isofit 4.0.0a0

Imaging Spectrometer Optimal FITting

ISOFIT — Imaging Spectrometer Optimal FITting

ISOFIT is a Python toolkit for retrieving surface reflectance, atmospheric state, and instrument calibration from imaging spectrometer radiance data. It is built around the optimal estimation framework and is designed for flexibility — users can combine different radiative transfer models, surface priors, and uncertainty models to suit their sensors and science questions. The effort began with a straightforward optimal estimation implementation, and has grown considerably - see the bibilography for the latest science.

ISOFIT is on version . Major revisions are not backwards compatible with one another due to updates to the configuration files. See Release History for older versions.

ISOFIT combines surface models, atmospheric radiative transfer, and instrument characteristics into a forward model. This forward model is then solved by inverting relative to a measurement, via optimization or MCMC, to generate the Maximum a Posterior estimate of the joint surface and atmospheric state vector.

+-------------------------------------------------------------------------+
|                                 isofit                                  |
|                                                                         |
|  +--------+      +---------------+      +----------------------------+  |
|  |        |      |               |      |          forward           |  |
|  |   io   | ===> |   inversion   | <=== |                            |  |
|  |        |      |               |      | +----------+  +----------+ |  |
|  +--------+      +---------------+      | |atmosphere|  | surface  | |  |
|                          |              | +----------+  +----------+ |  |
|                          |              |                            |  |
|                          |              |       +------------+       |  |
|                          |              |       | instrument |       |  |
|                          |              |       +------------+       |  |
|                          |              +----------------------------+  |
|                          |                                              |
+--------------------------|----------------------------------------------+
                           |
                           | produces
                           v
   +-------------------------------------------------------------+
   |    Joint Surface, Instrument, & Atmospheric State Vector    |
   +-------------------------------------------------------------+

---

Contents

• Features
• Installation
• Quick Start
• Examples
• Atmospheric Radiative Transfer Models
• Documentation
• Contributing
• Citation
• License

---

Components

Capability	Details
Radiative transfer	MODTRAN, 6S, LibRadTran, sRTMnet neural-network emulator
Surface models	Multicomponent Gaussian, Glint, LUT, Emissive (thermal)
Instrument models	SNR, parametric noise, nonlinear response, and more

---

Installation

Conda (recommended)

mamba create -n isofit_env -c conda-forge isofit
mamba activate isofit_env

Full Installation Instructions

For all options including developer installs, see the Installation guide.

---

Quick Start

1. Download supporting data and build examples

isofit download all
isofit build

This fetches the sRTMnet emulator, 6S, surface libraries, and pre-configured example datasets into ~/.isofit/. For users that want a smaller or mosre custom download, see Additional Data for customization.

2. Run an example to test

cd $(isofit path examples)/image_cube/small
bash default.sh

3. Start using!

Leverage the command line interface to build surface models, run isofit on scenes, or perform a wavelength calibration.

To get going, you'll need a radiance image, a location file (with lat, lon, and elevation), and an observation file with sensor and solar geometry. To get running, ISOFIT ships with both examples (all command line scripts) and tutorials (jupyter notebooks that annotate calls in python and the command line):

Tutorials

Tutorial	Description
apply_oe	Full walkthrough of inputs and outputs of apply_oe
single pixel inversion	Learn how to use isofit modules in your own code by building and executing individual modules on a single pixel

Examples

Example	Description
image_cube/small	Fast analytical retrieval — good first run
20151026_SantaMonica	AVIRIS-NG scene with 6S radiative transfer
20171108_Pasadena	Urban hyperspectral scene
20190806_ThermalIR	Thermal infrared joint VSWIR+TIR retrieval
20231110_Prism_Multisurface	Multi-component surface model
NEON	National Ecological Observatory Network data
SeaBASS_prism_001	Coastal / water target single-pixel case

Full instructions for running examples.

---

Atmospheric Radiative Transfer Models

ISOFIT supports several atmospheric RTM backends. The recommended starting point for new users is sRTMnet, a neural-network emulator that requires no licensed software:

• sRTMnet quickstart — automatic install via isofit download
• MODTRAN quickstart — requires a MODTRAN 6 licence
• 6S — Second Simulation of a Satellite Signal in the Solar Spectrum vector code. Automatic download and build with isofit download.
• libRadtran — library for radiative transfer. Automatic download and build with isofit download.
• Bring Your Own — isofit can run with any LUT that complies with the luts/reader.py standards. Adding a new radiative transfer model is also straightforward by extending the base BaseAtmosphere class.

Contributing

Contributions are welcome — bug reports, tests, documentation, and new features.
Please read the Contributing guide before opening a pull request.

Active development happens on the dev branch.
Issues and feature requests are tracked in the issue tracker.

---

Citation

If you use ISOFIT in your work, please cite the relevant papers listed in the Bibliography.
Package citation information is in the CITATION.cff file.

---

License

Apache License v2 — see LICENSE.

Images in this repository are licensed under CC0.