Publication Details

Snow-covered Landsat time series stacks improve automated disturbance mapping accuracy in forested landscapes

Stueve, Kirk M.; Housman, Ian W.; Zimmerman, Patrick L.; Nelson, Mark D.; Webb, Jeremy B.; Perry, Charles H.; Chastain, Robert A.; Gormanson, Dale D.; Huang, Chengquan; Healey, Sean P.; Cohen, Warren B.

Year Published

2011

Publication

Remote Sensing of Environment. 115: 3203-3219

Abstract

Accurate landscape-scale maps of forests and associated disturbances are critical to augment studies on biodiversity, ecosystem services, and the carbon cycle, especially in terms of understanding how the spatial and temporal complexities of damage sustained from disturbances influence forest structure and function. Vegetation change tracker (VCT) is a highly automated algorithm that exploits the spectral-temporal properties of summer Landsat time series stacks (LTSSs) to generate spatially explicit maps of forest and recent forest disturbances. VCT performs well in contiguous forest landscapes with closed or nearly closed canopies, but often incorrectly classifies large patches of land as forest or forest disturbance in the complex and spatially heterogeneous environments that typify fragmented forest landscapes. We introduce an improved version of VCT (dubbed VCTw) that incorporates a nonforest mask derived from snow-covered winter Landsat time series stacks ( LTSSw) and compare it with VCT across nearly 25 million ha of land in the Lake Superior (Canada, USA) and Lake Michigan (USA) drainage basins. Accuracy assessments relying on 87 primary sampling units (PSUs) and 2640 secondary sampling units (SSUs) indicated that VCT performed with an overall accuracy of 86.3%. For persisting forest, the commission error was 14.7% and the omission error was 4.3%. Commission and omission errors for the two forest disturbance classes fluctuated around 50%. VCTw produced a statistically significant increase in overall accuracy to 91.2% and denoted about 1.115 million ha less forest (- .371 million ha disturbed and -0.744 million ha persisting). For persisting forest, the commission error decreased to 9.3% and the omission error was relatively unchanged at 5.0%. Commission errors decreased considerably to near 22% and omission errors remained near 50% in both forest disturbance classes. Dividing the assessments into three geographic strata demonstrated that the most dramatic improvement occurred across the southern half of the Lake Michigan basin, which contains a highly fragmented agricultural landscape and relatively sparse deciduous forest, although substantial improvements occurred in other geographic strata containing little agricultural land, abundant wetlands, and extensive coniferous forest. Unlike VCT, VCTw also generally corresponded well with field-based estimates of forest cover in each stratum. Snow-covered winter imagery appears to be a valuable resource for improving automated disturbance mapping accuracy. About 34% of the world's forests receive sufficient snowfall to cover the ground and are potentially suitable for VCTw; other season-based techniques may be worth pursuing for the remaining 66%.

Citation

Stueve, Kirk M.; Housman, Ian W.; Zimmerman, Patrick L.; Nelson, Mark D.; Webb, Jeremy B.; Perry, Charles H.; Chastain, Robert A.; Gormanson, Dale D.; Huang, Chengquan; Healey, Sean P.; Cohen, Warren B. 2011. Snow-covered Landsat time series stacks improve automated disturbance mapping accuracy in forested landscapes. Remote Sensing of Environment. 115: 3203-3219.

Last updated on: March 14, 2013