Disentangling the Impact of Forest Management Intensity Components on Soil Biological Processes

ABSTRACT

Temperate forests cover 25% of the world's forest area and most of them are managed for timber production. To increase yields, native deciduous trees have been commonly replaced by fast‐growing conifers, especially in Western and Central Europe. Despite the importance of forest soils for a variety of ecosystem functions, the effects of forest management intensity on soil biological processes have not yet been sufficiently understood. Using a standardized sampling protocol that covers 200 plots across four regions with different abiotic site conditions, our study aims at disentangling the effects of individual components of forest management intensity such as (i) share of native deciduous trees, (ii) timber harvesting volume, and (iii) natural deadwood volume on soil biological processes. Our findings indicate that the effects of management practices on soil biological processes are as important as abiotic factors, such as subsoil pH value and soil texture. Piecewise structural equation modeling revealed that forest management has both direct and indirect effects on soil biological processes via humus form and topsoil pH. Generally, the ratio of deciduous and coniferous trees had the most significant effect on nutrient cycling and soil properties, followed by nutrient extraction through timber harvesting and natural deadwood volume. The strength of the observed effects was modified by regional variations in climate, topography, and bedrock. Our findings underline the high diagnostic value of humus form as an indicator of biological processes in mineral topsoil (particularly pH, C/N, and microbial characteristics can be predicted) that can guide forest managers in evaluating soil quality and identifying management impacts.

We analyzed the effects of forest management components (proportion of natural deadwood, deciduous trees, and timber extraction) on soil biological processes in temperate forests. The topsoil of 200 plots across four regions with different abiotic site conditions was examined. Piecewise structural equation modeling revealed that forest management exerts direct and indirect effects on soil biological processes through humus form and topsoil pH. Overall, the ratio of deciduous and coniferous trees had the most notable influence on nutrient cycling (e.g., microbial biomass and enzyme activities) and soil properties (e.g., ratio of C to N and extractable total P), followed by the impact of nutrient extraction through timber harvesting and natural deadwood volume.

1. Introduction

Europe's temperate forests are predominantly shaped by human activity, with only 2% remaining pristine old growth. Most forests are actively managed and, while often seminatural, their tree composition differs from that of potential natural forest communities, with a predominance of fast‐growing conifers (Forest Europe ²⁰²⁰; Gilliam ²⁰¹⁶). The multiple demands on forests and their ability to adapt to future environmental conditions in the face of climate change, the biodiversity crisis, and a globally increasing demand for timber (FAO ²⁰²²; Forest Europe ²⁰²⁰) pose major challenges to sustainable forest management: A pure focus on timber production often leads to an intensification of forest management (McGrath et al. ²⁰¹⁵) that may strongly hamper other ecosystem services. Management intensification is typically expressed by an increase in the proportion of non‐native, mostly coniferous tree species, and an increase in the frequency of harvesting and thinning, accompanied by a decline in natural deadwood. Changes in tree species composition, increased harvest volumes, and removal of deadwood alter the quantity and quality of biomass returning to the soil, as well as microclimatic conditions, and potentially affect soil (micro)organisms and soil functions. Soil biological processes support multiple ecosystem services provided by forests including climate regulation, flood protection, cycling of water, carbon and nutrients, biodiversity preservation, and timber production (Ali 2023). The integration of soil biological processes into decision‐making is thus becoming increasingly recognized, not only in agriculture but also in forest management (Arredondo‐Ruiz et al. ²⁰¹⁴; Creamer et al. ²⁰²²; Vogel et al. ²⁰²⁴).

Overall, the effect of forest management intensity on soil microbial activities and nutrient cycling has not yet been comprehensively investigated. Most studies either focus on a single management practice (Danise et al. 2022; Hassett and Zak ²⁰⁰⁵; Nacke et al. ²⁰¹⁶) or forest management types in general (Goldmann et al. 2015; Purahong et al. ²⁰¹⁴; Schall et al. ²⁰¹⁸). Yet, forest management is typically composed of multiple management components, whose individual and combined effects help to understand mechanisms and consequences of management on forest ecosystems (as it has been shown for grasslands; Socher et al. ²⁰¹³). Even studies quantifying forest management intensity using recently developed indices (Kahl and Bauhus 2014; Schall and Ammer ²⁰¹³) did typically not analyze the influence of single components (Kaiser et al. 2016; Richter et al. ²⁰¹⁸; Wilcke et al. ²⁰²⁴). Existing studies either focus on soil organic layers (Hobbie et al. 2006; Ribbons et al. ²⁰¹⁶; Wilcke et al. ²⁰²⁴) or on mineral topsoil (Awad et al. 2019; Danise et al. ²⁰²²; Goldmann et al. ²⁰¹⁵), but few combine both approaches (Bayranvand et al. 2021; Richter et al. ²⁰¹⁸; Zederer et al. ²⁰¹⁷). Furthermore, many studies are confined to a single local or regional environmental setting, which makes it difficult to judge the generality and context dependency of findings (Bastida et al. 2017; Cardenas et al. ²⁰¹⁵; Scheibe et al. ²⁰¹⁴) while only few provide a broader view (Gan et al. 2020; Richter et al. ²⁰¹⁸; Wilcke et al. ²⁰²⁴). To date, a comprehensive analysis of forest management effects on soil biological processes that consider gradients of forest management intensity across various regional environmental settings has not been conducted.

Existing studies generally indicate strong tree species effects on soil biological processes (e.g., decomposition and humus forms) via shifts in litter quality and quantity (Hobbie et al. 2006; Lorenz et al. ²⁰²⁰; Prescott ²⁰¹⁰). Thereby, they influence nutrient supply (Awad et al. 2019; Danise et al. ²⁰²²; Pollierer et al. ²⁰²¹) and pH levels (Antisari et al. 2011; Hagen‐Thorn et al. ²⁰⁰⁴; Nacke et al. ²⁰¹⁶) as well as soil microbial biomass carbon (MBC) that is typically higher under deciduous trees than under conifers (Bauhus, Paré, and Côté ¹⁹⁹⁸; Bayranvand et al. ²⁰²¹). Forest management also mediated changes in the soil microbial composition of fungi and bacteria, with associated shifts in the functions these species provide. Ectomycorrhizal fungal biomass decreases, and saprotrophic fungal biomass increases with increasing conifer presence (Awad et al. 2019) and overall fungal species richness is higher in native forests than in monoculture settings (Buée et al. 2011). Belowground bacterial species richness declines and the proportion of specialists rises with rising share of conifers (Penone et al. 2019). Changes in microbial community composition also affect microbial activity and extracellular enzyme activities, with peroxidases being more active under broadleaf trees and cellulases under conifers (Baldrian et al. 2013; Bayranvand et al. ²⁰²¹; Danise et al. ²⁰²²). Overall, more conifers result in lower microbial abundance and activity and thus slower litter decomposition processes, which can lead to an accumulation of thick carbon‐rich organic layers.

Timber harvesting is widely recognized to have a negative impact on soil physical, chemical, and biological quality through the compaction of the forest floor (Cambi et al. 2015; Klein‐Raufhake et al. ²⁰²⁴). In addition to mechanical stress, timber extraction has further significant impacts on forest soils through the removal of organic carbon and changes in the stand structure: Timber harvesting practices that open the tree canopy cause increased temperature and water supply in the forest floor. This accelerates organic matter decay and may lead to a decline in soil carbon and nitrogen, particularly in the organic horizon (Bastida et al. 2017; Nave et al. ²⁰¹⁰; von Wilpert ²⁰²²). While the microbial community composition may remain unchanged, timber harvesting often reduces microbial biomass and activity (Bastida et al. 2017; Hassett and Zak ²⁰⁰⁵), ultimately reducing microbially released nutrients. In addition, forest management affects the forest floor through the quantity and quality of deadwood (Kahl et al. 2017)—managed forests often retain only stumps and small branches after large timber extraction, while unmanaged forests have much higher levels of (natural) deadwood (Blaser et al. 2013). The importance of lying deadwood has recently been recognized as a “pedogenic hotspot of concentrated biogeochemical and ‐physical processes” (Stutz and Lang 2017) with particularly concentrated soil biological activity close to deadwood (Wambsganss, Stutz, and Lang ²⁰¹⁷), reflected in increased microbial biomass and activity, increased C‐, N‐, and P contents, a wider C/N ratio, higher base saturation, increased C mineralization, and soil organic matter stabilization in the mineral topsoil (Minnich et al. 2021; Stutz and Lang ²⁰¹⁷; Wambsganss, Stutz, and Lang ²⁰¹⁷).

Studies using the forest management intensity index (ForMI) as integrated measure of tree species identity, timber harvest volume, and deadwood volume have shown that a high ForMI is associated with lower litterfall, slower litter decomposition, and lower nitrogen, phosphorus, and potassium contents in the organic layer. Effects are also seen in the mineral soil, with increased fungal biomass and activity of enzymes involved in carbon and phosphorus cycling (Felipe‐Lucia et al. ²⁰²⁰; Richter et al. ²⁰¹⁸). Despite such general findings of forest management effects on soil biological processes, several authors, such as Richter et al. (2018), also emphasized that regional environmental characteristics may overrule the effects of forest management intensity on microbial community structure and biomass. Similarly, Gan et al. (2020) found that the effects of tree species on mineralization processes differed between nutrient‐rich and nutrient‐poor regions. As soil nutrient levels strongly determine ecosystem productivity and litter quality, they also affect belowground bacterial and fungal growth (Awad et al. 2019; Snajdr et al. ²⁰¹²) and community composition and activity (Felsmann et al. 2015; Richter et al. ²⁰¹⁸; Tedersoo et al. ²⁰¹⁵). Strong regional variations are related to differences in soil parent material which determines soil texture and pH. Both in turn influence belowground species richness (Tedersoo et al. 2015), microbial abundance, and community structure (Felsmann et al. 2015; Kaiser et al. ²⁰¹⁶; Richter et al. ²⁰¹⁸). Soil texture and subsoil pH value are closely related (Ponge and Chevalier 2006); however, particularly topsoil pH is also affected by the litter quality of the stand (Danise et al. 2022; Pollierer et al. ²⁰²¹), thus serving as a mediator between forest management and soil (microbial) properties. Likewise, the humus form plays a mediating role as its development depends on both initial soil conditions (Andreetta et al. 2016) and litter quality (Bayranvand et al. 2021). Given that all above‐ and belowground interactions take place in the humus profile (Ponge 2013), both stand and soil properties are closely related to the humus form, making it particularly suitable for assessing site quality (Ponge and Chevalier 2006).

Our study aims to disentangle the effects of forest management intensity components on soil biological processes across temperate forest sites. To assess forest management intensity in a standardized way, we used the ForMI (Kahl and Bauhus 2014). We quantified the ForMI components (proportions of (i) native deciduous trees, (ii) timber harvesting, and (iii) natural dead wood), determined humus forms, and measured nutrient contents and microbial activity, as well as microbial biomass in the mineral topsoil of 200 forest plots, across four sites differing in climate and topography (lowland vs. low mountain) and bedrock properties (base poor vs. base rich). Forest plots ranged from unmanaged stands dominated by native European broadleaved tree species (beech or oak) to intensively managed stands with non‐native conifers (mainly Norway spruce and Scotch pine).

We hypothesized that:

1. A high forest management intensity, as measured by the ForMI, decreases soil microbial biomass and nutrient turnover processes in the mineral topsoil.

2. The share of native deciduous trees versus conifers has the strongest impact on soil biological processes followed by timber harvest volume and the proportion of natural deadwood.

3. Management effects on soil properties are mediated indirectly through the humus form and pH of the topsoil.

4. Regional variation in abiotic site conditions modulates the strength of management impacts: Base‐poor sites will be stronger affected by nutrient export, for example, via timber harvest, than base‐rich sites.

5. The humus form is a suitable indicator, independent of site conditions, to assess and predict the influence of forest management intensity and its components on soil biological processes.

2. Methods

2.1. Sites, Study Design, and ForMI

Our study focused on four forest areas in Northwest Germany (Figure 1a), representing the most common forest types in Central Europe: mixed oak–beech forests on strongly acidic sandy–loamy Stagnic/Haplic Cambisols in the lowlands (1, Lower Rhine), mixed oak–hornbeam forests on more base‐rich Spodic/Haplic Stagnosols in the lowlands with soil texture varying from clayey to sandy (2, Münsterland), beech forests on strongly acidic loamy Stagnic/Haplic Cambisols in the low mountain range (3, Arnsberger Wald), and beech forests on base‐rich loamy/clayey Haplic Cambisols in the low mountain range (4, Egge‐Vorberge). A detailed site description can be found in Table S1. Differences in acidity and nutrient levels are shown in Figure 1c,d.

FIGURE 1.

(a) Location of the four study areas in North Rhine‐Westphalia, Germany. Map modified according to Klein‐Raufhake et al. (2024). Map lines delineate study areas and do not necessarily depict accepted national boundaries. (b) In each of the four regions, 50 plots were sampled. The composite soil samples were taken from five points within a 10 × 10 m area, and forest management variables were recorded within 0.1 ha sample circle. (c) pH values in the organic layer and the mineral soil at different depths. Dots correspond to the mean values and lines to the SD. The pH values from 0 to 5 and 30 to 60 cm mineral soil depth were used for further analyses. (d) Differences between the study sites in a set of variables: The forest management intensity index (ForMI) and its components. (1) Native deciduous trees, (2) timber harvesting, and (3) natural dead wood (all three given in proportions), the humus form (scaled from mor (1) to mull (9)), ratio of C to N (C/N), content of total extractable phosphorus (TP), and microbial biomass carbon (MBC) in 0–5 cm soil depth. Black dots with whiskers represent mean ± SD, and colored dots represent individual plots. Letters indicate significant group differences, determined by one‐way ANOVAs, followed by the post hoc Tukey HSD test.

Overall, a total of 200 plots of 0.1 ha were chosen, with n = 50 plots in each study region. The selection of plots was stratified according to the main soil types of the respective region. Plots were randomly selected in the following four categories: (i) unmanaged forests in protected areas where management ceased completely 10–50 years ago (10 plots), managed forest stands with predominantly native trees in the (ii) 40‐ to 80‐year age class (~ 15 plots) and (iii) in the > 80‐year age class (~ 15 plots), and (iv) managed forest stands dominated by non‐native conifers (10 plots). The categorization has been made on the basis of data provided by the state forestry administration, who typically use the abovementioned age classes. A representation of the precise stand age for all plots can be found in the Appendix, Figure S2. To ensure spatial independence and minimize edge effects, the distance between the plots and the forest edge was at least 100 m.

For each plot, we determined the ForMI as described by Kahl and Bauhus (2014) to warrant a uniform assessment of management intensity. The ForMI is a number that theoretically ranges between 0 and 3. It is calculated by summing up three factors: (i, Inonat) the volume fraction of non‐native trees (0–1); (ii, Iharv) the volume fraction of harvested wood in relation to the total potential wood volume on the plot (0–1); and (iii, Idwcut) the fraction of dead wood which is not of natural origin (0–1). By stratifying the plot selection into these four management categories (see above), we were able to represent a wide range of forest management intensities. In our study, the ForMI ranged from a minimum of 0, indicative of an almost natural forest with native tree species and no observable effects of management on the forest structure, to a maximum of 2.3, corresponding to an intensely managed stand with non‐native tree species (Table S1). For our analyses, we translate the three factors of the ForMI (i) to the share of native deciduous tree species (since no native conifer trees and no relevant proportion of non‐native deciduous trees occur in our plots), (ii) to the share of timber harvest volume, as well as (iii) to the share of naturally occurring deadwood.

2.2. Humus Forms and Soil Sampling

Humus form description and soil sampling were carried out on each plot within a 10 m x 10 m square (Figure 1b). Sampling was carried out over short periods (2–3 weeks) in spring to minimize seasonal variation (Baldrian 2017). Low mountain forests were sampled in March/April 2021 and lowland forests in March/April 2022. We sampled the organic layers (litter, Of‐, and Oh layer) at three different sampling points per plot in 20 × 20 cm metal frames and determined type of litter, the horizons' thickness, and its stratification to identify the present humus form according to the second national forest soil inventory (Wellbrock 2006). In the statistical analysis, an ordinal scale of 1–9 was used, corresponding to the nine distinguishable aeromorphic humus forms (from mor (1) to moder (4) to mull (9) and corresponding intermediate forms (Wachendorf et al. 2023)). The humus form thus represents the degree of decomposition of the organic layer from thick and not decomposed to almost completely decomposed.

After careful removal of the organic layer, mineral topsoil samples were taken from a depth of 0 to 5 cm using a volumetric sampling ring with a diameter of 8 cm. In addition to the three rings taken at the humus form sampling points, two further mineral topsoil samples were taken at the edges of the 10 × 10 m². All five replicates were pooled, weighed, and stored at 7°C until further processing. Within 1–3 days, all mineral topsoil samples were sieved (< 5 mm), and remaining roots were carefully removed by hand. A part of the sieved samples was frozen at −20°C for later analysis of microbial soil parameters. This procedure is often recommended to minimize storage effects (Lee et al. 2007). For analysis of physical and chemical soil parameters, another part was dried at 40°C and sieved through a 2 mm sieve. All results are expressed on a soil dry weight basis (105°C).

2.3. Physical and Chemical Properties of the Soil

A detailed description of soil property analyses is given in Klein‐Raufhake et al. (2024). In brief, soil pH (KCl) was measured potentiometrically with a 1:5 ratio of soil to potassium chloride solution (1 M) (Gutachterausschuss Forstliche Analytik ²⁰⁰⁵). The content of soil organic carbon (SOC) and total nitrogen (TN) was analyzed using an EA 3000 elemental analyzer (Eurovector/Hekatech, Italy). We obtained information about the respective soil texture from the 1:50.000 soil map (Geologischer Dienst NRW ²⁰¹⁹). The soil texture is ordinally scaled from 1 to 5 and represents the coarseness, ranging from sand (1) to clay (5).

2.4. Soil Microbial Properties

Three days before analysis of microbial biomass, soil respiration, and enzyme activity, samples were thawed in a refrigerator at around 7°C. The chloroform fumigation extraction (CFE) method was used to quantify the content of C and N in microbial biomass (MBC and MBN) (Vance, Brookes, and Jenkinson (1987)). For a detailed description, see Klein‐Raufhake et al. (2024). The filtrates of the nonfumigated extracts were used for the determination of extractable organic carbon (EOC) and extractable TN (EN) with a TOC/TN analyzer (Total Organic Carbon Analyzer TOC‐L CPH/CPN, Shimadzu Corporation, Japan) and the determination of inorganic N‐forms (NH₄ and NO₃) with a photometric approach (Skalar Analytik GmbH, Germany).

The CFE method was also used to determine phosphate in microbial biomass (MBP). According to Chen and He (2004), 0.03 M NH₄F and 0.025 M HCl were used as extractants. Phosphate (PO₄‐P) was measured with a continuous‐flow autoanalyzer at 880 nm (Skalar Analytik GmbH, Germany) and total phosphorus (TP) with ICP‐OES (Spectro Analytical Instruments, Germany) in the nonfumigated extracts.

Soil respiration was determined during 2 weeks of incubation at 20°C by measuring CO₂ evolution (SRI 8610c gas chromatograph; SRI Instruments, USA). The proportion of respired CO₂‐C from the total SOC pool was calculated (CO₂/SOC) as well as the metabolic quotient qCO₂ (CO₂‐C/MBC per hour).

The activity of four extracellular hydrolytic enzymes (EHEs) was determined: ß‐glucosidase (BG), acid phosphomonoesterase (AP), N‐acetylglucosaminidase (NAG), and xylanase (XYL). Fluorescently labeled (4‐MUF) substrates were used (e.g., German et al. (2011)). Tetramethylbenzidine (TMB) was used to measure the activities of phenoloxidases and peroxidases. For a detailed description, see Klein‐Raufhake et al. (2024). Specific enzyme activities were calculated by dividing enzyme activities by the SOC content.

2.5. Statistical Analysis

All statistical analyses were performed in R (R Core Team ²⁰²⁴, version 4.3.3). For graphical visualization, we used the packages ggplot2 (Wickham 2016), ggfortify (Horikoshi and Tang 2018), ggdag (Barrett 2024), ggally (Schloerke et al. 2024), and multcompView (Graves et al. 2024). Differences between study sites were identified using one‐way ANOVAs followed by the post hoc Tukey HSD (functions aov and TukeyHSD from R package stats; R Core Team ²⁰²⁴).

We analyzed the effects of forest management (the ForMI and its components indirectly derived: the proportion of native deciduous trees, timber extraction, and natural dead wood volume) and abiotic site properties (soil texture and soil pH of 30–60 cm depth) as fixed factors on a range of soil parameters using linear mixed‐effect models with the R function lmer from R package lme4 (Bates et al. 2015). The study site was included as a random factor and the respective soil parameter as a response variable (with n = 200). Separate models were also calculated for each subset of study sites (with n = 50 each). As some plots in the base poor lowlands have been plowed in the past, plowing was included as random factor in the models of this study site. All response variables were scaled and centered prior to analysis and log‐transformed within the models. To test whether model requirements were met, we inspected the residuals using the function simulateResiduals from the DHARMa package (Hartig 2022). Effect sizes were estimated by calculating 95% confidence intervals around the mean value. The upper and lower limits of these intervals were determined using the formula: mean ± (1.96 × SE), where the mean is the estimated mean and SE is the standard error derived from the linear models.

To quantify the relative contribution of the individual predictors (forest management components and abiotic site properties) to the explainable variance in our soil parameters, we used linear mixed‐effects models (lmer function, lme4 package; Bates et al. ²⁰¹⁵) in which we included all predictors as fixed effects and the study area as a random effect. We then conducted a Type III ANOVA (anova function, stats package; R Core Team ²⁰²⁴) to assess the unique contribution of each predictor while controlling for others. The proportion of variance explained by each predictor was calculated by dividing its sum of squares by the total sum of squares, which is a standardized measure of the importance of each predictor in explaining the observed variability (Legendre and Legendre 2012).

To disentangle the effects of the three management components, we tested a conceptual model with all predictors on a set of five key response variables reflecting nutrient levels, microbial quantity, quality, and activity using individual piecewise structural equation modeling (pSEM; function psem) in the piecewiseSEM package (Lefcheck 2016). We fitted separate linear mixed‐effects models for each of the paths in our conceptual model and used pSEM to merge the models and hypotheses into one single model. Exogeneous explanatory variables that were considered in the model are the proportions of native deciduous trees, natural dead wood, and timber extraction. Endogenous response variables are C/N ratio, MBC, MBC/MBN, TP, and enzyme activities (EA). The latter variable EA results from the PC1 of a principal component analysis (PCA), using the prcomp function from the stats package (R Core Team ²⁰²⁴) over all scaled and centered six specific enzyme activities analyzed (Figure S1). Humus form and soil pH of 0–5 cm depth were incorporated as mediating explanatory variables that might link the management variable to the soil parameters investigated. Within the pSEM, a separate linear mixed effects model (lmer function, lme4 package; Bates et al. ²⁰¹⁵) was calculated for each endogenous variable (as response variable), with all exogenous and mediator variables as fixed factors. All endogenous variables were log‐transformed. As pSEM is designed to model acyclic relationships, the model for the soil pH value also includes the humus form as an explanatory variable, but the model with the humus form as response variable does not include the pH value. The interactions between pH and humus form were considered by assuming d‐connectedness (see below). In all models, the study area was included as a random factor, which allows us to determine how much of the explained variance was due to fixed and random effects. The variance explained solely by the fixed factors is referred to as R² marginal (R²m), while the variance explained by both the fixed and random effects is referred to as R² conditional (R²c). Except for MBC/MBN and MBC, all endogenous variables are also considered d‐connected, that is, we assume direct effects between these variables. The global goodness of fit of the pSEM was assessed using Fisher's C and its p‐value as described by Lefcheck (2021). Since in the global model across all study sites marginal and conditional R ² deviated strongly from each other and thus implied strong effects of the random factor (the study site), we also tested the conceptual model separately for all four study sites. The variable reduction of the specific enzyme activities via the PCA was also performed with the respective subsets per study site (Figure S1). Scores of the base‐rich lowland and the base‐rich low mountain forests were multiplied by −1 to ensure that increasing PCA scores correspond to increasing specific enzyme activities, making results more intuitive and comparable across regions. In order to remove the influence of plowed plots, plowing was included as random factor in the pSEMs of the base‐poor lowland forest.

To test the suitability of humus form as a predictor for soil parameters, linear mixed‐effect models were calculated with the respective soil property as response variable, humus form as fixed factor, and study site as a random factor (lmer function, lme4 package; Bates et al. ²⁰¹⁵).

3. Results

3.1. Analysis Across all Study Sites

Across all study sites, the range of forest management intensity (ForMI) and its components was quite similar (Figure 1d). Only timber extraction was significantly highest in the base‐poor low mountain site (Figure 1d). Across all study sites (n = 200) we found significant effects of the ForMI in 15 of the 20 investigated parameters (Figure 2a). With increasing ForMI, litter turnover rate decreased, reflected in thicker organic layers; pH values declined; and C/N, EOC/EN, and MBC/MBN ratios got wider. Microbial growth decreased with increasing ForMI, as visible in both lower MBC and MBC/SOC ratios. While we found no difference in total extractable phosphorous (TP) and plant‐available phosphate (PO₄), phosphorous in the microbial biomass (MBP) decreased. The metabolic quotient (qCO₂) was significantly higher with increasing ForMI, and five out of six specific enzyme activities were significantly lower.

FIGURE 2.

Effect of (a) forest management intensity (ForMI), (b) proportion of deciduous trees, (c) timber harvest volume, and (d) natural dead wood proportion on topsoil properties (0–5 cm). Shown are the effect sizes, that is, linear model estimates (mean ± 1.96 SE) for the effect of the predictors on the investigated soil parameters. The latter were measured on 200 plots across four study sites. Differences between the study sites were taken into account by considering them as a random factor in the models. The points represent the estimate of the model effect of the measured soil parameter, and the lines reflect their 95% confidence intervals. Filled points show a significant effect, and hollow points show no effect based on ANOVA results.

The proportion of native deciduous trees in the stand affected the same soil parameters, but in opposite direction compared to ForMI (Figure 2b). Additionally, a positive relationship was found between native deciduous tree proportion and plant available nitrate (NO₃). The timber harvest volume was significantly related to eight of our analyzed parameters (Figure 2c). An increasing timber harvest volume led to significantly lower pH values, lower MBC/SOC, and lower specific enzyme activities in four of six enzymes measured, whereby the other two enzyme activities also tended to be lower. At the same time, the TP and PO₄ content increased with increasing timber harvest volume. The proportion of naturally occurring deadwood did not affect the analyzed soil parameters (Figure 2d). To disentangle the effects of the three ForMI components, we used variance partitioning and structural equation modeling (see below).

The variance analysis illustrates that of all three forest management variables, the proportion of native deciduous trees explained the largest share of the variance in the measured variables in mineral topsoil, with proportions of more than 25% explainable variance in 11 of 21 variables and over 50% explainable variance in MBC/MBN and MBP (Figure 4a). The timber harvest volume also had a relatively large influence, especially on TP and PO₄ contents and most specific enzyme activities. The influence of natural deadwood volume was almost negligible, except on nitrate levels, CO₂/SOC, and the metabolic quotient qCO₂. The soil pH value from 30 to 60 cm mineral soil depth explained more than 25% of the explainable variance in 11 of 21 cases and particularly a large proportion of variation in topsoil pH, ammonium contents, EOC/EN, TP, and PO₄‐P contents and certain specific enzyme activities. The soil texture explained a considerable proportion of the variance in SOC, EOC, MBC, and NO₃‐N contents, the C/N ratio, the metabolic quotient, and the specific enzyme activities of XYL. For an illustration of the effect direction of pH and soil texture on all parameters, see Figure 3a,b.

FIGURE 4.

(a) Relative contribution of the predictors to the variance in response variables, calculated using linear mixed‐effects models and Type III ANOVA. Bars represent the proportion of variance explained by each predictor, derived from the ratio of its sum of squares to the total sum of squares. (b) Diagram of the piecewise structural equation model (pSEM) for all 200 plots, linking forest management measures to soil ecological processes. Exogeneous variables that were considered in the model are the proportions of deciduous trees, natural dead wood, and timber extraction. Mediator variables are humus form and soil pH (0–5 cm mineral topsoil). Endogenous variables are C/N ratio, MBC, MBC/MBN, TP, and enzyme activities (EA), all measured in 0–5 cm mineral topsoil. Within the pSEM, a separate lmer model was calculated for each endogenous variable (as response variable), with all exogenous and mediator variables as fixed factors and study area as a random factor. All endogenous variables were log‐transformed. In the model for humus form as response variable, only exogeneous variables were included as fixed factors but for soil pH as response variable, additionally, the humus form was included as fixed factor. Except for MBC/MBN and MBC, all endogenous variables are considered d‐connected, that is, we assume direct effects between these variables. Only significant paths are shown. The color represents the effect direction (green = positive, purple = negative), and the transparency represents the effect strength (the darker, the greater).

FIGURE 3.

Effect of (a) soil texture (ordinally scaled from coarse to fine grained) and (b) subsoil pH (measured in KCl in 30–60 cm mineral topsoil depth) (b) on soil properties. Shown are the effect sizes, that is, linear mixed model estimates (mean ± 1.96 SE) for the effect of the predictors on the investigated soil parameters. The latter were measured on 200 plots across four study sites. Differences between the study sites were taken into account by considering them as a random factor in the models. The points represent the estimate of the model effect of the measured soil parameter, and the lines reflect their 95% confidence intervals. Filled points show a significant effect, and hollow points show no effect based on ANOVA results.

Structural equation modeling was used to disentangle the relationships between individual forest management components and soil biological processes. A visualization of the conceptual model can be found in Figure S11 and coefficients of all paths are considered in Table S2. The pSEM showed that tree species selection was the forest management component that had the most significant effect on soil biological processes. These effects were strongly mediated by soil pH (0–5 cm mineral topsoil) and notably the humus form (Figure 5). Direct effects of the proportion of native deciduous trees could only be detected on the C/N ratio and the enzyme activities (narrower/higher with increasing share). Positive effects mediated via the humus form were found on MBC and TP, and negative effects on C/N, MBC/MBN, and enzyme activities. The timber harvest intensity had an impact on the topsoil parameters, particularly via its effect on the pH value. The higher the timber harvest volume, the higher the TP content and the lower the pH value as well as specific enzyme activities. The pH value, in which 19% of its variance could be explained by the fixed factors of the model, had a significant negative effect on the C/N ratios in the soil and the microbial biomass, as well as on the TP content in the soil. The volume of natural dead wood showed only one significant correlation in the global model with a positive effect on the C/N ratio. The large differences between R ² m and R ² c (marginal and conditional R ²) showed that there were strong effects of the random factor study site on the parameters under consideration. Therefore, we calculated the structural equation model also for the individual study areas.

FIGURE 5.

Diagrams of the piecewise structural equation model (pSEM) for each of the four study sites (a–d). All paths of the conceptual model (Figure S10) are included in the model, but only the significant paths are shown. In the pSEM of the base‐poor lowlands (a), plowing was included as a random factor. Specific estimates of all paths can be taken from the model results in Tables S2–S6. The color represents the effect direction (green = positive, purple = negative), and the transparency represents the effect strength (the darker, the greater).

3.2. Effects Specific to the Study Area

The SEMs for the individual study sites differed considerably in their significant paths of influence. In the base‐poor lowland forest, the explainable variance was rather low, and only one significant path was detected for each ForMI component (Figure 5a). In the base‐rich lowlands, the model explained a relatively high proportion of the variance. The proportion of native deciduous trees with positive effects on soil pH and humus form was the most important component affecting soil parameters. These effects, however, were mainly mediated via the positive influence on the humus form and the pH value. Additionally, the proportion of natural dead wood had a significant positive effect on the C/N ratio (Figure 5b). The proportion of native deciduous trees induced strong direct and indirect effects via humus form in the base‐poor low mountain forest with direct and indirect effects via humus form (Figure 5c). In the base‐rich low mountain forest, however, the influence of the timber harvest volume was far more decisive (Figure 5d). The latter had a significantly negative effect on the humus form and enzyme activities and a positive influence on TP contents. Effects of the proportion of natural deadwood and native deciduous trees on soil response variables were only indirect via the humus form and its influence on the pH value of the topsoil. The coefficients of all paths considered in the models can be found in the Appendix in Tables S3–S6.

Overall, stronger effects of the ForMI and its components were visible in the low mountain regions (see Figures S4–S7).

3.3. Humus Form as Predictor for Soil Properties

The higher the proportion of native deciduous trees in the stand, the faster litter decomposition was, as reflected in a thinner organic layer and a more frequent occurrence of the humus forms moder and mull. Apart from single outliers, mull was only found in stands with > 80% deciduous trees (Figure 6a). In 12 of 22 cases, we were able to demonstrate a significant correlation between humus form and mineral topsoil properties (Figure 6b, Figure S11): In particular, soil acidity (R ²m = 0.14), C/N ratio (R ²m = 0.33), and the microbial abundance (MBC, R ²m = 0.11), quality (MBC/MBN, R ²m = 0.14), and activity (MBC/SOC, R ²m = 0.3) were well represented by the humus form. However, soil phosphorous contents (TP and PO4) and specific enzyme activities (apart from the XYL, see Figure S11) could not be estimated with humus form as single information.

FIGURE 6.

(a) Correlation between the proportion of native deciduous trees in the stand and humus form. Humus forms were grouped into three main types. (b) Correlation between humus form and a selection of soil response variables. The relationship was analyzed using linear mixed‐effect models with study site as a random factor. Significant correlations are represented by solid lines, while nonsignificant ones are shown with dashed lines. Dots depict the correlation between the proportion of deciduous trees and the response variables, with colors ranging from purple (pure conifer stand) to light green (pure deciduous tree stand). The Pearson correlation coefficients for the relationship between the proportion of deciduous trees and soil parameters are provided in Table S5 in the Appendix. For a visual representation of the remaining analyzed soil parameters, see Appendix Figure S11.

4. Discussion

Across all study sites, we could show clear effects of forest management intensity, expressed as the ForMI index, on most of the soil properties measured (Figure 2a). High management intensity was associated with an accumulation of organic matter in the humus layer, acidification of the topsoil, and generally slower metabolic processes, as indicated by the overall lower specific enzyme activities and the relatively lower nitrogen contents (increased C/N, EOC/EN, and MBC/MBN) in intensively managed stands. The overall fewer (lower MBC) and less active (lower MBC/SOC and specific enzyme activities) microorganisms respired more CO₂ (higher qCO₂) than they used to build microbial biomass, possibly indicating stress (Anderson and Domsch 1993). Our results are contradictory to those of Felipe‐Lucia et al. (2020), who found higher enzyme activities with increasing ForMI, and those of Richter et al. (2018), who detected only minor impacts of ForMI on microbial community structure and function.

The effects of increasing ForMI on soil properties were largely shaped by the dominant tree species (compare Figures 2a,b and 4a), which is in line with previous findings (Richter et al. 2018). In conifer stands (high ForMI), litter decomposition was significantly reduced, resulting in thicker organic layers. This was largely due to the chemical composition of conifer needles, which contain less N in relation to C, are tougher to decompose, and possess inhibitors that further slow microbial activity (Prescott 2010).

In contrast to Gan et al. (2020), we detected effects of forest management not only through tree species composition but also through harvest volume. With increasing timber harvest volume, topsoil acidification proceeded and specific enzyme activities decreased (Figure 2c). The former contradicts the findings of Clarke et al. (2021), who found no effect of timber harvesting on mineral soil pH but the latter is consistent with the findings by Hassett and Zak (2005). Unlike Clarke et al. (2021) and Vos et al. (2023), we found increasing TP and PO₄ contents with increasing timber harvest volume, which can be explained by the fact that fewer trees in the stand consume less phosphorus, which remains in the soil. Consequently, strong investment in the production of acid phosphatase is not necessary. In terms of TP and PO₄ contents as well as most enzyme activities, the effect of timber harvest volume was as great or even greater than that of the proportion of native deciduous trees (Figure 4a). Contrary to our expectations and existing studies (Hassett and Zak 2005; von Wilpert ²⁰²²), we found no changes in humus forms, or decreased microbial biomass caused by increased timber harvesting. A significant influence of the natural deadwood volume on soil properties was also not detected (Figures 2d and 4a). Whether the deadwood is of natural origin (as proposed by ForMI) or comes from a sawn‐off branch or stump is probably irrelevant to soil processes and nutrient recycling which is in sharp contrast to aboveground biodiversity (Lassauce et al. 2011; Seibold et al. ^{2016, 2021}). However, the analysis of the total deadwood volume on the soil properties also showed only minor effects: Nitrate and MBC contents, as well as specific activity of the acid phosphatase (AP/SOC), increased with increasing deadwood volume (data not shown).

A finer soil texture was associated with a more decomposed organic layer, soil carbon enrichment (higher SOC, EOC, and MBC contents and narrower C/N), and lower XYL enzyme activities (XYL/SOC). Our results are in line with previous findings, showing increased C mineralization in the forest floor (Côté et al. 2000) and higher MBC contents in the mineral soil (Bauhus, Paré, and Côté ¹⁹⁹⁸) of more clayey soils. Unexpected was the extremely high explanatory power of soil texture in these variables, particularly with more than 70% in NO₃ and specific XYL activity (XYL/SOC) (Figure 4a). One explanation for this could be that in sandy soils with many coarse pores nitrate is leached more quickly (Gaines and Gaines 1994). While, based on PLFA analyses, Richter et al. (2018) concluded that soil texture was the factor that strongly influenced the microbial community, we found no significant effects on the MBC/MBN ratio, which is often used to assess the ratio of bacteria to fungi. In addition to microbial community structure, reduced substrate availability, such as hemicellulose availability, could be a driver of reduced XYL production by soil microbes in clay‐rich soils (Tischer et al. 2019). The pH of the subsoil (30–60 cm) had an influence on most of the measured soil parameters (Figure 3b) and explained about as much of the total variance as the proportion of native deciduous trees for the majority of variables (Figure 4a). In particular, the topsoil pH, NH₄, TP, and PO₄ content and specific enzyme activities of BG and peroxidases were to more than 50% explained by the pH value of the subsoil, reflecting a major impact of bedrock chemistry on soil biological processes.

Our structural equation model indicated by the number of significant paths and the strength of their effects that the proportion of native deciduous trees was the most important management component by which foresters can influence soil biological processes. With an increasing proportion of native deciduous trees, topsoil properties are affected in a positive way both directly and, to a greater extent, indirectly via the positive effects on pH and, in particular, humus form (Figure 4b), which is in line with the findings by Andreetta et al. (2016) and Bayranvand et al. (2021). The volume of timber harvested and the associated nutrient and base cation extraction also had a direct effect on the considered soil response variables, as well as an indirect effect via a reduction in pH value. Despite the large influence of regional environmental factors on the decomposition of deadwood as described by Edelmann et al. (2023), our pSEM showed that an increasing proportion of natural deadwood led to a wider C/N ratio in the upper mineral topsoil, independent of site‐specific effects (R ²m = 0.44). Considering all pathways in the SEM, including the second stage with the mediating variables, the humus form had the greatest effect strength on soil properties (i.e., the highest explanatory value). However, only 6% of the variance in humus form was explained by management factors. The large difference between the marginal and conditional R ² of the response variables, particularly not only for humus form but also for MBC, MBC/MBN, and TP, suggests that the overall patterns observed are strongly modified by region‐specific variation in abiotic environmental settings.

For instance, in the base‐poor lowlands, the overall explanatory power of the management components was rather low, as indirect effects via pH and humus form were absent (Figure 5a), probably due to high nitrogen depositions in this region (LANUV ²⁰²⁴) that may obscure management impacts. Former plowing still had a strong influence on soil C/N ratio and MBC/MBN here, as indicated by the difference between R ²m and R ²c, but almost no effect on MBC, TP, and enzyme activities. In three of four regions, there was a significant yet modest impact of natural deadwood volume on soil parameters (Figure 5a,b,d). A significant effect of increased timber harvesting was only detected in the pSEMs for the base‐poor lowland and the base‐rich low mountain forest (Figure 5a,d). In the latter, high proportions of timber extraction decreased soil pH, microbial growth (reflected in lower MBC/SOC), and specific enzyme activities, especially of enzymes involved in the P‐ and N‐cycle, but increased the amount of carbon and phosphorus fractions (Figure S6). Against our hypothesis, we found no effect of timber harvesting volume in the base‐poor low mountain site. One reason for this may be that base saturation is already so low (we found the lowest pH values in the subsoil here, Figure 1c) so that additional base cation removal by increased timber harvesting has no visible effect on the system, which is already in the aluminum buffer range. In contrast, in the base‐rich low mountain ranges, three of five significant pathways originating from the management variables are attributed to timber harvesting, which exerts both direct and indirect effects through its negative impact on the humus form on all considered response variables. The influence of the proportion of native deciduous trees and natural deadwood volume was small. In comparison, in the base‐poor lowland (Figure 5c), management influenced soil properties solely via the tree species composition. Thus, while we found a significant influence of native deciduous tree species proportion in all four study areas, we also found that the strength of this effect varied by region, which is in line with previous studies (Gan et al. 2020). Considering the R ² s, it is clear that the variance of soil biological response variables explained by management varied greatly between study regions. For example, as much as 63% of the observed variance in MBC could be explained by management in the base‐rich lowlands compared to only 17% in the base‐poor low mountain region. The ratio of MBC/MBN also ranged from only 19% explained by management in the base‐rich lowlands to 78% in the base‐rich low mountain region. However, in the pSEMs, the humus form consistently played a crucial role in transferring management effects to the mineral soil via shifts in pH and biological activity.

Humus form was closely related to the proportion of native deciduous trees in the stand (Figure 6a). In line with Ponge (2003), we found mull, a humus system particularly valued for its rapid nutrient cycling (Zanella et al. 2018), exclusively in stands dominated by deciduous trees, most obvious due to their higher litter quality (Andreetta et al. 2016). We were able to predict 12 of 22 topsoil parameters by humus form (Figure 6a; Figure S11), particularly pH (0–5 cm), C/N, and microbial characteristics (MBC, MBC/MBN, and MBC/SOC), but with varying explanatory power (R ²s ranging from below 0.1 for NO₃, NH₄, and EOC to above 0.3 for C/N and MBC/SOC). Thus, we complement the findings of Ponge, Chevalier, and Loussot (2002) who showed significant correlations with base cations (Ca, Mg, and K), cation exchange capacity, and exchange acidity. Our results clearly support the idea of Ponge and Chevalier (2006) that stocking‐regulated soil properties can be effectively evaluated by practitioners through an assessment of the humus form.

Author Contributions

Theresa Klein‐Raufhake: data curation, formal analysis, investigation, visualization, writing – original draft. Jens Wöllecke: project administration. Lea Santora: data curation, writing – review and editing. Katharina Rentemeister: data curation, writing – review and editing. Ute Hamer: conceptualization, funding acquisition, investigation, methodology, resources, supervision, writing – original draft. Lena Neuenkamp: validation, writing – review and editing. Michael Elmer: conceptualization, funding acquisition, project administration. Jens Jakob Schaper: data curation, writing – review and editing. Norbert Hölzel: conceptualization, funding acquisition, methodology, project administration, resources, supervision, writing – original draft. Michael Meyer: validation, writing – review and editing. Britta Linnemann: conceptualization, funding acquisition, project administration. Max Fornfeist: data curation, writing – review and editing.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1.

Data Availability Statement

Data will be stored on Dryad.