Contemporary and historical detection of small lakes using cross-sensor super resolution Landsat imagery

Abstract

Landsat is the longest-running environmental satellite program and has been used for surface water mapping of large water bodies since its launch in 1972. Remote sensing image resolution is increasingly being enhanced through single image super resolution (SR), a machine learning task typically performed by neural networks. Here, we show that a 10x SR model (Enhanced Super Resolution GAN, or ESRGAN) trained entirely with Planet SmallSat imagery (3 m resolution) can be applied to 30 m Landsat imagery to produce 3 m Landsat SR images with preserved radiometric properties. We test the utility of these Landsat SR images for small lake detection by applying a simple water classification to SR and native Landsat imagery and comparing to independent, high-resolution water maps. SR images appear realistic and have fewer missed detections (type II error) compared to LR, but exhibit errors in lake location and shape, and yield increasing false detections (type I error) with decreasing lake size. SR enhancement improves detection of small lakes sized several Landsat pixels or less, with a minimum mapping unit (MMU) of ~ 2/3 of a Landsat pixel. We also apply the SR model to a historical Landsat 5 image and find similar performance gains, using an independent 1985 air photo map of 242 small Alaskan lakes. This demonstration of retroactively generated 3 m imagery dating to 1985 has exciting applications beyond water detection. Yet, much work remains to be done surrounding technical and ethical guidelines for the creation, use, and dissemination of SR satellite imagery.