LiDAR Based Street Tree Inventory for New York City

The vendor must use ground-based LiDAR equipment to scan trees on city rights-of-way and apply machine learning to produce a street tree inventory and 3D digital model. The scanner must have range accuracy ≤ 10 mm and range precision ≤ 7 mm at 30 m. Point density must meet minimum specifications at 20mph at 30 meters. Positional accuracy of the point cloud must be ≤ 10cm RMSE in open-sky conditions and ≤ 25 cm RMSE in degraded-GNSS areas. Metric accuracy thresholds include tree height ≤ 3ft, DBH ≤ 2in, crown height ≤ 3ft, lowest limbs ≤ 2ft, crown projection area ≤ 15% relative RMSE, and leaf area index ≤ 0.5 LAI units or 15% relative RMSE. The vendor's QA/QC plan must include a field-verification sample of at least 200 trees per borough. Deliverables are accepted when all thresholds are met; affected areas that fail must be reprocessed or re-captured at the vendor's cost. The vendor must provide a comprehensive report on discrepancies and corrective actions. Trees matching the existing inventory must be assigned to the species in the existing data; for new trees, Parks staff will coordinate with the vendor's arborist. Health condition must be a composite measure using leaf area index and dieback normalized to the species. All required metrics in Table 2 must be provided for all trees, including those already in the ForMS database.