Exploitation of Sentinel-2 Time Series for Horticulture Crops Inventory

Horticulture crops play an important commercial and economic role, providing employment and food security. A sustainable horticulture production requires updated and accurate statistics in terms of area and production. High temporal resolution remote sensing can be used to identify horticulture crops, especially vegetables that have shorter production cycles. Normalized Difference Vegetation Index (NDVI) bands generated from Sentinel-2A data are used to define the growth cycle of different vegetable types. NDVI time series allow the identification of several parameters, such as planting and maturation dates and crop cycle duration, that enable the characterization of each crop. A curve-matching algorithm, based on a set of NDVI curve parameters, were used to identify horticulture parcels. Two approaches were considered, one considering the total overlapping area and other considering a minimum of 1 pixel of overlap with the ancillary parcels delimitation. Results show that the latter approach allows the identification of possible horticulture crops with an accuracy higher than 80% while the former returns a lower accuracy of around 50%.