{
  "fetched_at": "2026-10-19T06:38:46Z",
  "source_url": "https://www.kaggle.com/competitions/eurecom-aml-2022-challenge-1/data",
  "pages": [
    {
      "name": "data-description",
      "content": "The data consists of pairs of meteorological features and target values at a particular\nlatitude/longitude and time. The target value is air temperature measurements at 2 metres above the\nground. The feature vectors include both weather-related features such as sun evaluation at\nthe current location, climate values of temperature, pressure and topography, and meteorological\nparameters on different pressure and surface levels from weather forecast model predictions. Weather\nforecast model predictions are values produced by the following weather forecast models: Global\nForecast System (GFS), Global Deterministic Forecast System from the Canadian Meteorological\nCenter (CMC), and the Weather Research and Forecasting (WRF) Model. Each model returns\nthe following predicted values: wind, humidity, pressure, clouds, precipitation, dew point, snow\ndepth, air and soil temperature characteristics. Where applicable, the predictions are given at different\nisobaric levels from 50 hPa (≈ 20 km above ground) to the ground level. \nAltogether, there are 111 features in total. It is important to note that the features are\nhighly heterogeneous, i.e., they are of different types and scales. \n\n## Files\n\n*   **train.csv** - the training set\n*   **test_feat.csv** - the features for the test set\n*   **example_preds.csv** - a sample submission file in the correct format\n\n## Columns\n\n* `fact_time`: time, unix timestamp\n* `fact_latitude`: geographical latitude, degrees\n* `fact_longitude`: geographical longitude, degrees\n* `topography_bathymetry\n* `sun_elevation\n* `climate_temperature`: climate temperature, C\n* `climate_pressure`: climate pressure, mmHg\n* `cmc_0_0_6_2`: dew point temp at 2m, K\n* `cmc_0_0_7_1000`: dew point depression at 1000 hPa isobaric level, K\n* `cmc_0_0_7_2`:  dew point depression at 2m, K\n* `cmc_0_0_7_500`: dew point depression at 500 hPa isobaric level, K\n* `cmc_0_0_7_700`: dew point depression at 700 hPa isobaric level, K\n* `cmc_0_0_7_850`: dew point depression at 850 hPa isobaric level, K\n* `cmc_0_0_7_925`: dew point depression at 925 hPa isobaric level, K\n* `cmc_0_1_0_0`:  absolute humidity from 0 to 1\n* `cmc_0_1_11_0`: snow depth, m\n* `cmc_0_1_65_0`: rain accumulated, mm\n* `cmc_0_1_66_0`: snow accumulated, mm \n* `cmc_0_1_67_0`: iced rain accumulated, mm\n* `cmc_0_1_68_0`: iced graupel accumulated, mm\n* `cmc_0_1_7_0`:  instant precipitation intensity, mm/h\n* `cmc_0_2_2_10`: wind U component at 10m, m/s\n* `cmc_0_2_2_500`: wind U component at 500 hPa isobaric level, m/s\n* `cmc_0_2_2_700`: wind U component at 700 hPa isobaric level, m/s\n* `cmc_0_2_2_850`: wind U component at 850 hPa isobaric level, m/s\n* `cmc_0_2_2_925`: wind U component at 925 hPa isobaric level, m/s\n* `cmc_0_2_2_1000`: wind U component at 1000 hPa isobaric level, m/s\n* `cmc_0_2_3_10`: wind V component at 10m, m/s\n* `cmc_0_2_3_500`: wind V component at 500 hPa isobaric level, m/s\n* `cmc_0_2_3_700`: wind V component at 700 hPa isobaric level, m/s\n* `cmc_0_2_3_850`: wind V component at 850 hPa isobaric level, m/s\n* `cmc_0_2_3_925`: wind V component at 925 hPa isobaric level, m/s\n* `cmc_0_2_3_1000`: wind V component at 1000 hPa isobaric level, m/s\n* `cmc_0_3_0_0`:  surface pressure, Pa\n* `cmc_0_3_0_0_next`:  next horizon surface pressure, Pa\n* `cmc_0_3_1_0`: sea level pressure, Pa\n* `cmc_0_3_5_500`: geopotential height at 500 hPa isobaric level, gpm\n* `cmc_0_3_5_700`: geopotential height at 700 hPa isobaric level, gpm\n* `cmc_0_3_5_850`: geopotential height at 850 hPa isobaric level, gpm\n* `cmc_0_3_5_925`: geopotential height at 925 hPa isobaric level, gpm\n* `cmc_0_3_5_1000`: geopotential height at 1000 hPa isobaric level, gpm\n* `cmc_0_6_1_0`: cloudiness, % from 0 to 100\n* `cmc_available`: is there any data from cmc\n* `cmc_horizon_h`:  cmc horizon, h\n* `cmc_precipitations`: avg precipitations rate between adjacent horizons, mm/h\n* `cmc_timedelta_s`: difference between cmc and forecast time, s\n* `gfs_2m_dewpoint`: dew point temperature at 2m, C\n* `gfs_a_vorticity`: absolute vorticity at height 1000 hPa, 1/s\n* `gfs_available`: is there any data from gfs\n* `gfs_cloudness`: sum of 3 level cloudiness, from 0 to 3\n* `gfs_clouds_sea`:  cloud mixing ratio at level 1000 hPa, \n* `gfs_horizon_h`:  gfs horizon, h\n* `gfs_humidity`: relative humidity at 2m, %\n* `gfs_precipitable_water`:  total precipitable water, kg/m^2\n* `gfs_precipitations`: avg precipitations rate between adjacent horizons, mm/h\n* `gfs_pressure`: surface pressure, mmHg\n* `gfs_r_velocity`:  vertical velocity at 1000 hPa, Pa/s\n* `gfs_soil_temperature`:  soil temperature at 0.0-0.1 m, C\n* `gfs_soil_temperature_available`: is there gfs soil temp data\n* `gfs_temperature_5000`: temperature at vertical level at 50 hPa, C\n* `gfs_temperature_7000`: temperature at vertical level at 70 hPa, C\n* `gfs_temperature_1000`: temperature at vertical level at 100 hPa, C\n* `gfs_temperature_1500`: temperature at vertical level at 150 hPa, C\n* `gfs_temperature_2000`: temperature at vertical level at 200 hPa, C\n* `gfs_temperature_2500`: temperature at vertical level at 250 hPa, C\n* `gfs_temperature_3000`: temperature at vertical level at 300 hPa, C\n* `gfs_temperature_3500`: temperature at vertical level at 350 hPa, C\n* `gfs_temperature_4000`: temperature at vertical level at 400 hPa, C\n* `gfs_temperature_4500`: temperature at vertical level at 450 hPa, C\n* `gfs_temperature_5000`: temperature at vertical level at 500 hPa, C\n* `gfs_temperature_5500`: temperature at vertical level at 550 hPa, C\n* `gfs_temperature_6000`: temperature at vertical level at 600 hPa, C\n* `gfs_temperature_6500`: temperature at vertical level at 650 hPa, C\n* `gfs_temperature_7000`: temperature at vertical level at 700 hPa, C\n* `gfs_temperature_7500`: temperature at vertical level at 750 hPa, C\n* `gfs_temperature_8000`: temperature at vertical level at 800 hPa, C\n* `gfs_temperature_8500`: temperature at vertical level at 850 hPa, C\n* `gfs_temperature_9000`: temperature at vertical level at 900 hPa, C\n* `gfs_temperature_9250`: temperature at vertical level at 925 hPa, C\n* `gfs_temperature_9500`: temperature at vertical level at 950 hPa, C\n* `gfs_temperature_9750`: temperature at vertical level at 975 hPa, C\n* `gfs_timedelta_s`: difference between gfs and forecast time, s\n* `gfs_total_clouds_cover_high`: cloud coverage (between horizons, divisible by 6) at high level, %\n* `gfs_total_clouds_cover_low`:  cloud coverage (between horizons, divisible by 6) at low level, %\n* `gfs_total_clouds_cover_middle`:  cloud coverage (between horizons, divisible by 6) at middle level, %\n* `gfs_u_wind`:  10 meter U wind component, m/s\n* `gfs_v_wind`:  10 meter V wind component, m/s\n* `gfs_wind_speed`: wind velocity, sqrt(gfs_u_wind^2 + gfs_v_wind^2), m/s\n* `wrf_available`: s there any data from wrf\n* `wrf_t2_next`:  next horizon temperature at 2m, K\n* `wrf_psfc`: pressure, Pa\n* `wrf_rh2`: relative humidity at 2m, from 0 to 1\n* `wrf_wind_u`: wind U component, m/s\n* `wrf_wind_v`: wind V component, m/s\n* `wrf_rain`: avg rain rate between two horizons, mm/h\n* `wrf_snow`: avg snow rate between two horizons, mm/h\n* `wrf_graupel`: avg graupel rate between two horizons, mm/h\n* `wrf_t2_interpolated`: wrf_t2_interpolated between horizons, K\n* `wrf_t2_grad`: difference between temperatures at 2m on adjacent horizons, K\n* `cmc_0_1_65_0_grad`: rain accumulated from cmc gentime difference between adjacent horizons, mm\n* `cmc_0_1_65_0_next`: rain accumulated from cmc gentime for next horizon, mm\n* `cmc_0_1_66_0_grad`: snow accumulated from cmc gentime difference between adjacent horizons, mm\n* `cmc_0_1_66_0_next`: snow accumulated from cmc gentime for next horizon, mm\n* `cmc_0_1_67_0_grad`: ice rain accumulated from cmc gentime difference between adjacent horizons, mm\n* `cmc_0_1_67_0_next`: ice rain accumulated from cmc gentime for next horizon, mm\n* `cmc_0_1_68_0_grad`: iced graupel accumulated from cmc gentime difference between adjacent horizons, mm\n* `cmc_0_1_68_0_next`: iced graupel accumulated from cmc gentime for next horizon, mm\n* `gfs_2m_dewpoint_grad`: dew point temperature at 2m difference between horizons, C\n* `gfs_2m_dewpoint_next`: dew point temperature on next horizon, C\n* `gfs_total_clouds_cover_low_grad`: difference between low level cloud coverage on adjacent horizons, %\n* `gfs_total_clouds_cover_low_next`:  next horizon cloud coverage (between horizons, divisible by 6) at low level, %\n* `fact_temperature`: temperature, C [targets]\n\n\n## Copyright \nThis data is created starting from the available dataset used for the **Shift Competition**, held during the [NeurIPS conference 2021](https://neurips.cc/Conferences/2021/CompetitionTrack).\nThe original Shifts Weather Prediction Dataset is released under CC BY NC SA 4.0 license. This dataset was constructed by combining features from publicly available weather prediction services and models. Specifically, we combined data from NOAA/NWS servers, data generated by the WRF model from NCAR/UCAR, and data from the Meteorological Service of Canada. Ground station readings were taken from [NOAA](https://www.weather.gov/disclaimer). \nThe data was further cleaned and subsampled to meet the requirements of the AML challenge.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "In this challenge, you will build models to predict the weather (specifically the temperature) at different locations around the globe. \n\nIn this challenge, you will be exposed to the concept of covariate shift. The mismatch between training and deployment data adversely impacts ML models and is ubiquitous in real, industrial applications. \n\nTo analyze the robustness of learned models to climate shifts, we use the Koppen climate classification which provides data to map latitudes and longitudes to one of five main climate types: Tropical, Dry, Mild Temperate, Snow and Polar. The dataset is augmented such that each sample has an associated climate type. \n\nThe training data consists of ~2M measurements made from September 1st, 2018 till February 28th, 2019 for climate types Dry and Mild Temperate (note that the climate type is not used as a training feature).\nThe evaluation data will also include measurements for Tropical climate. ",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "Submissions will be ranked based on the test RMSE.\n\n## Submission Format\n\nThe file should contain a header and have the following format:\n```\nindex,fact_temperature\n0,1.0\n```",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "rules",
      "content": "This is a page where you can include rules that participants must accept before joining. You may wish to include rules like:\n\n - Don't cheat!\n - Apply yourself!\n - Have fun!",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}