndex

time = DatetimeIndex([
 '2020-01-01 00:00:00',
 '2020-01-01 12:00:00',
 '2020-01-02 00:00:00',
 '2020-01-02 12:00:00',
 '2020-01-03 00:00:00',
 '2020-01-03 12:00:00',
 '2020-01-04 00:00:00',
 '2020-01-04 12:00:00',
 '2020-01-05 00:00:00',
 '2020-01-05 12:00:00',
 '2020-12-31 00:00:00',
 '2020-12-31 12:00:00'
], dtype='datetime64[ns]', name='time', length=732, freq=None))
```

--------------------------------

### Access Pangu-Weather Model Information

Source: https://weatherbench2.readthedocs.io/en/latest/_sources/data-guide

This section provides information about the Pangu-Weather model, including its origin, GitHub repository link, and the locations of its datasets stored in Google Cloud Storage. It also lists the available data files for different resolutions and time periods.

```markdown
<!-- (pangu)= -->
### Pangu-Weather

We ran the [Pangu](https://www.nature.com/articles/s41586-023-06185-3) model using the code available on [GitHub](https://github.com/198808xc/Pangu-Weather).  

Location: [`gs://weatherbench2/datasets/pangu/`](https://console.cloud.google.com/storage/browser/weatherbench2/datasets/pangu/)

Files:
* `2018-2022_0012_0p25.zarr`
* `2018-2022_0012_240x121_equiangular_with_poles_conservative.zarr`
* `2018-2022_0012_64x32_equiangular_conservative.zarr`
```

--------------------------------

### Evaluate Probabilistic Climatology vs. ERA5 (1440x721)

Source: https://weatherbench2.readthedocs.io/en/latest/official-evaluation

This command evaluates probabilistic climatology against ERA5 reanalysis data for a 1440x721 resolution. It specifies forecast, observation (ERA5), and climatology paths, along with output configurations. This example enables the evaluation of probabilistic climatology and sets the start and end years for the climatology period. It also configures input chunking, fanout, regions, and the list of variables.

```python
python evaluate.py -- \
  --forecast_path=gs://weatherbench2/datasets/ens/2018-2022-1440x721.zarr \
  --obs_path=gs://weatherbench2/datasets/era5/1959-2022-6h-1440x721.zarr \
  --climatology_path=gs://weatherbench2/datasets/era5-hourly-climatology/1990-2019_6h_1440x721.zarr \
  --output_dir=$OUTDIR/1440x721/probabilistic/ \
  --output_file_prefix=climatology_vs_era_2020_ \
  --input_chunks=init_time=1,lead_time=1 \
  --fanout=27 \
  --regions=all \
  --eval_configs=probabilistic \
  --evaluate_probabilistic_climatology=True \
  --probabilistic_climatology_start_year=1990 \
  --probabilistic_climatology_end_year=2019 \
  --time_start=2020-01-01 \
  --time_stop=2020-12-31 \
  --variables=geopotential,temperature,u_component_of_wind,v_component_of_wind,specific_humidity,2m_temperature,10m_u_component_of_wind,10m_v_component_of_wind,mean_sea_level_pressure,total_precipitation_6hr,total_precipitation_24hr,10m_wind_speed,wind_speed \
  --use_beam=True

```

--------------------------------

### Import Evaluation Functions in Python

Source: https://weatherbench2.readthedocs.io/en/latest/evaluation

Imports the necessary functions to run the evaluation process: `evaluate_in_memory` for smaller datasets and `evaluate_with_beam` for larger datasets using Apache Beam.

```python
from weatherbench2.evaluation import evaluate_in_memory, evaluate_with_beam
```

--------------------------------

### Import WeatherBench 2 Libraries

Source: https://weatherbench2.readthedocs.io/en/latest/evaluation

Imports essential libraries for using the WeatherBench 2 framework, including apache_beam, weatherbench2, and xarray. Importing apache_beam separately is noted as a workaround for potential TypingErrors.

```python
import apache_beam   # Needs to be imported separately to avoid TypingError
import weatherbench2
import xarray as xr
```

--------------------------------

### Define Time Index in Pandas

Source: https://weatherbench2.readthedocs.io/en/latest/_sources/data-guide

This code illustrates the creation of a Pandas DatetimeIndex, representing the time dimension of the dataset. It includes the start and end dates, the data type, and the name of the index. This is fundamental for time-series analysis.

```python
PandasIndex(DatetimeIndex(['2020-01-01 00:00:00', '2020-01-01 12:00:00',
       '2020-01-02 00:00:00', '2020-01-02 12:00:00',
       '2020-01-03 00:00:00', '2020-01-03 12:00:00',
       '2020-01-04 00:00:00', '2020-01-04 12:00:00',
       '2020-01-05 00:00:00', '2020-01-05 12:00:00',
       ...
       '2020-12-27 00:00:00', '2020-12-27 12:00:00',
       '2020-12-28 00:00:00', '2020-12-28 12:00:00',
       '2020-12-29 00:00:00', '2020-12-29 12:00:00',
       '2020-12-30 00:00:00', '2020-12-30 12:00:00',
       '2020-12-31 00:00:00', '2020-12-31 12:00:00'],
              dtype='datetime64[ns]', name='time', length=732, freq=None))
```

--------------------------------

### Evaluate Pangu-HRES-Init vs. Analysis (Python)

Source: https://weatherbench2.readthedocs.io/en/latest/official-evaluation

This command evaluates the Pangu-HRES-Init model against analysis data, likely from HRES. It specifies paths for forecast, observation (analysis), land surface model datasets, and climatology. The evaluation focuses on deterministic and temporal configurations using Apache Beam.

```python
python evaluate.py -- \
  --forecast_path=gs://weatherbench2/datasets/pangu_hres_init/2020_0012_0p25.zarr \
  --obs_path=gs://weatherbench2/datasets/hres_t0/2016-2022-6h-1440x721.zarr \
  --lsm_dataset=gs://weatherbench2/datasets/era5/1959-2022-6h-1440x721.zarr \
  --climatology_path=gs://weatherbench2/datasets/era5-hourly-climatology/1990-2019_6h_1440x721.zarr \
  --output_dir=$OUTDIR/1440x721/deterministic/ \
  --output_file_prefix=pangu_hres_init_vs_analysis_2020_ \
  --input_chunks=init_time=1,lead_time=1 \
  --fanout=27 \
  --regions=all \
  --eval_configs=deterministic,deterministic_temporal \
  --evaluate_climatology=False \
  --evaluate_persistence=False \
  --time_start=2020-01-01 \
  --time_stop=2020-12-31 \
  --variables=geopotential,temperature,u_component_of_wind,v_component_of_wind,specific_humidity,2m_temperature,10m_u_component_of_wind,10m_v_component_of_wind,mean_sea_level_pressure \
  --add_land_region=False \
  --use_beam=True
```

--------------------------------

### Pandas Indexing: Datetime

Source: https://weatherbench2.readthedocs.io/en/latest/_sources/data-guide

This snippet shows the creation of a Pandas DatetimeIndex. It represents a series of dates and times, starting from '2018-01-01 00:00:00' and including hourly and daily intervals. This index is useful for time-series data analysis.

```python
PandasIndex(DatetimeIndex(['2018-01-01 00:00:00', '2018-01-01 12:00:00',
               '2018-01-02 00:00:00', '2018-01-02 12:00:00',
               '2018-01-03 00:00:00', '2018-01-03 12:00:00',
               '2018-01-04 00:00:00', '2018-01-04 12:00:00',
               '2018-01-05 00:00:00', '2018-01-05 12:00:00',
               ...
               '2022-12-27 00:00:00', '2022-12-27 12:00:00',
               '2022-12-28 00:00:00', '2022-12-28 12:00:00'],
              dtype='datetime64[ns]', name='time', freq=None))
```

--------------------------------

### Configure Data Paths

Source: https://weatherbench2.readthedocs.io/en/latest/_sources/evaluation

Defines the file paths for forecast data, observational data, and the output directory for evaluation results. This configuration is essential for specifying where the data resides and where the results should be saved.

```python
paths = config.Paths(
    forecast=forecast_path,
    obs=obs_path,
    output_dir='./'   # Directory to save evaluation results
)
```

--------------------------------

### Pandas DatetimeIndex for Time in Python

Source: https://weatherbench2.readthedocs.io/en/latest/data-guide

This code snippet shows the 'time' index using Pandas DatetimeIndex. It represents daily and sub-daily timestamps starting from '2016-01-01 00:00:00'. This index is fundamental for time-series analysis and forecasting.

```python
PandasIndex(DatetimeIndex(['2016-01-01 00:00:00', '2016-01-01 12:00:00',
               '2016-01-02 00:00:00', '2016-01-02 12:00:00',
               '2016-01-03 00:00:00', '2016-01-03 12:00:00',
               '2016-01-04 00:00:00', '2016-01-04 12:00:00',
               '2016-01-05 00:00:00', '2016-01-05 12:00:00',
               ... (truncated for brevity)
               '2023-01-06 00:00:00', '2023-01-06 12:00:00',
               '2023-01-07 00:00:00', '2023-01-07 12:00:00',
               '2023-01-08 00:00:00', '2023-01-08 12:00:00',
               '2023-01-09 00:00:00', '2023-01-09 12:00:00',
               '2023-01-10 00:00:00', '2023-01-10 12:00:00'],
              dtype='datetime64[ns]', name='time', freq=None))
```

--------------------------------

### Evaluate Forecasts with Specific Parameters (Shell)

Source: https://weatherbench2.readthedocs.io/en/latest/official-evaluation

This command line shows how to execute an evaluation script with various parameters. It includes settings for output file prefixes, input chunking, fanout, regions, evaluation configurations, time range, variables, and Beam usage.

```shell
--output_file_prefix=era5-forecasts_vs_era_2020_ \
  --input_chunks=init_time=1 \
  --fanout=27 \
  --regions=all \
  --eval_configs=deterministic,deterministic_temporal \
  --evaluate_climatology=False \
  --evaluate_persistence=False \
  --time_start=2020-01-01 \
  --time_stop=2020-12-31 \
  --variables=geopotential,temperature,u_component_of_wind,v_component_of_wind,specific_humidity,2m_temperature,10m_u_component_of_wind,10m_v_component_of_wind,mean_sea_level_pressure,10m_wind_speed,wind_speed \
  --use_beam=True
```

--------------------------------

### PandasIndex for Time Series Data

Source: https://weatherbench2.readthedocs.io/en/latest/data-guide

This code demonstrates a Pandas DatetimeIndex for time series data, starting from '2016-01-01 00:00:00' with a frequency of 6 hours. It covers a significant period, suitable for climate analysis and model evaluation.

```python
PandasIndex(DatetimeIndex(['2016-01-01 00:00:00', '2016-01-01 06:00:00',
               '2016-01-01 12:00:00', '2016-01-01 18:00:00',
               '2016-01-02 00:00:00', '2016-01-02 06:00:00',
               '2016-01-02 12:00:00', '2016-01-02 18:00:00',
               '2016-01-03 00:00:00', '2016-01-03 06:00:00',
               ... (10268 values total)
               '2023-01-08 12:00:00', '2023-01-08 18:00:00',
               '2023-01-09 00:00:00', '2023-01-09 06:00:00',
               '2023-01-09 12:00:00', '2023-01-09 18:00:00',
               '2023-01-10 00:00:00', '2023-01-10 06:00:00',
               '2023-01-10 12:00:00', '2023-01-10 18:00:00'],
              dtype='datetime64[ns]', name='time', length=10268, freq=None))
```

--------------------------------

### Evaluate Pangu-HRES-Init vs. ERA5 (Python)

Source: https://weatherbench2.readthedocs.io/en/latest/official-evaluation

This script evaluates the Pangu-HRES-Init model against ERA5 reanalysis data. It configures input and output paths, evaluation settings, and the specific variables to be analyzed. Apache Beam is used for efficient data processing.

```python
python evaluate.py -- \
  --forecast_path=gs://weatherbench2/datasets/pangu_hres_init/2020_0012_0p25.zarr \
  --obs_path=gs://weatherbench2/datasets/era5/1959-2022-6h-1440x721.zarr \
  --climatology_path=gs://weatherbench2/datasets/era5-hourly-climatology/1990-2019_6h_1440x721.zarr \
  --output_dir=$OUTDIR/1440x721/deterministic/ \
  --output_file_prefix=pangu_hres_init_vs_era_2020_ \
  --input_chunks=init_time=1,lead_time=1 \
  --fanout=27 \
  --regions=all \
  --eval_configs=deterministic,deterministic_temporal \
  --evaluate_climatology=False \
  --evaluate_persistence=False \
  --time_start=2020-01-01 \
  --time_stop=2020-12-31 \
  --variables=geopotential,temperature,u_component_of_wind,v_component_of_wind,specific_humidity,2m_temperature,10m_u_component_of_wind,10m_v_component_of_wind,mean_sea_level_pressure \
  --add_land_region=False \
  --use_beam=True
```

--------------------------------

### Evaluate Pangu-HRES Initialization Forecasts (Python)

Source: https://weatherbench2.readthedocs.io/en/latest/official-evaluation

This script evaluates forecasts from Pangu-HRES with specific initialization settings against ERA5 data. It configures the evaluation for the year 2020, specifying Zarr paths for forecasts and observations, climatology, and output settings. It also includes options for land region analysis.

```python
python evaluate.py -- \
 --forecast_path=gs://weatherbench2/datasets/pangu_hres_init/2020_0012_240x121_equiangular_with_poles_conservative.zarr \
 --obs_path=gs://weatherbench2/datasets/era5/1959-2022-6h-240x121_equiangular_with_poles_conservative.zarr \
 --climatology_path=gs://weatherbench2/datasets/era5-hourly-climatology/1990-2019_6h_240x121_equiangular_with_poles_conservative.zarr \
 --output_dir=$OUTDIR/240x121/deterministic/ \
 --output_file_prefix=pangu_hres_init_vs_era_2020_ \
 --input_chunks=init_time=1,lead_time=10 \
 --fanout=27 \
 --regions=all \
 --eval_configs=deterministic,deterministic_temporal \
 --evaluate_climatology=False \
 --evaluate_persistence=False \
 --time_start=2020-01-01 \
 --time_stop=2020-12-31 \
 --variables=geopotential,temperature,u_component_of_wind,v_component_of_wind,specific_humidity,2m_temperature,10m_u_component_of_wind,10m_v_component_of_wind,mean_sea_level_pressure \
 --add_land_region=False \
 --use_beam=True
```

```python
python evaluate.py -- \
 --forecast_path=gs://weatherbench2/datasets/pangu_hres_init/2020_0012_240x121_equiangular_with_poles_conservative.zarr \
 --obs_path=gs://weatherbench2/datasets/hres_t0/2016-2022-6h-240x121_equiangular_with_poles_conservative.zarr \
 --lsm_dataset=gs://weatherbench2/datasets/era5/1959-2022-6h-240x121_equiangular_with_poles_conservative.zarr \
 --climatology_path=gs://weatherbench2/datasets/era5-hourly-climatology/1990-2019_6h_240x121_equiangular_with_poles_conservative.zarr \
 --output_dir=$OUTDIR/240x121/deterministic/ \
 --output_file_prefix=pangu_hres_init_vs_analysis_2020_ \
 --input_chunks=init_time=1,lead_time=10 \
 --fanout=27 \
 --regions=all \
 --eval_configs=deterministic,deterministic_temporal \
 --evaluate_climatology=False \
 --evaluate_persistence=False \
 --time_start=2020-01-01 \
 --time_stop=2020-12-31 \
 --variables=geopotential,temperature,u_component_of_wind,v_component_of_wind,specific_humidity,2m_temperature,10m_u_component_of_wind,10m_v_component_of_wind,mean_sea_level_pressure \
 --add_land_region=False \
 --use_beam=True
```
Tokens:
Raw