Clustered disturbances lead to bias in large-scale estimates based on forest sample plots

Abstract

Assessments from field plots steer much of our current understanding of global change impacts on forest ecosystem structure and function. Recent widespread observations of net carbon accumulation in field plots have suggested that terrestrial ecosystems may be a carbon sink, possibly resulting from climate change and/or CO₂ fertilization. We hypothesize that field plots may inadequately sample inherently rare mortality events, leading to bias when plot level measurements are scaled up to larger domains. In this study, we constructed a simple computer simulation model of forest dynamics to investigate the effects of disturbance patterns on landscape-scale carbon balance estimates. The model was constructed to be a balanced biosphere at the landscape-scale with a uniform spatial pattern of forest growth rates. Disturbance gap-size distributions across the landscape were modelled with a power-law distribution. Small and frequent disturbances result in a well-mixed heterogeneous forest where even small sample plots represented domain-wide behaviour. However, with disturbances dominated by large and rare events, sample plots as large as 50 ha displayed significant bias towards growth. We suggest that the accuracy of domain level estimates of carbon balance from sample plots are highly sensitive to the distribution of disturbance events across the landscape, and to the number, size and distribution of field plots that comprise the estimate. Assumptions that small clusters of field plots may be representative of domain-wide conditions should only be made very cautiously, and warrant further investigation for verification.