1 Introduction

In this module, we will discuss the following concepts:

  1. Learn about the types of data corrections commonly applied to remote sensing imagery.
  2. How to visually compare spatial data from different pre-processing levels within the same dataset.
  3. How to perform cloud masking and cloud masking assessments in Google Earth Engine for Landsat 8 surface reflectance imagery.

2 Background

What is Pre-Processing?
Much of the data you will find in Google Earth Engine (GEE) will have some level of pre-processing. This involves several different methods of quality control to ensure the highest levels of accuracy and consistency within raster collections. Depending on the collection, there may be a variety of pre-processing levels available and it is important to understand the differences to successfully integrate remotely sensed data into ecological studies. Three common sources of error for imagery products are consistently addressed by publishers before data are made available in GEE: atmosphere (i.e. air chemistry), topography (i.e. elevation), and geometry (i.e. pixel consistency).

Atmospheric Correction
As solar energy rebounds off the surface of the Earth and back towards our sensor in space, the atmosphere does a pretty good job of getting in the way. This takes place in the form of scattering and absorption (for more information, see Module 3). Identifying and correcting for these effects is important to accurately represent and interpret true surface conditions, like tree species leaf pigments or the differences between urban and agricultural pixels.

Topographic and Terrain Correction
Illumination effects from slope, aspect, and elevation pose additional challenges in collecting and processing remotely sensed data. Multiple correction methods have been developed, including the use of digital elevation models, to develop predictions of problematic topography. If your research is conducted at high altitudes or in areas of sheer relief, you will be relieved to know that pre-processing for terrain effects has been taken care of by specialists (though, for the careful and cautious, manual methods do exist).

Geometric Correction
This process makes sure the alignment of the raster images is systematic and consistent over time and relative to each other images. For Landsat, the processes of georeferencing and orthorectification are accomplished through independent ground control points and previously created digital elevation models. For an archival dataset like Landsat, ensuring that pixels line up pass after pass and year after year is paramount. Otherwise, remote sensing scientists and ecologist would have little ability to conduct multitemporal analyses.

It is important to remember that none of these quality assurance methods is 100% foolproof! Live by the motto of ‘Know Thy Data’ and carefully examine your images qualitatively and quantitatively. We will show a couple examples of this later in the module.

3 Pre-processing in Google Earth Engine with Landsat 8

It is an incredible advantage to have the (free!) dedicated support and behind-the-scenes work that happens before data becomes available in Google Earth Engine. However, you may still find it necessary to manipulate your dataset of interest to facilitate a particular research application. In this module, we will be working with Landsat 8 data and the image below details several use-cases for different levels of processing.

A decision workflow from Young et al, 2017 showing suggested use-cases for various levels of Landsat data pre-processing.

3.1 Example of Pre-processing Levels.

To give a qualitative sense of the difference between different pre-processing levels, we can look at several true-color images of southern Oregon, USA from late Summer, 2018. In this timeframe, our standard atmospheric interference has been worsened by smoke effects from the Carr Fire in northern California. To assess our initial imagery, we will load the ‘Raw’ Landsat 8 collection. Raw data (also called ‘at-sensor radiance’) has not be