Name
Creating harmonized Landsat 7 and 8 data for tracking LULC change
Date & Time
Friday, May 23, 2025, 3:45 PM - 4:00 PM
Description

The Landsat archive can be used to track long-term (e.g. >50 years) changes in land surface variables. Some of the sensors used in the Landsat mission are considered broadly equivalent, such as the TM and Enhanced Thematic Mapper Plus (ETM+) sensors, and the Operational Land Imager (OLI) and OLI-2 sensors. However, substantial differences in atmospheric correction algorithms, radiometric resolution, and spectral response functions, suggest that cross-sensor harmonization should be implemented when using data from OLI/ OLI-2 sensors with TM/ETM+ data. To facilitate effective land use land cover (LULC) change detection mapping using the majority of the Landsat archive, we created the Landsat ETM+ OLI Harmonization Script (LEOHS) using Google Earth Engine (GEE) to create ETM+/OLI harmonization functions over user-defined regions. LEOHS uses Collection-2 Landsat Tier 1 imagery in GEE and Theil-Sen regressions to create harmonization functions with ETM+ and OLI pixels that were observed within +/- one day. To illustrate the benefits of LEOHS, we conducted a LULC case study in Menorca, Spain. We trained a Random Forest (RF) classifier on a 2020 OLI image and 10,000 samples of three landcover classes (high vegetation, low vegetation, and urban). The trained RF model was used to predict LULC in a 1990 TM image that was harmonized using LEOHS, and a non-harmonized image. The maps generated from the harmonized and non-harmonized images had overall accuracies of 84% and 72% respectively, illustrating the substantial gains in regional LULC studies from using harmonized imagery produced by LEOHS. Landsat ETM+/OLI harmonization is essential for unbiased time series analysis using most of the Landsat collection.

Location Name
Canal (CB) 2104
Session Type
Oral Presentation
Abstract ID
CCA110
Speaker Name
Galen Richardson
Speaker Organization
University of Ottawa
Session Name
CS111 Satellite Earth Observation