Small area estimation of growing stock timber volume, basal area, mean stem diameter, and stem density for mountain forests in Austria

Arne Nothdurft, Valentin Sarkleti, Tobias Ofner-Graff, Andreas Tockner, Christoph Gollob, Tim Ritter, Ralf Kraßnitzer, Philip Svazek, Martin Kühmaier, Karl Stampfer, Andrew O. Finley · 2025

Read Small area estimation of growing stock timber volume, basal area, mean stem diameter, and stem density for mountain forests in Austria by Arne Nothdurft, Valentin Sarkleti, Tobias Ofner-Graff, Andreas Tockner, Christoph Gollob, Tim Ritter, Ralf Kraßnitzer, Philip Svazek, Martin Kühmaier, Karl Stampfer, Andrew O. Finley (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-05043.txt.

Introduction

Arne Nothdurft and ten co-authors estimated timber measures for Austrian mountain forest stands in a 2025 paper. Managers need stand-level numbers for volume, basal area, tree size, and density across 824 stands in the Brixen im Thale district. Field crews measured 146 plots with a handheld laser scanner.

Airborne laser data supplied the predictors, and a small-area framework turned plot models into stand estimates with uncertainty. The authors compared single- and multi-output models, with and without spatial effects, using blocked cross-checks. Despite hints that complex models should win, simpler spatial models, and sometimes plain ones, performed about as well. The paper gives forest planners a tested recipe and honest error bars.

Catalog record

CREATORArne Nothdurft, Valentin Sarkleti, Tobias Ofner-Graff, Andreas Tockner, Christoph Gollob, Tim Ritter, Ralf Kraßnitzer, Philip Svazek, Martin Kühmaier, Karl Stampfer, Andrew O. Finley
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE66 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-05043

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Small area estimation of growing stock timber volume, basal area, mean stem diameter, and stem density for mountain forests in Austria Arne Nothdurft, Valentin Sarkleti, Tobias Ofner-Graff, Andreas Tockner, Christoph Gollob, Tim Ritter, Ralf Kraßnitzer, Philip Svazek, Martin Kühmaier, Karl Stampfer, Andrew O. Finley Abstract Regression models were evaluated to estimate stand-level growing stock volume (GSV), quadratic mean diameter (QMD), basal area (BA), and stem density (N) in the Brixen im Thale forest district of Austria. Field measurements for GSV, QMD, and BA were collected on 146 inventory plots using a handheld mobile personal laser scanning system. Predictor variables were derived from airborne laser scanning (ALS)–derived normalized digital surface and terrain models. The objective was to generate stand-level estimates and associated uncertainty for GSV, QMD, BA, and N across 824 stands. A unit-level small area estimation framework was used to generate stand-level posterior predictive distributions by aggregating predictions from finer spatial scales. Both univariate and multivariate models, with and without spatially varying intercepts, were considered. Predictive performance was assessed via spatially blocked cross-validation, focusing on bias, accuracy, and precision. Despite exploratory analysis suggesting advantages of complex multivariate spatial models, simpler univariate spatial—and in some cases, non-spatial—models exhibited comparable predictive performance. Keywords: Bayesian inference, model-based, forest inventory, airborne laser scanning 1 Introduction Sustainable forest management requires precise information on available timber resources. Within forest management planning, this information is used to inform the decision-making process about silvicultural and harvesting activities for individual forest stands. Here, in accordance with Oliver et al., 1996 and Ashton and Kelty, 2018 , a stand is defined as the basic unit of management and identified as a spatially continuous extent in which trees are approximately the same age and structure characteristics are homogeneous. Forest operations that aim to meet desired goals are usually based on criteria associated with stand species composition and structure. For example, mean stand diameter at breast height (DBH) characterizes maturity ( Skovsgaard and Vanclay, 2007 ) and is used to differentiate between pre-commercial thinning and final harvesting operations ( Kojola et al., 2005 ) . Measures of DBH central tendency, such as mean quadratic DBH (QMD), might also determine technical restrictions for mechanized timber harvesting systems and can serve as a proxy for timber harvesting cost and productivity projections ( Eliasson, 1999 ; Louis et al., 2022 ) . Basal area (BA) and number of stems per unit area, i.e., stand density, (N) are used to characterize inter-tree competition ( Sterba, 1981 ; Sterba, 1985 ; Sterba, 1987 ) and silvicultural prescription harvest removals and residuals. Stand density also influences the efficiency of forest operations. Observed BA is often checked against reference values to estimate whether yield losses are expected due to under- or over-stocking ( Assmann, 1950 ; Assmann, 1954 ) . Finally, the total growing stock timber volume (GSV) is used for timber supply and above-ground carbon estimates ( Matala et al., 2009 ) . In summary, stand structure characteristics captured using GSV, QMD, BA, and N are key for defining silvicultural prescriptions, harvesting planning and implementation. Picchio et al., 2020 and Venanzi et al., 2023 show that including more precise data in planning and executing forest operations reduces costs, increases productivity, mitigates soil disturbance, promotes healthier forest ecosystems, and enhances safety and working conditions for forest workers. Our aim is to generate precise estimates for GSV, QMD, BA, and N at unobserved locations within stands, and summarize these predictions to arrive at stand-level parameter estimates appropriate for supporting silvicultural prescriptions, harvest planning and implementation. Given underlying tree- and stand-level allometric relationships, as well as common growing conditions and disturbance history, we expect these stand parameters to co-vary and exhibit spatial dependence. Chosen estimators should maintain biologically meaningful relationships among outcomes such that they are reflected in subsequent estimates. Planning and management occur at the stand-level; hence, they need to be informed by stand-level structure parameter estimates. Given cost constraints, installing sufficiently dense forest inventory plot networks to achieve the desired level of estimate precision using traditional design-derived estimators is typically not feasible. As a result, various approaches have been developed to estimate parameters in stands with insufficient numbers of observations. Such approaches commonly couple sparsely sampled forest inventory and remote sensing data within a non-parametric or parametric modeling framework. The efficiency of non-parametric nearest neighbor methods ( Moeur and Stage, 1995 ; Tomppo and Halme, 2004 ; LeMay and Temesgen, 2005 ; Maltamo et al., 2006 , e.g.,) , various interpretations of small area estimators ( Goerndt et al., 2011 ; Goerndt et al., 2012 ; Breidenbach and Astrup, 2012 ; Mandallaz et al., 2013 ; Steinmann et al., 2013 ; Mandallaz, 2014 ; Green et al., 2019 ; Katila and Heikkinen, 2020 ; Coulston et al., 2021 ; Green et al., 2022 , e.g.,) , and geostatistical methods ( Mandallaz, 2000 ; Tuominen et al., 2003 ; Zhang et al., 2004 ; Berterretche et al., 2005 ; Buddenbaum et al., 2005 ; Maselli and Chiesi, 2006 ; Babcock et al., 2018 ; Finley et al., 2024 , e.g.,) are explored in many studies and remain an active area of research. Within our current setting, an advantage of nearest neighbor (NN) methods is that simultaneous imputation of outcome variables/values preserves their covariance; however, a limitation of these methods is there are no exact variance estimators and therefore approximations must be used ( McRoberts et al., 2007 ; McRoberts et al., 2011 ; McRoberts et al., 2022 ; Magnussen, 2013 ) . Small area estimators that build on stand-level direct estimates (i.e., estimates from design-based methods) are prevalent in the literature; however, they require a sufficient number of observations within each stand to generate direct estimates. Given forest stands are typically only a few hectares in size, there are often too few (or no) observations to achieve the desired level of direct estimate accuracy and precision. Other approaches couple direct with indirect estimators via regression models or NN methods to form composite estimators ( Dettmann et al., 2022 ) . While composite estimators might inherit some attractive theoretical properties of the direct estimator used, they too often require prohibitively large samples sizes to deliver reliable estimates, and specification of the model component can be restrictive, especially when extended to multivariate outcome settings. Model-based gostatistical methods offer flexible possibilities to combine spatially independent effects and spatially dependent trends in a coherent model that can be used to predict outcome variables at unobserved locations ( Cressie, 1993 ; Schabenberger and Gotway, 2004 ) . Further flexibility is achieved when classical geostatistical methods are cast within a Bayesian hierarchical model ( Banerjee et al., 2014 ) . Under the Bayesian paradigm, asymptotic distribution assumptions about the parameters are relaxed, and model parameter and predictive inference proceed from posterior and posterior predictive distributions, respectively. Several studies assess univariate and multivariate gostatistical Bayesian hierarchical models for forest inventory applications. For example, Finley et al., 2013 and Babcock et al., 2013 compare such models for forest parameter point estimates and mapping. Their work builds on Finley et al., 2008 who detailed and applied a Bayesian multivariate hierarchical model for predicting forest biomass per unit area of multiple tree species in a 1,053 ha forest district in

Read the complete text