Geoderma 341 (2019) 138–147
Contents lists available at ScienceDirect
Geoderma journal homepage: www.elsevier.com/locate/geoderma
Prescribed fire affects the concentration and aromaticity of soluble soil organic matter in forest soils
T
⁎
Eleanor U. Hobleya, , Lena C. Zoora, Hari R. Shresthab, Lauren T. Bennettc, Christopher J. Westonc, Thomas G. Bakerb a
Soil Science, Emil-Ramann-Str. 2, Technische Universität München, 85354 Weihenstephan, Germany School of Ecosystem and Forest Sciences, The University of Melbourne, 500 Yarra Boulevard, Richmond, Victoria 3121, Australia c School of Ecosystem and Forest Sciences, The University of Melbourne, 4 Water Street, Creswick, Victoria 3363, Australia b
A R T I C LE I N FO
A B S T R A C T
Handling Editor: David Laird
Prescribed burning is used widely across Australia to reduce fuel load and associated wildfire hazard. However, prescribed burning can influence carbon (C) storage in affected ecosystems, potentially influencing C cycling. One important component of the C cycle is soluble C, which is available to microorganisms and therefore a key driver of nutrient cycling. However, studies of the effects of fire on soluble C in natural ecosystems are few. In this study, UV–Vis spectra of water extracts of soils from an Australian Eucalyptus forest were used to investigate soluble soil C characteristics to 30 cm depth after autumn burning every three or ten years for three decades contrasted with long non-burnt controls. A random forest prediction model was fit to the UV–VIS spectra and the wavelengths important to predicting soluble C investigated using conditional inference trees. The main absorbance of the UV–Vis spectra was in the aromatic region (~280 nm) indicating chemically complex soluble organic matter in both burnt and non-burnt soils. Water extractable organic carbon decreased significantly with depth and was reduced by 3-yearly burning in surface soils. Furthermore, burning, irrespective of treatment, led to significant shifts in the locations of soil spectra peaks, which were consistent with both lower quantities of non-aromatic substances and less substituted aromatic substances in the burnt than non-burnt sites. These results indicate that repeated prescribed fire over decades reduced the quantity of soluble C, and enhanced its aromaticity, potentially reducing biological availability and therefore nutrient cycling in the soils.
Keywords: Eucalypt forest Water-extractable organic matter UV–Vis spectra Random forest
1. Introduction Numerous studies indicate that, due to climate change, the number of days of extreme fire danger is expected to increase in south-eastern Australia and with it the likelihood of destructive wildfires (Middelmann, 2007). Prescribed fire is a widely used technique for managing fuel hazard over large areas, and assists wildfire suppression by reducing the intensity and severity of wildfires (Knapp et al., 2015; McCaw, 2013; Middelmann, 2007; Department of Sustainability and Environment, 2012). Additionally, prescribed fires are capable of reducing total carbon (C) emissions by up to 50% in the short-term (5 years) (Narayan, 2007; Bradstock et al., 2012). However, depending on burn frequency, this short-term reduction may be offset by greater emissions compared with natural fires over the same timeframe. Hence, in their review Gharun et al. (2017) underscored the need to consider the effects of prescribed burning on a wider range of environmental indicators (e.g. on C balances) than mere CO2 emissions, which might
⁎
be used to optimize fire management. Hazard reduction burning can conserve and improve the resilience of natural ecosystems and their ability to provide services such as water and C storage, production of forest goods, and maintenance of biodiversity (Department of Sustainability and Environment, 2003, 2012, 2015; Knapp et al., 2015; Gharun et al., 2017; Stephens et al., 2012). However, above- and below-ground C and nitrogen (N) stocks in soil can also be reduced by high-frequency prescribed burning (Bennett et al., 2014; Department of Sustainability and Environment, 2003; Muqaddas et al., 2015) and lead to a decline in soil fertility (Guinto et al., 1999). Reductions in above-ground biomass due to fire are welldocumented (Volkova et al., 2014; Fearnside et al., 1993), but effects in fire-adapted ecosystems are temporary (from years to decades) if there is sufficient time to recover before the next fire. In contrast, the effects of prescribed burning on soil C and N dynamics are less clear, with the effects of burning often site-specific and dependent upon many environmental factors (González-Pérez et al., 2004).
Corresponding author. E-mail address:
[email protected] (E.U. Hobley).
https://doi.org/10.1016/j.geoderma.2019.01.035 Received 19 September 2018; Received in revised form 6 January 2019; Accepted 13 January 2019 0016-7061/ © 2019 Elsevier B.V. All rights reserved.
Geoderma 341 (2019) 138–147
E.U. Hobley et al.
of the model results. However, despite their widespread adoption for investigating the solid phase of organic C in soils, to the best of our knowledge, random forest models and conditional inference trees have not yet been used to investigate soluble C based on UV–Vis spectra of soil extracts. In this study we investigated the effects of prescribed fire regimes on the quantity and quality of OC from forest soils in south-eastern Australia. We hypothesized that - concurrent with its effects on the chemical composition of soil organic carbon, namely the production of pyC with a recognizable ‘molecular fingerprint’ - fire induces a chemical shift to pyrogenic species in soluble OC compared with soluble OC of non-burnt soils. Additionally, due to the loss of soluble functional groups from pyC, DOC quantity is reduced in fire-affected soils, with the reduction in DOC amount inversely related to burn frequency. To investigate this, we extracted organic matter from the upper 30 cm of soils sampled in Australian Eucalyptus forests subjected to prescribed autumn burning every three or ten years over three decades, as well as a long non-burnt control site. We analyzed the quantity of water-extractable OC (WEOC) in the soil extracts and investigated its quality using UV–Vis spectra combined with a random forest model.
There are many potential impacts of prescribed burning on the properties of litter and soil, which vary with soil type, soil moisture content and vegetation (González-Pérez et al., 2004). Fire can increase water repellency of surface soil and change the structural composition of soil organic matter (Badía-Villas et al., 2014; Atanassova and Doerr, 2011), which can lead to increased surface runoff and soil erosion (Vega et al., 2005; Neary et al., 2005). In addition, run-off and drainage can export large quantities of dissolved organic matter (DOM), and the C within it (van Gaelen et al., 2014). However, C solubility in soils has been shown to decrease with increasing fire frequency due to changes in SOM properties (Preston and Schmidt, 2006). Nonetheless, while various studies have found significant decreases of dissolved organic C (DOC) content in peatland due to burning treatments (Worrall et al., 2007; Clay et al., 2012), others have detected short-term increases in soil DOC concentration after fire (Clay et al., 2009). In addition to changes in its quantity, fire can change the composition of DOC (Clay et al., 2012), for example reducing water-soluble fulvic acids (Almendros et al., 1990) or eliminating phenolic OH and COOH functional groups (Schnitzer and Hoffman, 1964). Importantly, fire results in the production of condensed aromatic ring structures, which have been labeled the ‘molecular fingerprint of black carbon degradation in soils’ (Hockaday et al., 2006). > 2% of the marine DOC pool has the signature of a heat-induced molecular species, indicating fluxes of dissolved pyrogenic carbon (pyC, also known as ‘black carbon’) from land to ocean (Kim et al., 2004; Dittmar et al., 2012). Although pyC is assumed to be a comparatively slow cycling component of the global carbon cycle (Schmidt and Noack, 2000), its biodegradability increases as its ages (Nguyen et al., 2010). Aging of pyC also leads to its oxidation and increased water solubility (Hockaday et al., 2006; Preston and Schmidt, 2006), resulting in a transfer of soil pyC to the aqueous phase (Abiven et al., 2011) and mobilization from soils (Jaffé et al., 2013; Ding et al., 2013). Investigating the molecular signatures of soluble pyC can therefore provide insights into its reactivity and fate in the environment. One simple but well-established method for investigating DOC properties is UV–Vis spectral analysis. Traditional investigations of DOC based on UV–Vis analyses have relied on simple statistics such as band intensities, ratios of band intensities or band intensity standardized to DOC quantity (Brandstetter et al., 1996; Peacock et al., 2014). In recent years, machine learning algorithms have been shown to produce robust predictions of soil organic C based on soil spectroscopic analyses (Viscarra Rossel and Behrens, 2010; Hobley et al., 2016). One such algorithm is the random forest (Breiman, 2001), which is able to model complex, non-linear relationships between covariant predictors, which is typically the case for spectral investigations of soil components. This stochastically driven algorithm recursively partitions data into nodes of ever increasing purity, producing individual decision trees which predict the target variable. Each tree is modelled based on a random subsample of target variables and predictors, with the final model comprised of multiple individual decision trees, over which prediction results are aggregated. One particular advantage of random forest modelling is that it enables data-mining of the spectra: rather than theory-driven interpretation of which spectral components could be important to predicting soil C components, the model learns from the spectra themselves without preconceptions of which spectral regions relate to the target predictor variable (i.e. soil C components). The regions important to prediction are identified by the model and can then be used to aid unbiased spectral interpretation for evaluating the C components of interest. Compared with many other predictive models (e.g. partial least squares, artificial neural networks), random forest modelling therefore has the added advantage of being able to identify variables important to model prediction, enabling interpretation of the model output (Hobley et al., 2015; Hobley et al., 2016). The variables important to predictions (‘predictors’) can then be modelled into a decision tree (or the closely related but statistically robust conditional inference tree, Hothorn et al., 2008), which enable visual interpretation
2. Methods 2.1. Study area characteristics Wombat State Forest is an open Eucalyptus forest in south-eastern Australia, which spans the northern and southern sides of Victoria's Great Dividing Range, about 75 km northwest of Melbourne, Victoria (37.4°S, 144.2°E). The climate of the area is temperate, with mean daily maximum temperatures varying from 8 °C in July to 24 °C in January (Australian Government Bureau of Meteorology, www.bom.gov.au/ climate/data). Mean annual precipitation varies from 900 mm to 1010 mm in the individual study areas. The elevation of the region varies from 590 m to 760 m above sea level and the topography varies from nearly flat to hills of low to moderate relief. The soils in the region developed on Ordovician sedimentary rocks. The predominant soil classes are ‘stony earths’ and ‘friable earths or mottled duplex soils’ (Kandosols and Dermosols respectively in the Australian Soil Classification (Isbell, 2016), which approximately correspond with Leptosols/Rudosols and Luvisols respectively in the FAO classification (WRB, 2015)). The vegetation of the study areas is dominated by open to tall open forest (E. obliqua L'Her., E. radiata Sieber ex DC. and E. rubida H. Deane and Maiden) (Specht, 1981; Bennett et al., 2013) of ~110–120 years age, with mature height 32–35 m (Tolhurst and Flinn, 1992). The understory is characterized by a sparse shrub layer 2–4 m in height and a ground layer dominated by Austral bracken, native perennial grasses, forbs, and rushes (Tolhurst, 2003). Detailed fire histories prior to establishment of the study were not available, although these forest types are prone to regular burning by wildfire. The last known wildfires in the study areas were between 1931 and 1974. 2.2. Burning treatments The study was established in 1985 as randomized block design consisting of a long non-burnt control and four prescribed fire treatments randomly allocated within each of five study areas (Table 1), located within a 25 km radius of each other. The prescribed fire treatments were a factorial combination of two fire seasons (autumn or spring) and two fire frequencies (nominally every 3 or 10 years). In this study we compare only the non-burnt control sites (NB) with autumn burning of either low (AL) or high (AH) frequency. The prescribed fire return intervals were nominally 3 or 10 years, but due to seasonal and operational limitations, mean prescribed fire intervals between 1987 (first autumn prescribed fires) and 2012 (this study's sampling) ranged from 9 to 16 years in the low fire frequency 139
Geoderma 341 (2019) 138–147
E.U. Hobley et al.
Table 1 Location, topography, and fire history of the five study areas in central Victoria, Australia. Study area
Latitude/longitude Elevation (m, above sea level)a Slope (°)a Aspect (°)a Last wildfireb Total experimental area (ha)b Mean fire interval (years): AHc Mean fire interval (years): ALc
Blakeville
Barkstead
Musk Creek
Burnt Bridge
Kangaroo Creek
37°31′S, 144°10′E 588–663 1–13 140–260 1935 81 3.0 (6, 1987–2007) 9.5 (3, 1987–2008)
37°29′S, 144°05′E 635–650 0–4 135–315 1931 19 4.0 (5, 1987–2007) 9.0 (3, 1987–2007)
37°28′S, 144°10′E 640–720 1–15 63–310 1974 78 4.0 (6, 1987–2008) 16.0 (2, 1987–2004)
37°25′S, 144°20′E 710–760 0–15 45–270 1953 62 5.7 (4, 1987–2007) 16.0 (2, 1987–2004)
37°19′S, 144°18′E 623–643 0–21 0–250 1944 128 3.4 (6, 1987–2009) 9.0 (3, 1987–2007)
a
Range from this study's non-burnt, AH and AL plots within each study area. Tolhurst and Flinn (1992). c Mean interval in years between successive prescribed fires during the experimental period (values in brackets indicate the number of prescribed fires, and the years of first and last prescribed fires); treatment abbreviations: ‘AH’ autumn high frequency, ‘AL’ autumn low frequency. b
treatments, and 3 to 6 years in the high fire frequency treatments. The date of last prescribed fire ranged from March 2007 to March 2009 for AH, and March 2004 to February 2008 for AL (Table 1). The prescribed fires in both autumn treatments were considered of low intensity, generally < 500 kW m−1, and with flame heights in the range 0.1 to 1.3 m. The prescribed fires were designed to cause negligible crown scorch (confirmed by post-fire observations), and to reduce surface fuel loads (leaves, twigs and barks of understorey and overstorey plants; Gould et al., 2011), which were decreased by 40% on average in both autumn treatments (Bennett et al., 2013).
Additionally, WEOC was normalized to the proportion of C in the soil samples (WEOC:C, unitless). Duplicate UV–Vis spectra of each soil extract were acquired in the region 190–1100 nm with a double-beam scanning UV–Vis spectrometer with split-beam technology and sipper system (SPECORD®50 PLUS, Analytik Jena AG, Jena, Germany). Six samples were only analyzed singularly due to limited extract. The spectra were recorded at 1 nm resolution using a scan speed of 3000 nm min−1, with scans integrated over 40 s while the extract was continually pumped through the analysis cell. Prior to acquisition, the system was referenced to a background measurement of deionized water.
2.3. Soil sampling and sample preparation 2.5. Data and spectral analysis Soils were sampled in autumn 2012, 25 to 27 years after the first experimental fire, and 3 to 8 years after the last. Within each fire treatment, three circular 0.1 ha sample plots were randomly designated at each site. Within each plot, twelve points were randomly sampled and samples composited per plot. The top 10 cm of soil was sampled after removal of litter using a 7.25 cm diameter corer containing stacked 2-, 3- and 5 cm-deep internal rings. These core samples were sub-divided into 0–2, 2–5, and 5–10 cm depths prior to compositing by depth interval. Soil from 10 to 20 cm depth was sampled using a corer in the same hole (core diameter 6.25 cm) and composited from 4 of the 12 points, and soil 20–30 cm depth was sampled at a single point. All samples were dried at 40 °C and gently sieved to < 2 mm to remove mineral and coarse organic fragments. Samples from the three plots within each fire treatment at each site were then composited (on equal mass basis), yielding 75 samples for the present investigation (5 depths from 3 treatments at 5 sites).
Prior to statistical analyses, the raw UV–Vis spectra were smoothed using a 5-point moving average filter to reduce noise. The region between 190 and 500 nm was particularly noisy and therefore smoothed a second time. Peak minima and maxima were detected based on the first and second derivatives of the spectra after simple smoothing with a 5point moving average filter. Peak areas were calculated by integration after subtracting a baseline from between the minima of the peaks. Effects of prescribed fire treatments and depth on soil C, N, C:N ratio, WEOC, WEOC:C and the spectral peak locations and integrated areas were tested using analysis of variances (ANOVA). Partial correlations were used to decipher fire treatment effects while accounting for depth effects. Games-Howell post-hoc tests were used to identify significant differences between individual treatments at specific depths. All data analyses were performed using the R software (version 3.4.2; R Development Core Team 2017).
2.4. Soil analyses
2.6. Random forest modelling of water-extractable organic carbon based on UV–Vis spectra
A subsample of each soil sample was finely ground in a ball mill and the total C and N concentrations determined in duplicate by dry combustion (Vario EL CN Analyzer, Elementar, Hanau, Germany). Given the acidic nature of the soils (pH < 5.5), the presence of carbonates can be excluded so that total C is equivalent to organic C. From the results, the C:N ratio of the soil samples was calculated. Soils (< 2 mm) were extracted by shaking 5 g of soil in 25 ml of deionized water for 1 h to obtain water-extractable organic carbon (WEOC). The suspensions were then centrifuged (3500 rpm, 5 min) and the supernatant pressure filtered through a 0.45 μm cellulose nitrate membrane filter. The soil extracts from 0–2 cm and 2–5 cm depths were diluted 1:2 by volume with deionized water prior to analysis. DOC in the extracts was quantified in duplicate by catalytic oxidation (Shimadzu TOC-5050A, Japan). Due to the acidic nature of the soils, total dissolved C is equivalent to dissolved organic C. WEOC was calculated as the mass (mg) of DOC extracted per soil mass (g).
Spectral treatment and random forest modelling followed Hobley et al. (2016). The UV–Vis absorbance spectra were standardized by zscoring (mean centering, i.e. subtracting the mean absorbance of the spectrum, and dividing by the standard deviation of the spectrum) prior to modelling. The random forest models were developed using the UV–Vis standardized absorbance spectra as predictor and mass proportion of WEOC in the soil extracts as response variables. The number of predictors used for modelling in each tree was set to 2 ∗ n spectral data points . For qualitative purposes, a random forest was grown using all replicates (N = 144), in which the minimum sum of weights in a terminal node was set to two (number of replicate scans per standard) and the minimum number of cases per terminal node to four. Relative variable importance (Hobley et al., 2015) was used to identify spectral regions of importance for WEOC, where important variables are those whose 140
Geoderma 341 (2019) 138–147
E.U. Hobley et al.
there was also a highly significant difference in WEOC between treatments (Table 2). Specifically, the higher frequency autumnal burning (AH) significantly reduced WEOC concentrations at the soil surface compared with the non-burnt control (Fig. 2). WEOC concentrations in the low-frequency treatment at the soil surface were intermediate between and statistically indistinct from the non-burnt control and highfrequency treatment. WEOC:C increased significantly with depth. The ANOVAs indicated significant treatment effects interacting with depth (Table 2) and the partial correlation indicated that the solubility of organic C in the low frequency burn treatment was significantly lower than the solubility of organic C in the non-burnt control.
influence to the model is greater than expected from a model in which all predictors are equally important. Additionally, a conditional interference tree (Hothorn et al., 2008) with Bonferroni correction and a significance level of p = 0.05 was fit to the spectra to investigate the relationships between WEOC and UV–Vis bands important to its prediction, with spectral regions interpreted based on literature studies of organic matter spectral characteristics. For quantitative predictions, the mean spectra per replicate were used to avoid inflation of goodness-of-fit statistics. In this model, the minimum number of cases per terminal node was set to 1. In each forest, 500 trees were grown. Model predictive performance was based on the coefficient of determination R2 = 1 – (MSE/Variance) (with MSE as the mean squared error of the model and Variance as the variance of the response variable modelled) of the out-of-bag estimates (i.e. the cases not used in fitting the data, which are unbiased prediction estimators, (Hobley et al., 2016, Strobl et al., 2009). Additionally, predictive performance was assessed via the root mean square error of prediction (RMSE, standard error of prediction), the mean absolute error (MAE) and the residual prediction deviation (RPD, standard deviation of samples divided by prediction error).
3.2. Fire and depth effects on UV–Vis spectra of water extractable organic carbon of the soils 3.2.1. Spectral characteristics All UV–Vis spectra for the soil extracts exhibited a large peak with a maximum of standardized absorbance between 263 nm and 280 nm (Fig. 3). Prescribed fire significantly shifted the location of the main peaks (AL peak location at 274 ± 3 nm, and AH peak location at 270 ± 5 nm) in the extracts from 0 to 2 cm soil compared with the control treatment (peak location at 280 ± 7 nm, Tables 1 & 2, Fig. 4). A shift in peak location was also apparent for the deeper soil extracts, though not statistically significant. Additionally, shoulder peaks were detected in the deeper soils (below 5 cm), in the region between 350 nm and 550 nm with a maximum in the region 361 ± 7–366 ± 1 nm (Fig. 3, Table 3), which were not apparent in the extracts from the surface soils. The peak maxima of the shallower samples (0–2, 2–5 cm) were greater than the peak maxima of the deeper samples (5–10, 10–20, 20–30 cm, Fig. 3). However, the main peak area showed no clear trend with depth (Table 2). In contrast, the shoulder peak area increased significantly with depth from 0.00 ± 0.001 to 0.2 ± 0.146 in Control, from 0.01 ± 0.005 to 0.35 ± 0.155 in AL and from 0.00 ± 0.004 to 0.15 ± 0.090 in AH treatments (Table 2). The significant differences between treatments indicated by the ANOVA were not reproducible using partial correlation to account for depth effects. The ratio between shoulder and peak area increased significantly with depth with linear models indicating interactions between depth and treatment effects. Tested by partial correlation, the ratio of shoulder peak to main peak
3. Results 3.1. Fire and depth effects on quantities of C, N and water extractable organic carbon in the soils Soil organic total C, total N, C:N ratio (Fig. 1) and water extractable organic C concentrations (Fig. 2, Table 2) were greatest at the soil surface and decreased significantly with depth. The depth distribution of C, N and WEOC were non-linear and were better described mathematically using a linear-logarithmic relationship. In contrast, soil C:N ratio and WEOC:C were, within uncertainties, approximately linearly correlated with depth. The C:N ratio decreased with depth, whereas WEOC:C increased with depth. Burning did not significantly influence soil total organic C and N concentrations and C:N across all depths in this present study. While the ANOVA indicated a highly significant effect of fire treatment on C:N ratios, which interacted with depth (p < 0.01), the post-hoc test failed to identify specific significant shifts in C:N ratios due to fire treatment after accounting for depth effects. In addition to the highly significant decrease in WEOC with depth,
Fig. 1. Depth distribution of total (organic) C and total N concentrations, and of total C: total N ratios in soils from non-burnt (NB), autumn low frequency (AL) and autumn high frequency (AH) prescribed burning treatments. Dots represent means, error bars represent one standard deviation (N = 5). 141
Geoderma 341 (2019) 138–147
E.U. Hobley et al.
Fig. 2. Depth distribution of water extractable organic carbon (WEOC) content and the ratio of WEOC to soil organic C. Dots represent means, error bars one standard deviation.
was significantly lower in AH than in the Control and AL treatments. 3.2.2. Random forest model and conditional inference tree of UV–Vis spectra The random forest model performed very well at fitting and predicting WEOC (Fig. 5), with coefficient of determinations of fit R2 = 0.96 and prediction R2 = 0.81. The standard error of prediction (RMSE) was 10.3 mg g−1, mean absolute error 8.1 mg g−1 and the residual prediction deviation 2.23. Model bias was low (−1.2 mg g−1) but model residuals indicated an increasing deviation between prediction and measured values with higher WEOC contents (Fig. 5). The relative variable importance extracted from the models indicated a central region of importance around 286 nm and another smaller region of importance around 236 nm (Fig. 6). Further evidence for the importance of the region around 286 nm was indicated by the conditional inference tree, which showed a primary split in this region to separate samples with higher WEOC contents from those with lower WEOC contents (Fig. 7). The model resulted in five terminal nodes, i.e. five statistically distinct end groups of samples obtained after recursive partitioning, with the highest WEOC (0.63 ± 0.06 mg g−1) indicated by highest standardized absorbance above 286 nm and 289 nm. Lowest WEOC concentrations (0.22 ± 0.04 mg g−1) were indicated by lowest standardized absorbance at 286 nm and 283 nm. The region at 227 nm
Fig. 3. Standardized UV–Vis spectra of aqueous extracts of 0–2 cm and 20–30 cm depth soils from non-burnt (NB), autumn low frequency (AL) and autumn high frequency (AH) prescribed burning treatments. Inset shows the peak maxima of the samples from 0 to 2 cm depth. Grey numbers indicate the location of the peak maxima.
Table 2 Significance of burning treatment and depth effects on the contents of water extractable organic C (WEOC), the ratio of WEOC to soil organic carbon (WOEC:C) and on peak locations and peaks areas detected with UV–Vis spectra. Partial correlation to identify treatment effects compared with NB sites Treatment Depth Treatment × Depth Partial correlation just treatment effect
WEOC
WEOC:C
Location main peak
Area main peak
⁎⁎
⁎⁎
⁎⁎⁎
⁎⁎⁎
⁎⁎⁎
⁎⁎
AH⁎⁎⁎
AL⁎⁎
- (⁎ in 0–2 cm) -
-
Location shoulder peak ⁎
-
Area shoulder peak
Ratio shoulder peak: main peak areas
⁎
⁎⁎
⁎⁎⁎
⁎⁎⁎
⁎⁎
⁎⁎⁎
-
AH⁎
NB: non-burnt sites, AH: sites burnt in autumn at high frequency, AL: sites burnt in autumn at low frequency. - not significant/no peak detected. AH*** - after partial correlation to account for depth effects, AH sites were significantly different at p < 0.001 from NB sites. AL** - after partial correlation to account for depth effects, AL site were significantly different at p < 0.01 from NB sites. ⁎⁎⁎ p < 0.001. ⁎⁎ p < 0.01. ⁎ p < 0.05. 142
Geoderma 341 (2019) 138–147
E.U. Hobley et al.
Fig. 5. Predicted (random forest) v. measured concentrations of water-extractable organic C (WEOC, orange), and model residuals (measured minus predicted, blue), in soils (all burning treatments, all depths). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 4. Shifts in (a) main peak locations (b) and water extractable organic carbon (WEOC) concentrations between treatments at 0–2 cm depth. ANOVAs indicated significant treatments effects on both main peak location and quantity of WEOC. Different uppercase letters indicate significant differences using the Games-Howell post-hoc test.
Fig. 6. Relative importance of the wavelengths used for water extractable organic carbon (WEOC) prediction in the random forest model with the dotted line showing the cut-off between importance and unimportance to the model, where importance is defined as the expected variable importance in a model in which all predictors are equally important (Hobley et al., 2015).
Table 3 Location and area of main and shoulder peaks in the UV–VIS spectra of aqueous extracts of soil from non-burnt (NB), autumn low frequency (AL) and autumn high frequency (AH) prescribed burning treatments. Values indicate mean (n = 5) ± standard deviation. - indicates no peak detected. Units are standardized spectral units (dimensionless) after mean centering. Depth (cm)
Treatment
0–2
NB AL AH NB AL AH NB AL AH NB AL AH NB AL AH
2–5
5–10
10–20
20–30
Location main peak (nm) 280 273 270 266 265 264 277 275 274 274 276 273 274 273 269
± ± ± ± ± ± ± ± ± ± ± ± ± ± ±
7 3 5 2 3 1 8 4 5 4 3 7 8 8 4
Area main peak
Location shoulder peak (nm)
Area shoulder peak
Ratio shoulder peak: main peak areas
121 ± 22 99 ± 13 89 ± 24 80 ± 16 69 ± 14 71 ± 11 120 ± 29 115 ± 20 111 ± 16 116 ± 14 112 ± 10 106 ± 35 110 ± 38 99 ± 27 89 ± 18
361 ± 7 361 ± 7 362 ± 10 364 ± 6 366 ± 1 364 ± 5 366 ± 1 365 ± 1 365 ± 1
± ± ± ± ± ± ± ± ±
0.0003 ± 0.006 0.0007 ± 0.007 0.0007 ± 0.004 0.0015 ± 0.001 0.0025 ± 0.0012 0.0012 ± 0.0008 0.0026 ± 0.0009 0.0035 ± 0.0010 0.0016 ± 0.0009
143
0.05 0.09 0.08 0.18 0.28 0.15 0.29 0.35 0.15
0.085 0.102 0.047 0.118 0.145 0.111 0.146 0.155 0.090
Geoderma 341 (2019) 138–147
E.U. Hobley et al.
Fig. 7. Conditional interference tree for quantifying water extractable organic carbon (WEOC) using UV–Vis spectra. ***: split is significant at p < 0.001, **split is significant at p < 0.01. The circled variables indicated the wavelength at which a split in the tree was made to statistically separate groups of samples based on WEOC content, the lines in the numbers indicted the cut-off point of standardized spectral response to separate the daughter nodes, the number of samples in each terminal node indicated above the node.
studies at the site. Further, we investigated C concentrations, not C stocks, which although highly covariate, are not linearly related due to potential effects of fire on soil bulk density. However, the general though not significant decline in surface soil C and N (0–5 cm) content associated with higher frequency fire intervals reflects results of other studies at the site (Department of Sustainability and Environment, 2003) and across the globe (Wang et al., 2013; Krishnaraj et al., 2016; Neary et al., 1999; Mataix-Solera et al., 2011). Moreover, the reduction in the quantities of litter, pyC, and soil C is influenced by the intensity and duration of fire (Knicker, 2007; Krishnaraj et al., 2016; Bennett et al., 2014), which can have naturally high variability depending on fuel loads and weather conditions, confounding detection of potentially small changes against high variance. Overall, we conclude that the investigated burning treatments in this study did not markedly affect concentrations of soil C and N in the solid phase.
was important to differentiating samples with intermediate WEOC contents.
4. Discussion 4.1. Fire effects on soil organic matter depth distribution in the solid phase The negative, non-linear relationship between soil C and N and depth is consistent with patterns of SOC depth distribution in Australia (Hobley and Wilson, 2016), including burned soils (e.g. Hobley et al., 2017). This reflects the non-linear relationship of soil organic C input (e.g. via roots and litter) and turnover with depth (Balesdent et al., 2018). In burned soils, the input of non-charred and charred material post-burning occurs primarily on the surface layer (Krishnaraj et al., 2016; Hobley et al., 2017), enhancing surface inputs of organic C and, potentially, the non-linearity of the relationship between soil C and depth. The significant decrease of the C:N ratio in soil with depth is consistent with a preferential metabolization of soil C compared with soil N by microbes (Kirkby et al., 2011). This indicates that a greater proportion of organic matter is microbially processed in the subsoil than at the surface. Increases in soil temperatures under prescribed fire are relatively small (cf. those under wildfire and/or heavy fuel loads), short-lived, and diminish steeply with depth (e.g. Penman and Towerton, 2008; Cawson et al., 2016) likely explaining the lack of any effect of fire treatments on soil C:N ratio. The lack of effect of fire treatments on C:N ratio is in contrast with the findings of Hobley et al. (2017), who attributed increased C:N ratio at the soil surface due to preferential loss of N during burning. Our results suggest that there was no preferential loss of C compared with N due to burning at the investigated sites. This is potentially the result of the different fire intensities between sites, indicated by the differences in flame height, scorch height, area burned and surface fuel consumed (Bennett et al., 2014). Similarly, although other studies have reported that higher burning frequency reduces soil C storage at these sites (Bennett et al., 2014; Williams et al., 2012), our results did not confirm these effects for soil C concentrations statistically. This we attribute to the smaller sample size and therefore lower statistical power in our investigation than in other
4.2. Fire effects on soluble soil organic matter depth distribution WEOM is a source of readily available C and nutrients for microbes and plays a key role in soil biochemical processes (e.g. Kalbitz et al., 2003; Chantigny et al., 2014). WEOC may not reflect the DOC currently available at any given point in time as soil moisture and therefore DOC quantity vary with climate. However, it is frequently investigated in studies of DOC or DOM in soils (e.g. Buzek et al., 2009; Ellerbrock and Kaiser, 2005) as it comparable to a standardized proxy for DOC, which can be viewed as an integrator of potentially extractable DOC in a soil. In contrast to the lack of detectable effects of fire on bulk C concentrations, the decrease in WEOC at the surface soil caused by burning indicates a shift in organic matter solubility. Soluble OC production depends on fresh litter contribution (Park et al., 2002) and litter-rich surfaces (Tu et al., 2011), which decrease with fire treatments (Krishnaraj et al., 2016; Hobley et al., 2017). Microbially processed SOM and the end products of microbial metabolism form another large proportion of soluble OC (Guggenberger et al., 1994), and these products may also be combusted during fire. Further, without input for metabolism, microbial biomass will decline (Metting, 1993). As such, microbial biomass is reduced in fire-affected soils compared with nonburnt soils (Choromanska and DeLuca, 2002). In combination, the reduced input of fresh litter for decomposition and reduced microbial 144
Geoderma 341 (2019) 138–147
E.U. Hobley et al.
conditional inference tree, which distinguished the highest WEOC contents (i.e. associated with non-burnt sites) based upon higher absorbance at 289 nm. In the region of 270 nm, the π_π* electron transition occurs for the functional groups eOH and eCOOH of aromatics in water (Pretsch et al., 2009). The shift in peak maxima location at the surface therefore suggests that there are more carboxylic functional groups in the WEOC of the burned soils than those of the non-burnt control. Hockaday et al. (2007) showed that the condensed aromatic ring structures of charcoal are functionalized over time with hydroxyl and carboxylic groups on their periphery during aging, which enables them to be partially aqueous soluble. Given the long fire-free history of the control sites (> 40 years non-burnt; Bennett et al., 2013), this suggests that the higher wavelengths of the WEOC may indicate ‘older’, highly altered, and therefore soluble, charcoal in the surface soil. However, stable WEOC has also been attributed to ‘lignocellulose-degradation products’ (Guggenberger et al., 1994). Thus, the higher wavelengths of the main peak in the control sites are also consistent with a greater portion of non-pyrogenic WEOC in the control sites compared with the burn sites. Overall, it can be said that WEOC of the soils was dominated by aromatic structures and very probably influenced by pyrogenic material. The appearance of the shoulder peak centered around 361 ± 7–366 ± 1 nm in the UV–Vis spectra in the deep soil (5–10, 10–20, 20–30 cm), which was not detected at the surface, likely indicates the presence of iron and soluble organic complexes in deeper soil samples (Carpenter and Smith, 1984). Kaiser (1998) reported that an increase of iron ions in solution leads to precipitation of organic complexes due to increasing hydrophobicity of the molecules. Extracts from soil samples below 10 cm formed a brown precipitate upon acidification with HCl, which is consistent with iron-organic complex precipitation. Thus, we believe that with increasing soil depth, iron-organic complexes are solubilized in this soil. However, the region of the shoulder peak (361–366 nm) was not selected as important to the random forest models or the conditional inference tree, which implies that the compounds responsible for the shoulder peaks are not a great influence on the content of WEOC in these soils. Nevertheless, the significantly smaller ratio of shoulder peak to main peak area in the high fire-frequency treatment compared with the less frequently and non-burned sites indicates that increased fire frequency also has an effect on iron-organo-complex concentrations. Litter leachates from eucalypt trees have been shown to mobilize iron in soils into solution (Bernhard-Reversat, 1999). The reduced litter input due to high fire frequency may therefore reduce this solubilization, influencing organic matter cycling in the subsoil. The decreasing WEOC with increasing depth is consistent with the soil C decrease with increasing depth. DOM components may bond to positively charged mineral surfaces like iron and aluminum hydrous oxides limiting its solubility (Kaiser et al., 1997; Jardine et al., 1989), and our results indicate that such compounds increase with depth in these soils. However, the ratio of WEOC to soil C increased with depth, associated with a greater reduction in soil C than WEOC with increasing depth. This indicates that WEOC was (bio)chemically altered with increasing depth, enhancing its solubility. Alternatively, the lower proportion of WEOC near the surface may indicate its loss as surface runoff or immobilization due to the higher pyC content at the surface.
biomass therefore contribute to the reduction in WEOC and with it a smaller bioavailability of organically bound nutrients like N, phosphorus (P), sulfur (S) (Kalbitz et al., 2000; Kaiser et al., 2001). WEOC quantity in soil has also been shown to depend on substrate properties (Kalbitz et al., 2000), which is also affected by burning (Atanassova and Doerr, 2011). Frequent burning increases input of charred material with a very high C:N ratio (Hobley et al., 2017; Williams et al., 2012) compared to long non-burnt sites, which could reduce mineralization of SOC and therefore the production of soluble SOC components. However, our C:N ratios were unchanged by burning, suggesting that C:N ratio was not an indicator of WEOC content. Furthermore, Choromanska and DeLuca (2002) also observed a very shortterm (14 day) increase in soluble sugars and a release of available C and N in forest mineral soil as a post-fire effect. Clay et al. (2009) too detected “short-lived” fire boosting effects on WEOC concentration but not in the long term (10–20 years). Hobley et al. (2017) suggested that turnover of dead roots post-fire contribute to a short-term (1–3 years) increase in soil C, which is consistent with a new source of easily degradable soil C and therefore WEOC production. As the last fire treatments at the site were 3–8 years prior to sampling, this short-term effect would not be detectable within the time frame of our study. Overall, we believe that the greater WEOC contents in the unburnt than AH soils are therefore consistent with a greater content of soluble, non-pyrogenic soil C. In contrast, the lower WEOC:C ratio in the low frequency burn site is likely related to overall higher C concentrations in the solid phase of the low frequency burn sites, which reduces the ratio of WEOC:C. 4.3. Fire effects on soluble soil organic matter spectral properties The random forest and conditional inference tree selection of important regions at 283–289 nm and 227–236 nm is consistent with a wide variety of studies investigating different absorbance bands in the UV–Vis spectra as proxies for WEOC, with absorbance at 230 nm and 263 nm (Peacock et al., 2014), 254 nm (Brandstetter et al., 1996), 270–330 nm (Causse et al., 2017), 285 nm (Giancoli Barreto et al., 2003) and 300 nm (McKnight et al., 1997) being used to detect WEOC. Due to different bands associated with different WEOC structures, Avagyan et al. (2014) advised the utilization of site-specific calibration models that include more than one wavelength to achieve the optimal accuracy of the proxy-based WEOC quantification. This is also consistent with the two proxies identified as relevant to WEOC prediction in the random forest models, and the suitability of the random forest modelling approach is also indicated by the very good model statistics and prediction results (residual prediction deviation > 2.0 supports the reliability of the model, Chang et al. (2001). Further, the conditional inference tree suggests that WEOC from these soils can be primarily distinguished based upon absorbance at 283–289 nm, with absorbance around 227 nm useful for distinguishing intermediate WEOC concentrations. The wavelengths selected as proxies for soluble C in the soils are in the range of aromatic compounds (Pretsch et al., 2009), suggesting that aromatic compounds are large contributors to WEOC in the soils. This is also supported by the findings of Kalbitz et al. (2003) who used the wavelength at 280 nm to detect aromaticity in DOM. Absorbance in this band is attributed to aromatic tannic acids and lignosulfonic acids (Lawrence, 1980) as well as aromatic phenols, benzoic acids, aniline derivatives, polyenes, and polycyclic aromatic hydrocarbons with two or more rings (Peuravuori and Pihlaja, 1997). Therefore, the UV–Vis spectra and the random forest model strongly suggest the presence of aromatic structures in the WEOC from the study areas, which indicates relatively stable WEOC (Kalbitz et al., 2003; Marschner and Kalbitz, 2003) and potentially a pyrogenic “molecular fingerprint” for the soils. The significant shift in the main peak location of the UV–Vis spectra of the upper (0–2 cm) soil samples from higher wavelengths in the control (280 nm), to smaller wavelengths (273–270 nm) in the burn sites indicates a shift in WEOC properties. This is also supported by the
5. Conclusion Higher fire frequency significantly decreased the solubility of soil organic carbon at the soil surface (0–2 cm) compared to low fire frequency and non-burnt soil. This is attributable to a greater proportion of more soluble non-pyrogenic and/or aged pyrogenic OM in the absence of fire compared to frequent fire. Nevertheless, the soluble C of all the investigated soils - control and prescribed burned - is dominated by aromatic structures, which is consistent with the presence of charred organic material in the soluble phase and their location in an 145
Geoderma 341 (2019) 138–147
E.U. Hobley et al.
environment with natural fire regimes spanning back millennia. However, the different locations of the standardized absorbance peak in the UV–Vis spectra between control and burning treatments in the upper layer (0–2 cm) are likely indicative of chemical differences of soluble C between sites of differing fire frequency. UV–Vis investigations of water-extractable organic C combined with machine learning and data mining methods therefore appear to be a useful tool to investigate integrated differences in chemical structure in fire-affected soils. Our approach to data analysis is applicable to a wide range of soils and treatments, so provides the possibility for novel investigations of soluble soil organic matter based on machine learning and data mining. Given the importance of soluble C for microbial turnover of organic matter, our results indicate the potential for impacts of frequent (~3yearly) prescribed burning on nutrient availability in fire-affected ecosystems. In contrast, the low frequency prescribed fires had a smaller but still significant effect on soluble C. From a management perspective, low fire frequency (~10 years return interval) will therefore have a smaller impact on soil-based ecosystem processes than high frequency fire treatments.
1016/S0003-2670(00)84304-4. Causse, J., Thomas, O., Jung, A.-V., Thomas, M.-F., 2017. Direct DOC and nitrate determination in water using dual pathlength and second derivative UV spectrophotometry. Water Res. 108, 312–319. https://doi.org/10.1016/j.watres.2016.11. 010. Cawson, J.G., Nyman, P., Smith, H.G., Lane, P.N.J., Sheridan, G.J., 2016. How soil temperatures during prescribed burning affect soil water repellency, infiltration and erosion. Geoderma 278, 12–22. Chang, C.-W., Laird, D.A., Mausbach, M.J., Hurburgh, C.R., 2001. Near-infrared reflectance spectroscopy–principal near-infrared reflectance spectroscopy–principal components regression analyses of soil properties. Soil Sci. Soc. Am. J. 65 (2), 480. Chantigny, M.H., Harrison-Kirk, T., Curtin, D., Beare, M., 2014. Temperature and duration of extraction affect the biochemical composition of soil water-extractable organic matter. Soil Biol. Biochem. 75, 161–166. Choromanska, U., DeLuca, T.H., 2002. Microbial activity and nitrogen mineralization in forest mineral soils following heating. Evaluation of post-fire effects. Soil Biol. Biochem. 34 (2), 263–271. https://doi.org/10.1016/S0038-0717(01)00180-8. Clay, G.D., Worrall, F., Fraser, E.D.G., 2009. Effects of managed burning upon dissolved organic carbon (DOC) in soil water and runoff water following a managed burn of a UK blanket bog. J. Hydrol. 367 (1–2), 41–51. https://doi.org/10.1016/j.jhydrol. 2008.12.022. Clay, G.D., Worrall, F., Aebischer, N.J., 2012. Does prescribed burning on peat soils influence DOC concentrations in soil and runoff waters? Results from a 10 year chronosequence. J. Hydrol. 448–449, 139–148. https://doi.org/10.1016/j.jhydrol. 2012.04.048. Department of Sustainability and Environment, 2003. Ecological Impacts of Fuel Reduction Burning in a Mixed Eucalypt Foothill Forest – Summary Report (1984–1999). vol. 57 Victorian Government, Melbourne, Australia. Department of Sustainability and Environment, 2012. Code of Practice for Bushfire Management on Public Land. Victorian Government, Melbourne, Australia. Department of Sustainability and Environment, 2015. Review of Performance Targets for Bushfire Fuel Management on Public Land. Victorian Government, Melbourne, Australia. Ding, Y., Yamashita, Y., Dodds, W.K., Jaffé, R., 2013. Dissolved black carbon in grassland streams. Is there an effect of recent fire history? Chemosphere 90 (10), 2557–2562. https://doi.org/10.1016/j.chemosphere.2012.10.098. Dittmar, T., Paeng, J., Gihring, T.M., Suryaputra, I.G.N.A., Huettel, M., 2012. Discharge of dissolved black carbon from a fire-affected intertidal system. Limnol. Oceanogr. 57 (4), 1171–1181. https://doi.org/10.4319/lo.2012.57.4.1171. Ellerbrock, R.H., Kaiser, M., 2005. Stability and composition of different soluble soil organic matter fractions–evidence from δ13C and FTIR signatures. Geoderma 128, 28–37. Fearnside, P., Leal, N., Fernandes, F.M., 1993. Rainforest burning and the global carbon budget: biomass, combustionefficiency, and charcoal formationin the Brazilian Amazon. J. Geophys. Res. Atmos. 98 (D9), 16733–16743. Gharun, M., Possell, M., Bell, T.L., Adams, M.A., 2017. Optimisation of fuel reduction burning regimes for carbon, water and vegetation outcomes. J. Environ. Manag. 203 (Pt 1), 157–170. https://doi.org/10.1016/j.jenvman.2017.07.056. Giancoli Barreto, S.R., Nozaki, J., Barreto, W.J., 2003. Origin of dissolved organic carbon studied by UV–vis spectroscopy. Acta Hydrochim. Hydrobiol. 31 (6), 513–518. https://doi.org/10.1002/aheh.200300510. González-Pérez, J.A., González-Vila, F.J., Almendros, G., Knicker, H., 2004. The effect of fire on soil organic matter—a review. Environ. Int. 30 (6), 855–870. https://doi.org/ 10.1016/j.envint.2004.02.003. Gould, J.S., Lachlan McCaw, W., Phillip Cheney, N., 2011. Quantifying fine fuel dynamics and structure in dry eucalypt forest (Eucalyptus marginata) in Western Australia for fire management. For. Ecol. Manag. 262, 531–546. Guggenberger, G., Zech, W., Schulten, H.-R., 1994. Formation and mobilization pathways of dissolved organic matter. Evidence from chemical structural studies of organic matter fractions in acid forest floor solutions. Org. Geochem. 21 (1), 51–66. https:// doi.org/10.1016/0146-6380(94)90087-6. Guinto, D.F., Saffigna, P.G., Xu, Z.H., APN House, Perera, M.C.S., 1999. Soil nitrogen mineralization and organic matter composition revealed by 13C NMR spectroscopy under repeated prescribed burning in eucalypt forests of south-east Queensland. Aust. J. Soil Res. 37, 123–135. Hobley, E.U., Wilson, B., 2016. The depth distribution of organic carbon in the soils of eastern Australia. Ecosphere 7 (1). https://doi.org/10.1002/ecs2.1214. Hobley, E.U., Wilson, B., Wilkie, A., Gray, J., Koen, T., 2015. Drivers of soil organic carbon storage and vertical distribution in Eastern Australia. Plant Soil 390, 111–127. Hobley, E.U., Brereton, A.J.L.E.G., Wilson, B., 2016. Soil charcoal prediction using attenuated total reflectance mid-infrared spectroscopy. Soil Res. 55 (1). https://doi.org/ 10.1071/SR16068. Hobley, E.U., Le Gay Brereton, A.J., Wilson, B., 2017. Forest burning affects quality and quantity of soil organic matter. Sci. Total Environ. 575, 41–49. https://doi.org/10. 1016/j.scitotenv.2016.09.231. Hockaday, W.C., Grannas, A.M., Kim, S., Hatcher, P.G., 2006. Direct molecular evidence for the degradation and mobility of black carbon in soils from ultrahigh-resolution mass spectral analysis of dissolved organic matter from a fire-impacted forest soil. Org. Geochem. 37 (4), 501–510. https://doi.org/10.1016/j.orggeochem.2005.11. 003. Hockaday, W.C., Grannas, A.M., Kim, S., Hatcher, P.G., 2007. The transformation and mobility of charcoal in a fire-impacted watershed. Geochim. Cosmochim. Acta 71 (14), 3432–3445. https://doi.org/10.1016/j.gca.2007.02.023. Hothorn, T., Bretz, F., Westfall, P., 2008. Simultaneous inference in general parametric models. Biom. J. 50 (3), 346–363. https://doi.org/10.1002/bimj.200810425. Isbell, R., 2016. The Australian soil classification. In: Australian Soil and Land Survey
Acknowledgements We thank the many staff of the Victorian state environment department (Department of Environment, Land, Water and Planning, DELWP) and at the University of Melbourne who have been responsible for maintaining the prescribed fire experiment for over thirty years. We also thank DELWP for supporting the re-assessment of the experiment in 2012. Special thanks to J. Najera, B. Smith, G. Szegedy and N. Ahmady for their roles in soil sampling and analysis. References Abiven, S., Hengartner, P., Schneider, M.P.W., Singh, N., Schmidt, M.W.I., 2011. Pyrogenic carbon soluble fraction is larger and more aromatic in aged charcoal than in fresh charcoal. Soil Biol. Biochem. 43 (7), 1615–1617. https://doi.org/10.1016/j. soilbio.2011.03.027. Almendros, G., Gonzalez-Vila, F.J., Martin, F., 1990. Fire-induced transformation of soil organic matter from an oak forest: an experimental approach to the effects of fire on humic substances. Soil Sci. 149, 158–168. Atanassova, I., Doerr, S.H., 2011. Changes in soil organic compound composition associated with heat-induced increases in soil water repellency. Eur. J. Soil Sci. 62 (4), 516–532. https://doi.org/10.1111/j.1365-2389.2011.01350.x. Avagyan, A., Runkle, B.R.K., Kutzbach, L., 2014. Application of high-resolution spectral absorbance measurements to determine dissolved organic carbon concentration in remote areas. J. Hydrol. 517, 435–446. https://doi.org/10.1016/j.jhydrol.2014.05. 060. Badía-Villas, D., González-Pérez, J.A., Aznar, J.M., Arjona-Gracia, B., Martí-Dalmau, C., 2014. Changes in water repellency, aggregation and organic matter of a mollic horizon burned in laboratory. Soil depth affected by fire. Geoderma 213, 400–407. https://doi.org/10.1016/j.geoderma.2013.08.038. Balesdent, J., Basile-Doelsch, I., Chadoeuf, J., Cornu, S., Derrien, D., Fekiacova, Z., Hatté, C., 2018. Atmosphere – soil carbon transfer as a function of soil depth. Nature. https://doi.org/10.1038/s41586-018-0328-3. Bennett, L.T., Aponte, C., Tolhurst, K.G., Löw, M., Baker, T.G., 2013. Decreases in standing tree-based carbon stocks associated with repeated prescribed fires in a temperate mixed-species eucalypt forest. For. Ecol. Manag. 306, 243–255. https:// doi.org/10.1016/j.foreco.2013.06.036. Bennett, L.T., Aponte, C., Baker, T.G., Tolhurst, K.G., 2014. Evaluating long-term effects of prescribed fire regimes on carbon stocks in a temperate eucalypt forest. For. Ecol. Manag. 328, 219–228. https://doi.org/10.1016/j.foreco.2014.05.028. Bernhard-Reversat, F., 1999. The leaching of Eucalyptus hybrids and Acacia auriculiformis leaf litter: laboratory experiments on early decomposition and ecological implications in congolese tree plantations. Appl. Soil Ecol. 12, 251–261. Bradstock, R.A., Boer, M.M., Cary, G.J., Price, O.F., Williams, R.J., Barrett, D., Cook, G., Gill, A.M., Hutley, L.B.W., Keith, H., Maier, S.W., Meyer, M., Roxburgh, S.H., RussellSmith, J., 2012. Modelling the potential for prescribed burning to mitigate carbon emissions from wildfires in fire-prone forests of Australia. Int. J. Wildland Fire 21 (6), 629. https://doi.org/10.1071/WF11023. Brandstetter, A., Sletten, R.S., Mentler, A., Wenzel, W.W., 1996. Estimating dissolved organic carbon in natural waters by UV absorbance (254 nm). Z. Pflanzenernähr. Bodenkd. 159 (6), 605–607. https://doi.org/10.1002/jpln.1996.3581590612. Breiman, L., 2001. Random forests. 45, 5–32. Buzek, F., Paces, T., Jackova, I., 2009. Production of dissolved organic carbon in forest soils along the north-south European transect. Appl. Geochem. 24 (9), 1686–1701. Carpenter, P.D., Smith, J.D., 1984. Simultaneous spectrophotometric determination of humic acid and iron in water. Anal. Chim. Acta 159, 299–308. https://doi.org/10.
146
Geoderma 341 (2019) 138–147
E.U. Hobley et al.
Temperature sensitivity of black carbon decomposition and oxidation. Environ. Sci. Technol. 44 (9), 3324–3331. https://doi.org/10.1021/es903016y. Park, J.-H., Kalbitz, K., Matzner, E., 2002. Resource control on the production of dissolved organic carbon and nitrogen in a deciduous forest floor. Soil Biol. Biochem. 34 (6), 813–822. https://doi.org/10.1016/S0038-0717(02)00011-1. Peacock, M., Evans, C.D., Fenner, N., Freeman, C., Gough, R., Jones, T.G., Lebron, I., 2014. UV–visible absorbance spectroscopy as a proxy for peatland dissolved organic carbon (DOC) quantity and quality: considerations on wavelength and absorbance degradation. Evnviron. Sci. Process. Impacts 16 (6), 1445–1461. https://doi.org/10. 1039/c4em00108g. Penman, T.D., Towerton, A.L., 2008. Soil temperatures during autumn prescribed burning: implications for the germination of fire responsive species? Int. J. Wildland Fire 17, 572–578. Peuravuori, J., Pihlaja, K., 1997. Molecular size distribution and spectroscopic properties of aquatic humic substances. Anal. Chim. Acta 337 (2), 133–149. https://doi.org/10. 1016/S0003-2670(96)00412-6. Preston, C.M., Schmidt, M.W.I., 2006. Black (pyrogenic) carbon: a synthesis of current knowledge and uncertainties with special consideration of boreal regions. Biogeosciences 3, 367–420. https://doi.org/10.1002/0471776688.ch1. Pretsch, E., Bühlmann, P., Badertscher, M., 2009. Structure Determination of Organic Compounds. Tables of Spectral Data. Springer Berlin Heidelberg, Berlin, Heidelberg. Schmidt, M.W.I., Noack, A.G., 2000. Black carbon in soils and sediments. Analysis, distribution, implications, and current challenges. Glob. Biogeochem. Cycles 14 (3), 777–793. https://doi.org/10.1029/1999GB001208. Schnitzer, M., Hoffman, I., 1964. Pyrolysis of soil organic matter. Soil Sci. Soc. Am. J. 28 (4), 520. https://doi.org/10.2136/sssaj1964.03615995002800040021x. Specht, R., 1981. Foliage Projective Cover and Standing Biomass. Stephens, S.L., McIver, J.D., Boerner, R.E.J., Fettig, C.J., Fontaine, J.B., Hartsough, B.R., Kennedy, P.L., Schwilk, D.W., 2012. The effects of forest fuel-reduction treatments in the United States. Bioscience 62 (6), 549–560. https://doi.org/10.1525/bio.2012.62. 6.6. Strobl, C., Malley, J., Tutz, G., 2009. An introduction to recursive partitioning: rationale, application, and characteristics of classification and regression trees, bagging, and random forests. Psychol. Methods 14, 323–348. Tolhurst, K.G., 2003. Effects of repeated low-intensity fire on the understory of a mixed eucalypt foothill forest in south-eastern Australia. In: Research Report/Fire Management, Dept. of Sustainability and Environment, No. 58. Department of Sustainability and Environment, Victorian Government, Melbourne, Australia. Ecological impacts of fuel reduction burning in dry sclerophyll forest: first progress report. In: Tolhurst, K.G., Flinn, D. (Eds.), Research Report No. 349. Forest Research Department of Conservation and Environment, Australia. Tu, C.-L., Liu, C.-Q., Lu, X.-H., Yuan, J., Lang, Y.-C., 2011. Sources of dissolved organic carbon in forest soils. Evidences from the differences of organic carbon concentration and isotope composition studies. Environ. Earth Sci. 63 (4), 723–730. https://doi. org/10.1007/s12665-010-0741-x. van Gaelen, N., Verschoren, V., Clymans, W., Poesen, J., Govers, G., Vanderborght, J., Diels, J., 2014. Controls on dissolved organic carbon export through surface runoff from loamy agricultural soils. Geoderma 226–227 (Supplement C), 387–396. https:// doi.org/10.1016/j.geoderma.2014.03.018. Vega, J.A., Fernández, C., Fonturbel, T., 2005. Throughfall, runoff and soil erosion after prescribed burning in gorse shrubland in Galicia (NW Spain). Land Degrad. Dev. 16 (1), 37–51. https://doi.org/10.1002/ldr.643. Viscarra Rossel, R., Behrens, T., 2010. Using data mining to model and interpret soil diffuse reflectance spectra. Geoderma 158 (1–2), 46–54. https://doi.org/10.1016/j. geoderma.2009.12.025. Volkova, L., Meyer, C.P., Murphy, S., Fairman, T., Reisen, F., Weston, C., 2014. Fuel reduction burning mitigates wildfire effects on forest carbon and greenhouse gas emission. Int. J. Wildland Fire. https://doi.org/10.1071/WF14009. Wang, F., Li, J., Zou, B., Xu, X., Li, Z., 2013. Effect of prescribed fire on soil properties and N transformation in two vegetation types in South China. Environ. Manag. 51 (6), 1164–1173. https://doi.org/10.1007/s00267-013-0044-6. Williams, R.J., Hallgren, S.W., Wilson, G.W.T., 2012. Frequency of prescribed burning in an upland oak forest determines soil and litter properties and alters the soil microbial community. For. Ecol. Manag. 265, 241–247. https://doi.org/10.1016/j.foreco.2011. 10.032. Worrall, F., Armstrong, A., Adamson, J.K., 2007. The effects of burning and sheep-grazing on water table depth and soil water quality in a upland peat. J. Hydrol. 339 (1–2), 1–14. https://doi.org/10.1016/j.jhydrol.2006.12.025. WRB, 2015. World Reference Base for Soil Resources 2014, Update 2015. International Soil Classification System for Naming Soils and Creating Legends for Soil Maps. Food and Agriculture Organization, Rome, Italy.
Handbooks, 2nd edn. CSIRO Publishing, Australia. Jaffé, R., Ding, Y., Niggemann, J., Vähätalo, A.V., Stubbins, A., Spencer, R.G.M., Campbell, J., Dittmar, T., 2013. Global charcoal mobilization from soils via dissolution and riverine transport to the oceans. Science (New York, N.Y.) 340 (6130), 345–347. https://doi.org/10.1126/science.1231476. Jardine, P.M., McCarthy, J.F., Weber, N.L., 1989. Mechanisms of dissolved organic carbon adsorption on soil. Soil Sci. Soc. Am. J. 53, 1378–1385. https://doi.org/10. 2136/sssaj1989.03615995005300050013x. Kaiser, K., 1998. Fractionation of dissolved organic matter affected by polyvalent metal cations. Org. Geochem. 28 (12), 849–854. https://doi.org/10.1016/S0146-6380(98) 00046-1. Kaiser, K., Guggenberger, G., Haumaier, L., Zech, W., 1997. Dissolved organic matter sorption on sub soils and minerals studied by 13C-NMR and DRIFT spectroscopy. Eur. J. Soil Sci. 48 (2), 301–310. https://doi.org/10.1111/j.1365-2389.1997.tb00550.x. Kaiser, K., Guggenberger, G., Zech, W., 2001. Organically bound nutrients in dissolved organic matter fractions in seepage and pore water of weakly developed forest soils. Acta Hydrochim. Hydrobiol. 28 (7), 411–419. https://doi.org/10.1002/1521-401X (20017)28:7<411:AID-AHEH411>3.0.CO;2-D. Kalbitz, K., Solinger, S., Park, J.-H., Michalzik, B., Matzner, E., 2000. Controls on the dynamics of dissolved organic matter ins soils. A review. Soil Sci. 165 (4). Kalbitz, K., Schmerwitz, J., Schwesig, D., Matzner, E., 2003. Biodegradation of soil-derived dissolved organic matter as related to its properties. Geoderma 113 (3–4), 273–291. https://doi.org/10.1016/S0016-7061(02)00365-8. Kim, S., Kaplan, L.A., Benner, R., Hatcher, P.G., 2004. Hydrogen-deficient molecules in natural riverine water samples—evidence for the existence of black carbon in DOM. New approaches in Marine Organic Biogeochemistry: A Tribute to the Life and Science of John I. Hedges. 92 (1), 225–234. https://doi.org/10.1016/j.marchem. 2004.06.042. Kirkby, C.A., Kirkegaard, J.A., Richardson, A.E., Wade, L.J., Blanchard, C., Batten, G., 2011. Stable soil organic matter: a comparison of C:N:P:S ratios in Australian and other world soils. Geoderma 163, 197–208. Knapp, B.O., Stephan, K., Hubbart, J.A., 2015. Structure and composition of an oakhickory forest after over 60 years of repeated prescribed burning in Missouri, U.S.A. For. Ecol. Manag. 344, 95–109. https://doi.org/10.1016/j.foreco.2015.02.009. Knicker, H., 2007. How does fire affect the nature and stability of soil organic nitrogen and carbon? A review. Biogeochemistry 85 (1), 91–118. https://doi.org/10.1007/ s10533-007-9104-4. Krishnaraj, S.J., Baker, T.G., Polglase, P.J., Volkova, L., Weston, C.J., 2016. Prescribed fire increases pyrogenic carbon in litter and surface soil in lowland Eucalyptus forests of south-eastern Australia. For. Ecol. Manag. 366, 98–105. https://doi.org/10.1016/ j.foreco.2016.01.038. Lawrence, J., 1980. Semi-quantitative determination of fulvic acid, tannin and lignin in natural waters. Water Res. 14 (4), 373–377. https://doi.org/10.1016/0043-1354(80) 90085-8. Marschner, B., Kalbitz, K., 2003. Controls of bioavailability and biodegradability of dissolved organic matter in soils. Geoderma 113 (3–4), 211–235. https://doi.org/10. 1016/S0016-7061(02)00362-2. Mataix-Solera, J., Cerdà, A., Arcenegui, V., Jordán, A., Zavala, L.M., 2011. Fire effects on soil aggregation. A review. Earth Sci. Rev. 109 (1–2), 44–60. https://doi.org/10. 1016/j.earscirev.2011.08.002. McCaw, W.L., 2013. Managing forest fuels using prescribed fire – a perspective from southern Australia. For. Ecol. Manag. 294, 217–224. https://doi.org/10.1016/j. foreco.2012.09.012. McKnight, D.M., Harnish, H., Wershaw, R.L., Baron, J.S., Schiff, S., 1997. Chemical characteristics of particulate, colloidal, and dissolved organic material in Loch Vale Watershed, Rocky Mountain National Park. Biogeochemistry 36, 99–124. Metting, F.B., 1993. Structure and physiological ecology of soil microbial communities. In: Soil Microbial Ecology. Middelmann, M.H., 2007. Natural Hazards in Australia. Identifying Risk Analysis Requirements. Geoscience Australia, Canberra, Australia. Muqaddas, B., Zhou, X., Lewis, T., Wild, C., Chen, C., 2015. Long-term frequent prescribed fire decreases surface soil carbon and nitrogen pools in a wet sclerophyll forest of Southeast Queensland, Australia. Sci. Total Environ. 536, 39–47. https://doi. org/10.1016/j.scitotenv.2015.07.023. Narayan, C., 2007. Review of CO2 Emissions Mitigation Through Prescribed Burning. European Forest Institute, Finland. Neary, D.G., Klopatek, C.C., DeBano, Leonard F., Ffolliott, Peter F., 1999. Fire effects on belowground sustainability: a review and synthesis. For. Ecol. Manag. 122, 51–71. Neary, D.G., Ryan, K.C., DeBano, L.F., 2005. Wildland Fire in Ecosystems. Effects of Fire on Soils and Water. U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station, Ft. Collins, CO. Nguyen, B.T., Lehmann, J., Hockaday, W.C., Joseph, S., Masiello, C.A., 2010.
147