Access the full text.
Sign up today, get DeepDyve free for 14 days.
Loading next page...
/lp/springer-journals/an-analysis-of-stand-level-size-distributions-of-decay-affected-norway-tKQ70iI0Ih
- Publisher
- Springer Journals
- Copyright
- Copyright © The Author(s) 2023
- ISSN
- 1286-4560
- eISSN
- 1297-966X
- DOI
- 10.1186/s13595-022-01170-8
- Publisher site
- See Article on Publisher Site
Abstract
Key message We studied size distributions of decay-affected Norway spruce trees using cut-to-length harvester data. The harvester data comprised tree-level decay and decay severity recordings from 101 final felling stands, which enabled to analyze relationships between size distributions of all and decay-affected trees. Distribution matching technique was used to transfer the size distribution of all trees into the diameter at breast height (DBH) distribution of decay-affected trees. Context Stem decay of Norway spruce (Picea abies [L.] Karst.) results in large economic losses in timber production in the northern hemisphere. Forest management planning typically requires information on tree size distributions. However, size distributions of decay-affected trees generally remain unknown impeding decision-making in forest management planning. Aims Our aim was to analyze and model relationships between size distributions of all and decay-affected Norway spruce trees at the level of forest stands. Methods Cut-to-length harvester data of 93,456 trees were collected from 101 final felling stands in Norway. For each Norway spruce tree (94% of trees), the presence and severity of stem decay (incipient and advanced) were recorded. The stand-level size distributions (diameter at breast height, DBH; height, H) of all and decay-affected trees were described using the Weibull distribution. We proposed distribution matching (DM) models that transform either the DBH or H distribution of all trees into DBH distributions of decay-affected trees. We compared the predictive performance of DMs with a null-model that refers to a global Weibull distribution estimated based on DBHs of all harvested decay-affected trees. Results The harvester data showed that an average-sized decay-affected tree is larger and taller compared with an average-sized tree in a forest stand, while trees with advanced decay were generally shorter and thinner compared with trees having incipient decay. DBH distributions of decay-affected trees can be matched with smaller error index (EI) values using DBH (EI = 0.14) than H distributions (EI = 0.31). DM clearly outperformed the null model that resulted in an EI of 0.32. Handling editor: John Lhotka *Correspondence: Janne Räty janne.raty@luke.fi; Johannes Breidenbach johannes.breidenbach@nibio.no Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. Räty et al. Annals of Forest Science (2023) 80:2 Page 2 of 15 Conclusions The harvester data analysis showed a relationship between size distributions of all and decay-affected trees that can be explained by the spread biology of decay fungi and modeled using the DM technique. Keywords Root and butt rot, Heterobasidion spp., Armillaria spp., Cut-to-length harvester, Forest management and planning 1 Introduction created by mechanical logging during snowless and Norway spruce (Picea abies [L.] Karst.) dominates the frost-free periods are potential infection routes as well boreal forests in Northern Europe and the subalpine (Metslaid et al. 2018). The risk of infection transfer areas of the Alps and Carpathian Mountains. Owing to between neighboring trees via root contacts is higher in its good performance in different site conditions, it was pure species forest stands in comparison with admixed also planted outside its natural distribution on lower stands (Lindén and Vollbrecht 2002; Möykkynen and elevations in more temperate forests (Caudullo et al. Pukkala 2010). 2016). Wood of Norway spruce has a low extractive con- While Heterobasidion species have benefited from tent, which makes it more prone to decay compared to summertime loggings, the species of Armillaria typi- Scots pine (Pinus sylvestris [L.]), the other widely dis- cally infect weakened trees, such as drought-stressed tributed conifer in Eurasia. Fungi in the genera of Heter- Norway spruce trees. From infected trees, or stumps of obasidion and Armillaria, the so-called white-rot fungi infected trees, Armillaria species spread to neighbor- capable of decomposing all structural polymers in wood, ing trees with the aid of rhizomorphs that grow freely are the most important root and stem wood-decaying in the soil. A single individual of an Armillaria species fungi of Norway spruce. In Norway, a nationwide survey can occupy Norway spruce trees at a distance of tens of of stumps in clear-cut forests concluded that 26.8% of meters (Prospero et al. 2003). the Norway spruce trees had stem decay that was most Wood decay causes considerable economic losses often (71% of the decay-affected stumps) associated with in timber production because decay arises normally the Heterobasidion species (Huse et al. 1994). The situa - from the roots and impairs the quality of the most valu- tion is similar in the other Nordic countries, where 20% able part of tree stem. While decay caused by Armil- or more of the trees in managed Norway spruce forests laria normally reaches a height of 1–2 m in a stem, the show decay caused by Heterobasidion species by the time heartwood decay column caused by Heterobasidion spe- of final harvest (Bendz-Hellgren and Stenlid 1995). cies can reach a height of 10–12 m in stems of mature In particular, the Heterobasidion species have ben- Norway spruce trees at a late stage of infection. Regard- efited from the current forest management practices less of stem decay, decay-affected trees can remain alive that involve logging operations year-round. Primary even for several decades and do not necessarily show any infection of forest stands by these fungi takes place obvious external signs of decay. The decay-affected part through the colonization of fresh stumps cut during the of the stem does not fulfill the quality criteria required active sporulation of the fungus. After spore infection for sawlogs but is instead used as pulpwood or energy of a fresh stump, the fungal mycelia rapidly colonize wood, depending on the degree of decay. Within the the dead root system. Once the next-generation plants European Union, the annual losses attributed to Heter- establish root contact with the pathogen-colonized obasidion spp. in timber production were estimated to stump roots, the infection is transferred to the next be approximately 800 million € (Woodward et al. 1998). tree generation (Piri 1996; Stenlid and Redfern 1998). Besides the losses due to timber wood quality, decay neg- A similar transfer process takes place between infected atively affects tree growth and carbon sequestration due mature Norway spruce trees and advanced regeneration to investment of energy resources to tree defense instead (Piri and Korhonen 2001). Owing to active tree defense, of growth (Bendz-Hellgren and Stenlid 1995; Oliva et al. the spread of Heterobasidion mycelia is slower in living 2012). In addition, decay-affected trees are also prone to roots than in roots of stumps. In general, the roots of other forest damages such as stem breakage due to wind the next-generation saplings can become infected by or snow. Considering the spread biology of Heteroba- the fungus after around 10 years of growth (Piri 2003). sidion, there are numerous reasons to anticipate that the Stem colonization usually initiates only after stem amount of decay and the inflicted economic losses will heartwood has started to develop, which in Norway increase along with climate warming (Müller et al. 2014). spruce begins between ages of 25 and 40 (Korhonen The anticipated increase of dry summers along the pro - and Stenlid 1998). Besides the infection routes through gression of climate change is likely to increase the occur- root contacts of neighboring trees, wounds on roots rence of decay caused by species of Armillaria. R äty et al. Annals of Forest Science (2023) 80:2 Page 3 of 15 Since decay fungi are generally confined to the physi - the economic value of timber resources and the growth ologically inactive heartwood in Norway spruce, the potential of forests at the level of forest stands. presence of decay in a tree and decay frequency at stand While the spread mechanisms of decay fungi are rela- level cannot be reliably determined without destruc- tively well understood, stand-level information on the tive sampling, e.g., drilling (Vollbrecht and Agestam size distribution of decay-affected trees is rarely avail - 1995). Unless one has prior knowledge of decay history able to support decision-making in forest management at the stand level, i.e., documented records of decay in planning. Harvester data with tree-level decay record- the previous tree generation or upon selective harvests ings enable to observe size distributions of all and (Müller et al. 2018), it is difficult to take decay into decay-affected trees at the level of entire forest stands. account when assessing the economic value of timber Stand-level data of the presence and extension of decay volume and when considering alternative forest man- in each tree are rare, since such data are too laborious agement scenarios at the level of forest stands. to collect using traditional field measurements. If such National Forest Inventories (NFI) typically collect data are accessible, the data allow to study the relation- information on the occurrence of stem decay, and NFI ship between the size distribution of decay-affected trees data have been used to model the risk of decay in for- and that of all trees at a forest stand. To this end, we used ests (Thor et al. 2005; Mattila and Nuutinen 2007; the distribution matching (DM) technique (Gonzáles and Hylen and Granhus 2018). Some previous studies have Woods 2002) to model the relationship between size dis- proposed mechanistic models for simulation of the tributions of all and decay-affected trees. The objectives spread of decay at the level of forest stands (Pukkala of this study were as follows: et al. 2005; Honkaniemi et al. 2014). The applicability of such models as decision tools in practical forest man- • To compare DBH and height (H) distributions of agement planning would ideally require knowledge of all trees with corresponding distributions of decay- a large group of parameters, such as localization and affected trees at the level of harvested stands. Our frequency of decay in the prior tree generation. Cur- aim was to scrutinize whether the size distributions rently, the most effective way to collect tree-level decay of all and decay-affected trees differ at stand level. information for complete forest stands is to utilize data • To analyze and compare size distributions of trees collected by cut-to-length harvesters, but this approach with incipient and advanced stem decay. Our main requires that the operator of a harvester observes and goal was to find out to what extent the degree of records the presence of decay during harvesting opera- decay influences the shape and location of tree size tion. Räty et al. (2021) utilized such harvester data and distributions. showed that stand-level timber volume affected by • To propose DM models that transfer DBH or H dis- decay can be mapped by means of airborne laser scan- tributions of all trees into DBH distribution of decay- ning data and environmental variables at satisfactory affected trees. error levels, given that harvester data are available close to the target forests. The planning of stand-level forest management treat - 2 Materials and methods ments typically requires simulations that use a group of 2.1 Study area models to account for growth dynamics of forests. In The forest stands from which the harvester data were order to apply tree-level models in growth simulators, the collected are located between the latitudes 59° and 65° description of tree sizes in a forest stand is a prerequisite. in Norway. Because harvests are typically carried out in The distribution of tree sizes is also needed in the cal - forests used for commercial timber production, the for- culation of timber assortment volumes, which form the est stands considered here are dominated by the boreal basis of the economic value of a forest stand. The most coniferous tree species Norway spruce or Scots pine. common way to describe the size distribution of all trees Broadleaved species, mostly birch (Betula spp. [L.]), may is to utilize a theoretical diameter at breast height (DBH) occur as mixtures in these conifer-dominated forests. distribution model, such as a two-parameter Weibull distribution, which can be estimated for forest stands 2.2 Harvester data by means of relatively rapid field measurements (Siipile - Harvester data were collected using five different hto and Mehtätalo 2013) or by coupling a sample of field machines between 2019 and 2021. The harvest operations plots and remotely sensed information (Gobakken and were planned and operated by different companies, and Næsset 2004). Tree size information of decay-affected the harvested sites are heterogenous units consisting of trees, combined with information on the number of several forest stands. The harvested sites were final-felled decay-affected trees, would facilitate the assessment of forests, which means that all merchantable trees, with the Räty et al. Annals of Forest Science (2023) 80:2 Page 4 of 15 exception possible retention and seed trees, were cut and 1967; Blingsmo 1985) that were calibrated according to recorded by the harvesters. the harvester-based diameter and corresponding height The harvesters were equipped with GNSS (global navi - measurements along stem (Hauglin et al. 2018; Räty et al. gation satellite system) devices, which recorded an XY 2021). location when a tree was felled. Three of the harvesters As part of a research project, the harvester opera- determined the position of the boom tip with sensors tors recorded the presence of visually observable decay that measure crane length and orientation. The rest of the at crosscut surfaces during the harvest operations. The harvesters only recorded the XY location of machine. The operators also evaluated the severity of decay based on machine-positioned trees followed the machine’s driving the width of the decay column. The harvested trees with routes and required post-processing prior to stand delin- decay were categorized into one of the following decay eation. The machine-positioned trees were distributed by classes: (1) incipient decay, trees with decay columns adding random deviations of 8 m to the x and y coordi- covering less than 50% of stem diameter at all crosscut- nates (Räty et al. 2021). The positioning errors of the har - tings, and (2) advanced decay, trees with decay columns vested trees are expected to vary between 5 and 20 m. covering 50% or more of stem diameter at least at a single The XY locations of the harvested trees were over - crosscut. The harvester data were stored in the Standard laid with segments provided by the Norwegian forest for Forest machine Data and communication (Stand- resource map SR16 (Astrup et al. 2019; Hauglin et al. ForD2010) format (Arlinger et al. 2012). Forest attributes 2021). The SR16 segments, based on numerous for - associated with the harvested segments are shown in est attributes derived from remotely sensed data, aim Table 1. to mimic actual forest stands. The harvested sites did Figure 2 shows the DBH and H distribution of healthy not usually follow the boundaries of the SR16 segments. and decay-affected trees in the harvester data. Figure 9 in Therefore, the XY locations of trees were used to delin - the Appendix shows proportions of decay-affected trees eate harvested sites by creating two-dimensional alpha by DBH and H classes. shapes (α = 25 m) using the alphahull package (Pateiro- Lopez and Rodriguez-Casal 2019) in the R environment 2.3 Analyzing tree size distributions (R Core Team 2022). For each alpha shape, a buffer of 2 2.3.1 A ssessing the similarity of the observed tree size m was added to approximately account for the distance distributions between crown edge and stem location. Finally, each The two-sample Kolmogorov-Smirnov (KS) test was used SR16 segment was cropped with the corresponding alpha to statistically test the similarity of the observed tree size shape polygon to establish a harvested segment. More distributions. The KS test was carried out to test the simi - information on the generation of harvested segments larity of size distributions associated with (1) healthy and can be found in Räty et al. (2021). We only selected har- decay-affected trees, (2) all and decay-affected trees, and vested segments that were larger than 0.5 ha and had a (3) trees with incipient and advanced decay. The compar - spruce volume proportion ≥ 90 %. Since our aim was to isons were carried out separately for each harvested seg- model the size distribution of decay-affected trees using ment. The null hypothesis (H0) is that the two tree size theoretical models, we selected harvested segments with distributions come from the same continuous distribu- a sufficient number of decay-affected trees (at minimum tion. The statistical significance level of 5% was used to 20 trees per segment). We wanted to focus on harvested test the null hypothesis. The KS test was carried out using segments in which the frequency of decay-affected trees the stats package in the R environment. was comparable to that found in prior studies in Norway (Huse et al. 1994). Therefore, we selected harvested seg -2.3.2 Weibull distribution ments with more than 15% of trees showing stem decay, A Weibull probability density function (pdf) was fit - which resulted in an average proportion of segment-level ted separately for both DBHs and Hs by maximizing the decay-affected trees of 22%. A total of 101 harvested seg - likelihood in each harvested segment for the following ments (n = 101) containing 93,456 harvested trees (94% groups of trees: (1) all trees, (2) decay-affected trees, (3) Norway spruce trees) were utilized in this study (Fig. 1). trees with incipient decay, and (4) trees with advanced The harvesters registered diameter in 10 cm intervals decay. The group all trees contains healthy and decay- along each stem, and these measurements were used to affected trees in a harvested segment, while the group of estimate DBH for each tree. The harvesters also regis - decay-affected trees contains trees with decay (independ - tered merchantable tree length, but they were not able ent on the degree of decay). The group trees with incipi - to register total tree length (Nordström and Hemmings- ent decay contains trees that have minor decay, whereas son 2018). Thus, tree height was predicted using spe - the group trees with advanced decay includes trees with cies-specific height-diameter curves (Eide 1954; Strand severe decay. R äty et al. Annals of Forest Science (2023) 80:2 Page 5 of 15 Fig. 1 Map of the study area and locations of harvested sites where c is the shape parameter, b is the scale parameter, Table 1 Means, standard deviations (SD), minimums (min), and and x is a harvester measured DBH or H. c, b, x > 0. maximums (max) of attributes associated with the harvested The likelihood function to be maximized is as follows: segments (n = 101) Mean SD Min Max l(θ |x ) = log f x |θ (2) i i ij i j=1 3 −1 Merchantable volume (m ·ha ) 229.62 104.72 90.74 775.87 Lorey’s height (m) 18.30 2.45 13.13 24.53 where θ is the vector of Weibull parameters b and c in −1 Stem frequency (stems·ha ) 714 214 228 1485 harvested segment i (i = 1, …, n) that are to be estimated, 2 −1 Basal area (m ·ha ) 29.88 9.86 13.07 82.7 x refers to the vector of harvester-based tree meas- Quadratic mean diameter (cm) 23.25 2.84 18.27 32.25 urements (DBH or H) in harvested segment i, k is the Proportion of decay-affected trees (%) 22.07 7.88 11.59 53.92 number of trees in harvested segment i, x is tree meas- ij Elevation above sea level (m) 430.34 304.31 7.92 931.57 urement j in harvested segment i (j = 1, …, k), and f (.) is Area (ha) 1.32 0.87 0.62 5.11 the Weibull pdf. The Weibull parameters were estimated using the maximum likelihood estimation (MLE) implemented in The two parameters of the Weibull function were esti - the ForestFit package (Teimouri 2021) in the R environ- mated for each harvested segment by maximizing the ment. No transformations were applied for the Weibull likelihood function. The two-parameter Weibull pdf is as parameters during the estimation. follows: c−1 c x −(x∕b) f(x c, b) = e (1) b b Räty et al. Annals of Forest Science (2023) 80:2 Page 6 of 15 Fig. 2 The diameter (A DBH) and height (B H) distribution of harvested healthy and decay-affected trees in the harvester data. The decay-affected trees were categorized into incipient and advanced decay classes 2.3.3 Comparing tree size distributions using deciles of fitted 2.4 Predicting DBH distributions of decay‑affected trees Weibull distributions2.4.1 Null model We compared tree size distributions of all and decay- A straightforward approach to estimate DBH distribu- affected trees in terms of their shape and horizontal tions of decay-affected trees at the level of harvested position using deciles associated with the fitted Weibull segments is to use a global estimate for all harvested seg- pdfs. We used Weibull pdfs, instead of raw distribu- ments. The global estimate was constructed by fitting a tions, in the comparison of tree size distributions of Weibull pdf using MLE and all harvested decay-affected all and decay-affected trees for two reasons. First, the trees in the study area. We call this approach “null focus was to investigate the general relationships, such model.” The null model is used as a reference approach in as location and width, between all and decay-affected the evaluation of the performance of DM for the predic- trees, and the continuous pdfs are more suitable for tion of DBH distributions of decay-affected trees. that purpose than the discrete raw distribution. Second, we wanted to keep the comparison part of this paper in 2.4.2 Distribution matching line with our ultimate goal to propose a smooth trans- DM is a well-known method in the field of digital image formation function to match size distributions of all processing, and DM has also been applied in various and decay-affected trees (Section 2.4). remote sensing-based forestry applications, such as the We compared (1) deciles of the DBH distribution calibration of predicted forest attribute maps (Baffetta of all trees with those of decay-affected trees and (2) et al. 2012) and transformation of crown-radii distribu- deciles of the H distribution of all trees with those tions to DBH distributions (Vauhkonen and Mehtätalo of decay-affected trees. To consider the relationship 2015). We used DM to transform DBH or H distributions between decay severity and tree size, we also com- of all trees (initial distribution) to DBH distributions of pared the size distributions of trees with incipient and decay-affected trees (target distribution). The initial dis - advanced decay. tributions are fitted Weibull pdfs (Section 2.3.2). We con- The comparison of distributions was carried out by cal - sidered H distributions here because they may be easier culating differences between deciles of two fitted Weibull to obtain than DBH distributions using remotely sensed distributions (ΔD10, ΔD20, …, ΔD90) for each harvested data in the future. DM pursues to predict the shape and segment (Fig. 3). Finally, ΔD10, …, and ΔD90 were visu- location of the DBH distribution of decay-affected trees, ally presented using box plots that show relationships which means that the actual number of decay-affected between the compared density distributions. trees was not predicted. R äty et al. Annals of Forest Science (2023) 80:2 Page 7 of 15 Fig. 3 An illustration on how two tree size distributions of a harvested segment were compared using the observed decile values (D10, …, D90) associated with the fitted Weibull distributions. x, height or diameter at breast height The transformation of initial distributions to target dis - the target segment. The predictive performances of the tributions was carried out using a linear mixed-effects DM and null models were evaluated against the observed model and distribution percentiles: DBH distribution of decay-affected trees using the error index (EI, Eq. 4). The EI is a variant of the well-known (1) (2) (1) (2) (3) (4) 2 y = + q + r + r + z + z r + il i il il il Reynold’s index (Reynolds et al. 1988). il i i (3) EI = f − f (4) i m m where y is the lth percentile of the target distribution il m=1 (l = 1, 2, …, 99) in harvested segment i, r is the corre- il where f and f refer to the observed and predicted sponding percentile of the initial distribution in har- m density of DBH class m in harvested segment i, respec- vested segment i, q refers to the ratio of the interquartile tively, and p is the number of DBH classes. The observed range to the median of the initial distribution of all trees, (1) (2) (.) DBH distribution of decay-affected trees refers to the z z β are the fixed model parameters, and are ran- i i Weibull pdf fit. A bin width of 2 cm and p = 30 were cho- dom intercept and slope parameters in harvested seg- sen, which resulted in the bin midpoints m = 1, 3, …, 59 ment i, and ε is the residual error for the lth percentile il cm. An EI value of 2 refers to a complete disagreement of the target distribution in harvested segment i. The ran - i of the distributions compared, whereas an EI value of 0 dom parameters were set to zero in the prediction phase. refers to a complete match of the compared distributions. We established two different DM setups to predict the Mean, minimum, standard deviation, and maximum of DBH distribution of decay-affected trees: (1) using the the EI values over all harvested segments were reported. DBH distribution of all trees as an initial distribution i Furthermore, the mean error of the mean DBH of (DM ), and (2) using the H distribution of all trees as DBH decay-affected trees (ME , Eq. 5) and the root mean an initial distribution (DM . The parameters of the DM MDBH H) squared error of the mean DBH of decay-affected trees models were estimated in the R environment using the (RMSE , Eq. 6) were calculated. nlme package (Pinheiro et al. 2020). MDBH � � MDBH − MDBH i i i=1 2.4.3 Performance assessments (5) ME = MDBH The performances of DM and null model were assessed in a leave-one-out cross-validation (LCV), where tar- get and neighboring harvested segments with a center � � closer than 500 m were omitted from the training data. MDBH − MDBH i i � i=1 (6) RMSE = The distance limit was used to avoid modeling with data MDBH that include potentially very similar forests compared to Räty et al. Annals of Forest Science (2023) 80:2 Page 8 of 15 3.2 Comparing size distributions of all and decay‑affected where MDBH and MDBH refer to the observed and trees predicted mean DBH of decay-affected trees in harvested Trees with decay showed a right-shifted DBH distribu- segment i, respectively, and n is the number of harvested tion in comparison with all trees (Fig. 4). This observa - segments. The MDBH and MDBH values were calcu- tion implies that stem decay is more likely to be present lated based on the midpoints of the 2 cm bins and cor- in larger trees than in smaller trees. Our data showed responding densities. that the median DBH of decay-affected trees is roughly 4 cm larger compared with the median DBH of all trees. In 3 Results addition, Fig. 4A shows the general trend that the decile 3.1 Testing the similarity of the observed size distributions values associated with DBH distributions of all trees were The KS test rejected the null hypothesis (5% statistical smaller compared with those of decay-affected trees. significance level) in 98% and 90% of the harvested seg - A similar observation can also be made in the case of H ments, when the DBH and H distributions of healthy and distributions (Fig. 4B), although the difference between decay-affected trees were compared, respectively. Corre - the H distribution of decay-affected trees and that of all spondingly, the null hypothesis was rejected in 92% and trees is not as large as in the case of DBH distributions. In 78% of the harvested segments, when DBH and H distri- the case of DBH distributions, the relationship between butions of all and decay-affected trees were compared, decile values of all trees and those of decay-affected trees respectively. The results of the KS test strongly indicated is rather stable, a slight trend of decreasing difference that the size distributions of decay-affected trees were being visible for deciles 60–90 (Fig. 4A). The correspond - different compared with those of healthy and all (all cat - ing differences associated with H distributions steadily egory includes decay-affected trees) trees. decreased towards large deciles, which indicates different The KS rejected the null hypothesis of no difference shapes of the compared distributions (Fig. 4B). (5% statistical significance level) in 67% and 45% of the harvested segments when DBH and H distributions of trees with incipient or advanced decay were compared, 3.3 Comparing size distributions of trees with incipient respectively. This means that that there was a larger dif - and advanced decay ference between the DBH distributions of trees with Trees with advanced decay were in general smaller in incipient and advanced decay compared with the corre- terms of DBH and H than trees with incipient decay. This sponding H distributions. Fig. 4 Differences in decile values of 101 harvested segments between A DBH distributions of all trees and DBH distributions of decay-affected trees and between B H distributions of all trees and H distributions of decay-affected trees in the harvested segments. DBH, diameter at breast height; H, height R äty et al. Annals of Forest Science (2023) 80:2 Page 9 of 15 trend was clearer in the case of H distributions in com- Table 2 Performance assessments of the null model and different distribution matching (DM) setups to predict DBH parison with the DBH distributions (Fig. 5A and B). distributions of decay-affected trees 3.4 Distribution matching Model Mean EI (min, SD, max) ME (cm) RMSE (cm) MDBH MDBH EI, ME , and RMSE values associated with the MDBH MDBH Null model 0.32 (0.03, 0.16, 0.94) 0.03 3.46 null model and DMs are presented in Table 2. The model DM 0.14 (0.02, 0.10, 0.68) 0.04 1.53 DBH parameters of DM , the best-performing model, are DBH DM 0.31 (0.02, 0.20, 1.02) −0.09 3.62 in Table 3 in the Appendix. DM outperformed the null DBH diameter at breast height, H height, SD standard deviation, EI error index, model when the initial distribution of the transformation ME mean error of mean DBH of decay-affected trees, RMSE root-mean- MDBH MDBH was the DBH distribution of all trees, whereas DM using square error of mean DBH of decay-affected trees the H distribution of all trees as an initial distribution did not generally outperform the null model. DM outperformed DM in terms of mean EI, ME DBH H M- large-scale studies on forest decay have typically relied on , and RMSE . Figures 6, 7, and 8 show examples DBH MDBH NFI data, where the presence of decay is assessed using of the harvested segments with a small, average, or large increment core samples taken at breast height (Thor EI value associated with DM , respectively. To avoid DBH et al. 2005; Mattila and Nuutinen 2007; Hylen and Gra- showing extreme cases, the examples were selected based nhus 2018). While it would, in principle, be possible to on the cumulative distribution of the EI values using 5%, drill all trees of a forest in some circumstances, incre- 50%, and 95% cumulative percentages for small, average, ment core samples tend to underestimate the frequency and large EI values, respectively. of decay-affected trees (Hylen and Granhus 2018). The underestimation results from the fact that short decay 4 Discussion columns cannot be observed at breast height (Tamminen Traditional field measurements are often carried out 1985). Short decay columns are typical at an early phase using field sample plots, because a census of DBHs for of stem decay. In addition, some wood decay fungi do not whole forest stands would be too laborious. On the con- cause high stem decay columns. Previous studies on the trary, harvesters automatically collect tree-level infor- decay presence in Norway spruce stands have also been mation from all merchantable trees in forest stands, and based on assessment of stumps after clear cutting (Huse stem decay can be simultaneously registered without et al. 1994). The harvester-based collection of decay any considerable reduction in cost efficiency. Previous Fig. 5 A Differences in decile values associated with diameter at breast height (DBH) distributions of trees with advanced and incipient decay in the harvested segments and B a corresponding comparison in terms of height (H) distributions Räty et al. Annals of Forest Science (2023) 80:2 Page 10 of 15 Fig. 6 Example of the harvested segment that produced a small error index (EI = 0.08) value associated with the distribution matching (DM) of diameter at breast height (DBH) distribution of decay-affected trees. The gray bars show the DBH distribution of all trees, whereas the bars with black borders refer to the DBH distribution of decay-affected trees. Note that density distributions were scaled using observed stem frequencies. MLE, maximum likelihood estimation information resembles stump surveying, but as a benefit, decay pathogens in Norway spruce, namely white-rot the decay observations can be linked to the correspond- fungi in the genera Heterobasidion and Armillaria, which ing harvester-based stem measurements. As a drawback, are considered to cause more than 90% of the decay in harvesters only operate in productive forests which intro- Norway spruce in Norway (Huse et al. 1994). Stem decay duce a selection bias in the data collection. In addition, caused by Heterobasidion and Armillaria generally origi- the smallest trees are typically underrepresented (DBH < nates from root infection, but due to active tree defense, 10 cm) in the harvested trees compared with the factual the spread of decay fungi is slow in living roots of young status in the forest, since the smallest trees may be too trees (Bendz-Hellgren et al. 1999). Therefore, it is rea - time-consuming to harvest in terms of their commercial sonable to assume that also many of the small trees had value or may not fulfill dimension requirements of mer - infection by wood decay fungi, although the decay was chantable logs. still confined to the roots, and thus not yet visible at the Our data showed that stem decay was absent or rare in stem base during harvesting. trees below 9 cm in DBH, while the proportion of decay- The observed increase in the proportion of decay- affected trees increased along with DBH classes, reaching affected trees along with tree size is in line with the find - a level of up to 40% in DBH classes ≥ 45 cm. As a result, ings of Hylen and Granhus (2018). This finding reflects the median DBH of decay-affected trees was roughly 4 the spread biology of the most common decay fungi of cm larger than that of all trees. We did not aim to deter- Norway spruce. The older the next-generation trees mine the causative agents of decay in this study, but the are, the more root contacts they have established with findings reflect the biology of the most common wood stumps of decay-affected previous-generation trees. R äty et al. Annals of Forest Science (2023) 80:2 Page 11 of 15 Fig. 7 Example of the harvested segment that produced an average error index (EI = 0.17) value associated with the distribution matching (DM) of diameter at breast height (DBH) distribution of decay-affected trees. The gray bars show the DBH distribution of all trees, whereas the bars with black borders refer to the DBH distribution of decay-affected trees. Note that density distributions were scaled using observed stem frequencies. MLE, maximum likelihood estimation The established root contacts increase the probability of Oliva et al. 2012; Nagy et al. 2022). Because the heights infection transmission (Pukkala et al. 2005). Trees receiv- of the harvested trees were predicted using conventional ing infection via root contacts in turn spread the infec- taper curves, the effect of decay on the growth rates tion further to neighboring trees once the trees establish was neglected, which may result in overpredictions of root contacts or grafts. It is also worth noting that myce- heights associated with decay-affected trees. We are not lial spread of decay fungi is faster in roots of stumps than aware of any study that would have considered whether, in living roots of trees owing to the absence of defense and to what extent, the heartwood decay of Norway responses in dead roots (Schönhar 1978). spruce affects the relation between diametric and height Based on the current understanding, the recorded growth. An additional factor is that trees with advanced trees with incipient stem decay represent more recent stem decay likely also have a less functioning root sys- infections than the trees with advanced stem decay. tem than trees with incipient decay, but this aspect is Our findings indicate that trees with incipient decay are not well established in previous studies. When consider- on average larger and taller compared with trees with ing the timing of final felling in a rotation forestry, decay advanced decay. This is expected, since the propor - may have implications on optimal rotation time from tion of water-conductive stem sapwood decreases along the perspective of carbon sequestration or economy as with advancement of decay, while at the same time, the suggested by Möykkynen and Pukkala (2010). The char - affected trees allocate more carbon to tree defense phe - acterization of the DBH distribution of decay-affected nolics at the interface between pathogen-colonized wood trees allows to better account for economic loss associ- and inner sapwood (Bendz-Hellgren and Stenlid 1995; ated with timber quality resulting from decay and even Räty et al. Annals of Forest Science (2023) 80:2 Page 12 of 15 Fig. 8 Example of the harvested segment that produced a large error index (EI = 0.36) value associated with the distribution matching (DM) of diameter at breast height (DBH) distribution of decay-affected trees. The gray bars show the DBH distribution of all trees, whereas the bars with black borders refer to the DBH distribution of decay-affected trees. Note that density distributions were scaled using observed stem frequencies. MLE, maximum likelihood estimation reduced biomass growth in the simulations of different (or H) distribution of merchantable trees in a forest stand management scenarios. is known. Our data consisted of spruce-dominated stands, DM, i.e., distribution matching using the observable and therefore, we were not able to study the effect of spe - DBH distribution of all trees in a stand as input, out- cies mixtures on the size distribution of decay-affected performed the null model assuming a fixed distribu - trees. Möykkynen and Pukkala (2010), among many others, tion in the prediction of the DBH distribution of the concluded that the spread rate of Heterobasidion decay is decay-affected trees. This suggests a predictable rela - significantly slower in admixed forests compared with pure tionship between the size distribution of all trees and spruce forests. It is noteworthy that this scenario does not that of decay-affected trees. DM predicts densities of apply to Armillaria, as the rhizomorphs of Armillaria spe- DBH classes, which means that the number of trees with cies causing decay in Norway spruce use also broadleaved decay must be derived from other sources. Potential trees or their stumps as a food base (Keča and Solheim data sources for the estimation of the number of decay- 2011). It should also be noted that our harvester data were affected trees are, for example, nation-wide stump sur - collected from mature forests that are used for commercial veys (Huse et al. 1994), NFI data-based models combined timber production. Therefore, our DM may not be appli - with climatic or environmental data (Hylen and Granhus cable in young forests or forests that are not managed for 2018), or a combination of harvester and remotely sensed the purposes of timber production. Regarding the silvicul- data (Räty et al. 2021). tural activity, a critical question associated with the spread The practical applicability of our current DM approach is of decay fungi is the history of silvicultural treatments, limited to certain forest types and requires that the DBH such as thinnings. Thinnings are not routinely carried out R äty et al. Annals of Forest Science (2023) 80:2 Page 13 of 15 in most forests in Norway, which may have a mitigating knowledge on stand-level timber decay damages in mature effect on the spread of decay compared with thinned for - forests. est stands (Metslaid et al. 2018). Thinnings typically affect the width and shape of DBH distribution of all trees, which 5 Conclusions may change the relationships between size distributions of We draw the following conclusions based on our analyses all and decay-affected trees. We did not have data about the carried out in mature spruce-dominated forests using cut- previous management history of the stands included in our to-length harvester data: study, and a history of thinning could be one reason for the poor predictive performance of our method in some stands (1) An average-sized decay-affected tree is larger and (Fig. 8). taller compared with an average-sized tree in a for- Future studies could supplement this study by account- est stand. ing for local differences in the prediction of the DBH dis - (2) Trees with incipient decay are on average larger and tribution of decay-affected trees, for example, by using taller compared with trees with advanced decay. various remotely sensed data and environmental, climatic, (3) The abovementioned findings reflect a relationship and historical data sources. The methodology presented between the DBH distribution of all trees and the here could also be adapted to stand-level forest invento- DBH distribution of decay-affected trees. ries, where the initial DBH distribution is predicted, for (4) Distribution matching that transforms the DBH example, by means of airborne laser scanning data or rapid distribution of all trees into the DBH distribution of field measurements. In addition, the potential of harvester decay-affected trees can be used to learn the rela - data for the modeling of decay column properties, such as tionship between all and decay-affected trees and height and width, should be investigated. Our study indi- to predict the DBH distribution of decay-affected cates that harvester data have a high potential to increase trees for forest stands. Appendix Fig. 9 Proportions of decay-affected trees by DBH (A) and height (B) classes in the harvester data. DBH, diameter at breast height Räty et al. Annals of Forest Science (2023) 80:2 Page 14 of 15 resources map SR16. Scand J For Res 34:484–496. https:// doi. org/ 10. Table 3 Parameter estimates of the model associated with the 1080/ 02827 581. 2019. 15889 89 distribution matching (DM ). For more information on model DBH Baffetta F, Corona P, Fattorini L (2012) A matching procedure to improve k-NN formulation and parameters, please refer to Section 2.4 estimation of forest attribute maps. For Ecol Manag 272:35–50. https:// doi. org/ 10. 1016/j. foreco. 2011. 06. 037 Bendz-Hellgren M, Brandtberg P-O, Johansson M et al (1999) Growth rate of Predictors Estimates Std. error p Heterobasidion annosum in Picea abies established on forest land and arable land. Scand J For Res 14:402–407. https:// doi. org/ 10. 1080/ 02827 Intercept −7.885 1.254 < 0.001 58995 01541 04 r 1.176 0.007 < 0.001 Bendz-Hellgren M, Stenlid J (1995) Long-term reduction in the diameter r −0.005 < 0.001 < 0.001 growth of butt rot affected Norway spruce, Picea abies. For Ecol Manag 74:239–243 q 19.248 2.170 < 0.001 Blingsmo K (1985) Avsmalingsfunksjoner og -tabeller for bjørk. Rapport fra Random effects Norsk institutt for skogforskning 10:35 (1) 2.315 Caudullo G, Tinner W, de Rigo D (2016) Picea abies in Europe: distribution, var( ) habitat, usage and threats. In: San-Miguel-Ayanz J, de Rigo D, Caudullo G, (2) 0.066 var( ) Houston Durrant T, Mauri A (eds) European Atlas of Forest Tree Species. (1) (2) Publ. O. EU ff , Luxembourg −0.76 cor( z , z ) l l Eide E (1954) Avsmalingstabell for granskog. Utvidet og revidert 1954, p 63 2 2 σ 0.231 Gobakken T, Næsset E (2004) Estimation of diameter and basal area distribu- tions in coniferous forest by means of airborne laser scanner data. Scand J For Res 19:529–542. https:// doi. org/ 10. 1080/ 02827 58041 00194 54 Acknowledgements Gonzáles RC, Woods RE (2002) Digital image processing, 3. ed., internat. ed. We thank Simon Berg for technical support with the harvester data. Prentice-Hall, Upper Saddle River Hauglin M, Hansen E, Sørngård E et al (2018) Utilizing accurately positioned Code availability harvester data: modelling forest volume with airborne laser scanning. The codes used to analyze the data were not disclosed. Can J For Res. https:// doi. org/ 10. 1139/ cjfr- 2017- 0467 Hauglin M, Rahlf J, Schumacher J et al (2021) Large scale mapping of forest Authors’ contributions attributes using heterogeneous sets of airborne laser scanning and Conceptualization, JR, AMH, JB, and RA; methodology, JR, JB, and AMH; writing National Forest Inventory data. For Ecosyst 8:65. https:// doi. org/ 10. 1186/ — original draft preparation, JR and AMH; writing — review and editing, JR, s40663- 021- 00338-4 AMH, JB, and RA; and funding acquisition, RA. The authors read and approved Honkaniemi J, Ojansuu R, Piri T et al (2014) Hmodel, a Heterobasidion the final manuscript. annosum model for even-aged Norway spruce stands. Can J For Res 44:796–809. https:// doi. org/ 10. 1139/ cjfr- 2014- 0011 Funding Huse K, Solheim H, Venn K (1994) Råte i gran registrert på stubber etter hogst This work was supported by the Norwegian Research Council through the vinteren 1992 [Stump inventory of root and butt rots in Norway spruce PRECISION project — precision forestry for improved resource use and cut in 1992]. Rapp Skogforsk 23:1–26 reduced wood decay in Norwegian forests (NFR 281140). Hylen G, Granhus A (2018) A probability model for root and butt rot in Picea abies derived from Norwegian national forest inventory data. Scand J For Availability of data and materials Res 33:657–667. https:// doi. org/ 10. 1080/ 02827 581. 2018. 14870 74 The dataset supporting the conclusions of this article is not publicly available Keča N, Solheim H (2011) Ecology and distribution of Armillaria species in due the private ownership of the harvester datasets. Norway. For Pathol 41:120–132. https:// doi. org/ 10. 1111/j. 1439- 0329. 2010. 00644.x Korhonen K, Stenlid R (1998) Biology of Heterobasidion annosum. In: Wood- Declarations ward S, Stenlid J, Huttermann A, Karjalainen R (eds) Heterobasidion anno- sum: biology, ecology, impact, and control. Cab International, Wallingford Ethics approval and consent to participate Lindén M, Vollbrecht G (2002) Sensitivity of Picea abies to butt rot in pure Not applicable. stands and in mixed stands with Pinus sylvestris in southern Sweden. Silva Fenn 36:767–778 Consent for publication Mattila U, Nuutinen T (2007) Assessing the incidence of butt rot in Norway Not applicable. spruce in southern Finland. Silva Fenn 41. https:// doi. org/ 10. 14214/ sf. 473 Metslaid M, Granhus A, Scholten J et al (2018) Long-term effects of single-tree Competing interests selection on the frequency and population structure of root and butt The authors declare that they have no competing interests. rot in uneven-sized Norway spruce stands. For Ecol Manag 409:509–517. https:// doi. org/ 10. 1016/j. foreco. 2017. 11. 050 Author details Möykkynen T, Pukkala T (2010) Optimizing the management of Norway spruce Natural Resources Institute Finland (LUKE), Yliopistokatu 6, 80100 Joensuu, and Scots pine mixtures on a site infected by Heterobasidion coll. Scand J Finland. Norwegian Institute of Bioeconomy Research (NIBIO), Høgskoleveien For Res 25:127–137. https:// doi. org/ 10. 1080/ 02827 58100 36673 22 8, 1433 Ås, Norway. Norwegian Institute of Bioeconomy Research (NIBIO), Müller MM, Kaitera J, Henttonen HM (2018) Butt rot incidence in the north- Skolegata 22, Building P1, 7713 Steinkjer, Norway. ernmost distribution area of Heterobasidion in Finland. For Ecol Manag 425:154–163. https:// doi. org/ 10. 1016/j. foreco. 2018. 05. 036 Received: 5 July 2022 Accepted: 9 December 2022 Müller MM, Sievänen R, Beuker E et al (2014) Predicting the activity of Heter- obasidion parviporum on Norway spruce in warming climate from its respiration rate at different temperatures. For Pathol 44:325–336. https:// doi. org/ 10. 1111/ efp. 12104 Nagy NE, Norli HR, Fongen M et al (2022) Patterns and roles of lignan and References terpenoid accumulation in the reaction zone compartmentalizing Arlinger J, Nordström M, Möller J (2012) StandForD 2010: Modern kommunika- pathogen-infected heartwood of Norway spruce. Planta 255:1–16 tion med skogsmaskiner Nordström M, Hemmingsson J (2018) Measure up! A Skogforsk guide to Astrup R, Rahlf J, Bjørkelo K et al (2019) Forest information at multiple scales: harvester measurement. Skogforsk, Uppsala development, evaluation and application of the Norwegian forest R äty et al. Annals of Forest Science (2023) 80:2 Page 15 of 15 Oliva J, Camarero JJ, Stenlid J (2012) Understanding the role of sapwood loss and reaction zone formation on radial growth of Norway spruce (Picea abies) trees decayed by Heterobasidion annosum s.l. For Ecol Manag 274:201–209. https:// doi. org/ 10. 1016/j. foreco. 2012. 02. 026 Pateiro-Lopez B, Rodriguez-Casal A (2019) alphahull: generalization of the convex hull of a sample of points in the plane Pinheiro J, Bates D, DebRoy S, et al (2020) nlme: linear and nonlinear mixed effects models Piri T (1996) The spreading of the S type of Heterobasidion annosum from Norway spruce stumps to the subsequent tree stand. Eur J For Pathol 26:193–204. https:// doi. org/ 10. 1111/j. 1439- 0329. 1996. tb008 39.x Piri T (2003) Early development of root rot in young Norway spruce planted on sites infected by Heterobasidion in southern Finland. Can J For Res 33:604–611. https:// doi. org/ 10. 1139/ x02- 200 Piri T, Korhonen K (2001) Infection of advance regeneration of Norway spruce by Heterobasidion parviporum. Can J For Res 31:937–942. https:// doi. org/ 10. 1139/ cjfr- 31-6- 937 Prospero S, Rigling D, Holdenrieder O (2003) Population structure of Armil- laria species in managed Norway spruce stands in the Alps. New Phytol 158:365–373. https:// doi. org/ 10. 1046/j. 1469- 8137. 2003. 00731.x Pukkala T, Möykkynen T, Thor M et al (2005) Modeling infection and spread of Heterobasidion annosum in even-aged Fennoscandian conifer stands. Can J For Res 35:74–84. https:// doi. org/ 10. 1139/ x04- 150 R Core Team (2022) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna Räty J, Breidenbach J, Hauglin M, Astrup R (2021) Prediction of butt rot volume in Norway spruce forest stands using harvester, remotely sensed and environmental data. Int J Appl Earth Obs Geoinf 105:102624. https:// doi. org/ 10. 1016/j. jag. 2021. 102624 Reynolds M, Burk T, Huang W (1988) Goodness-of-fit tests and model selection procedures for diameter distribution models. For Sci 34:373–399 Schönhar S (1978) Infektionsversuche mit Fomes annosus-Sporen an vitalen und absterbenden Fichtenwurzeln. Allgemeine Forst-und Jagdzeitung 149:73–74 Siipilehto J, Mehtätalo L (2013) Parameter recovery vs. parameter prediction for the Weibull distribution validated for Scots pine stands in Finland. Silva Fenn 47. https:// doi. org/ 10. 14214/ sf. 1057 Stenlid J, Redfern D (1998). In: Woodward S, Stenlid J, Karjalainen R, Hütter- mann A (eds) Spread within the tree and stand, w: Heterobasidion anno- sum, biology, ecology, impact and control. Cab International, Wallingford, pp 125–141 Strand L (1967) Avsmalingstabeller for furu sønnafjells. Medd norske Skogfor - sVes 84:431–482 Tamminen P (1985) Butt rot in Norway spruce in southern Finland. In: Com- municationes Instituti Forestalis Fenniae, p 52 Teimouri M (2021) ForestFit: statistical modelling for plant size distributions Thor M, Ståhl G, Stenlid J (2005) Modelling root rot incidence in Sweden using tree, site and stand variables. Scand J For Res 20:165–176. https:// doi. org/ 10. 1080/ 02827 58051 00083 47 Vauhkonen J, Mehtätalo L (2015) Matching remotely sensed and field-meas- ured tree size distributions. Can J For Res 45:353–363. https:// doi. org/ 10. 1139/ cjfr- 2014- 0285 Vollbrecht G, Agestam E (1995) Identifying butt rotted Norway spruce trees from external signs. Forest and Landscape Research, Denmark Woodward S, Stenlid J, Huttermann A, Karjalainen R (1998) Heterobasidion annosum: biology, ecology, impact, and control. Cab International, Wallingford Re Read ady y to to submit y submit your our re researc search h ? Choose BMC and benefit fr ? Choose BMC and benefit from om: : fast, convenient online submission Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in pub- thorough peer review by experienced researchers in your field lished maps and institutional affiliations. rapid publication on acceptance support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions
Journal
Annals of Forest Science – Springer Journals
Published: Jan 5, 2023
Keywords: Root and butt rot; Heterobasidion spp.; Armillaria spp.; Cut-to-length harvester; Forest management and planning