ProSAIL_forward

This repository allows to run the ProSAIL RTM in forward mode and generate Look-Up Tables (LUTs). It provides default data and parametrization for running the model for winter wheat in Switzerland and for the Sentinel-2 sensor.

Installation

git clone https://github.com/EOA-team/ProSAIL_forward.git
cd ProSAIL_forward
python -m venv my_venv
source my_venv/bin/activate
pip install -r requirements.txt

Running the code

To run the RTM, everything is set up in the simulate_S2_spectra_soil.py script.

First define some paths and parameters in config.yaml:

Parameter	Description
LUT	Dictionary with path to LUT input values stored in a .csv (lut_params) and LUT size (lut_size)
RTM	Dictionary containing sensor name (sensor), path to spectral response (fpath_srf), whether to remove invalid green peaks (remove_invalid_green_peaks). Sentinel-2A and 2B must be run separately!
out_dir	Directory where simulated spectra stored in .pkl file
soil_path	Optional. If you want the RTM to use custom soil samples, then pass the path to a .pkl dataframe containing each spectra in a row (2101 columns with reflectance values between 400 and 2500nm). If none (set to null) provided, it will use default soil spectra.
traits	Traits to include among LAI (lai), chlorophyll (cab), carotenoids (car), canopy chlorophyll content (ccc). Default is all four.

You can run it with the following command:

python simulate_S2_spectra_soil.py

The result will be a pickled dataframe containing a simulation per row. The columns will be the leaf/canopy parameters and the Sentinel2 bands' reflectance values:

import pandas as pd
file_path = 'path_to/your_results.pkl'
df = pd.read_pickle(file_path)
print(df.head()) 

Credits

The code provided here is built on existing code

• rtm_inv: A Python-backend for radiative transfer model inversion for crop trait retrieval. Code
• sentinel2_crop_traits: Sentinel-2 Crop Trait Retrieval Using Physiological and Phenological Priors from Field Phenotyping (Graf et al., 2023, RSE). Code