Assessing the performance of urban forest carbon sequestration models using direct measurements of tree growth

Across cities worldwide, people are recognizing the value of greenspace in ameliorating the health and well-being of those living there, and are investing significant resources to improve their greenspace. Although models have been developed to allow the quantification of ecosystem services provided by urban trees, refinement and calibration of these models with more accurate site-and species-specific data can increase confidence in their outcomes. We used data from two street tree surveys in Cambridge, MA, to estimate annual tree mortality for 592 trees and diameter growth rates for 498 trees. Overall tree turnover between 2012 and 2015 was relatively low (annualized 3.6% y(-1)), and mortality rate varied by species. Tree growth rates also varied by species and size. We used stem diameter (DBH) and species identity to estimate CO2 sequestration rates for each of 463 trees using three different model variations: (1) i-Tree Streets, (2) Urban Tree Database (UTD) species-specific biomass allometries and growth rates, and (3) empirically measured growth rates combined with UTD biomass allometries (Empirical + UTD). For most species, the rate of CO2 sequestration varied significantly with the model used. CO2 sequestration estimates calculated using i-Tree Streets were often higher than estimates calculated with the UTD equations. CO2 sequestration estimates were often the lowest when calculated using empirical tree growth estimates and the UTD equations (Empirical + UTD). The differences among CO2 sequestration estimates were highest for large trees. When scaled up to the entire city, CO2 sequestration estimates for the Empirical + UTD model were 49.2% and 56.5% of the i-Tree Streets and UTD estimates, respectively. We suggest future derivations of ecosystem service provision models allow localities to input their own species-specific growth values. By adding capacity to easy-to-use tools, such as i-Tree Streets, we can increase confidence in the model output. (C) 2017 Elsevier GmbH. All rights reserved.