Ding et al. (2025) Phenology-adapted potato mapping index (PMI) for ground sample-free identification

Identification

• Journal: International Journal of Applied Earth Observation and Geoinformation
• Year: 2025
• Date: 2025-11-30
• Authors: Peng Ding, Bingxue Zhu, Liwen Chen, Sijia Li, Kaishan Song
• DOI: 10.1016/j.jag.2025.104991

Research Groups

• State Key Laboratory of Black Soils Conservation and Utilization, Northeast Institute of Geography and Agroecology, Chinese Academy of Sciences (CAS), Changchun, China

Short Summary

This study proposes a novel Potato Mapping Index (PMI) leveraging Sentinel-2 temporal phenology for scalable, ground sample-free potato identification, achieving an average overall accuracy of 92% across six major potato-producing countries.

Objective

• To extract diagnostic optical features of potato growth from multi-temporal observations.
• To develop a robust and physically interpretable PMI for differentiating potatoes from other crops under diverse environmental settings.
• To evaluate the PMI’s applicability for extensive-area potato mapping across diverse terrains, farming systems, and climatic conditions, assessing its potential as a transferable framework for large-scale potato mapping.

Study Configuration

• Spatial Scale: Eight study sites across six countries (Belgium, China, Denmark, Germany, the Netherlands, United States) spanning three continents, covering approximately 970,000 square kilometers. Sentinel-2 data at 10 meter spatial resolution.
• Temporal Scale: Multiple years (2022 for S1-S4, 2023 for S5-S7, 2024 for S8). Sentinel-2 acquisitions between May and October, with NDVI time series analysis from April to October.

Methodology and Data

• Models used:
  • Potato Mapping Index (PMI)
  • PMIsoybean (adjusted PMI for soybean-dominated regions)
  • Random Forest (RF) as a benchmark supervised classifier
• Data sources:
  • Satellite: Sentinel-2 (COPERNICUS/S2SRHARMONIZED product from Google Earth Engine)
  • Reference data: European Union Crop Map 2022 (for S1-S4), United States Department of Agriculture (USDA) Cropland Data Layer (CDL) (for S5-S7), Field survey (for S8)
  • Validation samples: Land Use/Cover Area Frame Survey (LUCAS) dataset augmented by EU Crop Map (for S1-S4), random sampling from CDL (for S5-S7), comprehensive field campaign (for S8)
  • Auxiliary: European Space Agency (ESA) WorldCover 10 meter 2021 product (agricultural mask)

Main Results

• The PMI achieved an average overall accuracy (OA) of 92% and Kappa coefficients exceeding 0.8 across the eight study areas (970,000 square kilometers).
• The PMI consistently outperformed the traditional supervised classifier, Random Forest (RF), which yielded an average OA of 90% and an average Kappa coefficient of 0.72.
• The PMI demonstrated robust spatiotemporal generalization, maintaining high accuracy across diverse geographic settings, crop compositions, and climatic conditions without extensive training samples.
• An adjusted version, PMIsoybean, proved effective in regions with high soybean cultivation, showcasing the index's adaptability.
• The method effectively distinguished potato fields from adjacent crop types, enhanced boundary delineation, and mitigated spectral interference from mixed pixels and shadow effects in complex agricultural landscapes.

Contributions

• Proposes a novel, phenology-adapted Potato Mapping Index (PMI) for ground sample-free, large-scale potato identification.
• Achieves higher classification accuracy (92% OA) compared to traditional supervised methods (RF, 90% OA) without requiring extensive training samples.
• Demonstrates strong spatiotemporal transferability and adaptability across diverse environmental conditions, crop compositions, and cultivation practices in six countries.
• Offers a computationally efficient and unsupervised implementation, enhancing user-friendliness and scalability.
• Provides a robust framework that is phenological stage-independent and resilient to spatiotemporal growth asynchrony.

Funding

• National Key Research and Development Program of China (2024YFD1501100)
• Strategic Priority Research Program of the Chinese Academy of Sciences (XDA28050400)
• National Earth System Science Data Center, China (www.geodata.cn)

Citation

@article{Ding2025Phenologyadapted,
  author = {Ding, Peng and Zhu, Bingxue and Chen, Liwen and Li, Sijia and Song, Kaishan},
  title = {Phenology-adapted potato mapping index (PMI) for ground sample-free identification},
  journal = {International Journal of Applied Earth Observation and Geoinformation},
  year = {2025},
  doi = {10.1016/j.jag.2025.104991},
  url = {https://doi.org/10.1016/j.jag.2025.104991}
}