Multi-frequency SAR time series for the detection of seedbed preparation and early-season crop classification

Details

Supervisors

Defourny, Pierre ; Bouchat, Jean

Faculty

Faculté des bioingénieurs

Degree label

Master [120] : bioingénieur en sciences et technologies de l'environnement, à finalité spécialisée

Abstract

In the context of the sustainable development goals driving 21st-century debates, agriculture faces the challenge of reducing environmental impact while meeting global consumption needs. This study focuses on the use of multi-frequency SAR imagery, more specifically X-band and C-band data, to evaluate their individual and combined roles in achieving seedbed preparation detection and early season classification for common spring crops cultivated in the southern region of Belgium, including beet, potato, and maize. A seedbed preparation detection method is developed based on the date associated to the minimum VV coherence point between mid-March and late May. These first results are promising and allow to highlight homo- and heterogeneity in spatial and temporal distribution of sowing events between the different cultures. Concerning crop classification, combining C-band and X-band data yields the best results, achieving reliable classification of beet fields in the beginning of April, potato fields by late April, and maize fields by late May. Balanced class distribution, sowing date, and temporal resolution are identified as crucial factors for early classification performance. Seedbed preparation detection and early-season crop classification can serve as valuable tools to build a more sustainable agricultural sector by predicting yield shortenings through agricultural models, providing farmers with helpful management practices and monitor soil preparation activities.