Category: Research article
article id 7721, category Research article
Hyperspectral UAV-imagery and photogrammetric canopy height model in estimating forest stand variables. Silva Fennica vol. 51 no. 5 article id 7721. https://doi.org/10.14214/sf.7721
Highlights: Hyperspectral imagery and photogrammetric 3D point cloud based on RGB imagery were acquired under weather conditions changing from cloudy to sunny; Calibration of hyperspectral imagery was required for compensating the effect of varying weather conditions; The combination of hyperspectral imagery and photogrammetric point cloud data resulted in accurate forest estimates, especially for volumes per tree species.
Remote sensing using unmanned aerial vehicle (UAV) -borne sensors is currently a highly interesting approach for the estimation of forest characteristics. 3D remote sensing data from airborne laser scanning or digital stereo photogrammetry enable highly accurate estimation of forest variables related to the volume of growing stock and dimension of the trees, whereas recognition of tree species dominance and proportion of different tree species has been a major complication in remote sensing-based estimation of stand variables. In this study the use of UAV-borne hyperspectral imagery was examined in combination with a high-resolution photogrammetric canopy height model in estimating forest variables of 298 sample plots. Data were captured from eleven separate test sites under weather conditions varying from sunny to cloudy and partially cloudy. Both calibrated hyperspectral reflectance images and uncalibrated imagery were tested in combination with a canopy height model based on RGB camera imagery using the k-nearest neighbour estimation method. The results indicate that this data combination allows accurate estimation of stand volume, mean height and diameter: the best relative RMSE values for those variables were 22.7%, 7.4% and 14.7%, respectively. In estimating volume and dimension-related variables, the use of a calibrated image mosaic did not bring significant improvement in the results. In estimating the volumes of individual tree species, the use of calibrated hyperspectral imagery generally brought marked improvement in the estimation accuracy; the best relative RMSE values for the volumes for pine, spruce, larch and broadleaved trees were 34.5%, 57.2%, 45.7% and 42.0%, respectively.
article id 1567, category Research article
Nationwide airborne laser scanning based models for volume, biomass and dominant height in Finland. Silva Fennica vol. 50 no. 4 article id 1567. https://doi.org/10.14214/sf.1567
Highlights: Pooled data from nine inventory projects in Finland were used to create nationwide laser-based regression models for dominant height, volume and biomass; Volume and biomass models provided regionally different means than real means, but for dominant height the mean difference was small; The accuracy of general volume predictions was nevertheless comparable to relascope-based field inventory by compartments.
The aim of this study was to examine how well stem volume, above-ground biomass and dominant height can be predicted using nationwide airborne laser scanning (ALS) based regression models. The study material consisted of nine practical ALS inventory projects taken from different parts of Finland. We used field sample plots and airborne laser scanning data to create nationwide and regional models for each response variable. The final models had one or two ALS predictors, which were chosen based on the root mean square error (RMSE), and cross-validated. Finally, we tested how much predictions would improve if the nationwide models were calibrated with a small number of regional sample plots. Although forest structures differ among different parts of Finland, the nationwide volume and biomass models performed quite well (leave-inventory-area-out RMSE 22.3% to 33.8%, mean difference [MD] –13.8% to 18.7%) compared with regional models (leave-plot-out RMSE 20.2% to 26.8%). However, the nationwide dominant height model (RMSE 5.4% to 7.7%, MD –2.0% to 2.8%, with the exception of the Tornio region – RMSE 11.4%, MD –9.1%) performed nearly as well as the regional models (RMSE 5.2% to 6.7%). The results show that the nationwide volume and biomass models provided different means than real means at regional level, because forest structure and ALS device have a considerable effect on the predictions. Large MDs appeared especially in northern Finland. Local calibration decreased the MD and RMSE of volume and biomass models. However, the nationwide dominant height model did not benefit much from calibration.
article id 1087, category Research article
Tree species identification in aerial image data using directional reflectance signatures. Silva Fennica vol. 48 no. 3 article id 1087. https://doi.org/10.14214/sf.1087
Highlights: Multispectral reflectance data showed a strong and spectrally correlated tree effect; There was no gain in species classification from using species-specific differences of directional reflectance in real data and only a marginal improvement in simulated data; The directional signatures extracted in multiple images are obscured by the intrinsic within-species variation, correlated observations and inherent reflectance calibration errors.
Tree species identification using optical remote sensing is challenging. Modern digital photogrammetric cameras enable radiometrically quantitative remote sensing and the estimation of reflectance images, in which the observations depend largely on the reflectance properties of targets. Previous research has shown that there are species-specific differences in how the brightness observed changes when the viewing direction in an aerial image is altered. We investigated if accounting for such directional signatures enhances species classification, using atmospherically corrected, real and simulated multispectral Leica ADS40 line-camera data. Canopy in direct and diffuse illumination were differentiated and species-specific variance-covariance structures were analyzed in real reflectance data, using mixed-effects modeling. Species classification simulations aimed at elucidating the level of accuracy that can be achieved by using images of different quality, number and view-illumination geometry. In real data, a substantial variance component was explained by tree effect, which demonstrates that observations from a tree correlate between observation geometries as well as spectrally. Near-infrared band showed the strongest tree effect, while the directionality was weak in that band. The gain from directional signatures was insignificant in real data, while simulations showed a potential gain of 1–3 percentage points in species classification accuracy. The quality of reflectance calibration was found to be important as well as the image acquisition geometry. We conclude that increasing the number of image observations cancels out random observation noise and reflectance calibration errors, but fails to eliminate the tree effect and systematic calibration inaccuracy. Directional reflectance constitutes a marginal improvement in tree species classification.
article id 204, category Research article
Using the process-based stand model ANAFORE including Bayesian optimisation to predict wood quality and quantity and their uncertainty in Slovenian beech. Silva Fennica vol. 43 no. 3 article id 204. https://doi.org/10.14214/sf.204
The purpose of this study was to expand an existing semi-mechanistic forest model, ANAFORE (ANAlysing Forest Ecosystems), to allow for the prediction of log quality and the accompanying uncertainty as influenced by climate and management. The forest stand is described as consisting of trees of different cohorts, either of the same or of different species (deciduous or coniferous). In addition to photosynthesis, transpiration, total growth and yield, the model simulates the daily evolution in vessel biomass and radius, parenchyma and branch development. From these data early and latewood biomass, wood tissue composition, knot formation and density are calculated. The new version presented here, includes the description of log quality, including red heart formation of beeches. A Bayesian optimisation routine for the species parameters was added to the stand model. From a given range of input parameters (prior), the model calculates an optimised range for the parameters (posterior) based on given output data, as well as an uncertainty on the predicted values. A case study was performed for Slovenian beech forests to illustrate the main model functioning and more in particular the simulation of the wood quality. The results indicate that the ANAFORE model is a useful tool for analyzing wood quality development and forest ecosystem functioning in response to management, climate and stand characteristics. However, the Bayesian optimization showed that the remaining uncertainty on the input parameters for the chosen stand was very large, due to the large number of input parameters in comparison to the limited stand data.
article id 334, category Research article
Linear prediction application for modelling the relationships between a large number of stand characteristics of Norway spruce stands. Silva Fennica vol. 40 no. 3 article id 334. https://doi.org/10.14214/sf.334
The aim was to produce models for a large number of stand characteristics of Norway spruce dominated stands. A total of 227 national forest inventory based permanent stand plots, dominated by Norway spruce (Picea abies), were used in modelling eight stand variables as a function of the stand mean biological age and site characteristics. The basic models were able to characterize the average development of the modelled stand variables, but resulted in a relatively high RMSE. Basal area (G) and stem number (N) were the most inaccurate, having a RMSE of 34–41%, while that of mean diameter and height characteristics varied between 16–20%. The expectations and error variances of the basic models were calibrated with known stand variables using linear prediction theory. The best linear unbiased predictor (BLUP) with a single stand variable used for calibration proved to be ineffective for unknown G and N, but relatively effective for the unknown mean characteristics. However, calibration with one sum and one mean characteristic proved to be effective, and additional calibration variables enhanced the precision only marginally. The BLUP method provided a flexible approach when characterizing the relationships between a large number of stand variables, thus enabling multiple use of these models because they were not fixed to a specific inventory system.
article id 332, category Research article
Calibrating predicted tree diameter distributions in Catalonia, Spain. Silva Fennica vol. 40 no. 3 article id 332. https://doi.org/10.14214/sf.332
Several probability density functions have been used in describing the diameter distributions of forest stands. In a case where both the stand basal area and number of stems per hectare are assessed, the fitted or predicted distribution is scaled using only one of these variables, with the result that the distribution often gives incorrect values for the other variable. Using a distribution that provides incorrect values for known characteristics means wasting information. Calibrating the distribution so that it is compatible with the additional information on stand characteristics is a way to avoid such wasting. This study examined the effect of calibration on the accuracy of the predicted diameter distributions of the main tree species of Catalonia. The distributions were calibrated with and without considering the prediction errors of the frequencies of diameter classes. When prediction errors were assumed, the calibration was done with and without making allowance for estimation errors in the stand level calibration variables. Calibrated distributions were more accurate than non-calibrated in terms of sums of different powers of diameters. The set of calibration variables that gave the most accurate results included six stand variables: number of trees per hectare, stand basal area, basal-area-weighted mean diameter, non-weighted mean diameter, median diameter, and basal area median diameter. Of the tested three-variable combinations the best was: number of trees per hectare, stand basal area, and basal-area-weighted mean diameter. Means were more useful calibration variables than medians.
article id 368, category Research article
Generating a raster map presentation of a forest resource by solving a transportation problem. Silva Fennica vol. 39 no. 4 article id 368. https://doi.org/10.14214/sf.368
Necessary tools for raster map generation, for the approach based on the calibration estimator, were developed and implemented. The allocation of the area weight of each pixel to sample plots was formulated as a transportation problem, using a spectral distance measure as a transportation cost, and solved using the transportation simplex algorithm. Pixel level accuracy was calculated for the methods based on the calibration estimator so that the results could be compared with the results of the nearest neighbour estimation, the reference sample plot method (RSP) at pixel level. Local averaging in a 3 x 3 window was performed for each generated raster map as a postprocessing phase to smooth the map. Test plot results were calculated both for the unfiltered raster map and the filtered raster map. RSP produced the smallest RMSE in the pooled test data. Local averaging with a 3 x 3 filter decreased the pixel level error – and the bias – and the differences between the methods are smaller. Without local averaging, the pixel level errors of the methods based on solving the transportation problem were high. Raster map generation using the methods of this study forms an optional part – followed possibly by the classification of the pixel level results – of the whole computation task, when the area weight computation is based on the calibration estimation. For larger areas than in the present study, such as municipalities, the efficiency of the method based on the transportation model must be improved before it is a usable tool, in practice, for raster map generation. For nearest neighbour methods, the area size is not such a problem, because the inventory area is processed pixel by pixel.
article id 394, category Research article
Multilevel linear mixed model for tree diameter increment in stone pine (Pinus pinea): a calibrating approach. Silva Fennica vol. 39 no. 1 article id 394. https://doi.org/10.14214/sf.394
Diameter increment for stone pine (Pinus pinea L.) is described using a multilevel linear mixed model, where stochastic variability is broken down among period, plot, tree and within-tree components. Covariates acting at tree and stand level, as breast height diameter, density, dominant height or site index are included in the model as fixed effects in order to explain residual random variability. The effect of competition on diameter increment is expressed by including distance independent competition indices. The entrance of regional effects within the model is tested to determine whether a single model is sufficient to explain stone pine diameter increment in Spain, or if, on the contrary, regional models are needed. Diameter increment model can be calibrated by predicting random components using data from past growth measurements taken in a complementary sample of trees. Calibration is carried out by using the best linear unbiased predictor (BLUP) theory. Both the fixed effects model and the calibrated model mean a substantial improvement when compared with the classical approach, widely used in forest management, of assuming constancy in diameter increment for a short projection period.
article id 522, category Research article
Anticipating the variance of predicted stand volume and timber assortments with respect to stand characteristics and field measurements. Silva Fennica vol. 36 no. 4 article id 522. https://doi.org/10.14214/sf.522
Several models and/or several variable combinations could be used to predict the diameter distribution of a stand. Typically, a fixed model and a fixed variable combination is used in all conditions. The calibration procedure, however, makes it possible to choose the measurement combination from among many possibilities, although the model used is fixed. In this study, the usefulness of utilizing additional stand characteristics for calibrating the predicted diameter distribution is examined. Nine measurement strategies were tested in predicting the total stand volume, sawlog volume and pulpwood volume. The observed errors of these variables under each strategy were modeled as a function of basal area, basal area median diameter and number of stems. The models were estimated in three steps. First, an Ordinary Least Squares (OLS) model was fitted to the observed errors. Then, a variance function was estimated using the OLS residuals. Finally, a weighted Seemingly Unrelated Regression (SUR) analysis was used to model the observed errors, using the estimated variance functions as weights. The estimated models can be used to anticipate the precision and accuracy of predicted volume characteristics for each stand with different variable combinations and, consequently, to choose the best measurement combination in different stands.
article id 620, category Research article
Performance of percentile based diameter distribution prediction and Weibull method in independent data sets. Silva Fennica vol. 34 no. 4 article id 620. https://doi.org/10.14214/sf.620
Diameter distribution is used in most forest management planning packages for predicting stand volume, timber volume and stand growth. The prediction of diameter distribution can be based on parametric distribution functions, distribution-free parametric prediction methods or purely non-parametric methods. In the first case, the distribution is obtained by predicting the parameters of some probability density function. In a distribution-free percentile method, the diameters at certain percentiles of the distribution are predicted with models. In non-parametric methods, the predicted distribution is a linear combination of similar measured stands. In this study, the percentile based diameter distribution is compared to the results obtained with the Weibull method in four independent data sets. In the case of Scots pine, the other methods are also compared to k-nearest neighbour method. The comparison was made with respect to the accuracy of predicted stand volume, saw timber volume and number of stems. The predicted percentile and Weibull distributions were calibrated using number of stems measured from the stand. The information of minimum and maximum diameters were also used, for re-scaling the percentile based distribution or for parameter recovery of Weibull parameters. The accuracy of the predicted stand characteristics were also compared for calibrated distributions. The most reliable results were obtained using the percentile method with the model set including number of stems as a predictor. Calibration improved the results in most cases. However, using the minimum and maximum diameters for parameter recovery proved to be inefficient.
article id 682, category Research article
The use of digitized aerial photographs and local operation for classification of stand development classes. Silva Fennica vol. 32 no. 3 article id 682. https://doi.org/10.14214/sf.682
The increasing capacity of modern computers has created the opportunity to routinely process the very large data sets derived by digitizing aerial photographs. The very fine resolution of such data sets makes them better suited than satellite imagery for some applications; however, there may be problems in implementation resulting from variation in radial distortion and illumination across an aerial photograph. We investigated the feasibility of using local operators (e.g., non-overlapping moving window means and standard deviations) as auxiliary data for generating stand development classes via three steps: (i) derive 6 local operators intended to represent texture for a 16 by 16 m window corresponding to a forest inventory sampling unit, (ii) apply a calibration process (e.g., accounting for location relative to a photo's principal point and solar position) to these local operators, and (iii) apply the calibrated local operators to classify the forest for stand development. Results indicate that calibrated local operators significantly improve the classification compared to what is possible using uncalibrated local operators and satellite images.
Category: Research note
article id 1114, category Research note
Measurement errors in the use of smartphones as low-cost forestry hypsometers. Silva Fennica vol. 48 no. 5 article id 1114. https://doi.org/10.14214/sf.1114
Highlights: We analysed two smartphones (HTC Desire and Samsung Galaxy Note) to determine the errors in the height measurements; The calibration included with the Android applications is insufficient; After appropriate calibration, the smartphone errors are similar to other forest hypsometers (Blume Leiss and Vertex).
Various applications currently available for Android allow the estimation of tree heights by using the 3D accelerometer on smartphones. Some make the estimation using the image on the screen, while in others, by pointing with the edges of the terminal. The present study establishes the measurement errors obtained with HTC Desire and Samsung Galaxy Note compared to those from Blume Leiss and Vertex IV. Six series of 12 measurements each were made with each hypsometer (for heights of 6 m, 8 m, 10 m and 12 m). A Kruskall Wallis test is applied to the relative errors to determine whether there are significant differences between the devices. The results indicate that the errors of the uncalibrated smartphones significantly exceed those of traditional forestry apparatus. However, calibration is a very easy procedure that can be done by means of a linear regression line between real angles (obtained with a Digital Angle Finder or with a series of measurements taken independently of the experiment), and the angles of the accelerometer. With this adjustment, the smartphones achieve adequate quality levels although the bias was not totally eliminated. The relative errors when pointing with the edges of the terminal show no significant differences compared to Blume Leiss. Applications that use the screen image give better results (no significant differences were detected with Vertex). There is currently no application that offers calibration of the linear regression slope, which is an essential requirement for ensuring the accuracy of height measurements obtained with smartphones.
article id 310, category Research note
Comparison of approaches to integrate energy wood estimation into the Finnish compartment inventory system. Silva Fennica vol. 41 no. 1 article id 310. https://doi.org/10.14214/sf.310
The harvesting of energy wood from young stands is increasing as the demand for renewable wood fuel is growing. Energy wood consists of stems, tree tops, branches and needles, depending on the size of the trees and the logging method used. The current forest inventory and planning systems used in private forests in Finland do not produce estimates of energy wood components. In stands typical for energy wood harvesting, a large share of energy wood consists of trees smaller than the minimum size for pulpwood. In this study, energy wood was included into the calculation system of compartment inventory, and a procedure for simulating the thinning treatments in young stands was developed. The results for six inventory alternatives and prediction of energy wood were compared with the use of inventory material from 37 young stands that have plenty of energy wood. The measurement of additional stand characteristics and the use of a calibration estimation method was tested, as well as the use of plot-level inventory data instead of stand level data. The results showed that the measurement of the number of trees per hectare, in addition to stand basal area and mean diameter, improved the energy wood estimates. The additional minimum and maximum diameters improved the precision of the estimates, but did not affect bias. The removal estimates were more precise when plot-level data was used, rather than stand-level data. The removal estimates were higher with plot-level data. The results suggest that, in heterogeneous young stands, plot by plot prediction would give more accurate removal estimates than the calculation of a corresponding prediction at the stand-level.
article id 7524, category Article
Change detection and controlling forest information using multi-temporal Landsat TM imagery. Acta Forestalia Fennica no. 258 article id 7524. https://doi.org/10.14214/aff.7524
A method was developed for relative radiometric calibration of single multitemporal Landsat TM image, several multitemporal images covering each other, and several multitemporal images covering different geographical locations. The radiometrically calibrated different images were used for detecting rapid changes on forest stands. The nonparametric Kernel method was applied for change detection. The accuracy of the change detection was estimated by inspecting the image analysis results in field.
The change classification was applied for controlling the quality of the continuously updated forest stand information. The aim was to ensure that all the manmade changes and any forest damages were correctly updated including the attribute and stand delineation information. The image analysis results were compared with the registered treatments and the stand information base. The stands with discrepancies between these two information sources were recommended to be field inspected.