Monitoring vegetation regeneration using multiple remotely piloted aircraft system sensors and methodologies
No Thumbnail Available
Date
2024
Authors
Pearse, Ben
University of Lethbridge. Faculty of Arts and Science
Journal Title
Journal ISSN
Volume Title
Publisher
Lethbridge, Alta. : University of Lethbridge, Dept. of Geography and Environment
Abstract
Effective restoration of vegetation following mining or other anthropogenic disturbances requires the ability to accurately measure indicators of progress towards benchmarks and the desired restoration end points. In this thesis, remotely piloted aircraft systems (RPAS) were used to collect data for estimating ecosystem proxies for productivity and measures of vegetation diversity at reference and reclamation sites in the Yukon and Alberta, with a focus on developing or confirming existing methods and testing their use on an operational level. In the first case study, the ability to estimate Leaf Area Index (LAI) and classify plant functional type using Object-Based Image Analysis in mixed species communities was evaluated. It was found that both conceptual and regression models were robust enough after two years of data collection to estimate LAI across the range of sites sampled (r2 ranging from 0.73 - 0.86 and RMSE from 0.29 - 0.38 m2/m2), showing spatiotemporal transferability of the models. Plant functional types (shrub, herbaceous, grass, and moss) were classified with high accuracy (F-scores ranging from 0.95 - 1.0).
The second case study assessed the potential for lidar to be used as a stand-alone sensor to monitor vegetation regeneration of a post-wildfire study site by estimating biomass and LAI and classifying woody and herbaceous vegetation. Furthermore, the ability to classify vegetation species was evaluated using object-based image analysis, multi-temporal data, and a fusion of multiple sensor types. The results show that average height was best for estimating biomass (R2 = 0.76, RMSE = 254 g/m2) at 1m2 plots. Woody and herbaceous vegetation were poorly classified using the lidar point clouds, however, the addition of spectral (NDVI) and moisture information (distance to a stream) improved the classifications. Object-based image analysis using a single data acquisition during a period of maximum foliage was unable to comprehensively classify species. However, the addition of a second data acquisition during the fall capitalized on spectral diversity of different species during different phenophases and improved the classification. This research demonstrates the unique potential of RPAS to be used in restoration monitoring with its ability to utilize different sensors and collect datasets dependent on user needs. The methods developed here for estimating productivity and species diversity can potentially be incorporated into long-term industry-based monitoring programs and can help decision-makers learn from current restoration efforts and apply successes to new areas.
Description
Keywords
vegetation regeneration , RPAS remote sensing , remote sensing , structure from motion , lidar remote sensing , mine reclamation