Using met.no forecasts with weather-tools¶

This notebook-style script shows how to fetch met.no forecasts and combine them

with SILO observations using the new weather_tools APIs. Open it directly in

JupyterLab or VS Code with Jupytext to run the cells interactively.

Prerequisites¶

• Network access to https://api.met.no and a valid contact e-mail for the User-Agent header.

• Set SILO_API_KEY (your registered SILO e-mail) or pass api_key= directly to SiloAPI.

• Install weather-tools in your environment (either via pip install weather-tools or the local checkout).

• Optional: enable logging (e.g. logging.basicConfig(level=logging.INFO)) to inspect requests.

In [1]:

import pandas as pd

from weather_tools.merge_weather_data import get_merge_summary, merge_historical_and_forecast
from weather_tools.metno_api import MetNoAPI
from weather_tools.metno_models import MetNoFormat, MetNoQuery
from weather_tools.silo_api import SiloAPI
from weather_tools.silo_models import AustralianCoordinates

import pandas as pd from weather_tools.merge_weather_data import get_merge_summary, merge_historical_and_forecast from weather_tools.metno_api import MetNoAPI from weather_tools.metno_models import MetNoFormat, MetNoQuery from weather_tools.silo_api import SiloAPI from weather_tools.silo_models import AustralianCoordinates

In [2]:

# silo_api = SiloAPI(log_level="INFO")
# station_meta = silo_api.search_stations("Northam").iloc[0]

station_meta = {
    "station_code": 10111,
    "name": "NORTHAM",
    "latitude": -31.651,
    "longitude": 116.659,
    "state": "WA",
    "elevation": 170.0,
}

# silo_api = SiloAPI(log_level="INFO") # station_meta = silo_api.search_stations("Northam").iloc[0] station_meta = { "station_code": 10111, "name": "NORTHAM", "latitude": -31.651, "longitude": 116.659, "state": "WA", "elevation": 170.0, }

1. Configure the met.no API client¶

met.no requires a descriptive User-Agent string that includes contact details.

Replace the placeholder below with your own application name and e-mail.

In [3]:

api = MetNoAPI(
    # user_agent="weather-tools-example/0.1 (contact: you@example.com)",
    enable_cache=True,
    log_level="DEBUG",
)

api = MetNoAPI( # user_agent="weather-tools-example/0.1 (contact: you@example.com)", enable_cache=True, log_level="DEBUG", )

2. Build a forecast query for your location¶

The AustralianCoordinates model validates that the latitude and longitude sit

within the range supported by SILO (GDA94 datum). Adjust the coordinates to

target your site of interest.

In [4]:

coordinates = AustralianCoordinates(
    latitude=station_meta["latitude"], longitude=station_meta["longitude"]
)  # Northam, WA
query = MetNoQuery(coordinates=coordinates, format=MetNoFormat.COMPACT)

response = api.query_forecast(query)
metadata = response.get_meta()

metadata

coordinates = AustralianCoordinates( latitude=station_meta["latitude"], longitude=station_meta["longitude"] ) # Northam, WA query = MetNoQuery(coordinates=coordinates, format=MetNoFormat.COMPACT) response = api.query_forecast(query) metadata = response.get_meta() metadata

[11/21/25 13:17:05] DEBUG    🌐 Constructed URL:                                                                   
                             https://api.met.no/weatherapi/locationforecast/2.0/compact?lat=-31.651&lon=116.659    

                    DEBUG    📋 User-Agent: weather-tools/0.0.0 (Python 3.12)                                      

                    DEBUG    Making request (attempt 1/3):                                                         
                             https://api.met.no/weatherapi/locationforecast/2.0/compact                            

[11/21/25 13:17:06] DEBUG    Cached response for: e9450af0091badb1a028399707618845                                 

                    DEBUG    Request successful on attempt 1                                                       

Out[4]:

{'updated_at': '2025-11-21T01:18:54Z',
 'units': {'air_pressure_at_sea_level': 'hPa',
  'air_temperature': 'celsius',
  'cloud_area_fraction': '%',
  'precipitation_amount': 'mm',
  'relative_humidity': '%',
  'wind_from_direction': 'degrees',
  'wind_speed': 'm/s'}}

3. Convert forecasts to daily and hourly tables¶

MetNoAPI.to_dataframe can aggregate the hourly GeoJSON payload to daily summaries

or return hourly data. Use the frequency parameter to control aggregation:

• frequency='daily' (default) - Daily aggregates

• frequency='hourly' - Raw hourly data

• frequency='weekly' or frequency='monthly' - Other time periods

In [5]:

daily_forecasts = api.to_dataframe(response, frequency="daily")
hourly_forecasts = api.to_dataframe(response, frequency="hourly")

daily_forecasts.head()

daily_forecasts = api.to_dataframe(response, frequency="daily") hourly_forecasts = api.to_dataframe(response, frequency="hourly") daily_forecasts.head()

Out[5]:

	date	min_temperature	max_temperature	total_precipitation	avg_wind_speed	max_wind_speed	avg_relative_humidity	avg_pressure	avg_cloud_fraction	dominant_weather_symbol
0	2025-11-21	17.1	29.3	0.0	3.426316	5.1	48.757895	1008.336842	41.810526	partlycloudy_night
1	2025-11-22	15.0	29.4	0.3	5.058333	7.9	65.912500	1006.600000	33.129167	rainshowers_day
2	2025-11-23	12.1	22.6	0.1	6.073684	8.8	50.389474	1010.536842	19.694737	partlycloudy_day
3	2025-11-24	12.8	21.7	0.1	4.850000	6.4	58.775000	1011.625000	49.600000	partlycloudy_day
4	2025-11-25	12.6	23.8	0.0	4.500000	5.2	50.475000	1013.075000	23.850000	partlycloudy_day

In [6]:

hourly_forecasts.head()

hourly_forecasts.head()

Out[6]:

	time	air_pressure_at_sea_level	air_temperature	cloud_area_fraction	relative_humidity	wind_from_direction	wind_speed	precipitation_amount	symbol_code
0	2025-11-21 05:00:00	1009.2	28.4	7.0	31.8	131.4	3.2	0.0	clearsky_day
1	2025-11-21 06:00:00	1008.4	29.0	5.5	29.8	136.2	2.4	0.0	clearsky_day
2	2025-11-21 07:00:00	1007.8	29.3	11.7	28.9	165.1	1.7	0.0	clearsky_day
3	2025-11-21 08:00:00	1007.5	29.2	26.6	28.6	200.4	2.1	0.0	fair_day
4	2025-11-21 09:00:00	1007.5	28.5	10.9	29.6	200.5	2.1	0.0	clearsky_day

Quick helper: get_daily_forecast¶

get_daily_forecast returns a pandas DataFrame directly with daily summaries,

In [7]:

daily_df = api.get_daily_forecast(
    latitude=coordinates.latitude,
    longitude=coordinates.longitude,
    days=5,
)

daily_df  # Already a DataFrame!

daily_df = api.get_daily_forecast( latitude=coordinates.latitude, longitude=coordinates.longitude, days=5, ) daily_df # Already a DataFrame!

                    DEBUG    🌐 Constructed URL:                                                                   
                             https://api.met.no/weatherapi/locationforecast/2.0/compact?lat=-31.651&lon=116.659    

                    DEBUG    📋 User-Agent: weather-tools/0.0.0 (Python 3.12)                                      

                    DEBUG    Cache hit for request: e9450af0091badb1a028399707618845                               

Out[7]:

	date	min_temperature	max_temperature	total_precipitation	avg_wind_speed	max_wind_speed	avg_relative_humidity	avg_pressure	avg_cloud_fraction	dominant_weather_symbol
0	2025-11-21	17.1	29.3	0.0	3.426316	5.1	48.757895	1008.336842	41.810526	partlycloudy_night
1	2025-11-22	15.0	29.4	0.3	5.058333	7.9	65.912500	1006.600000	33.129167	rainshowers_day
2	2025-11-23	12.1	22.6	0.1	6.073684	8.8	50.389474	1010.536842	19.694737	partlycloudy_day
3	2025-11-24	12.8	21.7	0.1	4.850000	6.4	58.775000	1011.625000	49.600000	partlycloudy_day
4	2025-11-25	12.6	23.8	0.0	4.500000	5.2	50.475000	1013.075000	23.850000	partlycloudy_day

4. Merge met.no forecasts with SILO history¶

Pull the last five days from the SILO DataDrill API for the same coordinates,

then merge that history with the met.no forecast. The helper automatically

converts column names, fills optional variables (when enabled), and annotates

the data source for each record.

In [ ]:

first_forecast_date = pd.to_datetime(daily_df["date"]).min()
# history_end = first_forecast_date - pd.Timedelta(days=1)
history_start = first_forecast_date - pd.Timedelta(days=4)
silo_api = SiloAPI(log_level="INFO")
silo_history, _ = silo_api.get_data_drill(
    latitude=coordinates.latitude,
    longitude=coordinates.longitude,
    start_date=history_start.strftime("%Y%m%d"),
    end_date=first_forecast_date.strftime("%Y%m%d"),
    variables=["daily_rain", "max_temp", "min_temp", "et_short_crop"],
)

merged = merge_historical_and_forecast(
    silo_data=silo_history,
    metno_data=daily_df,
    overlap_strategy="prefer_silo",
    return_cols="silo_only",
)

merged

first_forecast_date = pd.to_datetime(daily_df["date"]).min() # history_end = first_forecast_date - pd.Timedelta(days=1) history_start = first_forecast_date - pd.Timedelta(days=4) silo_api = SiloAPI(log_level="INFO") silo_history, _ = silo_api.get_data_drill( latitude=coordinates.latitude, longitude=coordinates.longitude, start_date=history_start.strftime("%Y%m%d"), end_date=first_forecast_date.strftime("%Y%m%d"), variables=["daily_rain", "max_temp", "min_temp", "et_short_crop"], ) merged = merge_historical_and_forecast( silo_data=silo_history, metno_data=daily_df, overlap_strategy="prefer_silo", return_cols="silo_only", ) merged

In [10]:

merge_summary = get_merge_summary(merged)
merge_summary

merge_summary = get_merge_summary(merged) merge_summary

Out[10]:

{'total_records': 9,
 'silo_records': np.int64(4),
 'metno_records': np.int64(5),
 'date_range': {'start': Timestamp('2025-11-17 00:00:00'),
  'end': Timestamp('2025-11-25 00:00:00'),
  'days': 9},
 'silo_period': {'start': Timestamp('2025-11-17 00:00:00'),
  'end': Timestamp('2025-11-20 00:00:00')},
 'metno_period': {'start': Timestamp('2025-11-21 00:00:00'),
  'end': Timestamp('2025-11-25 00:00:00')},
 'transition_date': Timestamp('2025-11-21 00:00:00')}

4b. Merge Forecast with SILO PatchedPoint¶

Pull the last four days from the SILO Patch Point API for the same coordinates, using the station code.

In [ ]:

first_forecast_date = pd.to_datetime(daily_forecasts["date"]).min()
history_end = first_forecast_date  # - pd.Timedelta(days=1) # Include day before forecast to check overlap handling
history_start = history_end - pd.Timedelta(days=4)


silo_station_data, _ = silo_api.get_patched_point(
    station_code=str(station_meta["station_code"]),
    start_date=history_start.strftime("%Y%m%d"),
    end_date=history_end.strftime("%Y%m%d"),
    variables=["daily_rain", "max_temp", "min_temp", "et_short_crop"],
)

merge_historical_and_forecast(
    silo_data=silo_station_data,
    metno_data=daily_df,
    return_cols="silo_only",
)

first_forecast_date = pd.to_datetime(daily_forecasts["date"]).min() history_end = first_forecast_date # - pd.Timedelta(days=1) # Include day before forecast to check overlap handling history_start = history_end - pd.Timedelta(days=4) silo_station_data, _ = silo_api.get_patched_point( station_code=str(station_meta["station_code"]), start_date=history_start.strftime("%Y%m%d"), end_date=history_end.strftime("%Y%m%d"), variables=["daily_rain", "max_temp", "min_temp", "et_short_crop"], ) merge_historical_and_forecast( silo_data=silo_station_data, metno_data=daily_df, return_cols="silo_only", )