SAR and Optical Dataset for Agriculture in Seville (SODAS)

- Motivation

The potential of using radar remote sensing data for agricultural applications has been demonstrated in recent years, but this type of data is largely underused due to the complexity of its pre-processing and to the non-obvious physical interpretation of the derived features. To address these challenges, pre-processed datasets including synthetic aperture radar (SAR) images in analysis ready format are welcome.

- Dataset

The SAR and Optical Dataset for Agriculture in Seville (SODAS) integrates time series of radar images (Sentinel-1), optical images (Sentinel-2), precipitation records, and crop-type maps. The radar and optical time series consist of georeferenced Sentinel-1/-2 images over an agricultural area in Seville, Spain, spanning five years, from 2017 to 2021. Crop types include 18 different classes and fallow. The SAR images are provided in the form of 1) dual-polarimetric covariance matrices, which include the backscattering coefficient, and 2) repeat-pass interferometric coherence (amplitude and phase) at VV and VH polarimetric channels. The optical images correspond to partially or fully cloud-free Sentinel-2 reflectivity at red, green, blue, and near infra-red bands, as well as normalized difference vegetation index (NDVI) images. All images and crop-type maps are represented in the same cartographical grid in UTM coordinates, and the dataset is provided in NetCDF4 format. 

- Application

This dataset has many potential uses, such as development of algorithms for crop-type mapping, retrieval of biophysical parameters, crop monitoring, and data fusion.

- Usage

A jupyter notebook for inspecting and illustrating the dataset features is provided, together with a python file which includes multiple functions to load the dataset, visualise images, derive additional features, and create and compare time series.

- Documentation

A journal paper for documenting the dataset (pre-processing, structure, and usage) is under preparation.

- Acknowledgments

All the data employed to prepare the annual crop-type reference maps were kindly provided by the Regional Government of Andalusia (Consejería de Agricultura, Pesca, Agua y Desarrollo Rural, Junta de Andalucía).

Daily rainfall data were obtained from the Agroclimatic Information System for Irrigation (SIAR) of the Government of Spain (Ministerio de Agricultura, Pesca y Alimentación).

All Sentinel-1 images were downloaded from the Alaska SAR Facility (https://search.asf.alaska.edu/).

All Sentinel-2 images were obtained through the French Theia Land Data Centre (https://www.theia-land.fr/en/homepage-en/). 

All data and images included in this dataset are open access and publicly available.

Versions:

• Version 1.0. 08-May-2025. Original files
• Version 1.1. 26-May-2025. Changes:

  - Dataset:

  1) The dates of the coherence and phase products with temporal baselines of 6 and 12 days have been fixed. In the previous version there were some dates wrongly assigned. Please note that the date of each interferometric product is defined as the date of the second (most recent) acquisition.

  2) The file name of the dataset has been changed to avoid confusion about the file format.

  - Scripts (functions.py):

  1) In the previous version, some plots of time series were saved as empty (totally white) PNG files. A bug has been fixed in the corresponding functions.

  2) The functions relevant to the dual-pol model-based decomposition have been updated to avoid some numerical exceptions.

  3) The colormap used for representing the crop-type map is now pre-assigned (not random) and is the same for all years.