Object-Based Classification of Forest Disturbance Types in the Conterminous United States

Abstract

Forest ecosystems provide critical ecosystem goods and services, and any disturbance-induced changes can have cascading impacts on natural processes and human socioeconomic systems. Forest disturbance frequency, intensity, and spatial and temporal scale can be altered by changes in climate and human activity, but without baseline forest disturbance data, it is impossible to quantify the magnitude and extent of these changes. Methodologies for quantifying forest cover change have been developed at the regional-to-global scale via several approaches that utilize data from high (e.g., IKONOS, Quickbird), moderate (e.g., Landsat) and coarse (e.g., Moderate Resolution Imaging Spectroradiometer (MODIS)) spatial resolution satellite imagery. While detection and quantification of forest cover change is an important first step, attribution of disturbance type is critical missing information for establishing baseline data and effective land management policy. The objective here was to prototype and test a semi-automated methodology for characterizing high-magnitude (>50% forest cover loss) forest disturbance agents (stress, fire, stem removal) across the conterminous United States (CONUS) from 2003–2011 using the existing University of Maryland Landsat-based Global Forest Change Product and Web-Enabled Landsat Data (WELD). The Forest Cover Change maps were segmented into objects based on temporal and spatial adjacency, and object-level spectral metrics were calculated based on WELD reflectance time series. A training set of objects with known disturbance type was developed via high-resolution imagery and expert interpretation, ingested into a Random Forest classifier, which was then used to attribute disturbance type to all 15,179,430 forest loss objects across CONUS. Accuracy assessments of the resulting classification was conducted with an independent dataset consisting of 4156 forest loss objects. Overall accuracy was 88.1%, with the highest omission and commission errors observed for fire (32.8%) and stress (31.9%) disturbances, respectively. Of the total 172,686 km2 of forest loss, 83.75% was attributed to stem removal, 10.92% to fire and 5.33% to stress. The semi-automated approach described in this paper provides a promising framework for the systematic characterization and monitoring of forest disturbance regimes.

Classifications of Forest Change by Using Bitemporal Airborne Laser Scanner Data

“…The overall accuracy of 89% across districts is in line with accuracies reported in many of those studies. Using the Landsat archive, high overall accuracies have been reported for stand-replacing forest disturbance classifications, for example 93% [71], 88% [72], 90% [73], and 88% [74]. However, studies that included non-stand replacing disturbances in the classification reported lower overall accuracies; 75% [73] and 80% [75], likely because subtle forest changes do not tend to display a clear spectral change that can be linked to a change in land cover class [76,77].…”

Random Forest Algorithm for Mapping Deforestation in the Ituri-Epulu-Aru Landscape (Democratic Republic of Congo)

“…Also, the geomorphological indices seemed to play a less important role. They improved the total performance of the models by bringing non-redundant information to the models as earlier found by Huo et al [51] and Oeser et al [26]. The unique aspect of this study lies in the use of topography information which mitigates the relief effect on the spectral bands reflectance [33].…”

Phenology Modelling and Forest Disturbance Mapping with Sentinel-2 Time Series in Austria

“…However, this information is very important, for example, to specifically assess the amount of damage caused by bark beetle infestation [11]. Previous studies have already tried to discriminate different categories of change by directly using various disturbance metrics or by using those metrics as input data for machine learning algorithms (e.g., Random Forest) [20,[77][78][79][80]. First tests based on our data suggest that there are specific patterns both in the single pixel courses as well as in the FDD maps that could help to categorize disturbances by the cause of disturbance.…”