Comparison of Nearest Neighbor Methods for Estimating Basal Area and Stems per Hectare Using Aerial Auxiliary Variables

By V. Lemay, Hailemariam Temesgen. Forestry sciences, 2005

105 citations recorded by the source corpus.

Simulations were used to compare variable-space nearest neighbor methods for imputing stems per ha and basal area per ha (ground measured) for complex stands (many species and sizes) of southeastern British Columbia, Canada. Species composition and other characteristics obtained for every stand by interpreting aerial photography were used as auxiliary variables. The simulations included three distance metrics (squared Euclidean distance, most similar neighbor, and absolute distance), three intensities of stands with full information (20%, 50%, and 80%), two sets of aerial variables (mixed versus moderately high correlations with ground variables), and three averaging methods (nearest neighbor, average of three nearest neighbors, and distance weighted average of three nearest neighbors weighted). Increasing the number of stands with full information to 50% from 20% resulted in increased accuracy, with no noticeable improvement with a further increase to 80%. Of the three distance metrics, the most similar neighbor measure gave good results in imputing stems per ha and basal area per ha, particularly when there was a mixture of correlations, high and moderate, between the auxiliary (aerial) variables, and the ground variables. No large gain was noted in using the average of three neighbors rather than a single neighbor. FOR. SCI. 51(2):109–119.

Subfields: Ecology, Environmental Engineering, Nature and Landscape Conservation

View Paper: https://www.semanticscholar.org/paper/084369c07edc83b46450d55496f6fc30846cea8b

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Ecology

In Environmental Engineering

In Nature and Landscape Conservation

Citations in the corpus

Variable selection strategies for nearest neighbor imputation methods used in remote sensing based forest inventory (2012, background): “Numerous distance metrics are used in NN imputation and many comparisons of them have been published recently (e.g., LeMay and Temesgen, 2005; Chirici et al., 2008; Hudak et al., 2008; Breidenbach et al., 2010; Latifi et al., 2010).”
Model-based mean square error estimators for k-nearest neighbour predictions and applications using remotely sensed data for forest inventories (2009, methodology): “Today knn is widely used for small AOI forest inventoryestimationproblems(Collinsetal.,2004;Franco-Lopezetal., 2001; LeMay & Temesgen, 2005; McRoberts et al., 2002; Reese et al., 2002; Tomppo, 2006).”
Optimizing the k-Nearest Neighbors technique for estimating forest aboveground biomass using airborne laser scanning data (2015, background): “Metrics based on canonical correlation and canonical correspondence can also be expressed in matrix form (LeMay & Temesgen, 2005; Ohmann, Gregory & Roberts, 2014).”
Nearest neighbor imputation of species-level, plot-scale forest structure attributes from LiDAR data (2008, background): “…algorithms can provide predictions of timber volume (Mäkelä & Pekkarinen, 2004), basal area (FrancoLopez et al., 2001; LeMay & Temesgen, 2005), stems per hectare (LeMay & Temesgen, 2005), timber yield (Maltamo & Eerikäinen, 2001), and species-level forest inventory data (Temesgen et al., 2003).”
Evaluation of most similar neighbour and random forest methods for imputing forest inventory variables using data from target and auxiliary stands (2012, background): “The second question was addressed by testing a selection of k > 1 because averaging the neighbours could likely be advantageous over the use of a single neighbour if more than two response variables are of interest (LeMay and Temesgen 2005).”
Predicting categorical forest variables using an improved k-Nearest Neighbour estimator and Landsat imagery (2009, methodology): “…al., 2001; McRoberts et al., 2007; Reese et al., 2003; Tokola et al., 1996; Tomppo, 1996; Tomppo & Halme, 2004), Mahalanobis distance, and metrics based on canonical correlation analysis (LeMay & Temesgen, 2005; Temesgen et al., 2003) and canonical correspondence analysis (Ohmann & Gregory, 2002).”
Parametric, bootstrap, and jackknife variance estimators for the k-Nearest Neighbors technique with illustrations using forest inventory and satellite image data (2011, methodology): “Distancemetrics that are optimized using observations of the response variables include the Canonical Correlation Analysis metric (LeMay & Temesgen, 2005; Moeur & Stage, 1995), the Canonical Correspondence metric (Ohmann & Gregory, 2002), and the Fuzzy, Multiple regression, and Nonparametric…”