Assessment of yield stability patterns using sentinel-2 over a 7-year period
This study employed Sentinel-2 satellite data spanning 2018–2024 to conduct a subfield-level yield stability assessment within the Danubian lowland, Slovakia. A robust spatiotemporal multi-crop framework was implemented based on Normalized Difference Vegetation Index (NDVI) classification, followed...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Taylor & Francis Group
2025-12-01
|
| Series: | Geocarto International |
| Subjects: | |
| Online Access: | https://www.tandfonline.com/doi/10.1080/10106049.2025.2532529 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This study employed Sentinel-2 satellite data spanning 2018–2024 to conduct a subfield-level yield stability assessment within the Danubian lowland, Slovakia. A robust spatiotemporal multi-crop framework was implemented based on Normalized Difference Vegetation Index (NDVI) classification, followed by multiyear stability analysis. This resulted in the identification of stable high, medium, low, and unstable yield productivity zones. Stable high yielding zones accounted for only 6.5% of the cropland, whereas unstable yield zones comprised almost 47%. The resulting map revealed scattered fine-scale variations in yield stability at the parcel level, while also identifying some general spatial trends related to climatic and elevation gradients. The key factors for distinguishing the two contrasting classes (stable high versus low yield zones) were analysed. Gradient boosting classification models (XGBoost) were implemented to distinguish these classes, achieving a good performance with a Receiver Operating Characteristic Area Under the Curve (AUC) of 0.86. Shapley additive explanation analysis (SHAP) was applied to enhance model interpretation. Notably, elevation, topsoil spectral reflectance (from a Sentinel-2-based bare soil mosaic), topography and field size have been identified as key predictive features for the classification model. |
|---|---|
| ISSN: | 1010-6049 1752-0762 |