http://www.nofc.forestry.ca/authors/read/20782?format=citation Publications by G. Daigle en-ca 2010-08-03 00:00:00 MST 2010-08-03 00:00:00 MST webmaster@nofc.cfs.nrcan.gc.ca http://www.nofc.forestry.ca/publications?id=31774 Estimation of forest attributes at the stand or polygon level across the forest domain is a basic component of forest inventory programs. We tested a “<em>k</em>-Nearest Neighbours” (<em>k</em>-NN)-based method for imputing merchantable volume. Our target dataset consisted of a discrete set of forest polygons within two large forest management units, and our reference dataset was a large historical database of temporary sample plots measured over the past three decades. The linkage between the target and reference datasets was provided by polygon-level photo-interpreted forest attributes. Measurements in temporary sample plots located in all target polygons enabled us to estimate fit statistics between imputed and measured merchantable volumes. A parallel imputation exercise was also done using the current operational method used by the Province of Québec to map forest attributes over the publicly owned forest lands. Results show that the volumes estimated using the historical <em>k</em>-NN method show fit statistics similar to those of the operational method, with a slightly higher bias that is largely within the error term of the estimates. For both methods, the coefficient of determination between measured and imputed merchantable volume is between 0.16 and 0.19 for total volume, increases substantially when the volume is partitioned between hardwoods and softwoods, but then decreases when the volume is further distributed among species. The results underline the importance of photo-interpretation uncertainties in bounding the accuracy of volume imputation as well as the value of the <em>k</em>-NN procedure for imputation purposes in the context of natural forests. Tue, 03 Aug 2010 http://www.nofc.forestry.ca/publications?id=31774 http://www.nofc.forestry.ca/publications?id=27698 This study proposes a within-subject variance-covariance (VC) structure to take into account repeated measurements and heteroscedasticity in a context of growth modeling. The VC structure integrates a variance function and a continuous autoregressive covariance structure. It was tested on a nonlinear growth model parameterized with data from permanent sample plots. Using a stand-level approach, basal area growth was independently modeled for red spruce (<em>Picea rubens</em> Sarg.) and balsam fir [<em>Abies balsamea</em> (L.) Mill.] in mixed stands. For both species, the implementation of the VC structure significantly improved the maximum likelihood of the model. In both cases, it efficiently accounted for heteroscedasticity and autocorrelation, since the normalized residuals no longer exhibited departures from the assumptions of independent error terms with homogeneous variances. Moreover, compared with traditional nonlinear least squares (NLS) models, models parameterized with this VC structure may generate more accurate predictions when prior information is available. This case study demonstrates that the implementation of a VC structure may provide parameter estimates that are consistent with asymptotically unbiased variances in a context of nonlinear growth modeling using a stand-level approach. Since the variances are no longer biased, the hypothesis tests performed on the estimates are valid when the number of observations is large. Tue, 30 Oct 2007 http://www.nofc.forestry.ca/publications?id=27698