Patch-Based Forest Change Detection from Landsat Time Series

Abstract

Abstract:In the species-rich and structurally complex forests of the Eastern United States, disturbance events are often partial and therefore difficult to detect using remote sensing methods. Here we present a set of new algorithms, collectively called Vegetation Regeneration and Disturbance Estimates through Time (VeRDET), which employ a novel patch-based approach to detect periods of vegetation disturbance, stability, and growth from the historical Landsat image records. VeRDET generates a yearly clear-sky composite from satellite imagery, calculates a spectral vegetation index for each pixel in that composite, spatially segments the vegetation index image into patches, temporally divides the time series into differently sloped segments, and then labels those segments as disturbed, stable, or regenerating. Segmentation at both the spatial and temporal steps are performed using total variation regularization, an algorithm originally designed for signal denoising. This study explores VeRDET's effectiveness in detecting forest change using four vegetation indices and two parameters controlling the spatial and temporal scales of segmentation within a calibration region. We then evaluate algorithm effectiveness within a 386,000 km 2 area in the Eastern United States where VeRDET has overall error of 23% and omission error across disturbances ranging from 22% to 78% depending on agent.

Google Earth Engine Cloud Computing Platform for Remote Sensing Big Data Applications: A Comprehensive Review

“…Vegetation analysis is also a popular subject in temporal analysis. Therefore, GEE has several algorithms, such as Vegetation Change Tracker (VCT) [33] and Vegetation Regeneration and Disturbance Estimates through Time (VERDET) [34], which are specifically developed for this purpose. VCT can automatically analyze Landsat time-series images to generate forest disturbance history.…”

“…VERDET categorizes forest change into three types, including disturbed, stable, and regenerating. The analysis is based on the total variation regularization in the spatial and temporal domain [34].…”

A Suite of Tools for Continuous Land Change Monitoring in Google Earth Engine

“…Rather than rely on a pair of observations before and after to detect a change, algorithms can now be used to monitor land change in a more continuous fashion (Woodcock et al, 2020). The first time-series approaches used a single observation for each year, enabling the detection of changes on an annual basis (Huang et al, 2010;Kennedy et al, 2010;Hughes et al, 2017). These algorithms were usually based on a single spectral index and were designed to find specific types of change, often change in forests.…”

No-Cost Landsat Data Redefines Student Research Projects in Applied Remote Sensing Classes at Uw

“…Studies have used the spectral information collected by Landsat satellites 5, 7 and 8 to derive secondary information, i.e. spectral indices, which are valuable to monitoring or tracking changes in crop growth (Rufin et al, 2019), forest (Hansen et al, 2013;Hughes et al, 2017), and water quantity and quality (Arvor et al, 2018). This paper describes how no-cost Landsat data has transformed undergraduate student research projects in an applied remote sensing course at the University of Wyoming.…”