Source: UNIVERSITY OF ALASKA submitted to
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
Funding Source
Reporting Frequency
Accession No.
Grant No.
Project No.
Proposal No.
Multistate No.
Program Code
Project Start Date
Mar 18, 2019
Project End Date
Sep 30, 2021
Grant Year
Project Director
Young-Robertson, JE.
Recipient Organization
Performing Department
Forest Science
Non Technical Summary
Appropriate forest harvest management and use of local wood could mitigate the effects of climate warming in various ways, such as creating fuel breaks, planting resilient genotypes and/or species, and producing energy using renewable local wood resources. Forest management in the interior Alaska has in general had low profit margins because of a small local demand and a long distant to major markets. However, new revenue sources are emerging, such as wood biomass for energy generation and carbon credits which also mitigate the effects of climate change by reducing carbon footprints. In order to supply wood biomass sustainably for energy generation or to successfully trade carbon credits, accurate and precise forest inventory is essential.Data acquisition using UAV has great advantages in data collection over other techniques (e.g. field measurement and airborne remote sensing or laser scanning), such as lower cost, faster data acquisition and flexibility (e.g. weather conditions). UAV photogrammetry and Lidar are new, emerging technologies. The goal of this project is to find the optimum method to estimate aboveground biomass and to develop a protocol to accomplish fast, accurate and precise forest inventory using UAV.Aerial images and Lidar data will be acquired using UAV. Aerial images will be processed using Structure from Motion algorithm to generate dense 3D point clouds. For photogrammetry, digital Surface Model (DSM) will be developed to obtain the elevation of the surface of forests. Canopy Height Model (CHM) will be acquired by subtracting Digital Terrain Model (DTM) from DSM. Individual trees will be detected to estimate height, crown size, and species.Lidar data will be processed in ArcGIS. Both DSM and DTM will be generated using Lidar data unlike photogrammetry which can only generate DSM in dense forest. CHM will be obtained by subtracting DTM from DSM. Individual trees will be detected to estimate height crown size, and species.Aboveground biomass will be predicted using individual tree-based (sum of individual tree biomass) and area-based (biomass estimated using site characteristics, such as maximum, mean and percentiles of height) approaches. The outcomes will be compared to the field measurement and among the different methods to evaluate accuracies. I will develop a protocol of forest inventory using UAV and host workshops and training in UAV operation and data analysis to help land owners and managers learn efficient forest inventory technique using UAV photogrammetry.
Animal Health Component
Research Effort Categories

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
Goals / Objectives
Goal 1 - Finding the optimum method to estimate aboveground biomassGoal 2 - Securing funding to further advance the research area, such as increasing accuracy and precision, application for forest regeneration survey, application for carbon credit trading, and accounting for belowground carbon stocksGoal 3 -Educating land owners and managers by developing protocol and hosting workshops and training
Project Methods
Aerial images Lidar data of a white spruce, a mixed white spruce and hardwood, and a hardwood stand will be acquired by an unmanned aerial vehicle on board with a consumer grade camera and a light Lidar sensor. In the field, coordinates, species, status (live or dead), DBH, height and crown diameters of every tree with diameter at breast height (DBH) ≥ 5 cm will be recorded in fixed-radius (11.28 m) circular plots immediately before or after image acquisition. Aerial images will be processed using Structure from Motion algorithm to generate dense 3D point clouds to develop Digital Surface Model (DSM) which is the elevation of the surface of forests or objects. Canopy Height Model (CHM) will then be acquired by subtracting Digital Terrain Model (DTM) acquired by Airborne Laser Scanning (ALS) from DSM. Lidar data will be processed in ArcGIS. Both DSM and DTM will be generated using Lidar data and then Lidar derived DTM will be subtracted from DSM to obtain CHM.Individual trees will be detected using local maxima and segmentation techniques and tree detection rates will be evaluated for all forest types and tree detection techniques. Tree heights and crown diameters for individual trees will be obtained from CHM. The outcomes will be compared to the field measurements and evaluated using coefficient of determination (R2), root mean square error (RMSE), and mean difference. Tree species will be predicted using Random Forest classification algorithm with independent variables selected from tree height, canopy area, and RGB values. The classification outcomes will be evaluated using confusion matrix. Finally, for carbon stock estimation, individual tree-based and area-based aboveground biomass will be predicted by nonparametric regression method using the variables derived from CHMs. The outcomes will be compared to the values derived from field measurement. The performances of predictions will be evaluated using R2, RMSE, and mean difference. The performances of photogrammetry and Lidar will be compared.This project will recognize priority research areas which need to be advanced. I will identify those areas and develop proposals to secure funding for future research.Finally, I will develop a protocol for forest inventory using UAV photogrammetry and UAV Lidar for land owners and managers who are in need of efficient forest inventory technique. I will host workshops and training in UAV operation and data analysis. Survey will be conducted after each workshop or training to evaluate the effectiveness and identify their needs. The education and outreach part of the project will be primarily done using Renewable Resources Extension Act grant.

Progress 10/01/19 to 09/30/20

Target Audience:In this reporting period, we conducted additional UAV flights to increase replication for statistical analyses over long-term vegetation monitoring plots in collaboration with the State Forestry Division. The UAV sensors and flights will be used to quantify tree height, species composition, and tree density in the plots. We worked with the UAF UAV group ACUASI to build a new UAF with multi-sensor capacity, and we received training on the new UAV to develop new protocols for plot sampling. Changes/Problems:The original PI has taken a job with State Forestry, but we are continuing to collaborate with her and get her input on data collection and analysis. Additionally, COVID restricted our ability to meet all the goals in the prior year but we were able to conduct some flights and develop the new UAV. What opportunities for training and professional development has the project provided?A technician on the project received training on how to operate the new UAV. How have the results been disseminated to communities of interest? Nothing Reported What do you plan to do during the next reporting period to accomplish the goals?We plan to fly a couple more plots and analyze the data and share the information with stakeholders. Once the data are analyzes, we plan to determine the next steps with this project to secure future funding. The original PI has taken a new position outside of the university.

What was accomplished under these goals? We focused on Goal 1 due to COVID restrictions limiting our research capacity over the past year. Under Goal 1, we worked with ACUASI to attach new sensors to the UAV with improved capacity to estimate aboveground biomass and learn to fly the new UAV.


    Progress 03/18/19 to 09/30/19

    Target Audience: Nothing Reported Changes/Problems:Morimoto encounteredtechnical challenges with building drone and equipping sensors on it. Morimoto originally planned to use a DJI S1000 octocopter. However, Morimoto ended up buying Phantom 4 Proto collect data for the reporting period. The DJI S1000 was given to Morimoto by another researcher at UAF who bought it about 5 years ago. This drone was not built at all and discontinued when given and the flight controller is outdated. As a result, Morimoto had issues with updating the firmware and such. Morimoto was not sure if she could get the drone going in time and collect data with it, so she bought much simpler drone (Phantom 4 Pro) which has a built-in camera. Morimoto also intended to collect LiDAR data using the DJI S1000. However, collecting useful LiDAR data turned out to be much more challenging than anticipated and required more parts. Phantom 4 Pro also can't carry other sensors than a built-in camera. As a result, Morimoto only collected aerial images during the reporting period. What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest? Nothing Reported What do you plan to do during the next reporting period to accomplish the goals?Goal 1 -Morimoto plan to collect more field data and aerial images. Field data will be collected in new plots and existing plots. In existing plots, Morimoto will collect location data of each tree in the plots at a higher accuracy to be able to georeference photogrammetry outcomes at a higher accuracy and precision. Additional sample will increase the accuracy and precision of aboveground biomass estimations. Morimoto will use other algorithms, such as watershed delineation, for individual tree detections and try plot level aboveground biomass estimations. Orthomosaics and their RGB data will be used for species identification which will then be used for aboveground biomass estimations. Although Phantom 4 Pro isinexpensive and reliable to collect aerial images, it cannot carry other sensors than the built-in camera. A drone like Tarot X4 needs to be built but is inexpensive andhas a larger payload capacity allowing more flexibility. As a result, Morimoto plans to build a Tarot X4 with a help of Professor Michael Hatfield at UAF who is very knowledgeable with drones andhas built Tarot X4.With this drone, thermal, normalized difference vegetation index (NDVI) and LiDAR data will be collected to increase the accuracy of aboveground biomass estimation. Goal 2 - Morimoto will write proposals to secure funding. One project will focus on post-harvest regeneration and a use of UAV photogrammetry for regeneration survey. Another project will look at forest hydrology with UAV remote sensing technology (e.g. thermal images and NDVI). Goal 3 - Morimoto will attend and present at a UAV workshop held locally (pre-conference event of National Indian Timber Symposium). Morimoto will also create digital contents, such as videos and online articles, which shows how to use UAV remote sensing for forest management.

    What was accomplished under these goals? In spring and early summer, Morimoto tried to build a drone (DJI S1000) and LiDAR system. However, due to technical difficulties, Morimoto decided to use Phantom 4 Pro which is much simpler to fly and collect aerial images. Morimoto sampled two broadleaf stands (mix of birch and other species) and two black spruce stands in Caribou Poker Creeks Research Watershed, and two white spruce stands along Parks Highway. In broadleaf and white spruce stands, a 17.84 m radius circular plot and in black spruce stands, a 12.62 m radius circular plot were used. In each plot, distance and azimuth from the center of the plot, species, status (live or dead), crown position (open, dominant, co-dominant, intermediate, and shaded), diameter at breast height (DBH) and height are recorded for all stems larger than 2 cm in DBH. Morimoto flew the Phantom 4 Pro over the two broadleaf and two black spruce plots a few times after the field measurements with Matthew Robertson's help. The flight path was either single or double grid pattern with over 70% side and front overlaps. Flight altitude was between 40 and 80m. Ground control points (GCPs; large and reflective materials) were used to georeference the images. The images acquired by the drone were aligned using Structure from Motion algorithm in Agisoft Metahsape.The images obtained from the first few flights, especially from the broadleaf stands, had issues with aligning. This is likely due to a low flight altitude which made the images too similar between each other for the algorithm to detect similarities and differences. Georeferencing using GCPs was challenging due to low accuracy of GPS data at each GCPs. If GCPs were used to georeference, the 3D point cloudand orthomosaic were significantly distorted resulting in large errors. As a result, Morimoto aligned the images without GCPs in Agisoft and then georeferenced the 3D point cloud manually in ArcGIS. Canopy height models were produced in ArcGIS followed by individual crown detections using local maxima algorithm. Individual tree detections appeared to work well for black spruce stands but not so well for broadleaf stands because of simpler crown shape of black spruce compared to broadleaf.Better tree location data will help evaluate the accuracy and precision of tree detection and height estimate using photogrammetry.