Remote estimation of wheat yield based on vegetation indices derived from time series data of landsat 8 imagery

Abstract

Crop yield estimation prior to harvest is an important for planning and taking various policy decisions. The recent development in satellite remote sensing technologies with their increased spatial and temporal resolution enabled their enormous application for many users with low cost. The current study was planned based on time series Landsat 8 remote sensing data for real time estimation of wheat (Triticum aestivum L.) yield based on vegetation indices and ground-truthing wheat yield data for growing season 2015-16 in district Chakwal, Pakistan. Wheat yield data were collected from an area of 10 m × 10 m at 43 sites along with GPS positions at farmer’s field from district Chakwal to develop regressional model. Different indices like, Green Normalized Difference Vegetation Index (GNDVI), Normalized Difference Vegetation Index (NDVI), Wide Dynamic Ratio Vegetation Index (WDRVI) and Enhanced Vegetation Index (EVI) were derived from time series Landsat 8 imagery throughout the growing season (2015-16). Linear regression models fitting were developed between all the indices and ground truthing wheat yield data were analyzed based on Coefficient of Determination (R2) and Root Mean Square Error (RMSE). The results revealed that EVI index showed higher values for the month of March 2016, compared with other months. This showed that crop was at the booting and anthesis stage in these months. The EVI and GNDVI indices showed better accuracy and precision with coefficient of determination (R2) 0.89 and 0.82 values with RMSE value of 203.83 and 224.67 respectively for the month of March-2016. This indicated that Landsat 8 imagery can be used for reliable estimation of wheat yield prior to harvest which can be useful for planning and maintaining national food security stock timely.