Estimation of Carbon Storage in Urban Trees Using Multispectral Airborne Laser Scanning Data
MetadataShow full item record
With the continued growth of global population, urbanization becomes an inevitable trend. As substantial urban expansion undergoes, ecosystem and global land cover have been altered consequently. Urban development becomes the biggest contributor to global carbon emissions while the process of urbanization results in urban heat islands, climate change, and losses of carbon sinks. Urban vegetation has drawn direct attention of city planners and policy makers by considering the importance of vegetation in urban climate modification and energy conservation in different ways. For instance, tree shading and wind shielding effects can attenuate the direct solar heat and air infiltration into individual houses. In city wide, vegetation contributes the largest proportion of carbon storage which reduces climate warming and urban heat island effects by sequestering CO2 and storing carbon in biomass. The carbon content stored in individual trees can be estimated by dendrometric parameters such as the diameter at breast height (DBH) using allometry-based models. With the development of airborne laser scanning (ALS) technology, ALS data and very high resolution multispectral imagery have proven to be promising tools for deriving dendrometric parameters in forest. With the emerging multispectral ALS technology, it became possible to obtain both the range and spectral information from a single source meanwhile the intensity of multispectral ALS showed its power in vegetation mapping. This study aims to develop a workflow that can quantify the carbon storage in urban trees using multispectral ALS data. The workflow consists of four steps: multispectral ALS data processing, vegetation isolation, dendrometric parameters estimation, and carbon storage modeling. First, the raw multispectral ALS data is intensity-rectified and filtered to generate a normalized Digital Surface Model (nDSM) and multispectral ALS intensity information at wavelengths: 532 nm (Green), 1064 nm (Near-infrared, NIR), and 1550 nm (Shortwave Infrared, SWIR), respectively. Vegetation covers are isolated by the support vector machine (SVM) classifier using multispectral ALS intensity information and nDSM in which total six classes including two vegetation classes (grass and tree) are classified. Individual tree crown is delineated by local maxima filtering and marker-controlled watershed segmentation. Tree height and crown width are derived from the crown segments and compared with field measurements. An ALS-DBH (diameter at breast height) multiple linear regression model is developed to predict field-measured DBH using ALS-derived tree height and crown width and assessed by cross validation. Then the carbon storage in individual trees is calculated by allometric equations using ALS-estimated DBH and height. A total of 40 trees are sampled in the field that four attributes: height, crown width, DBH, and biomass are recorded for each single tree. The results show that the land cover classification with multispectral ALS intensity images and nDSM achieves above 90% overall accuracy. The result of local maxima filtering is improved by using both multispectral ALS intensity and nDSM as input data. The ALS-derived tree height has a root mean square error (RMSE) of 1.21 m (relative RMSE = 6.8%) and the ALS-derived crown width has a RMSE of 1.47 m (relative RMSE = 16.4%). The prediction performance of the ALS-DBH model achieves R2 over 0.80 with a RMSE of 4.6 cm. The predicted carbon storage using ALS-modeled DBH corresponded to a RMSE of 142 kg (28.6%) and a bias of 14.4 kg. Results suggest that ALS-based dendrometric parameter estimation and allometric models can yield consistent performance and accurate estimation. Citywide carbon storage estimation is derived in this study by extrapolating the values within the study area to the entire city based on the specific proportion of each land cover type in the entire city. The proposed workflow also reveals the potential of multispectral ALS data in estimating carbon storage at individual-tree level and mapping vegetation in the urban environment.