Forest Management Inventories Research Articles

Volumes by tree species can be predicted using photogrammetric UAS data, Sentinel-2 images and prior field measurements

Photogrammetric point clouds obtained with unmanned aircraft systems (UAS) have emerged as an alternative source of remotely sensed data for small area forest management inventories (FMI). Nonetheless, it is often overlooked that small area FMI require considerable field data in addition to UAS data, to support the modelling of forest attributes. In this study, we propose a method whereby tree volumes by species are predicted with photogrammetric UAS data and Sentinel-2 images, using models fitted with airborne laser scanning data. The study area is in a managed boreal forest area in Eastern Finland. First, we predicted total volume with UAS point cloud metrics using a prior regression model fitted in another area with ALS data. Tree species proportions were then predicted by nearest neighbor (-NN) imputation based on bi-seasonal Sentinel-2 images without measuring new field plot data. Species-specific volumes were then obtained by multiplying the total volume by species proportions. The relative root mean square error (RMSE) values for total and species-specific volume predictions at the validation plot level (30 m × 30 m) were 9.0%, and 33.4–62.6%, respectively. Our approach appears promising for species-specific small area FMI in Finland and in comparable forest conditions in which suitable field plots are available.kk

From comprehensive field inventories to remotely sensed wall-to-wall stand attribute data — a brief history of management inventories in the Nordic countries

Forest management inventories (FMIs) provide critical information, usually at the stand level, for forest management planning. A typical FMI includes (i) the delineation of the inventory area to stands by applying auxiliary information; (ii) the classification of the stands according to categorical attributes such as age, site fertility, main tree species, and stand development; and (iii) measurement, modelling, and prediction of stand attributes of interest. The emergence of wall-to-wall remote-sensing data has enabled a paradigm change in FMIs from highly subjective, visual assessments to objective, model-based inferences. Previously, optical remote-sensing data were used to complement visual assessments, especially in stand delineation and height measurements. The evolution of airborne laser scanning (ALS) has made objective estimation of forest characteristics with known accuracy possible. New optical and Lidar-based sensors and platforms will allow further improvements of accuracy. However, there are still bottlenecks related to species-specific stand attribute information in mixed stands and assessments of tree quality. Here, we concentrate on approaches and methods that have been applied in the Nordic countries in particular.

Model-Based Estimation of Forest Inventory Attributes Using Lidar: A Comparison of the Area-Based and Semi-Individual Tree Crown Approaches

The use of individual tree detection methods to support forest management inventories has been a research topic for over two decades, but a formal assessment of these methods to produce stand-level and region-level predictions of forest attributes and measures of error is lacking. We employed model-based estimation methods in conjunction with the semi-individual tree crown approach (s-ITC) to produce predictions and measures of error for tree volume (VOL), basal area (BA), stem density (DEN), and quadratic mean diameter (QMD) at the scale of forest stands and the entire study region. We compared the s-ITC approach against the area-based approach (ABA) for predictions of region-level and stand-level attributes via model-based root mean squared errors (RMSEs). The study was conducted at the Panther Creek watershed in Oregon, USA using a set of 78 field plots and aerial lidar information. For region-level attributes, s-ITC RMSEs demonstrated changes between −31% and 17% relative to ABA models. At the stand level, median s-ITC RMSEs generally increased, with changes between −29% and 414% relative to ABA models, but demonstrated important reductions in stands where segmentation provided large increases in sample size and was less prone to extrapolation than ABA models. The ABA demonstrated smaller RMSEs in stands without sampled population units for all variables. Our findings motivate further research into niche applications where s-ITC models may consistently outperform ABA models.

Predicting stand age in managed forests using National Forest Inventory field data and airborne laser scanning

BackgroundThe aim of this study was to construct a nationwide stand age model by using National Forest Inventory (NFI) data and nationwide airborne laser scanning (ALS) data. In plantation forestry, age is usually known. While this is not the case in boreal managed forests, age is still seldom predicted in forest management inventories. Measuring age accurately in situ is also very laborious. On the other hand, tree age is one of the accurately measured sample tree attributes in NFI field data. Many countries also have a nationwide coverage of airborne laser scanning (ALS) data. In this study, we merged these data sources and constructed a nationwide, area-based model for stand age.ResultsWhile constructing the model, we omitted old forests from the data, since the correlation between ALS height metrics and stand age diminished at stands with age > 100 years. Additionally, the effect of growth conditions was considerable, so we also utilized different geographical and NFI variables such as site fertility and soil type in the modeling. The resultant nationwide model for the stand age of managed forests yielded a root mean square error (RMSE) of about 14 years. The model could be improved further by additional forest structure variables, but such information may not be available in practice.ConclusionsThe results showed that the prediction of stand age by ALS, geographical and NFI information was challenging, but still possible with moderate success. This study is an example of the joint use of NFI and nationwide ALS data and re-use of NFI data in research.

Airborne-laser-scanning-derived auxiliary information discriminating between broadleaf and conifer trees improves the accuracy of models for predicting timber volume in mixed and heterogeneously structured forests

Managing forests for ecosystem services and biodiversity requires accurate and spatially explicit forest inventory data. A major objective of forest management inventories is to estimate the standing timber volume for certain forest areas. In order to improve the efficiency of an inventory, field based sample-plots can be statistically combined with remote sensing data. Such models usually incorporate auxiliary variables derived from canopy height models. The inclusion of forest type variables, which quantify broadleaf and conifer volume proportions, has been shown to further improve model performance. Currently, the most common way of quantifying broadleaf and conifer forest types is by calculating the proportions of the corresponding areas of the canopy cover. This practice works well for single-layer forests with only a few species, but we hypothesized that this is not best practice for heterogeneously structured and mixed forests, where the area proportion does not accurately reflect the timber volume proportion. To better represent the broadleaf and conifer volume proportions, we introduced two new auxiliary variables in which the area proportion is weighted by height information from a canopy height model.The main objectives of this study were: (1) to demonstrate the advantage of including forest type (broadleaf/conifer distinction) information in ordinary least squares regression models for timber volume prediction using widely available data sources, and (2) to investigate the hypothesis that including the broadleaf and conifer proportions, weighted by canopy height information, as additional auxiliary variables is favourable over including simple area proportions. The study was conducted in three areas in Switzerland, all of which have heterogeneously structured and mixed forests. Our main findings were that the best model performance can generally be achieved: (1) by deriving conifer and broadleaf proportions from a high-resolution broadleaf/conifer map derived from leaf-off airborne laser scanning data, and (2) by using broadleaf/conifer proportions weighted by height information from a canopy height model. Incorporating the so-derived conifer and broadleaf proportions increased the model accuracy by up to 9 percentage points in root mean square error (RMSE) compared with models not using any forest type information, and by up to 2 percentage points in RMSE compared with models using conifer and broadleaf proportions based solely on the corresponding area proportions, as done in current practice. Our findings are particularly relevant for mixed and heterogeneously structured forests, such as those managed to achieve multiple functions or to adapt effectively to climate change.

Sensitivity Analysis of the DART Model for Forest Mensuration with Airborne Laser Scanning

Airborne Laser Scanning (ALS) measurements are increasingly vital in forest management and national forest inventories. Despite the growing reliance on ALS data, comparatively little research has examined the sensitivity of ALS measurements to varying survey conditions over commercially important forests. This study investigated: (i) how accurately the Discrete Anisotropic Radiative Transfer (DART) model was able to replicate small-footprint ALS measurements collected over Irish conifer plantations, and (ii) how survey characteristics influenced the precision of discrete-return metrics. A variance-based global sensitivity analysis demonstrated that discrete-return height distributions were accurately and consistently simulated across 100 forest inventory plots with few perturbations induced by varying acquisition parameters or ground topography. In contrast, discrete return density, canopy cover and the proportion of multiple returns were sensitive to fluctuations in sensor altitude, scanning angle, pulse repetition frequency and pulse duration. Our findings corroborate previous studies indicating that discrete-return heights are robust to varying acquisition parameters and may be reliable predictors for the indirect retrieval of forest inventory measurements. However, canopy cover and density metrics are only comparable for ALS data collected under similar acquisition conditions, precluding their universal use across different ALS surveys. Our study demonstrates that DART is a robust model for simulating discrete-return measurements over structurally complex forests; however, the replication of foliage morphology, density and orientation are important considerations for radiative transfer simulations using synthetic trees with explicitly defined crown architectures.

Forest inventories for small areas using drone imagery without in-situ field measurements

Drone applications are becoming increasingly common in the arena of forest management and forest inventories. In particular, the use of photogrammetrically derived drone-based image point clouds (DIPC) in individual tree detection has become popular. Use of an area-based approach (ABA) in small areas has also been considered. However, in-situ field measurements of sample plots substantially increase the cost of small area forest inventories. Therefore, we examined whether small-scale forest management inventories could be carried out without local field measurements. We used nationwide and regional ABA models for stem volumes fitted with airborne laser scanning (ALS) data to predict stem volumes using corresponding metrics calculated from DIPC data. The stem volumes were predicted at the cell level (15 × 15 m) and aggregated to test plots (30 × 30 m). Height metrics for the dominant tree layer from the DIPC data showed strong correlations with similar metrics computed from the ALS data. The ALS-based models applied with DIPC metrics performed well, especially if the ABA model was fitted in the same geographical area (regional model) and the inventory units were disaggregated to coniferous and deciduous dominated stands using auxiliary information from Multi-source National Forest Inventory data (root mean square error at 30 × 30 m level was 13.1%). The corresponding root mean square error associated with the nationwide ABA model was 20.0% with an overestimation (mean difference 9.6%).

Predicting and mapping site index in operational forest inventories using bitemporal airborne laser scanner data

Forest productivity reflects the wood production capacity of a given site and provides crucial information for forest management planning. The most widely accepted measure of forest productivity is site index (SI), defined as the average height of dominant trees at a given index age. In forest management inventories, SI is commonly interpreted manually from aerial images. While the use of airborne laser scanner (ALS) data has revolutionized operational practices for estimating many forest attributes relevant to forest management planning, practices for determining SI remain unchanged. The main objective of this study was to demonstrate a practical method for predicting and mapping SI in repeated ALS-based forest inventories. We used data acquired as part of three operational large-scale forest inventories in southeastern Norway. First, we identified areas in which forest growth had remained undisturbed since the initial inventory. We then regressed field predictions of SI against bitemporal ALS canopy metrics and used the regression models to predict SI for forest areas classified as undisturbed. The result was SI maps constructed with a spatial resolution of 15.81 m. User accuracies of class predictions of undisturbed forest in the three districts were 92%, 95% and 89%. Plot-level validation revealed root mean squared errors of SI predictions ranging from 1.72 to 2.84 m for Norway spruce, and 1.35 to 1.73 m for Scots pine. The method presented here can be used to map SI over large areas of forest automatically, depicting forest productivity at a much finer spatial resolution than what is common in operational inventories.

An allometric area-based approach—a cost-effective method for stand volume estimation based on ALS and NFI data

Abstract Reliable information concerning stand volume is fundamental to making strategic decisions in sustainable forest management. A variety of remotely sensed data and different inventory methods have been used for the estimation of forest biometric parameters. Particularly, airborne laser scanning (ALS) point clouds are widely used for the estimation of stand volume and forest biomass using an area-based approach (ABA) framework. This method relies on the reference measurements of field plots with the necessary prerequisite of a precise co-registration between ground reference plots and the corresponding ALS samples. In this research, the allometric area-based approach (AABA) is proposed in the context of stand volume estimation of Scots pine (Pinus sylvestris L.) stands. The proposed method does not require detailed information about the coordinates of the field plots. We applied Polish National Forest Inventory data from 9400 circular field plots (400 m2) to develop a plot level stand volume allometric model using two independent variables: top height (TH) and relative spacing index (RSI). The model was developed using the multiple linear regression method with a log–log transformation of variables. The hypothesis was that, the field measurements of TH and RSI could be replaced with corresponding ALS-derived metrics. It was assumed that TH could be represented by the maximum height of the ALS point cloud, while RSI can be calculated based on the number of tree crowns delineated within the ALS-derived canopy height model. Performance of the developed AABA model was compared with the semi-empirical ABASE (with two predictors: TH and RSI) and empirical ABAE (several point cloud metrics as predictors). The models were validated at the plot level using 315 forest management inventory plots (400 m2) and at the stand level using the complete field measurements from 42 Scots pine dominated forest stands in the Milicz forest district (Poland). The AABA model showed a comparable accuracy to the traditional ABA models with relatively high accuracy at the plot (relative root mean square error (RMSE) = 22.8 per cent; R2 = 0.63) and stand levels (RMSE = 17.8 per cent, R2 = 0.65). The proposed novel approach reduces time- and cost-consuming field work required for the classic ABA method, without a significant reduction in the accuracy of stand volume estimations. The AABA is potentially applicable in the context of forest management inventory without the necessity for field measurements at local scale. The transportability of the approach to other species and more complex stands needs to be explored in future studies.

Study on Stand Volume Models for Sub-compartment Forest Management Inventory in Beijing

Study on Stand Volume Models for Sub-compartment Forest Management Inventory in Beijing

FOREST COVER DIGITAL MAPPING OF THE KARELIAN PART OF THE WHITE SEA COASTAL ZONE BASED ON IMPROVED INTERPRETATION METHOD OF REMOTE SENSING DATA

A modified interpretation method based on a combination of unsupervised and supervised image classification has been applied to space medium-resolution images taken in wintertime (data of OLI device of the LandSat 8 satellite) to create a digital thematic map of coniferous vegetation (for the Karelian part of the White Sea coastal zone). The template for the classification was the digital forest management inventory database (DB) for part (7.6 %) of the study area. Since the forest management inventory DB does not cover the entire study area, it is particularly interesting to determine the feasibility of producing digital vector layers of coniferous stands through supervised classification of medium-resolution remotely sensed data based on small amounts of template inventory information. To create training sets/supervised classification signatures, bitmap layers were formed from a color RGB composite of the source multi-spectral satellite image for sets of inventory units of each coniferous species. Unsupervised classification by the K-means method was performed for each prevalent species with a division into 5/8/10 clusters. Analysis of the findings revealed that the optimal number of clusters corresponds to 5 groups. Weighted average inventory parameters of template units were calculated to identify correlations with training sets. As a result of refining the technique for DB data classification, a set of digital thematic vector layers in GIS format, each containing coniferous stands as polygonal objects reliably identified by the main species and the growing stock, was produced. The set of digital layers of coniferous stands created using medium-resolution remotely sensed data can be used for the purposes of environmental monitoring and forecasting of human impact on the natural environment in the north-eastern part of the Republic of Karelia.

Forest information at multiple scales: development, evaluation and application of the Norwegian forest resources map SR16

ABSTRACTThis paper describes the development and utility of the Norwegian forest resources map (SR16). SR16 is developed using photogrammetric point cloud data with ground plots from the Norwegian National Forest Inventory (NFI). First, an existing forest mask was updated with object-based image analysis methods. Evaluation against the NFI forest definitions showed Cohen's kappa of 0.80 and accuracy of 0.91 in the lowlands and a kappa of 0.73 and an accuracy of 0.96 in the mountains. Within the updated forest mask, a 16×16 m raster map was developed with Lorey's height, volume, biomass, and tree species as attributes (SR16-raster). All attributes were predicted with generalized linear models that explained about 70% of the observed variation and had relative RMSEs of about 50%. SR16-raster was segmented into stand-like polygons that are relatively homogenous in respect to tree species, volume, site index, and Lorey's height (SR16-vector). When SR16 was utilized in a combination with the NFI plots and a model-assisted estimator, the precision was on average 2–3 times higher than estimates based on field data only. In conclusion, SR16 is useful for improved estimates from the Norwegian NFI at various scales. The mapped products may be useful as additional information in Forest Management Inventories.

Individual Tree Detection and Crown Delineation with 3D Information from Multi-view Satellite Images

Individual tree detection and crown delineation (<small>ITDD</small>) are critical in forest inventory management and remote sensing based forest surveys are largely carried out through satellite images. However, most of these surveys only use 2D spectral information which normally has not enough clues for <small>ITDD</small>. To fully explore the satellite images, we propose a <small>ITDD</small> method using the orthophoto and digital surface model (<small>DSM</small>) derived from the multi-view satellite data. Our algorithm utilizes the top-hat morphological operation to efficiently extract the local maxima from <small>DSM</small> as treetops, and then feed them to a modified superpixel segmentation that combines both 2D and 3D information for tree crown delineation. In subsequent steps, our method incorporates the biological characteristics of the crowns through plant allometric equation to falsify potential outliers. Experiments against manually marked tree plots on three representative regions have demonstrated promising results – the best overall detection accuracy can be 89%.

A STUDY OF USING FULLY CONVOLUTIONAL NETWORK FOR TREETOP DETECTION ON REMOTE SENSING DATA

Abstract. Individual tree detection and counting are critical for the forest inventory management. In almost all of these methods that based on remote sensing data, the treetop detection is the most important and essential part. However, due to the diversities of the tree attributes, such as crown size and branch distribution, it is hard to find a universal treetop detector and most of the current detectors need to be carefully designed based on the heuristic or prior knowledge. Hence, to find an efficient and versatile detector, we apply deep neural network to extract and learn the high-level semantic treetop features. In contrast to using manually labelled training data, we innovatively train the network with the pseudo ones that come from the result of the conventional non-supervised treetop detectors which may be not robust in different scenarios. In this study, we use multi-view high-resolution satellite imagery derived DSM (Digital Surface Model) and multispectral orthophoto as data and apply the top-hat by reconstruction (THR) operation to find treetops as the pseudo labels. The FCN (fully convolutional network) is adopted as a pixel-level classification network to segment the input image into treetops and non-treetops pixels. Our experiments show that the FCN based treetop detector is able to achieve a detection accuracy of 99.7 % at the prairie area and 66.3 % at the complicated town area which shows better performance than THR in the various scenarios. This study demonstrates that without manual labels, the FCN treetop detector can be trained by the pseudo labels that generated using the non-supervised detector and achieve better and robust results in different scenarios.

Direct and indirect site index determination for Norway spruce and Scots pine using bitemporal airborne laser scanner data

Forest site productivity, usually represented by site index (SI), is a fundamental resource variable in forest management planning as it is a quantitative measure of the production capacity of forest land. Site index is usually derived from estimates of dominant height (Hdom) at a given reference age using empirical age-height curves. However, it is commonly quantified with large uncertainty in forest management inventories, resulting in economic losses due to incorrect management decisions. In this study, we used bitemporal airborne laser scanner (ALS) data acquired for a study area in southeastern Norway with a time interval of 15 years to estimate SI by means of an area-based approach. We present two practical methods for SI determination, i.e., the (1) direct and (2) indirect method. With the direct method, we regressed field observations of age-height SI against canopy height metrics derived from ALS data from the first point in time and changes in ALS metrics reflecting canopy height growth during the observation period. With the indirect method, we first modelled Hdom for the two points in time using the respective ALS metrics as predictors. We then derived SI from the initial Hdom, the estimated Hdom increment, and the length of the observation period using empirical SI curves. We used bitemporal field data collected from 80 georeferenced sample plots of size 232.9 m2 to fit the species-specific regression models for SI and Hdom. We then applied the models to an independent dataset comprising 42 georeferenced validation plots of size ∼3700 m2, for which ground reference values were collected at both points in time, to assess the precision of both methods. Both the proposed methods produced SI estimates with satisfactory precision. For the direct method, the independent validation revealed root mean squared errors (RMSE) of 1.78 and 1.08 m for Norway spruce and Scots pine, respectively, compared to 1.82 m obtained for both tree species using the indirect method. The indirect method can provide a good alternative to the direct method as field observations of SI are not required to calibrate the regression models.

Incorporating tree- and stand-level information on crown base height into multivariate forest management inventories based on airborne laser scanning

Incorporating tree- and stand-level information on crown base height into multivariate forest management inventories based on airborne laser scanning

Analysis of spatial correlation in predictive models of forest variables that use LiDAR auxiliary information

Accounting for spatial correlation of LiDAR model errors can improve the precision of model-based estimators. To estimate spatial correlation, sample designs that provide close observations are needed, but their implementation might be prohibitively expensive. To quantify the gains obtained by accounting for the spatial correlation of model errors, we examined (i) the spatial correlation patterns of residuals from LiDAR linear models developed to predict volume, total and stem biomass per hectare, quadratic mean diameter (QMD), basal area, mean and dominant height, and stand density and (ii) the impact of field plot size on the spatial correlation patterns in a standwise managed Mediterranean forest in central Spain. For all variables, the correlation range of model residuals consistently increased with plot radius and was always below 60 m except for stand density, where it reached 85 m. Except for QMD, correlation ranges of model residuals were between 1.06 and 8.16 times shorter than those observed for the raw variables. Based on the relatively short correlation ranges observed when the LiDAR metrics were used as predictors, the assumption of independent errors in many forest management inventories seems to be reasonable and appropriate in practice.

Imputing plant community classifications for forest inventory plots

Native plant community (NPC) classifications typically require on-site visits and in-depth observations by trained ecologists. The goal is to identify unique floristic and environmental characteristics indicative of a particular plant community, ecosystem, or demographic condition. Such data are often desired to inform management decisions on sustainable timber and ecosystem services production over local to large landscapes. Yet, the time and funding needed to identify, assess, catalogue, and map these communities is often limited. Lacking these classifications, we rely on imprecise determinations of the prevalence of various NPCs. Further, extrapolating statewide NPC extent from previously imputed classifications for state managed stands is difficult without a representative sampling design including all ownerships. As a solution to the NPC sample coverage limitation, we describe an extension of a previously reported imputation model to provide the desired statewide classifications and corresponding estimates of the ecological landscape state indicator provided by NPC extent.First, NPC observations from the Minnesota Department of Natural Resources (MNDNR) Division of Ecological and Water Resources for 1964–2015 were linked with MNDNR Forest Inventory Management (FIM) stand data to provide a set of observed polygons for training the imputation model. Then, USDA Forest Service Forest Inventory and Analysis (FIA) plot data, were associated with the observed stands to provide NPC classifications for a subset of plots (e.g., training plots) contained in the FIA database for Minnesota. NPC information was then linked to forest inventory and physiographic layers via spatial association techniques in a geographic information system. Soils data describing drainage, productivity, thickness of the rooting zone, and land position were also used. Finally, validation of resulting imputed classifications shows that application of the model to the statewide FIA inventory will result in an error rate between 8% and 30% with a mean of 83% of imputations correct at the class level.We then updated the publicly accessible FIA database for Minnesota with imputed NPC classifications and scripted labeling schemes integrated with the EVALIDator report building tool to produce estimates of forestland extent. Here, we focus on estimates of NPC class by FIA Survey Unit and inventory year. Finally, quantified estimates of landscape state (e.g., NPC extent and condition) are enabled for inventories ending between 1977 and 2014. Imputed data from this series of statewide inventories enables the analysis of landscape change, and facilitates strategic planning to move the bioregional landscape in a desired ecological direction, or to provide specific ecosystem services.

Distance-independent tree basal area growth models for Norway spruce, Douglas-fir and Japanese larch in Southern Belgium

This paper presents new harmonized distance-independent individual tree basal area growth models for Norway spruce, Douglas-fir and Japanese larch in pure even-aged stands in Southern Belgium. The selected model was originally developed for Norway spruce and Douglas-fir in neighboring France. New formulations are proposed for some of the model components in order to lower the number of fitted parameters and facilitate the fitting procedure. The resulting models integrate the most recent corresponding top-height growth models and use four simple and usually collected explanatory variables: stand age, top-height, total basal area and tree girth at breast height. The modified formulations maintain similar fitting performances and make it easier to interpret the influence of the explanatory variables on tree growth. Parameters estimates were fitted on thousands of growth measurements gathered from several monitoring plots, forest management inventories and silvicultural field experiments that represent the wide range of site conditions and of forest management scenarios applied to coniferous stands in Southern Belgium. Cross-validation of the models revealed no bias and highlighted their consistent behavior over the entire range of girth at breast height, age, top-height, site index and density represented in our dataset. Combining utility and robust performances, these models represent useful forest management tools, purposely ideal for forest simulation software development. Moreover, the flexibility and generic capabilities of the model formulation should make it easily adjustable for other species in even-aged stands.

Maximizing the Diversity of Ensemble Random Forests for Tree Genera Classification Using High Density LiDAR Data

Recent research into improving the effectiveness of forest inventory management using airborne LiDAR data has focused on developing advanced theories in data analytics. Furthermore, supervised learning as a predictive model for classifying tree genera (and species, where possible) has been gaining popularity in order to minimize this labor-intensive task. However, bottlenecks remain that hinder the immediate adoption of supervised learning methods. With supervised classification, training samples are required for learning the parameters that govern the performance of a classifier, yet the selection of training data is often subjective and the quality of such samples is critically important. For LiDAR scanning in forest environments, the quantification of data quality is somewhat abstract, normally referring to some metric related to the completeness of individual tree crowns; however, this is not an issue that has received much attention in the literature. Intuitively the choice of training samples having varying quality will affect classification accuracy. In this paper a Diversity Index (DI) is proposed that characterizes the diversity of data quality (Qi) among selected training samples required for constructing a classification model of tree genera. The training sample is diversified in terms of data quality as opposed to the number of samples per class. The diversified training sample allows the classifier to better learn the positive and negative instances and; therefore; has a higher classification accuracy in discriminating the “unknown” class samples from the “known” samples. Our algorithm is implemented within the Random Forests base classifiers with six derived geometric features from LiDAR data. The training sample contains three tree genera (pine; poplar; and maple) and the validation samples contains four labels (pine; poplar; maple; and “unknown”). Classification accuracy improved from 72.8%; when training samples were selected randomly (with stratified sample size); to 93.8%; when samples were selected with additional criteria; and from 88.4% to 93.8% when an ensemble method was used.