Introduction

Soil respiration is the second-largest terrestrial carbon (C) flux. Global soil respiration is estimated to range from 68–101 petagram C yr−1 (Jian et al. 2021), which is about ten times larger than annual C emissions from fossil fuel combustion (Bond-Lamberty and Thomson 2010). Therefore, soil respiration has a strong influence on atmospheric CO2 concentrations and soil C budgets (Davidson and Janssens 2006; Lu et al. 2021; Rustad et al. 2000). Soil respiration comprises autotrophic and heterotrophic respiration. Autotrophic respiration originates from plant roots, mycorrhizal fungi, and the rhizosphere, while heterotrophic respiration refers to the CO2 flux emitted from soil due to organic matter decomposition by soil microorganisms (Bond-Lamberty and Thomson 2010; Hanson et al. 2000). The proportion of autotrophic and heterotrophic respiration of total soil respiration vary temporally and spatially within and between forest ecosystems (14–73%; Subke et al. 2006).

Abiotic and biotic factors drive temporal and spatial variability in soil respiration (Bond-Lamberty et al. 2018; Raich and Schlesinger 1992; Rubio and Detto 2017). The main abiotic factors that influence soil respiration rates are soil temperature (Bond-Lamberty et al. 2018; Bond-Lamberty and Thomson 2010) and soil water content (Davidson et al. 1998; Reichstein et al. 2003). Many studies report a positive exponential relationship between soil temperature and soil respiration (e.g., Davidson et al. 1998). The effect of soil water content on soil respiration is complex and inconsistent across temperature ranges. Some studies have shown a parabolic relationship between soil respiration and soil water content (Davidson et al. 1998; Fischer et al. 2024; Gutierrez del Arroyo and Wood 2021). Soil water content affects soil respiration by modifying the connectivity of soil pores, which controls gas diffusivity (Millington 1959). Soil substrate quantity and quality, such as soil organic C and nitrogen (N) concentration and soil C:N ratio, are also factors driving soil respiration rates in forest ecosystems (Ngao et al. 2012; Saiz et al. 2006).

Biotic factors, such as vegetation type and restoration strategies (i.e., passive (natural regeneration) versus active (e.g., tree planting), Brancalion and Holl 2020) are also important determinants of soil respiration because they control the quantity and quality (chemical composition) of litterfall and root production, and mediate the soil microclimate (Raich and Tufekciogul 2000; Sayer et al. 2011) and soil microbial activity (Chen et al. 2019). For example, rapid stand development in planted forests may produce higher fine root biomass, leading to higher autotrophic respiration (Wei et al. 2022). Further, differences in forest characteristics between natural succession and plantations, such as density, species richness and soil properties, may result in differences in litter production and carbon sequestration (Cifuentes-Croquevielle et al. 2020; Hardiman et al. 2013). Tree density can also differ between forest types, with higher density in planted than naturally regenerating forests (Hickey et al. 2019; Machado et al. 2021), partly explaining higher litterfall and litterfall nutrient inputs (Machado et al. 2021; Williams-Linera et al. 2021). Studies comparing planted and naturally regenerating forests have mainly focused on aboveground C sequestration (Bukoski et al. 2022). Yet, the understanding of how changes in biotic and abiotic conditions associated with different reforestation strategies affect litterfall C and soil respiration is limited.

Studies have shown that forest regrowth (i.e., secondary or successional forests) is largely defined by land-use history (Cramer et al. 2008; Jakovac et al. 2021; Nytch et al. 2023). Land-use history (e.g., old pastoral fields, abandoned crop fields, logging) determines soil nutrient conditions, physical properties, and soil microbial communities (Delcourt et al. 2022; Felzer 2023; Peddle et al. 2024; Standish et al. 2006). For example, the fertilization history of old fields (i.e., ecosystems that form on previously human-managed land after management has ceased, Clark et al. 2019) can affect the N cycles of forest regrowth (Felzer 2023; Richter et al. 2000). Empirical analysis has shown that current soil organic carbon stocks are related to historical land use (Schulp and Verburg 2009). Under reforested sites in the northeast United States, soil carbon stocks are driven by interactions between individual site histories (e.g., disturbance, past agricultural management practices) and place-based factors (e.g., physiography, soil type) (Nave et al. 2024).

In New Zealand, natural succession on abandoned pastures or marginal land is often dominated by Leptospermum scoparium (mānuka; Myrtaceae) (Stephens et al. 2005; Wardle 1991). Naturally regenerating forests dominated by L. scoparium also occur on infertile soils formerly dominated by conifers such as Agathis australis (Enright 1989; Esler and Astridge 1974). Leptospermum scoparium is also commonly planted in forest restoration programmes. Naturally regenerating and planted forests of L. scoparium have environmental and social benefits, particularly in minimizing soil erosion (Ross et al. 2009) and in sequestering C (Carswell et al. 2009; Scott et al. 2000).

In this study, we measured C fluxes (litterfall and soil respiration) at locations that differed in land-use history and forest restoration strategy (here: planted and naturally regenerating L. scoparium forests). The specific objectives were: (1) to quantify the C input from litterfall, (2) to measure total soil respiration and estimate its component fluxes, and (3) to identify abiotic and biotic factors influencing litterfall C and total soil respiration. Our hypotheses were: (1) litterfall C input increases with increasing soil nutrient concentrations, (2) total soil respiration increases with increasing litterfall input, and (3) soil water content mediates the effect of soil temperature on total soil respiration.

Materials and methods

Study area

The study was conducted at the end of the Whangaparāoa peninsula (Shakespear Regional Park; 377 ha; 36°36′S, 174°49.5′E), north of Auckland, New Zealand (Supplementary Material, Figure S1). The annual average temperature of the area is 15–16 °C and the annual total rainfall is 900–1000 mm (Chappell 2014). The underlying bedrock consists of layered yellow sandstone and softer grey mudstone (both Waitematā Group) (Auckland Regional Council 1991, 2003). The predominant soil type is yellow ultic soils (New Zealand Soil Classification, Hewitt 2010), which are characterised by low pH and low nutrient concentrations (Hewitt 2013).

Before human settlement in the thirteenth century and subsequent deforestation, warm temperate rainforests covered the study area (Newnham et al. 2018). In the 1860s, the park was covered in shrubland of thick L. scoparium (mainly on the eastern tip), and fern (primarily Pteridium esculentum on the sheltered southern side) (Beever 1981). This vegetation may have been the result of Māori (indigenous people of Aotearoa New Zealand) burning for cultivation (Newnham et al. 2018). Europeans subsequently cleared areas of regenerating forests for pastoral farming, which was particularly intense from 1945 to 1967. Whangaparāoa (Shakespear Regional Park) was nearly entirely deforested when it was purchased as a reserve in 1967, except for a few remnants of mature, warm temperate rainforest in gullies. The current vegetation is dominated by grazed pastures, gully forest remnants, naturally regenerating forests, and planted forests of native trees for restoration. Restoration plantings, mainly of L. scoparium and Kunzea ericoides, started in 1967 with 2000–5000 plants per year until 2010. Approximately 15,000 plants (1.5 ha) per year have been planted since 2010 (Griffiths et al. 2021).

Plot selection and set-up

Four plots (20 m × 20 m) in planted forests and four plots in naturally regenerating L. scoparium forests were established. Plot locations were selected using the following criteria: plots were located within large areas of continuous forest cover (~ 20,000 m2 for planted forests and ~ 100,000 m2 for naturally regenerating forests); the proportion of L. scoparium stems were > 50%; soil types were equivalent (yellow ultic soils); and tree age was similar (Table 1). The planted L. scoparium forests were all established on former pasture. Information on the management of pastures is lacking but nitrogen and phosphorus fertilisation is common regionally especially on low fertility ultic soils (Hewitt et al. 2021). Naturally regenerating L. scoparium forests are often found where agriculture (pasture) was attempted but not viable. The land use immediately before the development of the naturally regenerating forests is unclear, except that the trees are of a similar age to the planted sites, suggesting the areas were all free of substantial woody growth c. 30 years ago. Throughout the paper we use the terms planted and naturally regenerating forests but acknowledge that differences in site history mean that our results are not solely due to differences in restoration strategies.

Table 1 Vegetation and soil characteristics of planted and naturally regenerating L. scoparium forests at Whangaparāoa, New Zealand
Full size table

Plot set-up and survey followed the Permanent Plot method for measuring and describing New Zealand vegetation (Hurst et al. 2022). Within each plot, measurements and sample collection (n = 5 locations per plot) were conducted at centre of each plot and the four corner points of the 10 m × 10 m core area (Figure S1c).

To age the trees, discs from immediately above the buttress were taken from four trees adjacent to the plot. To measure the likely age of establishment, trees within the 80th and 90th percentile (diameter at breast height) of those found within the plot were selected. Discs were oven-dried (40 ℃) and sanded to a fine polish to clearly show the individual tree rings. Tree rings were counted across three cross-sections per disc using a low-power binocular microscope.

Organic layer and mineral soil characteristics

Within each plot, organic layer (i.e., forest floor; including organic and litter layer) samples and mineral soil samples (Figure S1c) were taken in September 2020. Organic layer samples were collected using a 50 cm × 50 cm quadrat. After removal of the organic layer, mineral soil samples (0–10, 10–20 and 20–30 cm depth) were taken using a soil auger (3.5 cm in diameter). In addition, five mineral soil samples (0–10 cm depth) were collected for bulk density analysis using a PVC cylinder (5 cm in radius, 10 cm in height). Bulk density was determined by drying undisturbed soil samples at 105 °C for 72 h and then weighing them.

Organic layer samples were dried at 60 °C to constant weight and weighed. Organic layer pH measurement was conducted by mixing 3 g of organic layer samples with 30 ml of distilled water. Mineral soil samples were oven dried (40 ℃) for 72 h, roots and organic material removed, and the samples sieved at 2 mm. Then, 10 g of processed mineral soil was mixed with 25 ml of distilled water for the pH measurement, which was conducted using a portable pH meter (SensION + , Hach Company, Loveland, CO, US).

To determine total C and N concentrations, organic layer and mineral soil was ground into a fine powder using an Ultra Centrifugal Mill (ZM 200, Retsch GmbH, Haan, Germany) and 5 mg were analysed using an elemental analyser (Vario El Cube Elementar Analysensysteme, Langenselbold, Germany). The combustion and reduction temperature were 950 °C and 600 °C, respectively. Of these, 10% of samples were measured twice for analytical replication. The analytical error was < 1%.

Mineral soil samples were extracted for dissolved C (HWEC) and organic N (HWEON) using the hot water extraction method developed by Ghani et al. (2003). 3 g of each sample was transferred into a 50 ml test tube, and 30 ml of distilled water was added. The tube was placed on a vortex shaker for 10 s and then in a hot water bath at 80 ℃ for 16 h. The extract was centrifuged at 3500 rpm for 20 min and filtered through a 0.45 m nitrocellulose membrane filter. The filtrate was analysed for C on a TOC analyser (Sievers Innovox TOC, GE Instruments). The samples were diluted for N. Inorganic N was analysed by flow injection analysis (QC8500, Lachat). Total N was determined by potassium persulfate digestion followed by flow injection analysis (QC8500, Lachat).

Olsen-available phosphorus (P) was determined by extracting 5 g soil (1:20 ratio) with 0.5 M sodium hydrogen carbonate solution adjusted to pH 8.5 for 30 min at 25 °C (Blakemore et al. 1987). Samples were filtered, and PO4-P was analysed using a QuikChem 8500 flow injection analyser.

Litterfall

Litterfall was collected from five litter traps (Garden PE Pop-up Bag, 63 cm in diameter, 30 cm in height) within each plot (Figure S1c). The lower part of the trap was 50 cm off the ground. Litterfall was collected every four weeks from October 2020 to September 2021. Each litterfall sample was dried at 60 °C to constant mass and weighed. Litterfall collected in October 2020 (austral spring), January 2021 (austral summer), April 2021 (austral autumn), and August 2021 (austral winter) was sorted into four fractions (leaf, twig and bark, reproductive material of L. scoparium, and other (non-L. scoparium) material) and weighed. Total C and N concentrations in these four litterfall fractions were determined by grinding oven-dry material into fine powder using an Ultra Centrifugal Mill (ZM 200, Retsch GmbH, Haan, Germany), 5 mg of which was combusted at 950 °C using an elemental analyser (Vario El Cube Elementar Analysensysteme, Langenselbold, Germany). Around 10% of samples were measured twice. The analytical error was < 1%.

Total litterfall input was calculated by dividing total dry biomass (g) from each trap by the number of days between two litter collection dates to determine the litterfall input per day per m2. Litterfall biomass was converted into litterfall C using fraction-specific C concentrations for a given season. Since non-L. scoparium litter consisted mainly of Kunzea ericoides (kānuka) and Metrosideros excelsa (pōhutukawa) leaf litter, C concentrations of these species were used (K. ericoides, Schwendenmann and Mitchell 2014; M. excelsa, Nguyen, unpublished data).

Root biomass

Five mineral soil samples (0–10 cm depth) were collected within each plot using PVC cylinders (5 cm in radius, 10 cm in height) to quantify root (< 10 mm) biomass. Roots were then separated from the soil by washing the samples through a series of sieves (2 mm, 1 mm, 250 µm) with running tap water. Roots retained by the sieves were collected with tweezers. Clean roots were oven-dried at 60 °C to constant mass and weighed.

Total soil respiration and its component fluxes, soil temperature and soil water content

Within each plot, five PVC collars (5 cm in radius, 10 cm in height) were inserted into the ground to approximately 2–5 cm depth and left in place for the duration of the study (October 2020 to August 2021) (Figure S1c). Soil respiration was measured with a portable CO2 gas analyser (EGM-5, PP Systems, Amesbury, MA, USA) connected to an SRC-2 soil respiration chamber (PP Systems, Amesbury, MA, USA). Soil respiration was measured between 10:00 AM and 3:00 PM (defined as time) every four weeks (defined as month). For each measurement, the change in CO2 concentration over 120 s was recorded, and total soil respiration (μmol m−2 s−1) was calculated as follows:

$$\text{Soil}\;\text{respiration}\;\left(\mathrm\mu\text{mol}\;\text{m}^{-2}\text{s}^{-1}\right)=\left({\mathrm\Delta\text{CO}}_2/\mathrm\Delta\text{t}\right)\times\left(\textit{P}\times\textit{V}\right)/\left(\textit{R}\times\textit{T}\times\textit{A}\right)$$
(1)

where ΔCO2 is the change in CO2 concentration (μmol mol−1) over time (t in s), calculated as the slope of the linear regression; P is the atmospheric pressure (Pa), V is the volume of the chamber including PVC tube (m3), R is the universal gas constant (8.314 m3 Pa K−1 mol−1), T is the temperature (K), and A is the surface area of ground covered by the chamber (0.007854 m2).

The root biomass regression method was used to estimate the contribution of heterotrophic and autotrophic respiration to total soil respiration (Kucera and Kirkham 1971; Kuzyakov 2006). The heterotrophic respiration was derived by plotting total soil respiration (dependent variable) versus root biomass (roots < 10 mm, independent variable). The y-intercept of the linear regression between root biomass and total soil respiration estimates respiration in the absence of roots (i.e., heterotrophic respiration). Autotrophic respiration was then calculated by subtracting the heterotrophic respiration from total soil respiration. The root biomass regression method is based on the assumption that heterotrophic respiration is independent of root biomass. However, heterotrophic respiration is influenced by root exudates and turnover (Kuzyakov et al. 2000). Thus, the regression method may overestimate autotrophic respiration with increasing root biomass. Another caveat of the regression technique is that high variation of root biomass and total soil respiration results in low regression coefficient (Kucera and Kirkham 1971).

Soil temperature and volumetric soil water content were measured next to each of the PVC collars using a soil temperature probe (Novel Ways Ltd, Hamilton, New Zealand) and soil moisture sensor (Hydrosense II, 12 cm probe, Campbell Scientific Inc., Logan, UT, USA).

Statistical analysis

All statistical analyses were performed using R (version 4.1.0) (R Core Team 2021) and RStudio (version 1.4.1717). The Shapiro–Wilk test showed that variables were normally distributed at the forest type level for a given month (p > 0.05). To determine the effect of (A) forest type and month and (B) plot (variation within each forest type) and month on litterfall C, soil respiration, soil temperature, and soil water content (response variables) we used linear mixed-effects models. For model A, forest type and month were included as fixed effects and Plot ID and time as nested random effects. For model B, plot ID and month were included as fixed effects and time as a random effect. The mixed effects models were determined using the lmer() in the lme4 package (Bates et al. 2014). T-test (significant at p < 0.05) was used to test for differences between months. The p-values were adjusted using the Bonferroni multiple-testing correction method.

Mixed-effects models were also used to identify how soil water content and soil temperature (with and without interactions) affect total soil respiration. Soil water content and soil temperature were included as fixed effects (with/without interaction term) and Plot ID nested within month and time were used as random effects. To compare models, the second-order Akaike's information criterion (AICc) values were calculated for each fitted model with the lowest AICc considered as having the highest explanatory power. Function visreg() in the visreg package (Breheny and Burchett 2017) was used to plot model predictions for best-fitting models to test the effect of the interdependent interactions of soil temperature and soil moisture content on total soil respiration.

We used principal component analysis (PCA) to explore the relationships between soil respiration, litterfall, soil, and vegetation characteristics followed by Pearson product-moment correlation test to determine the relationship between soil and vegetation variables with litterfall C and total soil respiration by forest type.

Results

Soil physical–chemical characteristics

Soil bulk density, soil C stocks, and hot water extractable C were 10%, 22%, and 45% lower in planted forests than in naturally regenerating forests (Table 1). Although total soil N stocks did not differ between them, pH, hot water extractable N, and mineral soil Olsen P were 13%, 26%, and 71% higher in planted forests than in naturally regenerating forests (Table 1). The soil C:N ratio was significantly higher in naturally regenerating forests than in planted forests (Table 1).

Litterfall C input, soil respiration, soil temperature, and soil water content

Annual litterfall C in planted forests (258 g C m−2) was more than double that in naturally regenerating forests (123 g C m−2) (Table 2); likewise monthly litterfall C in planted forests was significantly higher (Fig. 1a, Table 3). Litterfall fractions from both forest types were dominated by leaf input. Annual leaf litter C production accounted for 38% and 43% of total litterfall C in planted and naturally regenerating forests, respectively. In planted forests, reproductive litter C was the second most significant contributor (25%) followed by twig and bark input (18%). In naturally regenerating forests, the contribution of twig and bark litter C (20%) was equal to reproductive litter C (20%) (Table S1).

Table 2 Litterfall, total soil respiration, autotrophic respiration, soil temperature, soil water content, and root carbon stocks of planted versus naturally regenerating L. scoparium forests at Whangaparāoa, New Zealand. Values are means ± standard error. Statistically significant differences between planted and naturally regenerating forests are denoted with different letters (unpaired t-tests, p < 0.05)
Full size table
Fig. 1
figure 1

Monthly litterfall carbon input (a), total soil respiration (Rs) (b), soil temperature (c) and soil water content (d) in planted (PF) versus naturally regenerating (NRF) L. scoparium forests at Whangaparāoa, New Zealand. Boxes represent the 25th and 75th percentile with a horizontal line at the median. Whiskers above and below the box indicate the largest and smallest observations falling within the 10th and 90th percentile. * Indicates significant differences between PF and NRF at p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001

Full size image
Table 3 The p-values (Mixed-effects models) for litterfall, soil respiration, soil temperature, and soil water content in planted versus naturally regenerating L. scoparium forests at Whangaparāoa, New Zealand
Full size table

The mean total soil respiration rate was 28% higher in planted (2.83 μmol m−2 s−1) than in naturally regenerating forests (2.08 μmol m−2 s−1); likewise monthly soil respiration in planted forests was significantly higher, except in November (late spring) and March (late summer) (Fig. 1b, Table 3).

The y-intercepts (= heterotrophic respiration) and regression coefficients were 1.42 μmol m−2 s−1 (r2 = 0.745; p < 0.001) and 0.72 μmol m−2 s−1 (r2 = 0.455; p = 0.001) for planted and naturally regenerating forests, respectively. There was no significant difference in autotrophic respiration rates between planted (1.47 μmol m−2 s−1) and naturally regenerating forests (1.39 μmol m−2 s−1) (Table 2). However, the percent contribution of autotrophic respiration to total soil respiration was 13% lower in planted than in naturally regenerating forests (Table 2).

Annual mean soil temperature was lower in planted than in naturally regenerating forests, whereas no differences in soil water content were found (Tables 2 and 3).

Relationships between soil respiration, soil temperature, and soil water content

Mixed-effects models without an interaction between soil temperature and soil water content had less explanatory power (higher AICc: 640.7 and 516.3 for planted and naturally regenerating forests, respectively; Table 4) than models with an interaction term (593.4 and 513.4 for planted and naturally regenerating forests, respectively; Table 4). Total soil respiration was controlled by the significant interaction between soil temperature and soil water content (Table 4; Fig. 2). In planted forests total soil respiration increased with increasing temperature (r2 = 0.62, p < 0.001) when soil water content was > 36%. Under drier conditions (soil water content < 22%), total soil respiration decreased with increasing temperature (r2 = 0.32, p = 0.03) (Fig. 2a). A different pattern was observed in naturally regenerating forests (Fig. 2b). A weak positive relationship was found between soil temperature and total soil respiration at medium soil water content (14–35%). A strong positive relationship between soil temperature and total soil respiration (r2 = 0.63, p < 0.001) was observed at soil water contents above 36% (Fig. 2b).

Table 4 The p-values and AICc scores comparing the mixed-effects model fit with and without interaction terms between soil temperature and soil moisture measured in planted and naturally regenerating L. scoparium forests, Whangaparāoa, New Zealand
Full size table
Fig. 2
figure 2

Plots of the modelled relationship between soil respiration, soil temperature and soil water content in planted (a) and naturally regenerating (b) L. scoparium forests at Whangaparāoa, New Zealand

Full size image

Relationships between litterfall C, soil respiration, and soil characteristics

The PCA showed that organic layer and mineral soil characteristics, root C stock, and litterfall C explained 45.6% and 51.5% of the variation in total soil respiration in planted and naturally regenerating forests, respectively (Fig. 3).

Fig. 3
figure 3

Principal component analysis (PCA) biplot of litterfall, root, and soil characteristics for a. planted (PF) and b. naturally regenerating (NRF) L. scoparium forests, Whangaparāoa, New Zealand. The black arrows show the direction and magnitude of the variables, which were total soil respiration (Rs, µmol m−2 s−2), soil temperature (Ts, °C), soil water content (SWC, %), litterfall carbon stock (LFCstock, kg Cm−2 yr−1), organic layer carbon stock (OLCstock, kg C m−2), organic layer pH (OLpH), organic layer C:N ratio (OLCNRatio), bulk density (BD, kg m−3), mineral soil pH (MSpH), mineral soil carbon stock (MSCstock, kg C m−2), mineral soil available phosphorus concentration (OlsenP, mg kg−1), mineral soil hot water extractable carbon (HWEC, mg kg−1), mineral soil C:N ratio (MSCNRatio), root carbon stock (RootCStock, kg C m−2)

Full size image

Pearson product-moment correlation test revealed that litterfall C in planted forests was positively correlated with mineral soil Olsen P and mineral soil pH (Table 5). The only variable associated with total soil respiration was root C stock (Table 5). In naturally regenerating forests, litterfall C was positively correlated with available C (HWEC) and nutrients (Olsen P and HWEON), and soil water content. Total soil respiration in naturally regenerating forests was positively correlated with root C stock and soil water content (Table 5).

Table 5 Pearson product-moment correlation coefficients (r) between soil and vegetation characteristics and litterfall C and total soil respiration (Rs) in planted and naturally regenerating L. scoparium forests, Whangaparāoa, New Zealand. HWEC–hot water extractable carbon; HWEON–hot water extractable organic nitrogen
Full size table

Discussion

Our study quantified litterfall and soil respiration at locations differing in land-use history and restoration strategies. Our hypotheses were: (1) litterfall C input increases with increasing soil nutrient concentrations, (2) total soil respiration increases with increasing litterfall input, and (3) soil water content mediates the effect of soil temperature on total soil respiration. We found that annual litterfall C in planted forest on former pasture was more than double that of naturally regenerating forest on land abandoned after brief use in agriculture. Annual mean soil respiration in planted forests was 28% higher than in naturally regenerating forests.

Litterfall

Annual litterfall in naturally regenerating forests at Whangaparāoa (248 g m−2 yr–1) was lower than that recorded in other naturally regenerating forests containing L. scoparium, i.e., 356 g m−2 yr–1 (Lambie and Dando 2020) and 780 g m–2 yr–1 (Wassilieff 1982). As was the case for the planted forests at Whangaparāoa (annual litterfall = 535 g m−2 yr–1) these other studies occurred in mixed forests with another, faster-growing species, Kunzea ericoides (also Myrtaceae), present which may partly explain the higher litterfall.

Litterfall biomass and litterfall C in planted was more than double that of naturally regenerating L. scoparium forests (Table 2). This is consistent with previous studies showing that active forest restoration strategies (e.g., planting) lead to higher litter production than passive restoration (Celentano et al. 2011; Gazell et al. 2012; Williams-Linera et al. 2021). In support of our first hypothesis, we found a positive correlation between mineral soil Olsen P and litterfall C in both forest types (Table 5). Phosphorus limitation of aboveground plant production has been observed across ecosystems (Hou et al. 2020). Most New Zealand soils are naturally low in P, in particular ultic soils likely without P fertilisation such as the naturally regenerating forest plots (Table 1). This may explain the higher sensitivity between mineral soil Olsen P and litterfall C across the less fertile naturally regenerating forest plots. Further, our findings suggest that litterfall C production across naturally regenerating forests was co-limited by C and N, and soil water content (Table 5). While mean soil water content did not differ between planted and naturally regenerating forests, soil water content in naturally regenerating forests tended to be lower during austral spring and summer. Studies have shown that enhanced soil moisture during summer months stimulated annual stem primary production and litterfall input (Cotrufo et al. 2011). Litterfall C across plots was also positively correlated with total stand basal area (r = 0.905, p = 0.002). Previous studies have shown that litterfall production is determined by factors such as basal area, vegetation density and species composition (Machado et al. 2021; Meentemeyer et al. 1982).

Soil respiration

Annual total soil respiration was 28% and heterotrophic respiration was 49% higher in planted than in naturally regenerating forests. Higher heterotrophic respiration could imply that planted forest soil is undergoing somewhat greater decomposition activity than the naturally regenerating forest soils. Lower litterfall C:N ratio, higher soil pH, and 3.5-fold higher mineral soil Olsen P in planted forests (Table 1, Table S1) may accelerate the decomposition of soil organic matter through stimulation of microbial activity as shown in other studies (e.g., Blagodatskaya and Anderson 1998; Fontaine et al. 2004). In contrast, higher litterfall C:N ratio and lower soil pH in naturally regenerating forests could negatively affect organic matter mineralisation, resulting in lower soil respiration rates and higher soil C stocks.

Contrary to our second hypotheses, total soil respiration did not increase with increasing litterfall C input (Table 5). Litterfall input has been shown to affect soil respiration through direct (e.g., C and nutrient input) and indirect effects (e.g., soil structure, microclimate) and positive relationships between litterfall input and soil respiration have been reported across forest ecosystems (e.g., Bowden et al. 1993; Davidson et al. 2002; Litton et al. 2007). Besides litter quantity, litter quality (e.g., lignin and nitrogen content) is an important factor driving soil respiration rates because they affect decomposition rates (e.g., Gholz et al. 2000). All litter fractions in planted forests had higher N concentrations and lower C:N ratios (Table S1) that may enhance decomposition in planted forests, resulting in higher soil respiration rates and lower organic layer and mineral soil carbon stocks.

Further, studies suggest that belowground C supply from plant roots has an important effect on soil respiration (Raich and Tufekciogul 2000; Jackson et al. 2019). We found a strong positive relationship between root C stocks and total soil respiration in both planted and naturally regenerating forests (Table 5). Root exudates and root turnover may accelerate the decompostion of soil organic matter through the the rhizosphere priming effect resulting in higher heterotrophic and total soil respiration (Kuzyakov et al. 2000).

Our third hypothesis that soil water content mediates the effect of soil temperature on total soil respiration was supported (Fig. 2). Soil water content and soil temperature tend to co-vary which makes it difficult to separate their effects on soil respiration (Davidson et al. 1998; Reichstein et al. 2003) and likely explains the poor performance of temperature-only models in this study. Higher explanatory power of soil temperature/soil water content interaction models as shown in this study (Table 4) has also been found in other forest ecosystems (e.g., Lellei-Kovács et al. 2011; Weissert et al. 2016). In this study, soil temperature exerted a significant positive control on total soil respiration when soil water content was above 23% in naturally regenerating forests and above 36% in planted forests (Fig. 2). Several authors have reported threshold values in soil water content affecting temperature sensitivity across ecosystems (e.g., Davidson et al. 1998; Emmett et al. 2004). Threshold values, however, seem to be ecosystem specific as the factors influencing soil water content (e.g., soil physical characteristics, water requirements by plants) differ between sites and ecosystems (e.g., Chang et al. 2014; Lellei-Kovács et al. 2011). In this study, soil water content became a controlling factor when mean soil water content dropped below 15% (Fig. 1b). This suggests that soil water content below 15% limited vegetation and microbial activity and thus autotrophic and heterotrophic respiration, in particular during the austral summer and early autumn (February/March 2021).

Conclusions

We discovered differences in litterfall C input and total soil respiration between planted and naturally regenerating L. scoparium forests, with greater litterfall production in the planted forest being linked with higher soil nutrient availability, which is likely a legacy of its past land-use in pastoral agriculture. Total soil respiration in both planted and naturally regenerating L. scoparium forests were associated with belowground C allocation (i.e., root C stocks). Soil temperature sensitivity of total soil respiration was influenced by soil water content. Although, the low number of plots in our study may limit its generalisability, our study is an important step towards addressing knowledge gaps. Naturally regenerating L. scoparium forests and L. scoparium restoration plantings are widespread in New Zealand, including for C offsets, yet we know too little about them throughout their latitudinal range. Our study lays the foundation for further comparative work between planted and naturally regenerating forest systems.