Multi-temporal lidar test of chronosequence assumptions in secondary tropical forest

Secondary forests make up more than half of all tropical forests and are a globally significant carbon sink. However, nearly everything known about secondary forest regeneration comes from chronosequence studies that substitute space for time to approximate long-term secondary succession. Here we examine the efficacy of chronosequence predictions over 11 years of forest regrowth using two lidar datasets collected over the La Selva Biological Station in 1998 and 2009, each covering 381 ha of secondary forest and 803 ha of mature forest. We use these data to ask: 1) Do lidar waveforms from different age classes predict forest structure changes from repeated measurements at the same location? 2) Do simulated chronosequences predict the landscape mean biomass change? 3) How do differences in plot size and number affect the accuracy and precision of chronosequence based biomass recovery estimates? Lidar waveforms indicate that tree height and forest structure was similar between 1998 and 2009 for any given age class. For example, an 11-20 year old forest in 1998 had similar lidar returns to an 11-20 year old forest in 2009. Simulated chronosequences predict the landscape mean biomass change, but the accuracy of predictions depends on the size and number of plots used in the chronosequence. In forest with 0-10 years in 1998, 86 to 99% of 1000 simulated chronosequences predict the landscape mean biomass change within 20 Mg/ha depending on the plot size and number. However, predictions in forests with 11-20 years in 1998 are less accurate with 60-71% of predictions within 20 Mg/ha of the landscape mean. With area kept equal, chronosequences with many small plots, rather than fewer larger plots, have a higher probability of accurately predicting the landscape mean biomass change over the 11 year period. Overall, our results suggest both deterministic and stochastic controls on biomass accumulation in these secondary forests.