Quick Start

This page shows one canonical minimal path for new users.

Colab notebooks

For the fastest start, use one of the maintained notebooks:

Minimal working example

  1. Set up project context

from bakaano.core.project import ProjectContext

working_dir = "/path/to/working_dir"
study_area = "/path/to/basin.shp"

project = ProjectContext(
    working_dir=working_dir,
    study_area=study_area,
    climate_data_source="ERA5",
)

  1. Run preprocessing and hydrology modules

Run these modules in order:

  • bakaano.data.dem.DEM

  • bakaano.data.tree_cover.TreeCover

  • bakaano.data.ndvi.NDVI

  • bakaano.data.soil.Soil

  • bakaano.data.alpha_earth.AlphaEarth

  • bakaano.data.meteo.Meteo

  • bakaano.hydrology.veget.VegET

Use Inputs and Outputs for the exact files each module reads and writes.

  1. Train a model

from bakaano.neuralnet.train import train_streamflow_model

project.validate_project(for_task="train")

model_path = train_streamflow_model(
    working_dir=working_dir,
    study_area=study_area,
    train_start="1981-01-01",
    train_end="2020-12-31",
    grdc_netcdf="/path/to/GRDC.nc",
    batch_size=32,
    num_epochs=300,
    learning_rate=1e-3,
    routing_method="mfd",
    area_normalize=True,
    log_transform=True,
)

  1. Evaluate or simulate

from bakaano.neuralnet.simulate import (
    evaluate_streamflow_model_interactively,
    simulate_grdc_csv_stations,
    simulate_streamflow,
)

evaluate_streamflow_model_interactively(
    working_dir=working_dir,
    study_area=study_area,
    model_path=model_path,
    val_start="2013-01-01",
    val_end="2020-12-31",
    grdc_netcdf="/path/to/GRDC.nc",
)

simulate_grdc_csv_stations(
    working_dir=working_dir,
    study_area=study_area,
    model_path=model_path,
    sim_start="1981-01-01",
    sim_end="2020-12-31",
    grdc_netcdf="/path/to/GRDC.nc",
)

simulate_streamflow(
    working_dir=working_dir,
    study_area=study_area,
    model_path=model_path,
    sim_start="1981-01-01",
    sim_end="1990-12-31",
    latlist=[13.8, 13.9],
    lonlist=[3.0, 4.0],
)

Before you run expensive steps

  • project.validate_project(for_task="train") before training

  • project.validate_project(for_task="evaluate") before interactive evaluation

  • project.validate_project(for_task="simulate") before simulation

Important notes

  • Simulation outputs begin after a one-year warmup period.

  • If rainfall_sparse_arrays.pkl exists for the requested dates, Bakaano uses routed rainfall automatically as an extra temporal predictor.

  • New models save area_normalize and log_transform beside the Keras checkpoint, and inference loads those settings automatically.

Where to go next