Analysis of unit level models for small area estimation in crop statistics assisted with satellite auxiliary information.

This article focuses on estimating block-level wheat crop yields in the Hisar and Sirsa districts of Haryana, India, using classical and robust Small Area Estimation (SAE) methods. The study compares the Empirical Best Linear Unbiased Prediction (EBLUP) technique based on maximum likelihood estimation with a robust random-effect predictor employing a Huber-type M-estimation to address the influence of outliers. Results indicate that while both methods yield similar estimates for Sirsa, the robust approach provides more accurate and reliable estimates for Hisar due to the presence of influential outliers in its data. The study highlights the utility of integrating remote sensing data, such as the Normalized Difference Vegetation Index (NDVI), with administrative records to improve micro-level agricultural yield estimation, supporting resource allocation and planning at smaller administrative levels.