Dataset Card for MODCiX - Mowing Detection Intercomparison Exercise

Dataset Summary

The MODCiX (Mowing detection intercomparison exercise) dataset is an unprecedented, independent reference dataset designed to evaluate and benchmark grassland mowing detection algorithms across Europe.

Compiled by multiple European research groups, it contains information on approximately 5,600 grassland mowing events collected between 2017 and 2021 across more than 3,000 parcels in eight European countries. The dataset was harmonized, anonymized, and split into training and validation sets to establish a standardized framework for evaluating remote-sensing-based algorithms utilizing optical (Landsat 8–9, Sentinel-2) and SAR (Sentinel-1) time series.

---

Dataset Structure and Contents

This repository is structured to cleanly separate the vector reference data from the heavy satellite image time series (approx. 274 GB). The imagery originates from a datacube created and managed with FORCE and has been masked using grassland parcels including a buffer.

Repository Tree

modcix/
├── data/
│   ├── imagery/                                              # Directory: All 30x30km FORCE satellite tiles
│   │   ├── X0045_Y0045/                                      # Sub-directory: Example tile coordinate
│   │   │   ├── 2017-2021_001-365_HL_TSA_LNDLG_BLU_TSS.tif    # Optical: Blue reflectance (all years)
│   │   │   ├── 2017-2021_001-365_HL_TSA_LNDLG_GRN_TSS.tif    # Optical: Green reflectance (all years)
│   │   │   ├── ...                                           # Optical: Other bands (all years)
│   │   │   ├── 2017-2021_001-365_HL_TSA_VVVHP_BVH_TSS.tif    # SAR: Backscatter coefficient VH (conditional years)
│   │   │   ├── 2017-2021_001-365_HL_TSA_VVVHP_BVV_TSS.tif    # SAR: Backscatter coefficient VV (conditional years)
│   │   │   ├── 2017-2021_001-365_HL_TSA_VVVHP_CVH_TSS.tif    # SAR: Coherence VH (conditional years)
│   │   │   ├── 2017-2021_001-365_HL_TSA_VVVHP_CVV_TSS.tif    # SAR: Coherence VV (conditional years)
│   │   │   └── *.tif.aux.xml                                 # Metadata: XML files containing band dates
│   │   ├── X0046_Y0045/         
│   │   └── ...
│   └── reference/                                            # Directory: Ground-truth mowing events
│       ├── modcix_reference_data_2017-2021_public_train.gpkg # Split: Training dataset
│       └── modcix_reference_data_2017-2021_public_test.gpkg  # Split: Validation/Test dataset
└── README.md                                                 # Hugging Face dataset card

1. Satellite Imagery (data/imagery/)

The imagery is divided into unique tile folders matching the 30x30 km FORCE tile grid system (naming convention: X00**_Y00**). Each tile folder contains individual GeoTIFF files for each feature and accompanying .xml metadata files providing the acquisition date of each band.

Global Raster Constants

• Pixel Size / Spatial Resolution: 10m x 10m
• Projection: Lambert-Azimuthal Equal Area (EPSG:3035)
• Data Type: 16-bit signed integer
• NoData Value: -9999

Data Domains & Preprocessing

Feature Domain	Available Bands / Identifiers	Temporal Coverage	Scaling Factor	Unit / Details
Optical	BLU (Blue), GRN (Green), RED (Red), NIR (Near-Infrared), SW1 (Shortwave IR 1), SW2 (Shortwave IR 2), NDV (NDVI), EVI (Enhanced Vegetation Index)	Continuous (All years 2017–2021)	1e-4 (divide by 10,000)	Platforms: Sentinel-2A/B (Level-1C), Landsat 8/9 (Level-1TP). Irregular interval. Steps: Preprocessing and atmospheric correction with FORCE
SAR Gamma Naught Backscatter	BVV (VV polarized), BVH (VH polarized)	Only years where reference data is present	1e-2 (divide by 100)	Platforms: Sentinel-1A/B (GRD). 6-day interval, ascending direction. Unit: dB. Steps: ThermalNoiseRemoval -> Remove-GRD-Border-Noise -> Calibration -> Speckle-Filter (Boxcar 3x3 px) -> Terrain-Correction (SRTM 1Sec HGT).
Interferometric Coherence	CVV (VV polarized), CVH (VH polarized)	Only years where reference data is present	1e-4 (divide by 10,000)	Platforms: Sentinel-1A/B (SLC). 6-day interval, ascending direction. Steps: Apply-Orbit-File -> Back-Geocoding (SRTM 1Sec HGT) -> Coherence (2x10 px, subtractTopographicPhase) -> TOPSAR-Deburst -> Terrain-Correction (SRTM 1Sec HGT).

2. Reference Data (data/reference/)

The reference datasets are delivered as GeoPackages (.gpkg) split into standalone training and test partitions. They include the following attributes:

• MOD_ID: Unique MODCiX identifier.
• Mow_1 - Mow_6: Dates of reported or observed mowing activity in the format YYYY-MM-DD.
• NMow: Total number of mowing events.
• Region: Label for each region.
• Year: Label for each year.
• Label: Quality label based on the initial method of data acquisition and/or other known issues (1: High certainty — 3: Confounding factors).
• Comment: Additional comments.

---

Dataset Creation

Source Data

Reference data on grassland management was acquired from various reliable sources, including strictly documented farmer logs, daily webcam time series, and direct field observations. Participating European research groups shared their regional data to provide a common baseline for the MODCiX exercise. This data was subsequently centralized, harmonized, and stripped of sensitive personal information.

Additional Information

Dataset Curators

Marcel Schwieder, Felix Lobert, Dominique Weber, Sophie Reinermann, Sarah Asam, Filippo Sarvia, Samuele De Petris, Enrico Borgogno-Mondino, Arnab Muhuri, Natascha Oppelt, Clement Atzberger, Iason Tsardanidis, Charalampos Kontoes, François Godechal, Cozmin Lucau-Danila, Viviane Planchon, Anatol Garioud, Célestin Huet, Silvia Valero, Clément Mallet, Julien Morel, Mattia Rossi, Francesco Vuolo, Aleksandar Dujakovic, Andreas Schaumberger, Andreas Klingler, Ann-Kathrin Holtgrave, Zander Venter, Ruth Sonnenschein, Mathilde De Vroey, Julien Radoux, Oliver Buck, Anna Katharina Franke, Uta Schumacher, Andreas Ostrowski, Patrick Hostert, Stefan Erasmi.

Licensing Information

This dataset is distributed under the Creative Commons Attribution 4.0 International (CC-BY-4.0) license.

Citation Information

If you use this dataset in your research, please cite the corresponding paper and Zenodo record:

@article{modcix2026a,
  title = {Mowing detection intercomparison exercise (MODCiX) – Evaluation of grassland mowing detection algorithms across Europe},
  year = {2026},
  journal = {Remote Sensing of Environment},
  volume = {342},
  pages = {115466},
  issn = {0034-4257},
  doi = {https://doi.org/10.1016/j.rse.2026.115466},
  url = {https://www.sciencedirect.com/science/article/pii/S0034425726002361},
  author = {Marcel Schwieder and Felix Lobert and Dominique Weber and Sophie Reinermann and Sarah Asam and Filippo Sarvia and Samuele {De Petris} and Enrico Borgogno-Mondino and Arnab Muhuri and Natascha Oppelt and Clement Atzberger and Iason Tsardanidis and Charalampos Kontoes and François Godechal and Cozmin Lucau-Danila and Viviane Planchon and Anatol Garioud and Célestin Huet and Silvia Valero and Clément Mallet and Julien Morel and Mattia Rossi and Francesco Vuolo and Aleksandar Dujakovic and Andreas Schaumberger and Andreas Klingler and Ann-Kathrin Holtgrave and Zander Venter and Ruth Sonnenschein and Mathilde {De Vroey} and Julien Radoux and Oliver Buck and Anna Katharina Franke and Uta Schumacher and Andreas Ostrowski and Patrick Hostert and Stefan Erasmi},
}

@article{modcix2026b,
  title={Mowing detection intercomparison exercise (MODCiX) – Evaluation of grassland mowing detection algorithms across Europe},
  year={2026},
  publisher={Zenodo},
  doi={10.5281/zenodo.18834294},
  url={https://doi.org/10.5281/zenodo.18834294}
  author = {Marcel Schwieder and Felix Lobert and Dominique Weber and Sophie Reinermann and Sarah Asam and Filippo Sarvia and Samuele {De Petris} and Enrico Borgogno-Mondino and Arnab Muhuri and Natascha Oppelt and Clement Atzberger and Iason Tsardanidis and Charalampos Kontoes and François Godechal and Cozmin Lucau-Danila and Viviane Planchon and Anatol Garioud and Célestin Huet and Silvia Valero and Clément Mallet and Julien Morel and Mattia Rossi and Francesco Vuolo and Aleksandar Dujakovic and Andreas Schaumberger and Andreas Klingler and Ann-Kathrin Holtgrave and Zander Venter and Ruth Sonnenschein and Mathilde {De Vroey} and Julien Radoux and Oliver Buck and Anna Katharina Franke and Uta Schumacher and Andreas Ostrowski and Patrick Hostert and Stefan Erasmi},
}