Theses and Dissertations

ORCID

https://orcid.org/0009-0003-5574-8593

Advisor

Poudel, Krishna P.

Committee Member

Adhikari, Ram K.

Committee Member

Polinko, Adam D.

Date of Degree

12-12-2025

Original embargo terms

Visible MSU Only 6 months

Document Type

Graduate Thesis - Campus Access Only

Major

Forestry

Degree Name

Master of Science (M.S.)

College

College of Forest Resources

Department

Department of Forestry

Abstract

We evaluated small area estimation (SAE) methods for county-level forest biomass across four Forest Inventory and Analysis (FIA) regions in the contiguous United States using the coefficient of variation and relative root mean squared error. The Standard Fay–Herriot (FH) model was compared with multivariate, spatial, and measurement-error extensions, while the unit-level Battese–Harter–Fuller (BHF) model was compared with mixed-effects random forests (MERF) with auxiliary data from remote sensing. Spatial FH improved precision when spatial dependence was strong; otherwise, Standard FH was more reliable. All FH models were effective with fewer than 100 plots. Measurement-Error FH produced more accurate estimates than FIA county means, on average. Multivariate FH improved estimation of coarse woody debris in its bivariate form, but yielded unstable errors when volume was added. MERF outperformed BHF in most regions. Overall, an effective SAE method for biomass depends on data structure, quality, and objectives.

Download

Share

COinS