Uncertainly Assessment of Remote Sensing Derived Variables for Agricultural Applications

This research evaluates the performance of various open-access, remote-sensing derived agricultural variables, including ET0, ET, and NPP in different datasets through validation against ground-based measurements across different regions, climates, and management practices. The study will employ statistical analysis to systematically investigate the factors and sources of uncertainty, including model choice, parameterisations, input data quality, and spatial resolution, that influence their accuracy and applicability for agricultural applications and decision-making. This research includes four specific objectives. 1) To validate and assess the performance of multiple open access spatial ET0 datasets against ET0 estimations from ground stations, quantify the uncertainty and evaluate how dataset’s performance varies with different climate, elevation, spatial resolution, input data and model choices. 2)To conduct a comparative performance assessment and validation of multiple widely used NPP datasets against ground measurements on global croplands, quantify their uncertainty and perform a sensitivity analysis to determine how input data and parameterization choices affect the resulting error for applications in crop yield estimation. 3)To develop a high-resolution (10m) ET dataset using spatio-temporal machine learning models and validate its performance to quantify the reduction of uncertainty as an improved alternative to existing coarse-resolution datasets for field-scale water consumption assessments. 4)To integrate remote sensing data with field measurements to quantify the water consumption and yield of Direct-Seeded Rice (DSR) versus Transplanted Rice (TPR), and quantify the uncertainty associated with these final estimates to support water-management policy in Punjab.