Prototyping a Generic Algorithm for Crop Parameter Retrieval across the Season Using Radiative Transfer Model Inversion and Sentinel-2 Satellite Observations

• Published in: Remote sensing (Basel, Switzerland) Vol. 13; no. 18; p. 3659
• Main Authors: Tomíček, Jiří, Mišurec, Jan, Lukes, Petr
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.09.2021
• Subjects: Alfalfa; Algorithms; Artificial neural networks; Barley; Beta vulgaris; Brassica; Brassica napus; Chlorophyll; Corn; Crop development; Crops; Czech Republic; data collection; Farms; fertilizers; Hordeum vulgare; In situ measurement; irrigation; Laboratories; Leaf area; Leaf area index; leaf chlorophyll content; leaf water content; Leaves; Lookup tables; Mathematical models; Medicago sativa; Moisture content; Neural networks; Nitrogen; Optimization; Parameters; Phenology; Precision agriculture; PROSAIL; Prototyping; Radiative transfer; Rapeseed; Remote sensing; Retrieval; Satellite observation; satellites; Sentinel-2; spring barley; sugar beet; Triticum aestivum; Variables; Vegetation; Water content; Wheat; Winter; Winter wheat; Zea mays; Zea mays subsp. mays; Czech Republic
• ISSN: 2072-4292; 2072-4292
• DOI: 10.3390/rs13183659

In this study, Sentinel-2 data were used for the retrieval of three key biophysical parameters of crops: leaf area index (LAI), leaf chlorophyll content (LCC), and leaf water content (LWC) for dominant crop types in the Czech Republic, including winter wheat (Triticum aestivum), spring barley (Hordeum vulgare), winter rapeseed (Brassica napus subsp. napus), alfalfa (Medicago sativa), sugar beet (Beta vulgaris), and corn (Zea mays subsp. Mays) in different stages of crop development. Artificial neural networks were applied in combination with an approach using look-up tables that is based on PROSAIL simulations to retrieve the biophysical properties tailored for each crop type. Crop-specific PROSAIL model optimization and validation were based upon a large dataset of in situ measurements collected in 2017 and 2018 in lowland of Central Bohemia region. For LCC and LAI, respectively, low relative root mean square error (rRMSE; 25%, 37%) was achieved. Additionally, a relatively strong correlation with in situ measurements (r = 0.80) was obtained for LAI. On the contrary, the results of the LWC parameter retrieval proved to be unsatisfactory. We have developed a generic tool for biophysical monitoring of agricultural crops based on the interpretation of Sentinel-2 satellite data by inversion of the radiation transfer model. The resulting crop condition maps can serve as precision agriculture inputs for selective fertilizer and irrigation application as well as for yield potential assessment.