RADET - beta

WARNING: This code is in development and may change without notice.

This repository provides a Google Earth Engine (Python API) implementation of the RADET model (Radiation Advection Diffusivity-independent Evapotranspiration) for estimating actual evapotranspiration (ET). RADET estimates ET based on the Diffusivity-Independent Flux hypothesis and conditionally incorporates Penman’s aerodynamic term when and where advection is expected to be significant (Kim et al., 2026). The RADET-beta implementation here is designed to be consistent with the OpenET Python pipeline to facilitate interoperability and integration within existing workflows.

Model Design

The primary component of the RADET model is the Image() class. The Image class can be used to compute a single ET image from a single input image. The Image class should generally be instantiated from an Earth Engine Landsat image using the collection specific methods listed below. ET image collections can be built by computing ET in a function that is mapped over a collection of input images.

Input Collections

RADET can currently be computed for Landsat Collection 2 Level 2 (SR/ST) images from the following Earth Engine image collections:

• LANDSAT/LT05/C02/T1_L2
• LANDSAT/LE07/C02/T1_L2
• LANDSAT/LC08/C02/T1_L2
• LANDSAT/LC09/C02/T1_L2

Landsat Collection 2 SR/ST Input Image

To instantiate the class for a Landsat Collection 2 SR/ST image, use the Image.from_landsat_c2_sr method.

The input Landsat image must have the following bands and properties:

SPACECRAFT_ID	Band Names
LANDSAT_5	SR_B1, SR_B2, SR_B3, SR_B4, SR_B5, SR_B7, ST_B6, ST_EMIS, QA_PIXEL
LANDSAT_7	SR_B1, SR_B2, SR_B3, SR_B4, SR_B5, SR_B7, ST_B6, ST_EMIS, QA_PIXEL
LANDSAT_8	SR_B1, SR_B2, SR_B3, SR_B4, SR_B5, SR_B6, SR_B7, ST_B10, ST_EMIS, QA_PIXEL
LANDSAT_9	SR_B1, SR_B2, SR_B3, SR_B4, SR_B5, SR_B6, SR_B7, ST_B10, ST_EMIS, QA_PIXEL

Property	Description
system:index	- Landsat Scene ID - Must be in Earth Engine format (e.g. LC08_044033_20170716)
system:time_start	Image datetime in milliseconds since 1970
SPACECRAFT_ID	- Used to determine Landsat sensor type - Must be one of: LANDSAT_5, LANDSAT_7, LANDSAT_8, LANDSAT_9

Model Output

The primary output of the RADET model is the actual ET (ETa) in millimeters.

Examples

The examples/ folder contains the following:

• radet_single_image.ipynb — Compute RADET for a single Landsat image
• radet_collection_interpolate.ipynb — Build a RADET image collection and interpolate

Project Structure

radet-beta/
├── radet/
│   ├── __init__.py
│   ├── collection.py      # ET image collection builder
│   ├── image.py           # Core Image class for single ET computation
│   ├── interpolate.py     # Temporal interpolation utilities
│   ├── landsat.py         # Landsat-specific preprocessing
│   ├── meteorology.py     # Meteorology-specific preprocessing
│   ├── model.py           # RADET model implementation
│   └── utils.py           # Helper functions
├── examples/
│   ├── README.md
│   ├── radet_single_image.ipynb
│   └── radet_collection_interpolate.ipynb
├── .gitignore
├── LICENSE
└── README.md

Dependencies

• earthengine-api # main RADET model dependency
• openet-core # main RADET model dependency
• openet-refet-gee # main RADET model dependency
• pandas # For example notebooks (optional)
• notebook # For example notebooks (optional)

Installation

1. Download and Install Anaconda/Miniconda

Either Anaconda or Miniconda is required for managing Python packages (Python >= 3.10 recommended).

Windows users: After installation, open Anaconda Prompt and run conda init powershell to add conda to PowerShell.

Linux/Mac users: Ensure conda is added to your PATH (typically automatic). Restart your shell if needed.

Update conda: conda update conda

2. Create the Conda Environment

Create and activate a new conda environment:

conda create -y -n radet python=3.12
conda activate radet

Navigate to the radet-beta directory and install the package:

cd /path/to/radet-beta

Option A — Install the core RADET model only:

pip install -e .

Option B — Install with notebook dependencies (includes pandas):

pip install -e .[notebooks]

Note: Use Option B if you plan to run the example notebooks.

Google Earth Engine Authentication

This project uses the Google Earth Engine (GEE) Python API for geospatial data extraction.

1. Install Google Cloud CLI
2. Create a GCloud project (e.g., gee-radet) with GEE API enabled at https://console.cloud.google.com/
3. Configure the project:

   gcloud config set project gee-radet
   gcloud auth application-default set-quota-project gee-radet  # if prompted
   earthengine authenticate

See the Earth Engine Python installation guide for details.

Related Repositories

• radet-analysis — Analysis and evaluation workflows for RADET

References

Kim, Y., Huntington, J. L., Comini de Andrade, B., Johnson, M. S., Volk, J. M., Majumdar, S., Morton, C., & ReVelle, P. (2026). Thermodynamically constrained surface energy balance using medium-resolution remote sensing for efficient evapotranspiration mapping. EarthArXiv (preprint). https://doi.org/10.31223/X51B4P