Comparing parametric and non-parametric methods of predicting Site Index for radiata pine using combinations of data derived from environmental surfaces, satellite imagery and airborne laser scanning

Abstract

Site Index (SI) is one of the main measures of forest productivity used throughout the world. For even-age plantations Site Index is defined as the height of dominant trees at a given reference age. Site Index is normally determined from field measurements and expressed from these measurements at the resolution of the stand. Development of fine resolution spatial surfaces describing variation in productivity across broad landscapes would be of considerable use in improving stand management. Using data obtained from a large Pinus radiata D. Don forest located in the central North Island, New Zealand, the objective of this study was to compare the precision of parametric and non-parametric models of Site Index that included explanatory variables extracted from aerially acquired Light Detection and Ranging (LiDAR), satellite imagery or environmental surfaces and combinations of these three data sources. Models were constructed both with and without age as an explanatory variable as managers may not always have access to stand age. A total of 32 models (16 data sources×two model methods) were constructed using data from 484 plots. Validation methods used to examine precision and bias of these models included leave one out cross validation and k-fold analysis.For all but one of the 16 data sources parametric models were found to be more precise than non-parametric models. Inclusion of stand age as an explanatory variable improved the precision of all but one model. For parametric models that included stand age, the R2 and RMSE (in brackets) for models with (i) all metrics derived from satellite imagery, (ii) environmental surface variables, (iii) variables derived from satellite imagery and environmental surfaces, (iv) LiDAR metrics and (v) all available variables were, respectively, 0.237 (2.850m), 0.613 (2.267m), 0.716 (2.025m), 0.883 (1.378m) and 0.801 (1.672m). These results show that LiDAR was the most useful data source for precise and unbiased prediction of Site Index. The parametric model created using variables derived from environmental surfaces and satellite imagery was also very precise showing that, in combination, these datasets may provide a useful alternative for predictions of Site Index when LiDAR data are not available.