Image-based tea yield estimation using Landsat-8 OLI and Sentinel-2B images

Remote sensing imagery poses a considerable potential for estimating agricultural and plantation production, one of which is tea yields. Estimations using this technology provide explicit spatial information of the total area and production of tea yield. This study aims to assess the effect of spatial resolution (pixel size) of the images on tea yield estimates in the plantation area of PT (limited company) Pagilaran, Central Java, Indonesia. The estimation was carried out semi-empirically using fractional canopy cover (FC) as a proxy, which was applied to Landsat-8 OLI (30m) and Sentinel-2B (10m) images. Statistical correlation and regression analyses were performed to identify the relationship (1) between the Soil-Adjusted Vegetation Index (SAVI) and actual tea FC and (2) between fractional canopy cover of tea trees (FCtea) and actual tea yields. These analyses produced regression equations to build models of tea yield estimation based on the images. The results showed that the relationship between the two pairs of variables was linear and positive. Both images can be used to efficiently estimate total tea yield with a difference of merely 6.18 kg/ha/month. The large pixel size of Landsat-8 OLI causes a large number of mixed pixels of tea plants and other objects so that the accuracy of its tea production estimate is lower than that of the Sentinel-2B image. © 2020 Elsevier B.V.