Integration of Remote Sensing Data with GIS for Wheat Acreage Assessment in Akola District of Maharashtra, India

Abstract

Accurate and timely assessments of crop area coverage are crucial for efficient agricultural planning and policymaking. These assessments play a vital role in decision-making related to procurement, storage, public distribution, export, import and other agricultural matters. In the state of Maharashtra,India,wheat is the predominant crop grown during the winter season (rabi season), making it especially important for acreage estimation. Traditionally, crop acreage estimates have been obtained through various methods, including comprehensive surveys conducted by government agencies, sample surveys and personal assessments by local officials. However, the application of remote sensing and Geographic Information Systems (GIS) technology can offer more precise and timely acreage estimates. This research paper focuses on this methodology employed to classify wheat crop using Normalized Difference Vegetation Index (NDVI) time series data derived from multi-date Sentinel-2A satellite imagery. The study area encompasses the wheat-producing Akola district in Maharashtra state. Data from the 2022-2023 rabi season were utilized for analysis. The classification process involved two stages: first, the dataset was subjected to unsupervised Iterative Self-Organizing Data Analysis Technique (ISODATA) for clustering, followed by the labelling of these clusters based on the temporal spectral profiles of wheat and other competing crops. The subsequent steps included vectorization of the classified image, manual editing, labelling of mixed-class polygons and the application of decision rules for integration. This hybrid classification technique utilized the inherent clustering tendencies of land use and land cover classes in the feature space, enhanced by the inclusion of temporal data in the form of NDVI time series. Additionally, known crop signatures were utilized for labelling the clusters. The wheat acreage estimated using this approach for the study area was determined to be 19,553 hectares, deviating by 4.36% from the reference data reported by the Government of Maharashtra. This technique is characterized by its simplicity, time efficiency, reduced subjectivity and lower expertise requirements when compared to hierarchical classification techniques.