Publication Date:
2024-06-12
Description:
This dataset contains detailed inventories of 7 large plots of mangrove forests in the Utría National Park in the Colombian Pacific Coast. The inventory consists of individual geo-referenced tree masks for the endemic Pelliciera rhizophorae species (〈site_name〉/pelliciera_trees/Pelliciera.shp), and area coverages for the Rhizophora mangle species, as well as Mud and Water areas (〈site_name〉/other_classes_coverage/〈class_name〉.tiff). For each individual tree of the Pelliciera rhizophorae species we provide the predicted height, crown diameter and crown area (〈site_name〉/pelliciera_trees/trees.csv). We also provide the cover area of the other predicted classes (〈site_name〉/other_area_coverage/area_coverages.csv). The inventories were automatically produced with trained Artificial Intelligence (AI) algorithms. The algorithms were trained with orthomosaic images and digital surface models (DSMs) produced from Unoccupied Aerial System (UAS) imagery with Structure-from-Motion software, both paired with expert annotations of the trees and areas (〈site_name〉/annotations/〈class_name〉.shp). In this dataset we provide all the input data for the algorithms, as well as the predicted geo-referenced data products, such as: predicted Pelliciera rhizophorae tree masks, Rhizophora mangle areas, Water areas, Mud areas, canopy height models (CHM), digital elevation models (DEM), digital terrain models (DTM) and various ancillary images. We also provide the initial orthomosaic files (〈site_name〉/orthomosaic.tif) and the DSM files (〈site_name〉/DSM.tif), that were produced with SfM software Agisoft Metashape v1.6.2 from the aerial footage captured in 2019 (19–22 February) using two consumer-grade UASs: the DJI Phantom 4 and DJI Mavic Pro (SZ DJI Technology Co., Ltd—Shenzhen, China). The DJI Phantom 4 has an integrated photo camera, the DJI FC330 and the DJI Mavic Pro was equipped with the integrated DJI FC220. The flights were programmed to follow the trajectories in an automated mode by means of the commercial application "DroneDeploy". Ground control points (GCPs) were positioned in the field, and their geographic location was acquired. We used two single-band global navigation satellite system (GNSS) receivers: an Emlid Reach RS+ single-band real-time kinematics (RTK) GNSS receiver (Emlid Tech Kft.—Budapest, Hungary) as a base station, and a Bad Elf GNSS Surveyor handheld GPS (Bad Elf, LLC—West Hartford, AZ, USA). RINEX static data from the base station was processed with the Precise Point Positioning Service (PPP) of the Natural Resources of Canada, while rover position was processed using the RTKLib software through a post processed kinematics (PPK) workflow. The final absolute positional accuracy of the products is below one meter because the results of the PPP workflow has a positional accuracy between 0.2 m and 1 m.
Keywords:
01_la_chunga_north_2019; 02_la_chunga_west_2019; 03_la_chunga_east_2019; 04_terron_colorado_2019; 05_estero_grande_shore_2019; 06_estero_grande_north_2019; 07_estero_grande_south_2019; 4D-REEF; Binary Object; Binary Object (File Size); Binary Object (MD5 Hash); Binary Object (Media Type); Camera, SZ DJI Technology, FC220; mounted on unoccupied aerial system (UAS), SZ DJI Technology, Phantom 4; Estero Grande North; Estero Grande Shore; Estero Grande South; forest inventory; Habitat Mapping; La Chunga East; La Chunga North; La Chunga West; machine learning; Mangrove forest; Past, present, and future of coral reefs in the Coral Triangle; Terron Colorado; UAS; UAV; UAV data; Unoccupied aerial system (UAS); followed by artificial intelligence (AI) algorithms; Utría National Park, Colombia
Type:
Dataset
Format:
text/tab-separated-values, 2 data points
Permalink
