Embracing Uncertainty and Probabilistic Outcomes for Ecological Critical Loads

Species are a sensitive gauge of air quality only if the “signal” of their response to atmospheric deposition is properly distinguished from the “noise” of model error, measurement error and ecological variation. Here, we quantified and mapped uncertainty in ten lichen-based critical loads (CLs) or exceedances for nitrogen and sulfur deposition in the USA. We tested the effects of model error by Monte Carlo resampling of model parameters, and the effects of measurement error in the number and identity of species using bootstrap resampling. Measurement error contributed more to uncertainty than model error. For nitrogen CLs, the average width of a 95% variability band (kg N ha −1 y −1 ) was 0.51–2.53 for model error, 2.42 for error in species number, and 3.22 for error in species identity. Variability bands for sulfur CLs were of similar magnitude. Despite its influential role, we found that measurement error was sufficiently small: > 84% of surveyed plots had more species than required to keep error below a stringent Measurement Quality Objective (SE < 1.5 kg ha −1 y −1 ), suggesting that the field sampling design is robust and defensible for assessing CL exceedances. Across the USA, uncertainty owing to species identities was greatest in species-poor areas where enhanced monitoring could reduce uncertainty. Overall, our findings establish ranges of plausible outcomes that communicate when and where to rely on estimates of air quality impacts. To avoid misplaced confidence, researchers and managers alike can embrace uncertainty to view critical loads not as sharp thresholds, but as a probabilistic distribution of plausible values.