Robust fixed-count density estimation with virtual plots

Autor(es)

Magnussen, Steen ^[1]

Afiliação do(s) autor(es)

^[1] Pacific Forestry Ctr, Canadian Forest Serv, Victoria, BC V8Z 1M5 - Canada

• Tipo de documento: Artigo Científico
• Fonte: CANADIAN JOURNAL OF FOREST RESEARCH-REVUE CANADIENNE DE RECHERCHE FORESTIER; v. 44, n. 4, p. 377-382, APR 2014.
• Citações Web of Science: 2

Resumo

Fixed-count sampling (plotless) remains attractive for forest inventories in difficult terrains and for their control of the number (k) of trees to measure. Although recent fixed-count estimators of density (PDE) are less biased than older ones, the risk of a nontrivial bias remains a deterrent. A recently published PDE based on a generic algorithm for predicting distances to the k + m nearest tree (m = 1, 2, ...) has attractive properties in terms of average bias and average root mean squared errors across a wide spectrum of spatial point patterns. However, the risk of a sizeable bias remains an issue. Sensitivity to spatial patterns is seen as its main weakness. It is hypothesized that a new PDE with robust properties will mitigate the bias issue and encourage wider use. To this end, a new PDE estimator is proposed. It builds on a mixture of observed and predicted distances to a set of k + m nearest trees to generate counts of actual and virtual trees inside a circle with a data-driven fixed radius. The proposed new robust fixed-count density estimator achieved an average absolute bias of 1.2% when tested across a wide range of point patterns (54 actual and four simulated). The maximum absolute bias was 4.4%, a significant reduction when compared with otherwise attractive alternative PDEs. Root mean squared errors and coverage of 95% confidence intervals were also encouraging. The deterrent bias issue in PDEs has been sharply reduced with the proposed estimator. (AU)

• Processo FAPESP: 03/12595-7 - Composição florística, estrutura e funcionamento da Floresta Ombrófila Densa dos Núcleos Picinguaba e Santa Virgínia do Parque Estadual da Serra do Mar, estado de São Paulo, Brasil
• Beneficiário: Carlos Alfredo Joly
• Modalidade de apoio: Auxílio à Pesquisa - Programa BIOTA - Temático