No abstract

Accurate measures of forest structural parameters are essential to forest inventory and growth models, managing wildfires, and modeling of carbon cycle. Terrestrial laser scanning (TLS) fills the gap between tree scale manual measurements and large scale airborne LiDAR measurements by providing accurate below crown information through non-destructive methods. This study developed innovative methods to extract individual tree height, diameter at breast height (DBH), and crown width of trees in East Texas. Further, the influence of scan settings, such as leaf-on/leaf-off seasons, tree distance from the scanner, and processing choices, on the accuracy of deriving tree measurements were also investigated. DBH was retrieved by cylinder fitting at different height bins. Individual trees were extracted from the TLS point cloud to determine tree heights and crown widths. The R-squared value ranged from 0.91 to 0.97 when field measured DBH was validated against TLS derived DBH using different methods. An accuracy of 92% (RMSE = 1.51 m) was obtained for predicting tree heights. The R-squared value was 0.84 and RMSE was 1.08 m when TLS derived crown widths were validated using field measured crown widths. Examples of underestimations of field measured forest structural parameters due to tree shadowing have also been discussed in this study. The results from this study will benefit foresters and remote sensing studies from airborne and spaceborne platforms, for map upscaling or calibration purposes, for aboveground biomass estimation, and prudent decision making by the forest management. OPEN ACCESSRemote Sens. 2015, 7 1878

Tidal wetlands contain large reservoirs of carbon in their soils and can sequester carbon dioxide (CO ) at a greater rate per unit area than nearly any other ecosystem. The spatial distribution of this carbon influences climate and wetland policy. To assist with international accords such as the Paris Climate Agreement, national-level assessments such as the United States (U.S.) National Greenhouse Gas Inventory, and regional, state, local, and project-level evaluation of CO sequestration credits, we developed a geodatabase (CoBluCarb) and high-resolution maps of soil organic carbon (SOC) distribution by linking National Wetlands Inventory data with the U.S. Soil Survey Geographic Database. For over 600,000 wetlands, the total carbon stock and organic carbon density was calculated at 5-cm vertical resolution from 0 to 300 cm of depth. Across the continental United States, there are 1,153-1,359 Tg of SOC in the upper 0-100 cm of soils across a total of 24 945.9 km of tidal wetland area, twice as much carbon as the most recent national estimate. Approximately 75% of this carbon was found in estuarine emergent wetlands with freshwater tidal wetlands holding about 19%. The greatest pool of SOC was found within the Atchafalaya/Vermilion Bay complex in Louisiana, containing about 10% of the U.S. total. The average density across all tidal wetlands was 0.071 g cm across 0-15 cm, 0.055 g cm across 0-100 cm, and 0.040 g cm at the 100 cm depth. There is inherent variability between and within individual wetlands; however, we conclude that it is possible to use standardized values at a range of 0-100 cm of the soil profile, to provide first-order quantification and to evaluate future changes in carbon stocks in response to environmental perturbations. This Tier 2-oriented carbon stock assessment provides a scientific method that can be copied by other nations in support of international requirements.

A plethora of information contained in full-waveform (FW) Light Detection and Ranging (LiDAR) data offers prospects for characterizing vegetation structures. This study aims to investigate the capacity of FW LiDAR data alone for tree species identification through the integration of waveform metrics with machine learning methods and Bayesian inference. Specifically, we first conducted automatic tree segmentation based on the waveform-based canopy height model (CHM) using three approaches including TreeVaW, watershed algorithms and the combination of TreeVaW and watershed (TW) algorithms. Subsequently, the Random forests (RF) and Conditional inference forests (CF) models were employed to identify important tree-level waveform metrics derived from three distinct sources, such as raw waveforms, composite waveforms, the waveform-based point cloud and the combined variables from these three sources. Further, we discriminated tree (gray pine, blue oak, interior live oak) and shrub species through the RF, CF and Bayesian multinomial logistic regression (BMLR) using important waveform metrics identified in this study. Results of the tree segmentation demonstrated that the TW algorithms outperformed other algorithms for delineating individual tree crowns. The CF model overcomes waveform metrics selection bias caused by the RF model which favors correlated metrics and enhances the accuracy of subsequent classification. We also found that composite waveforms are more informative than raw waveforms and waveform-based point cloud for characterizing tree species in our study area. Both classical machine learning methods (the RF and CF) and the BMLR generated satisfactory average overall accuracy (74% for the RF, 77% for the CF and 81% for the BMLR) and the BMLR slightly outperformed the other two methods. However, these three methods suffered from low individual classification accuracy for the blue oak which is prone to being misclassified as the interior live oak due to the similar characteristics of blue oak and interior live oak. Uncertainty estimates from the BMLR method compensate for this downside by providing classification results in a probabilistic sense and rendering users with more confidence in interpreting and applying classification results to real-world tasks such as forest inventory. Overall, this study recommends the CF method for feature selection and suggests that BMLR could be a superior alternative to classical machining learning methods.

No abstract