micro-weather-nature

This Zenodo record provides the prepared data used by the micro-weather-nature project for model training, inference, and analysis. What this dataset contains The archive includes prepared inputs and outputs for the full modeling workflow: ERA5 gridded meteorology (processed/interpolated products) MADIS station observations and station metadata Static geospatial node features (terrain, land cover, satellite, FM embeddings) Inference-target node features for arbitrary-location prediction Region shapefiles used to define training/inference domain Example model outputs and inference result files Directory structure (after extraction) Data/ ERA5/ Interpolated/ Processed/ madis/ processed/ stations/ NodeFeatures/ TerrainStats/ LCLUStats/ Satellite/ FoundationModelEmbedding/ NodeFeaturesInference/ TerrainStats/ LCLUStats/ Satellite/ FoundationModelEmbedding/ Shapefiles/ CONUS/ ECOSTRESS/ ModelOutputs/ InferenceResults/ Meaning of key folders ERA5 Background meteorological fields used as model context. Processed/ and Interpolated/ contain cached products used directly by training/inference. madis Station-based observations and station subsets used to build training and validation samples. NodeFeatures Static predictors at known stations/locations, including terrain, land-cover, satellite products, and foundation-model embeddings. NodeFeaturesInference Inference-target feature stacks used when predicting at arbitrary geolocations (file2infer-based workflows). Shapefiles Geographic boundaries; includes CONUS geometry used by the default launch settings. ECOSTRESS ECOSTRESS tif files for the qualitative comparison ModelOutputs Stored model artifacts and run outputs (e.g., model checkpoints, normalization objects, and metadata files). InferenceResults NetCDF prediction outputs organized by model and inference configuration. How to use with the codebase After extraction, set your code --data_path argument to the extracted Data root. Examples: python Source/Arg_Parser.py --data_path /path/to/Data --output_saving_path /path/to/output python Source/Inference.py --data_path /path/to/Data --trained_path /path/to/trained/outputs --output_saving_path /path/to/output