Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Titre : Integrating multispectral Sentinel-2 data, Google Earth Engine and Machine Learning techniques for the assessment of bean growth under different irrigation supplies

Auteur : Hachem, Ali

Etablissement de soutenance : Istituto Agronomico Mediterraneo di Bari (IAMB) - Centre international de hautes études agronomiques méditerranéennes (CIHEAM)

Grade : Master of science Sustainable water and land management in agriculture ; 2022

Résumé
Bean crop was grown in 2022 at the experimental fields of CIHEAM Bari (Southern Italy) to assess its response under full and deficit (50%) irrigation, and rainfed cultivation. Crop biophysical parameters (leaf area index, dry aboveground biomass, fraction of green canopy cover, chlorophyll content index and relative water content) were measured four times during the season in combination with satellite remote sensing data acquisition. Google Earth Engine (GEE) was applied to acquire Sentinel 2 spectral bands and integrate them with machine learning algorithms. Measured and estimated crop parameters were statistically compared and analysed inside the machine learning toolbox (ARTMO). Bean growth was strongly affected by reduction of water supply with a significant difference of crop biophysical parameters between irrigated and non-irrigated treatments. Satellite vegetation indices were able to detect crop response to different water regimes even with a low number of measurements. The highest correlation was observed for chlorophyll content index with a correlation coefficient R of 0.83 coupled with the best performing algorithm which is Random Forest (TreeBagger). The overall results affirmed the advantage of integrating remote sensing data, GEE and machine learning algorithms for the real time monitoring and assessment of crop growth.

Mots clés : AUTOMATED RADIATIVE TRANSFER MODELS OPERATOR (ARTMO) BEAN (PHASEOLUS VULGARIS L.), MACHINE LEARNING ALGORITHMS PRECISION AGRICULTURE REMOTE SENSING

Présentation (IAMB Library)