Date of Award

Spring 5-5-2023

Level of Access Assigned by Author

Open-Access Thesis

Degree Name

Master of Forest Resources (MFR)


Forest Resources


Dr. Daniel Hayes

Second Committee Member

Dr. Ivan Fernandez

Third Committee Member

Dr. Shawn Fraver

Additional Committee Members

Ian Prior

Dr. Aaron Weiskittel


Forests provide essential ecosystem services such as carbon sequestration, clean water, lumber, and more. It is important that foresters be able to collect accurate forest inventories, especially in a changing climate. Foresters need to know what is in the forest not only to manage for the economic benefits, but also to manage for social acceptability and ecological soundness to prevent further degradation of these ecosystem services. One way to collect accurate and precise forest inventories is through the utilization of remote sensing products. These enhanced forest inventories (EFIs) can be done at varying resolutions that are contingent on the plot design creating wall-to-wall raster data and thus, complete spatial knowledge of these estimates can be determined. A popular remote sensing product to be used to create EFIs is airborne laser scanning (ALS). Although best practices guides have been created in other countries, research on the best plot type and design has not been done for Maine’ structurally diverse and intensively managed forests.

The goal of this study was to investigate a range of forest designs to determine the best ground-based calibration plot specifications for developing EFI models from ALS data in Maine. We developed a model that compared fixed versus variable radius plots, sampling size and intensity, and sample design with ALS data to map EFI variables like percent softwood, volume, BA, and tree count. Data were collected from the Penobscot Experimental Forest (PEF) in summer 2022 that had two different plot types, two sample sizes and sampling intensities, and two different sample designs. Data from other study sites were provided to us from our partners that only included one plot type, sample size and intensity, and sample design each. For validation, we used data collected in the Demeritt Forest also in summer 2022. We assessed R2, root mean square error (RMSE), coefficient of variation (CV), and mean bias for models with varying forest inventory designs to establish the best calibration plot for ALS in our study areas. It was determined that a principal component analysis for plot placement gave better model results than randomly placed plots. Also, fixed radius plots (FRPs) and a smaller sample size generated better evaluation statistics when predicting percent softwood, volume, and tree count in the PEF. In contrast, VRPs with a smaller sample size provided better model outcomes when predicting basal area (BA). Once the best forest inventory calibration plot design was identified and validated, we applied it to the PEF to estimate aboveground biomass.

Although there were obvious trends in our results, there is still more research to be done to ensure that our potential recommendations are correct. It seems that there was better model performance in spruce-fir forest types than other forest types like oak-pine. Our results provide insights on an optimal approach for specific conditions and underscore the importance of future research to assist decision-making on plot type and sample design for the broad range of conditions on forested landscapes in Maine.