Model

The Model class provides access to weather forecast data from Jua's weather prediction models. It allows you to retrieve future weather predictions with flexible spatial and temporal selection options.

Historical weather data (known as hindcasts) are also available directly directly through the Model class. Hindcasts are past model runs that allow you to analyze historical weather patterns, evaluate model performance, or test your pipelines. All model runs that are older that 90 days (i.e. an having an init_time that is more than 90 days in the past) are classified as hindcasts.

Information about model/query pricing is available in the developer portal.

Overview

The Model class serves as your interface to Jua's weather forecasting capabilities. It provides methods to:

• Retrieve forecast data for global coverage or specific locations

• Access forecasts for specific initialization times or the latest available forecast

• Query metadata about available forecasts

• Check forecast availability and readiness

List Available Models

Code

from jua.weather import Models

print("Available models:")
for model in Models:
    print(f"  - {model}")

Output

Available models:
  - Models.EPT1_5
  - Models.EPT1_5_EARLY
  - Models.EPT2
  - Models.EPT2_E
  - Models.EPT2_EARLY
  - Models.EPT2_HRRR
  - Models.EPT2_RR
  - Models.EPT2_REASONING
  - Models.AIFS
  - Models.AURORA
  - Models.ECMWF_IFS_SINGLE
  - Models.ICON_EU
  - Models.ICON_GLOBAL
  - Models.NOAA_GFS_SINGLE
  - Models.ECMWF_AIFS_ENSEMBLE
  - Models.ECMWF_AIFS_SINGLE
  - Models.ECMWF_IFS_ENSEMBLE
  - Models.GFS_GLOBAL_ENSEMBLE
  - Models.GFS_GLOBAL_SINGLE
  - Models.GFS_GRAPHCAST
  - Models.ICON_D2
  - Models.KNMI_HARMONIE_AROME_EUROPE
  - Models.KNMI_HARMONIE_AROME_NETHERLANDS
  - Models.METEOFRANCE_AROME_FRANCE_HD
  - Models.UKMO_GLOBAL_DETERMINISTIC_10KM
  - Models.UKMO_UK_DETERMINISTIC_2KM

Request model metadata

Code

import json

from jua import JuaClient
from jua.weather import Models

client = JuaClient()
model = client.weather.get_model(Models.EPT1_5)
metadata = model.get_metadata()
print(json.dumps(metadata.model_dump(), indent=4, default=str))

Output

{
    "model": "ept1_5",
    "variables": [
        "air_temperature_at_height_level_2m",
        "wind_speed_at_height_level_10m",
        "wind_direction_at_height_level_10m",
        "wind_speed_at_height_level_100m",
        "wind_direction_at_height_level_100m",
        "air_pressure_at_mean_sea_level",
        "geopotential_at_pressure_level_50000Pa",
        "surface_downwelling_shortwave_flux_sum_1h"
    ],
    "grid": {
        "num_latitudes": 2221,
        "num_longitudes": 4440
    }
}

Getting the Forecasts Available for a Model

Code

from jua import JuaClient
from jua.weather import Models

client = JuaClient()
model = client.weather.get_model(Models.EPT2)

# The available forecasts will be returned in descending order,
#  from most recent to oldest; this returns the 10 most recent forecasts
available_forecasts = model.get_available_forecasts(limit=10)
for forecast in available_forecasts:
    print(f" - {forecast.init_time}")

Output

Available init times for ept2
 - 2025-10-30 06:00:00
 - 2025-10-30 00:00:00
 - 2025-10-29 18:00:00
 - 2025-10-29 12:00:00
 - 2025-10-29 06:00:00
 - 2025-10-29 00:00:00
 - 2025-10-28 18:00:00
 - 2025-10-28 12:00:00
 - 2025-10-28 06:00:00
 - 2025-10-28 00:00:00

The get_available_forecasts forecast response is paginated. So to obtain all available forecasts for a model, you need to iterate through the responses.

Code

from jua import JuaClient
from jua.weather import Models

client = JuaClient()
model = client.weather.get_model(Models.EPT2)

# Paginate through the response to get all available init_times
result = model.get_available_forecasts(limit=1000)
available_forecasts = list(result.forecasts)
while result.has_more:
    result = result.next()
    available_forecasts.extend(result.forecasts)

print(f"Available init times for {model}: {len(available_forecasts)}")

Output

Available init times for ept2: 4134

You can also specify a range of dates for which you want to get available init_times.

Code

from datetime import datetime

from jua import JuaClient
from jua.weather import Models

client = JuaClient()
model = client.weather.get_model(Models.EPT2)

# get the available hindcasts from EPT-2 for January 2025
available_forecasts = model.get_available_forecasts(
    since=datetime(2025, 1, 1),
    before=datetime(2025, 2, 1),
    limit=200,
)
for forecast in available_forecasts:
    print(f" - {forecast.init_time}")

Output

 - 2025-02-01 00:00:00
 - 2025-01-31 18:00:00
 - 2025-01-31 12:00:00
 - 2025-01-31 06:00:00
 - 2025-01-31 00:00:00
 - 2025-01-30 18:00:00
 - 2025-01-30 12:00:00
 - 2025-01-30 06:00:00
 - 2025-01-30 00:00:00
 - 2025-01-29 18:00:00
 - 2025-01-29 12:00:00
 - 2025-01-29 06:00:00
 - 2025-01-29 00:00:00
 - 2025-01-28 18:00:00
 - 2025-01-28 12:00:00
 - 2025-01-28 06:00:00
 - 2025-01-28 00:00:00
 - 2025-01-27 18:00:00
 - 2025-01-27 12:00:00
 - 2025-01-27 06:00:00
 - 2025-01-27 00:00:00
 - 2025-01-26 18:00:00
 - 2025-01-26 12:00:00
 - 2025-01-26 06:00:00
 - 2025-01-26 00:00:00
 - 2025-01-25 18:00:00
 - 2025-01-25 12:00:00
 - 2025-01-25 06:00:00
 - 2025-01-25 00:00:00
 - 2025-01-24 18:00:00
 - 2025-01-24 12:00:00
 - 2025-01-24 06:00:00
 - 2025-01-24 00:00:00
 - 2025-01-23 18:00:00
 - 2025-01-23 12:00:00
 - 2025-01-23 06:00:00
 - 2025-01-23 00:00:00
 - 2025-01-22 18:00:00
 - 2025-01-22 12:00:00
 - 2025-01-22 06:00:00
 - 2025-01-22 00:00:00
 - 2025-01-21 18:00:00
 - 2025-01-21 12:00:00
 - 2025-01-21 06:00:00
 - 2025-01-21 00:00:00
 - 2025-01-20 18:00:00
 - 2025-01-20 12:00:00
 - 2025-01-20 06:00:00
 - 2025-01-20 00:00:00
 - 2025-01-19 18:00:00
 - 2025-01-19 12:00:00
 - 2025-01-19 06:00:00
 - 2025-01-19 00:00:00
 - 2025-01-18 18:00:00
 - 2025-01-18 12:00:00
 - 2025-01-18 06:00:00
 - 2025-01-18 00:00:00
 - 2025-01-17 18:00:00
 - 2025-01-17 12:00:00
 - 2025-01-17 06:00:00
 - 2025-01-17 00:00:00
 - 2025-01-16 18:00:00
 - 2025-01-16 12:00:00
 - 2025-01-16 06:00:00
 - 2025-01-16 00:00:00
 - 2025-01-15 18:00:00
 - 2025-01-15 12:00:00
 - 2025-01-15 06:00:00
 - 2025-01-15 00:00:00
 - 2025-01-14 18:00:00
 - 2025-01-14 12:00:00
 - 2025-01-14 06:00:00
 - 2025-01-14 00:00:00
 - 2025-01-13 18:00:00
 - 2025-01-13 12:00:00
 - 2025-01-13 06:00:00
 - 2025-01-13 00:00:00
 - 2025-01-12 18:00:00
 - 2025-01-12 12:00:00
 - 2025-01-12 06:00:00
 - 2025-01-12 00:00:00
 - 2025-01-11 18:00:00
 - 2025-01-11 12:00:00
 - 2025-01-11 06:00:00
 - 2025-01-11 00:00:00
 - 2025-01-10 18:00:00
 - 2025-01-10 12:00:00
 - 2025-01-10 06:00:00
 - 2025-01-10 00:00:00
 - 2025-01-09 18:00:00
 - 2025-01-09 12:00:00
 - 2025-01-09 06:00:00
 - 2025-01-09 00:00:00
 - 2025-01-08 18:00:00
 - 2025-01-08 12:00:00
 - 2025-01-08 06:00:00
 - 2025-01-08 00:00:00
 - 2025-01-07 18:00:00
 - 2025-01-07 12:00:00
 - 2025-01-07 06:00:00
 - 2025-01-07 00:00:00
 - 2025-01-06 18:00:00
 - 2025-01-06 12:00:00
 - 2025-01-06 06:00:00
 - 2025-01-06 00:00:00
 - 2025-01-05 18:00:00
 - 2025-01-05 12:00:00
 - 2025-01-05 06:00:00
 - 2025-01-05 00:00:00
 - 2025-01-04 18:00:00
 - 2025-01-04 12:00:00
 - 2025-01-04 06:00:00
 - 2025-01-04 00:00:00
 - 2025-01-03 18:00:00
 - 2025-01-03 12:00:00
 - 2025-01-03 06:00:00
 - 2025-01-03 00:00:00
 - 2025-01-02 18:00:00
 - 2025-01-02 12:00:00
 - 2025-01-02 06:00:00
 - 2025-01-02 00:00:00
 - 2025-01-01 18:00:00
 - 2025-01-01 12:00:00
 - 2025-01-01 06:00:00
 - 2025-01-01 00:00:00

Check if the forecasted data is available

Code

from jua import JuaClient
from jua.weather import Models

client = JuaClient()
model = client.weather.get_model(Models.EPT1_5)

# init_time might be any valid datetime object or datetime string
if model.is_ready(forecasted_hours=480, init_time="latest"):
    print(f"{model.name} has at least 480 hours of forecast data available")
else:
    print(f"{model.name} does not yet have 480 hours of forecast data available")

Output

ept1_5 has at least 480 hours of forecast data available

Requesting the forecast data

Point-Based

from jua import JuaClient
from jua.types.geo import LatLon
from jua.weather import Models, Variables

# Setup JuaClient
client = JuaClient()

# Choose the model
model = client.weather.get_model(Models.EPT1_5)

# Defining the sites that we are interested in
sites = [
    # label is optional
    LatLon(55.06, 13.00, label="Kriegers Flak"),
    LatLon(54.04, 5.96, label="Gemini Wind Farm"),
    LatLon(51.71, 2.91, label="Borssele III & IV"),
    LatLon(53.89, 1.79, label="Hornsea Two"),
]

# Define the variables we are interested in
# If set to `None` all available variables will be loaded
variables = [
    Variables.WIND_SPEED_AT_HEIGHT_LEVEL_100M,
]

site_forecasts = model.get_forecasts(
    init_time="latest", # or any valid datetime / datetime string
    points=sites,
    # The below parameters are optional
    variables=variables,
    min_lead_time=0, # hours
    max_lead_time=72, # hours
    method="bilinear",
)

# Convert the response to an xarray dataset
print(site_forecasts.to_xarray()) 

<xarray.Dataset> Size: 2kB
Dimensions:                          (points: 4, init_time: 1,
                                      prediction_timedelta: 73)
Coordinates:
  * points                           (points) object 32B 'borssele_iii_&_iv' ...
  * init_time                        (init_time) datetime64[ns] 8B 2025-10-30...
  * prediction_timedelta             (prediction_timedelta) timedelta64[ns] 584B ...
    latitude                         (points) float32 16B 51.71 54.04 ... 55.06
    longitude                        (points) float32 16B 2.91 5.96 1.79 13.0
    requested_lat                    (points) float64 32B 51.71 54.04 ... 55.06
    requested_lon                    (points) float64 32B 2.91 5.96 1.79 13.0
Data variables:
    wind_speed_at_height_level_100m  (points, init_time, prediction_timedelta) float32 1kB ...

⚠️ When querying data with method="nearest" the returned coordinates will be slightly different to the requested ones, as it will return the coordinates of the nearest grid points for the model.

Region-Based

⚠️ Region-based requests can be expensive. For more information, see our pricing docs.

⚠️ Requests might be large and take a long time to load.

from datetime import datetime

from jua import JuaClient
from jua.weather import Models, Variables

client = JuaClient()
model = client.weather.get_model(Models.EPT2)
 
# Define the data we want to load
# In all cases scalars, lists and slices are supported
# If set to `None` the whole dimension will be loaded
europe_lat_slice = slice(36, 71) 
europe_lon_slice = slice(-15, 50)
forecast_window = [0, 12, 24]  # Hours
variables = [
    Variables.AIR_TEMPERATURE_AT_HEIGHT_LEVEL_2M,
    Variables.WIND_SPEED_AT_HEIGHT_LEVEL_100M,
]

forecast = model.get_forecasts(
    init_time=datetime(2025, month=7, day=1, hour=0),
    variables=variables,
    latitude=europe_lat_slice,
    longitude=europe_lon_slice,
    prediction_timedelta=forecast_window,
)

print(forecast.to_xarray())

Reading data... | 35.8 MB | 30.2 MB/s
<xarray.Dataset> Size: 8MB
Dimensions:                             (init_time: 1, prediction_timedelta: 3,
                                         latitude: 421, longitude: 781)
Coordinates:
  * init_time                           (init_time) datetime64[ns] 8B 2025-07-01
  * prediction_timedelta                (prediction_timedelta) timedelta64[ns] 24B ...
  * latitude                            (latitude) float32 2kB 36.0 ... 71.0
  * longitude                           (longitude) float32 3kB -15.0 ... 50.0
Data variables:
    air_temperature_at_height_level_2m  (init_time, prediction_timedelta, latitude, longitude) float32 4MB ...
    wind_speed_at_height_level_100m     (init_time, prediction_timedelta, latitude, longitude) float32 4MB ...

Requesting Hindcasts for Models

Hindcasts can be obtained directly through the same get_forecasts method.

Point-Based

from datetime import datetime
from jua import JuaClient
from jua.types.geo import LatLon
from jua.weather import Models, Variables

# Setup JuaClient
client = JuaClient()

# Choose the model
model = client.weather.get_model(Models.EPT2)

# Defining the sites that we are interested in
sites = [
    # label is optional
    LatLon(55.06, 13.00, label="Kriegers Flak"),
    LatLon(54.04, 5.96, label="Gemini Wind Farm"),
    LatLon(51.71, 2.91, label="Borssele III & IV"),
    LatLon(53.89, 1.79, label="Hornsea Two"),
]

# Define the variables we are interested in
# If set to `None` all available variables will be loaded
variables = [
    Variables.WIND_SPEED_AT_HEIGHT_LEVEL_10M,
]

site_forecasts = model.get_forecasts(
    init_time=slice(
        datetime(2025, month=1, day=1),
        datetime(2025, month=3, day=31, hour=23, minute=59),
    ),
    points=sites,
    # The below parameters are optional
    variables=variables,
    min_lead_time=0, # hours
    max_lead_time=72, # hours
    method="bilinear",
)

# Convert the response to an xarray dataset
print(site_forecasts.to_xarray()) 

<xarray.Dataset> Size: 424kB
Dimensions:                         (points: 4, init_time: 360,
                                     prediction_timedelta: 73)
Coordinates:
  * points                          (points) object 32B 'borssele_iii_&_iv' ....
  * init_time                       (init_time) datetime64[ns] 3kB 2025-01-01...
  * prediction_timedelta            (prediction_timedelta) timedelta64[ns] 584B ...
    latitude                        (points) float32 16B 51.71 54.04 53.89 55.06
    longitude                       (points) float32 16B 2.91 5.96 1.79 13.0
    requested_lat                   (points) float64 32B 51.71 54.04 53.89 55.06
    requested_lon                   (points) float64 32B 2.91 5.96 1.79 13.0
Data variables:
    wind_speed_at_height_level_10m  (points, init_time, prediction_timedelta) float32 420kB ...

⚠️ When querying data with method="nearest" the returned coordinates will be slightly different to the requested ones, as it will return the coordinates of the nearest grid points for the model.

Region-Based

⚠️ Region-based requests can be expensive. For more information, see our pricing docs. In this example, the request_credit_limit needs to be increased for the JuaClient or the request will fail.

⚠️ Requests might be large and take a long time to load.

⚠️ When pulling large grids, it's better to make smaller requests and combine the data locally (also so that your local memory does not run out)

europe_lat_slice = slice(36, 71) 
europe_lon_slice = slice(-15, 50)
variables = [
    Variables.AIR_TEMPERATURE_AT_HEIGHT_LEVEL_2M,
    Variables.WIND_SPEED_AT_HEIGHT_LEVEL_100M,
]

forecast = model.get_forecasts(
    init_time=slice(
        datetime(2025, month=1, day=1, hour=0),
        datetime(2025, month=1, day=1, hour=18),
    ),
    variables=variables,
    latitude=europe_lat_slice,
    longitude=europe_lon_slice,
    max_lead_time=2 * 24,
)

<xarray.Dataset> Size: 516MB
Dimensions:                             (init_time: 4,
                                         prediction_timedelta: 49,
                                         latitude: 421, longitude: 781)
Coordinates:
  * init_time                           (init_time) datetime64[ns] 32B 2025-0...
  * prediction_timedelta                (prediction_timedelta) timedelta64[ns] 392B ...
  * latitude                            (latitude) float32 2kB 36.0 ... 71.0
  * longitude                           (longitude) float32 3kB -15.0 ... 50.0
Data variables:
    air_temperature_at_height_level_2m  (init_time, prediction_timedelta, latitude, longitude) float32 258MB ...
    wind_speed_at_height_level_100m     (init_time, prediction_timedelta, latitude, longitude) float32 258MB ...