TorchRTM: A PyTorch-based Radiative Transfer Model

TorchRTM is a flexible and efficient package designed for simulating radiative transfer in the atmosphere, vegetation, and soil systems, based on the PROSAIL model. It leverages GPU-accelerated computations and integrates various atmospheric correction models, including SMAC for top-of-atmosphere reflectance estimation. This package is particularly suitable for remote sensing and environmental monitoring applications and is build for pyTorch integration.

---

Features

• GPU Acceleration: Built with PyTorch in mind, TorchRTM offers GPU inference to speed up simulations.
• SMAC Integration: Perform atmospheric correction using the Simplified Model for Atmospheric Correction (SMAC).
• PROSAIL Model: Integrates the PROSAIL radiative transfer model for canopy reflectance simulations.
• Flexible Inputs: Handles both batch and individual simulations of canopy and atmospheric parameters.
• Retrieval Component: Includes a fast and flexible trait retrieval system, offering both neural network (NN)-based inference and Look-Up Table (LUT)-based matching.

---

Installation

Prerequisites

Ensure that you have Python 3.9+ and PyTorch installed. If you need PyTorch, follow the installation instructions from the official site.

Install TorchRTM

You can install TorchRTM directly from GitHub:

pip install torchrtm

---

Usage

1. Simulating Canopy Reflectance

The prosail_shell_v2 function can be used to simulate the canopy reflectance based on input vegetation parameters.

<<<<<<< HEAD
from torchrtm.models import prosail
=======
from torchrtm.models import prosail_shell_v2
import torch
# Simulate canopy reflectance
B=10000
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

traits=torch.tensor([[40.0, 8.0, 0.0, 0.01, 0.008]] * B).to(device)
N=torch.tensor([1.5] * B).to(device)
LIDFa=torch.tensor([-0.3] * B).to(device)
LIDFb=torch.tensor([-0.1] * B).to(device)
lai=torch.tensor([3.0] * B).to(device)
q=torch.tensor([0.5] * B).to(device)
tts=torch.tensor([30.0] * B).to(device)
tto=torch.tensor([20.0] * B).to(device)
psi=torch.tensor([10.0] * B).to(device)
alpha=torch.tensor([40.0] * B).to(device)

psoil=torch.tensor([0.5] * B).to(device)
<<<<<<< HEAD
toc = prosail(traits,N,LIDFa,LIDFb,lai,q,tts,tto,psi,alpha,psoil,batch_size=100000,use_prospectd=False,lidtype=2)
=======
toc = prosail_shell_v2(traits,N,LIDFa,LIDFb,lai,q,tts,tto,psi,alpha,psoil,batch_size=100000,use_prospectd=False,lidtype=2)

2. Atmospheric Correction with SMAC

Use the smac function to apply atmospheric correction using the SMAC model.

from torchrtm.atmosphere.smac import smac
from torchrtm.data_loader import load_smac_sensor

# Load sensor-specific coefficients
sensor_name = "Sentinel2A-MSI"
coefs = load_smac_sensor(sensor_name)

# Set atmospheric parameters
tts = torch.tensor([30.0])
tto = torch.tensor([20.0])
psi = torch.tensor([10.0])

# Apply SMAC atmospheric correction
ttetas, ttetav, tg, s, atm_ref, tdir_tts, tdif_tts, tdir_ttv, tdif_ttv = smac(tts, tto, psi, coefs)

3. Converting TOC to TOA Reflectance

Use the toc_to_toa function to convert Top-of-Canopy (TOC) reflectance to Top-of-Atmosphere (TOA) reflectance.

from torchrtm.atmosphere.smac import toc_to_toa

# Convert TOC to TOA reflectance
R_TOC, R_TOA = toc_to_toa(toc, sm_wl, ta_ss, ta_sd, ta_oo, ta_do, ra_so, ra_dd, T_g, return_toc=True)

---

Tests

This package includes a set of unit tests to ensure the correctness of all key components. To run the tests, use pytest:

pytest

---

Contributing

Contributions to TorchRTM are welcome! Please fork this repository, create a new branch, and submit a pull request. Make sure to include tests for any new features or bug fixes.

---

License

This project is licensed under the MIT License.

---

Acknowledgments

To be added.