Factors influencing predictions of understory vegetation biomass from visual cover estimates

Abstract

AbstractAccurately estimating forage biomass is essential for understanding the nutritional value of vegetation communities for wild herbivores. However, clipping, drying, and weighing vegetation by species requires substantial labor, which restricts the size and number of plots that can be sampled over a large geographical area. Labor‐intensive sampling effort can be reduced by using a double sampling strategy that pairs visual estimates of horizontal cover of vegetation with a subset of clipped plots to create equations that convert horizontal cover to biomass. However, sampling strategy and environmental covariates have rarely been considered when developing cover‐biomass equations. We evaluated environmental and sampling factors that could affect the cover‐biomass relationship using 3,404 cover‐biomass pairs measured in the Clearwater River Basin in North‐central Idaho, USA, collected from May 31–October 4 in 2016 and 2017. Our study demonstrated that when carefully applied, double sampling techniques can streamline field sampling while providing cover‐biomass conversion equations with moderate to high predictive ability (pseudo R2 = 0.25–0.99, = 0.70) for both biomass of current annual growth and high‐quality (e.g., leaves of shrubs) portion of plants (pseudo R2 = 0.65–1). Based on our results, we recommend comparing fit and predictive ability of regression types (linear, log‐linear, Gamma) and collecting >20 cover‐biomass pairs spanning a wide range of horizontal cover (1 to >50%) for each plant species. We found that accounting for overstory canopy cover and season through interaction terms improved the fit of equations by as much as 20% for over half of the 81 species‐specific and 12 structure groups. We found a modest degree of observer bias; thus, training should reduce observer bias and improve consistency but may not alleviate inconsistencies among studies. Careful classification of plant species into plant structure groups can reduce the number of equations, but species‐specific equations for the most abundant and objectively relevant plant species should be developed.