Space2Ground 2.0 Dataset

Space2Ground 2.0 Dataset

The Space2Ground 2.0 dataset is a multi-source benchmark dataset for parcel-level agricultural monitoring in Cyprus during the 2022 growing season. It combines Sentinel-1 SAR, Sentinel-2 multispectral imagery, and crowdsourced street-level images with official agricultural parcel information  (geometries and crop type labels) to facilitate research on multimodal Earth Observation, crop classification, and agricultural monitoring.

The dataset was generated using the Space2Ground 2.0 processing framework, which automatically associates street-level observations and space-level time-series with agricultural parcels through viewpoint projection, quality filtering, and parcel-level annotation.

📄 Preprint: https://arxiv.org/abs/2607.28247

💻 GitHub: https://github.com/Agri-Hub/Space2Ground-2.0

🏛️ Accepted at: 45th EARSeL Symposium, Athens, Greece, 29 September – 2 October 2026

Dataset Contents

The dataset consists of three complementary components.

1. Reference Data

The file space2ground_2.0_cyprus_2022.gpkg contains two GeoPackage layers:

crop_parcels

• parcel geometries
• parcel identifiers
• crop codes
• crop names
• harmonized crop classes (level 2 taxonomy)
• higher-level crop groups (level 1 taxonomy)

street_photos_metadata

• image identifiers
• acquisition timestamps
• image URLs
• camera information
• associated parcel identifiers
• image locations

2. Satellite Component

Parcel-level satellite time series for the complete 2022 agricultural season, including:

• Sentinel-1 Ascending (VV/VH backscatter)
• Sentinel-1 Descending (VV/VH backscatter)
• Sentinel-2 multispectral observations

For each satellite acquisition date, pixel values intersecting each agricultural parcel were spatially aggregated using the mean value. The resulting monthly CSV files store parcel-level mean observations while preserving the original satellite acquisition timestamps.

3. Ground Component

The ground component consists of 46,050 crowdsourced street-level images collected through the Mapillary platform and associated with 8,581 agricultural parcels.

Each image is linked to the corresponding parcel through the metadata stored in the GeoPackage.

Applications

The dataset is intended for research on:

• Crop type classification
• Satellite and street-level data fusion
• Agricultural monitoring
• Computer vision
• Earth Observation
• Multimodal machine learning
• Parcel-level visual verification
• Benchmarking crop classification algorithms

Repository Structure

Space2Ground_2.0/
│
├── reference_data/
│   └── space2ground_2.0_cyprus_2022.gpkg
│
├── satellite_time_series/
│   ├── sentinel1_ascending/
│   ├── sentinel1_descending/
│   └── sentinel2/
│
└── street_level_images/

Citation

If you use this dataset, please cite:

Tsardanidis, I., Koukos, A., Choumos, G., Sitokonstantinou, V., & Kontoes, C. (2026). Space2Ground 2.0: A Multi-Source Dataset and Framework for Agricultural Monitoring through Fusion of Street-Level and Satellite Imagery.