A novel fusion of Sentinel-1 and Sentinel-2 with climate data for crop phenology estimation using Machine Learning
Authors: Shahab A. Shojaeezadeh et al.
Links: [DOI Not Provided]
Why this paper matters
-
[cite_start]
- Decision Making: Crop phenology provides valuable information for agricultural management strategies [cite: 125]. [cite_start]
- Big Data: Leverages the ubiquity of Earth Observation data (Big EO Data) [cite: 126].
Problem Formulation
The goal is to accurately detect crop phenology (physiological development stages from planting to harvest) using Remote Sensing. Challenge: Optical data (Sentinel-2) is often blocked by clouds.
Methodology
This study employs a Multi-sensor Fusion approach combined with Machine Learning:
- Sentinel-1: Synthetic Aperture Radar (SAR) - penetrates clouds.
- Sentinel-2: Optical multispectral imagery. [cite_start]
- Climate Data: High-resolution weather data [cite: 126].
Teaching Module
Concept 1: Optical vs. Radar
Exercise: Why is fusing Sentinel-1 (Radar) necessary for accurate phenology estimation in temperate or tropical climates?
In many agricultural regions, growing seasons coincide with rainy seasons. Sentinel-2 (Optical) cannot see through clouds. Sentinel-1 (C-band SAR) penetrates clouds, providing continuous data on crop structure and biomass even during bad weather.
Concept 2: Feature Engineering
Exercise: What derived indices would you extract from Sentinel-2 for this task?
NDVI (Normalized Difference Vegetation Index) and EVI (Enhanced Vegetation Index) are standard inputs to track greenness and biomass density over time.