Compositing high-resolution SDGSAT-1 nighttime light data by ranking of structural image features

To unlock the potential of high-resolution nighttime light (NTL) data, robust compositing methods are required. We present an automatic and scalable method for high-resolution NTL data compositing, demonstrated using SDGSAT-1 data. It operates at the pixel level by ranking structural image features across a temporal stack and selecting the observation acquired under the most favourable conditions. Unlike workflows that rely on external cloud masks and physical atmospheric correction models, it uses only internal image characteristics and stationarity metrics to mitigate contamination from clouds, haze, moonlight reflection, and sensor artefacts. The resulting composites show strong agreement with established, annual VIIRS products used as independent external benchmarks (correlation up to 0.95, R 2 consistently >0.80), while retaining a spatially crisper signal than VIIRS at a common aggregated scale (e.g. 800 m). The method also improves SDGSAT-1 usability by suppressing acquisition-dependent artefacts, including scene anomalies and inter-band RGB misregistration, and by improving spatial alignment: mean positional error is reduced from >340 m in the input data to 19.2 m in the composite. Internal geometric consistency improves markedly, with mean inter-band RGB misregistration decreasing from 47.6 m to 3.6 m. The workflow also introduces a taxonomy combining NTL brightness, temporal stationarity, and built-up area presence. For the Po Plain (Italy), the stationary NTL domain covers ∼10% of the area yet contains ∼60% of total light emissions; within it, 73.6% originate from built-up areas and 26.4% from non-built-up infrastructure (e.g. lit roads). Overall, the method supports integration of SDGSAT-1 data into global monitoring frameworks.