Question: Reliable estimates of understory (non-tree) plant cover following fire are essential to assess early forest community recovery. Photographic digital image analysis (DIA) is frequently used in seral, single-strata vegetation, given its greater objectivity and repeatability compared to observer visual estimation; however, its efficacy in multi-strata forest vegetation may be compromised, where various visual obstructions (coarse downed wood—CDW—conifer regeneration, and shadows) may conceal plant cover in the digital imagery. We asked whether vegetation complexity influences plant cover estimated by DIA relative to two visual methods: plot-level (20 m2) estimation (PLE) and quadrat-level (1 m2) estimation (QLE)?

Location: Greater Yellowstone Ecosystem, U.S.A.

Methods: We estimated understory plant cover in subalpine forest vegetation on permanent plots (n = 141) at two study areas ~30 years after the 1988 Yellowstone fires to 1) assess differences in visual obstructions between study areas in our digital imagery, 2) compare digital to visual estimates of plant cover, and 3) determine relationships between estimated plant cover and visual obstructions measured in situ.

Results: Percent conifer regeneration pixels differed significantly (odds ratio=8.34) between study areas, which represented the greatest difference in visual obstructions. At the study area with lower conifer pixels, DIA estimated 9% (95% confidence interval (CI)=3–14%) and 16% (95% CI=10–21%) more understory plant cover than PLE or QLE, respectively, but had comparable variability. At the study area with higher conifer pixels, DIA estimated 28% (95% CI=24–32%) and 22% (95% CI=18–26%) less understory plant cover than PLE or QLE, respectively, and had more variability. Furthermore, plot-level subcanopy regeneration (height>137 cm) density was negatively associated with digitally-derived plant cover but showed no relationship with visually-derived plant cover.

Conclusions: Post-fire conifer regeneration hindered the efficacy of DIA in estimating understory plant cover. Digital estimation is advantageous in single-strata vegetation but should not be used in complex, multi-strata vegetation.