In-Season Tobacco Crop Growth and Yield Monitoring Using UAV and Satellite Imagery

Abstract

Tobacco is one of the main cash crops in Pakistan living source about 4 million people. Due to its importance it is necessary to monitor the tobacco crop timely by incorporating new technologies as well as in season crop yield forecasting. The main objective of the study was, to predict tobacco crop yield using UAV and planet-scope imagery also growth monitoring using field data and Sentinel imagery at district Swabi. Average number of leaves, stem thickness and plant height were assessed for 8 plots and tobacco plant’s growth stages were analyzed. Micasence Red-edge imagery on board Matrice 100 was used to calculate different indices processed in Pix4D mapper software. These Indices were further classified and correlation with observed yield was carried out to estimate yield for each index. SAVI was derived from PlanetScope satellite imagery and then yield was estimated for that 8 field plots. Area estimated from sentinel imagery was 16380 ha over estimating cultivated area about 1253 ha from ground data. For crop growth monitoring the tobacco crop have period of total number of about 110-120 number of days. Correlation of observed yield with indices showed from moderate to high, similarly have high correlation with number of leaves and plant height while having low correlation with stem thickness. MCARI, SAVI and TGI classification results were very much identical, but MCARI and SAVI misclassify the leaves outer most edges of leaves, while TGI also misclassify the center of canopy as well. Per hectare yield SIPI estimate almost same as observed tobacco yield of 2563.22 kg, with residual of 0.26 kg, GNDVI overestimating yield with 124.24 kg, MCARI overestimating it by 89.41 kg, SAVI by 98.6 kg, LCI by 229.23 kg, BNDVI by 346.02 kg, NDRE by 276.24, and NDVI overestimating tobacco yield by 489.31 kg. It is recommended to monitor tobacco crop using UAV at developing stage and temporally so can monitor tobacco crop growth efficiently and estimate yield accurately.