Validation of allometric biomass models: How to have confidence in the application of existing models

Authors

Keryn I. Paul, CSIRO Agriculture and CSIRO Land and Water, ACT, Australia
Philip J. Radtke, Virginia Tech, Forest Resources & Environmental Conservation, Blacksburg, VA, United States
Stephen H. Roxburgh, CSIRO Agriculture and CSIRO Land and Water, ACT, Australia
John Larmour, CSIRO Agriculture and CSIRO Land and Water, ACT, Australia
Robert Waterworth, Mullion Group, 2a Fitzroy Rd, Forrest, ACT, Australia
Don Butler, Queensland Department of Science, Information Technology, Innovation and the Arts (DSITIA), Toowong, Qld, Australia
Kim Brooksbank, Department of Agriculture and Food, Western Australia, Esperance, AustraliaFollow
Fabiano Ximenes, Department of Primary Industries, Parramatta, NSW, Australia

Document Type

Article

Publication Date

3-15-2018

Journal Title

Forest Ecology and Management

ISSN

ISSN 0378-1127 eISSN 1872-7042

Keywords

Allometry, Above ground biomass, Bias, Carbon sequestration, Eucalyptus, Verification

Disciplines

Agribusiness | Agricultural Science | Agronomy and Crop Sciences | Biostatistics | Cellular and Molecular Physiology | Climate | Data Science | Environmental Indicators and Impact Assessment | Environmental Monitoring | Forest Biology | Forest Management | Natural Resources Management and Policy | Numerical Analysis and Scientific Computing | Plant Biology

Abstract

The development of biomass estimation models is highly resource intensive as it generally entails harvesting (or excavating) trees of a range of sizes to determine dry weight of above-ground (or below-ground) biomass. To maximise the cost effectiveness of such sampling, guidance is required on whether an allometric model that already exists is suitable for a new site or species, or whether further sampling and model development is necessary. With the aim to provide such guidance, we collated 12 pairs of well-sampled (N > 50) data sets of the same species at two sites, or two species at the same site. These provided case studies for: (i) assessing alternative statistical approaches to validate the application of a model developed using one data set to predict biomass of independent data from another site or species, and (ii) applying scenario analyses to explore the impact of sample size on uncertainty of validation, e.g. minimising type I and type II errors. Our results indicate that although an allometric model for a given species or plant functional type may be applied across multiple sites, validation will be important when an existing generic multi-site and multi-species model is applied to a new species. Results obtained demonstrated that an independent sample size of N ≤ 15 frequently (37–46% of the time) provides insufficient power to avoid incorrectly accepting “validation” (type II errors). Hence, to ensure a useful outcome from resources spent in sampling biomass, it is recommended that at least 50 trees be sampled for each species. An equivalence test may then be applied to determine if the minimum detectable negligible difference between the existing model and the new independent data is <25% (or whichever threshold is deemed acceptable). If so, the new data set may then be combined with existing data to refine a generalised model, which may then be applied with confidence. If not, then the resources expended need not be wasted as the sample size is sufficient to develop a new model suitable for application to the specific species sampled.

Recommended Citation

Keryn I. Paul, Philip J. Radtke, Stephen H. Roxburgh, John Larmour, Robert Waterworth, Don Butler, Kim Brooksbank, Fabiano Ximenes, Validation of allometric biomass models: How to have confidence in the application of existing models, Forest Ecology and Management, Volume 412, 2018, Pages 70-79, ISSN 0378-1127, https://doi.org/10.1016/j.foreco.2018.01.016.