Methodology for Cleaning, Quality, and Normalization of Vegetation Index Data Derived from Sentinel 2

This paper proposes a methodology for preprocessing, cleaning, and normalizing vegetation indices derived from Sentinel-2 remote sensing data (satellites 2A and 2B), with an initial focus on the Normalized Difference Vegetation Index (NDVI). However, the methodology can be easily adapted to other vegetation indices that can be generated from these sensors. The method describes steps for extractingand cropping areas of interest, removing invalid data via quality assurance (QA) bands, discarding outliers based on user-provided parameters, and using anomaly detection algorithms, particularly the Local Outlier Factor (LOF), to produce post-processed data suitable for precision agriculture analysis and phenological modeling. Preliminary results indicate that the proposed workflow removes common noise such as cloud cover and shadows, as well as non-physical or inconsistent NDVI values,enabling more robust and reliable analyses.