Deriving crop calendars from satellite phenology and unsupervised learning: an application to Andalusia (Spain)
Timely and accurate crop calendars are essential for optimizing agricultural practices and improving food security. Traditional calendar generation methods, based on field surveys, are often spatially limited and time-consuming. In this study, we present a replicable methodology to derive crop calen...
| Authors: | , |
|---|---|
| Format: | article |
| Status: | Published version |
| Publication Date: | 2026 |
| Country: | España |
| Institution: | Universidad de Sevilla (US) |
| Repository: | idUS. Depósito de Investigación de la Universidad de Sevilla |
| OAI Identifier: | oai:dnet:idus________::e7c3571d4d725e82c446deead7407e39 |
| Online Access: | https://hdl.handle.net/11441/184099 |
| Access Level: | Open access |
| Keyword: | Crop calendar Crop phenology Land Surface Phenology (LSP) Sentinel-2 Unsupervised classification |
| Summary: | Timely and accurate crop calendars are essential for optimizing agricultural practices and improving food security. Traditional calendar generation methods, based on field surveys, are often spatially limited and time-consuming. In this study, we present a replicable methodology to derive crop calendars using Land Surface Phenology (LSP) metrics extracted from Sentinel-2 imagery and unsupervised classification techniques. The analysis was conducted over Andalusia (Spain), a European NUTS-2 region characterized by high agroclimatic diversity and complex cropping systems. We computed Enhanced Vegetation Index 2 (EVI2) time series for more than one million agricultural plots and derived phenological metrics (start, middle, and end of season) using a double-logistic model. To address the circular nature of phenological data (expressed in Julian days), we applied sine and cosine transformations prior to clustering. Six phenological groups were identified using k-means, and a classification tree (CART) was used to interpret the clustering structure. The satellite-derived calendars were validated against official field-based calendars and showed high agreement for arable crops. The proposed approach allows for the generation of scalable, up-to-date, and spatially explicit crop calendars that can support agricultural monitoring systems, inform policy making, and enhance digital farming tools. © 2026 The Author(s) |
|---|