Improving Precision in Forest Inventory through Small Area Estimation for Loblolly Pine Plantations in Coastal Georgia
| dc.contributor.author | Subedi, Bipana | en |
| dc.contributor.committeechair | Radtke, Philip J. | en |
| dc.contributor.committeemember | Coulston, John W. | en |
| dc.contributor.committeemember | Thomas, Valerie Anne | en |
| dc.contributor.department | Forest Resources and Environmental Conservation | en |
| dc.date.accessioned | 2025-02-01T09:00:18Z | en |
| dc.date.available | 2025-02-01T09:00:18Z | en |
| dc.date.issued | 2025-01-31 | en |
| dc.description.abstract | The use of small area estimation (SAE) in forest inventory has shown promise for improving the precision of estimates needed for informed decision-making when sample data are sparse. We evaluated the potential of unit-level SAE for increasing the precision of stand-level estimates of basal area, volume, and above-ground biomass estimates in loblolly pine plantations in coastal Georgia. Following the unit-level approach, field plots sampled in plantations owned by Rayonier Inc. were georeferenced to aerial lidar data using high-quality GPS field coordinates. Results focused on A) gains in precision for stand-level basal area, volume, and above-ground biomass estimates achieved by combining data from field plots with lidar-derived canopy height models in a SAE framework, B) impacts of small sample sizes on the precision of estimated stand level attributes, and C) the effects of nonrandom field plot placement in stands of interest when using unit-level SAE. Findings indicate that higher precision is achievable with greater variance stability than what is possible from very small samples of field data alone. This was true for all three attributes of interest. With careful attention to checking assumptions of the unit-level SAE approach, the use of non-random sampling does not appear to impair SAE's ability to deliver unbiased estimates for forest plantation stands. Simulating the entire population's basal area to test for the effects of non-random plot placement showed that SAE is robust to the type of sampling technique used. However, results can be affected when sampling is intentionally biased. This work can be useful to landowners and forest managers working with southern loblolly pine plantations. By leveraging simulation techniques to generate non-random sampling data from the available random sampling data, this study attempted to bridge the gap between the available empirical data and the desired sampling framework, ultimately widening the applicability of SAE in forest inventory settings. | en |
| dc.description.abstractgeneral | Accurate forest inventory estimates are essential to make important decisions for forest management. Our research explored how advanced statistical methods, specifically small area estimation (SAE), can enhance forest inventories when only limited data is available. We focused on loblolly pine plantations in coastal Georgia, using data from field plots combined with aerial lidar technology to estimate important forest metrics: basal area (tree density), wood volume, and above-ground biomass. By pairing field and lidar data, we found that SAE significantly improved the accuracy of forest estimates, even when the number of field samples was very small. We also tested how different sampling strategies, such as non-random plot selection, affected the results. Our results showed that SAE proved resilient to non-random sampling as long as certain assumptions were met. However, deliberate biases in sampling could still lead to less reliable estimates. Our findings provide valuable tools for forest managers and landowners, especially those managing loblolly pine plantations in the Southeastern US. By applying simulation techniques to extend the use of existing data, this study showed how SAE can fill data gaps and provide more accurate forest measurements, helping to guide better management and conservation decisions. | en |
| dc.description.degree | Master of Science | en |
| dc.format.medium | ETD | en |
| dc.identifier.other | vt_gsexam:41892 | en |
| dc.identifier.uri | https://hdl.handle.net/10919/124468 | en |
| dc.language.iso | en | en |
| dc.publisher | Virginia Tech | en |
| dc.rights | In Copyright | en |
| dc.rights.uri | http://rightsstatements.org/vocab/InC/1.0/ | en |
| dc.subject | Forest inventory | en |
| dc.subject | remote sensing | en |
| dc.subject | small area estimation | en |
| dc.subject | unit-level SAE. | en |
| dc.title | Improving Precision in Forest Inventory through Small Area Estimation for Loblolly Pine Plantations in Coastal Georgia | en |
| dc.type | Thesis | en |
| thesis.degree.discipline | Forestry | en |
| thesis.degree.grantor | Virginia Polytechnic Institute and State University | en |
| thesis.degree.level | masters | en |
| thesis.degree.name | Master of Science | en |
Files
Original bundle
1 - 1 of 1