Comparing Remote Sensing and In Situ Analyses on assessing the Phenology of Test Wheat Plants as Means for Optimizing Data Collection in Precision Agriculture

Abstract

This study assessed the potential applications of open data source satellite images in estimating the phenology of the wheat crop on a study farm found in the village of Ovcha Mogila, Bulgaria. A Landsat-9 and Sentinel-2 satellite images were extracted from the open data sources. An Unmanned Aerial Vehicle (UAV) was used to capture the spectral response of plant leaves. In addition, SpectraVue 710s Leaf Spectrometer was used to measure the spectral response of the crop at five different locations. The soil samples were collected in eight spots within the farm plot. The physicochemical properties of the soil (pH, texture, N, P2 O5, and K2 O) were analyzed in the certified laboratory of AUP. The five broadband vegetation indices (VIs) have been estimated based on the reflectance wavelength range of remote sensing tools. A linear regression analysis was used along with the coefficient of determination (R2 ), Root Mean Square Error (RMSE), and correlation (r) matrix for comparing the performance of the sensors. The soil analysis revealed the study farm plot is slightly alkaline with a dominant soil texture of Clay and Clay Loam. The vegetation indices (VIs) increased linearly with crop development. Significant correlations were observed for most vegetation indices of Sentinel-2, Landsat-9, and the Buteo drone, with the highest correlation for NDVI of Sentinel-2 and Buteo drone (R2 of 0.37 and RMSE of 0.06). In relative terms, the Sentinel-2 VIs correlated better with the Buteo drone vegetation indices than the Landsat-9. The Landsat-9 VIs somewhat align better with the leaf spectrometer.