Wheat Acreage Mapping and Yield Prediction Using Landsat-8 OLI Satellite Data: a Case Study in Sahibganj Province, Jharkhand (India)

Abstract

The emergence of remote sensing technologies and availability of satellite data over three decades have facilitated to monitor and understand the agricultural systems in many intensive agricultural regions. Here, we performed a comprehensive study on utilization of multi-temporal satellite data (i.e., Landsat-8 and MODIS) for wheat acreage and yield estimation during winter season (2016–2017) over the Sahibganj District in Jharkhand (India). Phenological variables were derived using the time-series normalized difference vegetation index (NDVI), which helps to understand the phenological transitions of wheat. The NDVI profile was used to derive rules for decision tree classifier to map the acerage of wheat. The key findings indicate that the acerage of wheat was estimated as ~3870 ha. Further, the long-term wheat statistics data were used to derive a yield model. Based on this model, wheat production was predicted as ~4523 t for the winter season 2016–2017, while, the mean was 3482 t. Predicted wheat yield was as ~1.17 t/ha, which was underestimated by 0.07 t/ha. Thus, it can be concluded that the accuracy of yield prediction depends on the precision of wheat acerage map derived from remote sensing data. A significant challenge for accurate acerage mapping could be the coarser spatial resolution of satellite data as the average plot sizes of Indian farmers can be far smaller than pixel sizes of the satellite data. Nevertheless, this comprehensive case study inferred that satellite-derived wheat acerage can be preferred to predict yield instead of traditional-based survey estimates.