Although there is an extensive array of optical remote sensing sensors from a variety of satellites providing long time data records, their data are incapable of retrieving reliable land surface information when clouds, aerosols, shadows, and strong angular effects are present in the scenes. The mitigation of noise and gap filling of satellite data are preliminary and almost mandatory tasks for any remote sensing application aimed at effectively analyzing the earth's surface continuously through time.
In 2020, to tackle this challenging problem Moreno-Martinez et al. (2020) proposed using the Google Earth Engine (GEE) cloud computing platform to implement the HIghly Scalable Temporal Adaptive Reflectance Fusion Model (HISTARFM). This method generates reduced noise and gap-free estimates of Landsat reflectance values at vast scales. Despite the computational power of GEE and the optimizations of HISTARFM, the computational burden and memory costs of HISTARFM are too high to carry out any extra computations after the gap-filling process. Therefore, the data have to be pre-processed in different study areas. We have generated data for number of world regions already, and in this tutorial we will show how to use it and provide examples of how you can improve your research and applications with this enhanced Landsat-based dataset.
page:
HISTARFM - How to Work with Gap-Filled Imagery | Google Earth Engine | Google Developers