Continuous Change Detection and Classification (CCDC) Using NASA’s Harmonized Landsat and Sentinel-2 (HLS) Data in Google Earth Engine (GEE)

NASA’s Harmonized Landsat and Sentinel-2 (HLS) global surface reflectance products are generated by combining input data from OLI and MSI sensors aboard NASA/USGS’s Landsat 8/9 and ESA’s Sentinel-2A/B satellites, respectively. The analysis-ready HLS dataset is produced at a medium spatial resolution of 30m with a near-global coverage enabling land observation every 2-3 days. The production of harmonized surface reflectance on a common MGRS grid involves several processing steps, including atmospheric correction of Top of Atmosphere (TOA) data, cloud masking, normalizing bi-directional view angle effects and bandpass adjustment to account for sensor level differences with OLI as the reference. The dataset has undergone rigorous validation and consistency evaluation. The data harmonization is found to be efficacious and therefore the data products are suitable for quantitative analyses. Compared to the revisit times of individual constituent satellites, the HLS virtual constellation dataset offers significantly higher observational frequency. The HLS data archive exceeds 4 PB (and ~30M products) and extends nearly a decade (HLS Landsat component L30: April 2013 onwards, HLS Sentinel-2 component S30: Nov. 2015 onwards). This rich dataset is useful for many applications such as disaster response and vegetation monitoring. In particular, availability of highly frequent surface reflectance greatly benefits time series based analysis in uncovering seasonality and long-term trends. For geospatial analysis with large datasets and at global scales, Google Earth Engine (GEE) has emerged as a powerful platform which removes barriers to users by offering convenient tools and computing resources. HLS L30 data products are available on GEE and as of December 2024, HLS S30 data is being actively ingested. The goal of this study is to demonstrate the benefits of HLS data series compared to Landsat or Sentinel-2 alone data stacks by using the Continuous Change Detection and Classification algorithm available on GEE. The study will focus on a few key applications and highlight the ease of use at different scales by providing examples of pixel-based time series and spatial visualizations. These analyses can be further extended to other land cover applications to derive useful insights by leveraging benefits of HLS dataset and GEE.