Comparison of small area estimation methods applied to biopower feedstock supply in the Northern U.S. region. (February 2019)

• Record Type: Journal Article
• Authors: Goerndt, Michael E.; Wilson, Barry T.; Aguilar, Francisco X.
• Is Part Of: Biomass and bioenergy. Volume 121(2019)
• Page Start: 64
• Page End: 77
• Publication Date: 2019-02
• Subjects: Small area estimation -- Biomass -- Feedstock -- Biopower -- Imputation -- MODIS; Biomass energy -- Periodicals; Biomass -- Periodicals; Energy-Generating Resources -- Periodicals; Bioénergie -- Périodiques; 333.9539
• DOI: 10.1016/j.biombioe.2018.12.008
• Languages: English
• ISSNs: 0961-9534

Abstract

Abstract: Increasing interest in utilization of forest biomass for bioenergy has prompted extensive contemporary research regarding costs, supply and technology for efficiently producing electricity and other forms of renewable energy. One challenge facing both researchers and users is obtaining precise estimates of available forest biomass within plausible supply areas for individual power plants. Due to the wide distribution of power plants poised to co-fire with forest biomass, assessing its availability requires methods that can yield precise and low-bias estimates of aboveground forest biomass and other key attributes at varying spatial scales. Small area estimation (SAE) methods have high potential to accomplish this due to the availability of national forest inventory data, combined with satellite imagery and other forms of remotely-sensed auxiliary information. The study assessed several indirect, direct and composite estimators of four forest attributes: aboveground tree biomass, biomass of small-diameter trees, biomass of tops and limbs, and volume at the county-level and within the estimated supply areas around power plants across 20 states in the contiguous Northern U.S. Composite estimators using both k-nearest neighbors imputation and multiple linear regression provided superior estimates of indicators of forest biomass availability based on both precision and bias at the county-level at sampling intensities as low as 10–20%, compared to the other SAE methods examined. The composite estimator using k-nearest neighbors imputation was subsequently shown to produce precise estimates of forest biomass availability for selected power plant supply areas. Highlights: SAE methods can provide precise estimates of forest biomass using limited sample sizes. Composite predictors yield the most precise estimates of biomass at low sampling intensity. Composite KNN prediction was considered the optimal approach of the methods analyzed. Unlike estimators assuming uniformity, tested SAE methods capture variability across counties.