Land use and land cover (LULC) mapping initiatives are essential to support decision-making and implement policies. There is a need for timely and accurate LULC classifications. However, building them is challenging. The Sentinel-2/MultiSpectral Instrument (S2/MSI) expands the frequency of satellite observations, a relevant factor to elaborate detailed and timely LULC classifications. However, storing, managing, and processing big data is costly and challenging, inducing a dimensionality reduction by modeling images as composite products. Contrastingly, it obliterates the temporal resolution improvement. As LULC changes are subtle over time, little is said about how much detail we lost by degrading temporal resolution. Data cube architectures enable storing, accessing, and modeling big data, mitigating losses. Brazil Data Cube (BDC) produces multidimensional data cube collections from different medium-resolution satellite data for Brazil, including S2/MSI. Given the demand for optimizing the generation of accurate information and to answer how to detect subtle variations across harvest periods with high accuracy (improving monitoring initiatives), we present a low-cost, semi-automated, accessible, and robust processing chain (classification scheme) created to generate LULC classifications with few training samples/calibration. It congregates steps from the insertion of LULC samples to the accuracy assessment of the LULC classification, and the core is a dense time series analysis approach from EO data cubes and the Surface Reflectance to Vegetation Indexes (sr2vgi), a tool to automate spectral index calculation. The technical highlights of this processing chain are the detection of interannual changes in management and monitoring with crops still on the stands (intra-harvest monitoring approach). As a proof-of-concept, we used this processing chain to generate LULC maps for the Brazilian Cerrado agricultural belt. The results, presented in the papers entitled Improving crop mapping in Brazil’s Cerrado from a data cubes-derived Sentinel-2 temporal analysis and A semi-automated workflow for LULC mapping via Sentinel-2 data cubes and spectral indices, indicate its potential to detect subtle landscape variations and provide timely and accurate LULC mapping by detecting different vegetation patterns in S2/MSI-derived time series. Moreover, this processing chain was presented in the course Aggregating crop calendar knowledge and Sentinel-2/MSI big data for crop monitoring, ministered by Michel Chaves at the XX Brazilian Symposium on Remote Sensing (XX SBSR), 2023.

The development of this processing chain was supported by the São Paulo Research Foundation (FAPESP) (grant 2021/07382-2 - MC) and the National Council for Scientific and Technological Development (CNPq) (grant PQ-310042/2021-6 - IS).