Elevated mortality rates of large trees allow for increased frequency of intermediate trees: a hypothesis supported by demographic model comparison with plot and LiDAR data

Tree diameter distributions are important indicators of forest structure, are a principal element of forest carbon stock estimates, and are the outcome of forest demography. Hump-shaped tree diameter distributions, also known as rotated-sigmoid diameter distributions, are characterized by an increased frequency of intermediately-sized trees over what would be extrapolated from the small tree size distribution and are common in temperate forests. One hypothesis to explain hump-shaped tree size distributions is the U-shaped mortality curve, where the mortality rates of the largest canopy trees are higher than those of intermediately-sized canopy trees. However, studies that have directly tested this hypothesis by comparing tree diameter distributions predicted from forest demographic models with empirical diameter distributions are lacking, and how the U-shaped mortality curve influences other aspects of forest structure such as canopy height is not well understood. We used model-data comparisons to test the hypothesis that the U-shaped mortality curve generates the hump-shaped tree diameter distribution. We used two versions of a forest demographic model (the PPA model) to predict tree diameter distributions, one that assumed that tree demographic rates varied only between understory and canopy trees, and a second that assumed tree demographic rates varied among understory trees, small canopy trees, and large canopy trees, consistent with the U-shaped mortality curve. We compared predictions from each form of the model with empirical tree diameter distributions from the Wind River Forest Dynamics plot, a 27.2 ha forest dynamics plot with a hump-shaped diameter distribution. We found that the model that accounted for the higher mortality rates of larger canopy trees generated a hump-shaped distribution, whereas the model that did not account for the higher mortality rates of larger trees was not able to generate a hump-shaped diameter distribution. To understand how the assumptions of forest demographic patterns affected other aspects of forest structure that could be derived from remote sensing, we compared predictions of forest canopy height from each form of the model with measurements of forest canopy height derived from airborne LiDAR. We found that accounting for the higher mortality rates of larger trees became even more critical to accurate prediction of forest canopy height because intermediate-sized trees occupy large areas in the canopy. Our results demonstrate the link between the U-shaped mortality curve and hump-shaped diameter distributions, and suggest that accurate characterization of the demography of large trees may be particularly important for predicting forest canopy structure and carbon stocks.