Methods to estimate aboveground wood productivity from long-term forest inventory plots

Abstract

Forest inventory plots are widely used to estimate biomass carbon storage and its change over time. While there has been much debate and exploration of the analytical methods for calculating biomass, the methods used to determine rates of wood production have not been evaluated to the same degree. This affects assessment of ecosystem fluxes and may have wider implications if inventory data are used to parameterise biospheric models, or scaled to large areas in assessments of carbon sequestration. Here we use a dataset of 35 long-term Amazonian forest inventory plots to test different methods of calculating wood production rates. These address potential biases associated with three issues that routinely impact the interpretation of tree measurement data: (1) changes in the point of measurement (POM) of stem diameter as trees grow over time; (2) unequal length of time between censuses; and (3) the treatment of trees that pass the minimum diameter threshold ("recruits"). We derive corrections that control for changing POM height, that account for the unobserved growth of trees that die within census intervals, and that explore different assumptions regarding the growth of recruits during the previous census interval. For our dataset we find that annual aboveground coarse wood production (AGWP; in Mgha-1year-1 of dry matter) is underestimated on average by 9.2% if corrections are not made to control for changes in POM height. Failure to control for the length of sampling intervals results in a mean underestimation of 2.7% in annual AGWP in our plots for a mean interval length of 3.6years. Different methods for treating recruits result in mean differences of up to 8.1% in AGWP. In general, the greater the length of time a plot is sampled for and the greater the time elapsed between censuses, the greater the tendency to underestimate wood production. We recommend that POM changes, census interval length, and the contribution of recruits should all be accounted for when estimating productivity rates, and suggest methods for doing this. © 2014 Elsevier B.V.