Microbial soil characteristics of grassland and arable soils linked to thermogravimetry data: correlations, use and limits

Thermogravimetry (TG) is a simple method that enables rapid analysis of soil properties such as content of total organic C, nitrogen, clay and C fractions with different stability. However, the possible link between TG data and microbiological soil properties has not been systematically tested yet and limits TG application for soil and soil organic matter assessment. This work aimed to search and to validate relationships of thermal mass losses (TML) to total C and N contents, microbial biomass C and N, basal and substrate-induced respiration, extractable organic carbon content, anaerobic ammonification, urease activ5 ity, short-term nitrification activity, specific growth rate, and time to reach the maximum respiration rate for two sample sets of arable and grassland soils. Analyses of the training soil set revealed significant correlations of TML with basic soil properties such as carbon and nitrogen content with distinguishing linear regression parameters and temperatures of correlating mass losses for arable and grassland soils. In a second stage the equations of significant correlations were used for validation with an independent second sample set. This confirmed applicability of developed equations for prediction of microbiological proper10 ties mainly for arable soils. For grassland soils was the applicability lower, which was explained as the influence of rhizosphere processes. Nevertheless, the application of TG can facilitate the understanding of changes in soil caused by microorganism’s activity and the different regression equations between TG and soil parameters reflect changes in proportions between soil components caused by land use management. 1 https://doi.org/10.5194/soil-2021-109 Preprint. Discussion started: 10 November 2021 c © Author(s) 2021. CC BY 4.0 License.


TG analyses and TML determination
Prior to TG analyses the air dried and 2 mm sieved soils were stored in a a desiccator to equilibrate at 43 % relative humidity 80 (RH) for 3 weeks prior to analysis to insure comparable conditions for measurement of soils of different origins.
The correlations between TML and soil properties discussed in the introduction were obtained for air dried soils exposed to 76% relative air humidity (RH) prior to TG analysis (Kučerík et al., 2013). Recent results showed that correlations can also be observed for soils exposed to 43% RH; the RH closer to most laboratory conditions and easier to maintain . For this reason, the thermogravimetric experiments were carried out in air enriched to 43 % RH (at 25°C) by passing 85 over an oversaturated solution of potassium carbonate as a reactive gas was used. Around 0.2 g of sample was transferred to an alumina sample holder placed into the thermoscale equipped with an autosampler (TA Instruments Q550, New Castle, Delaware, USA) from laboratory temperature ( 20°C) to 950°C with a heating rate of 5°C min −1 . The flow rate of the reactive gas (air) was 100 mL min −1 . To maintain samples prior to analysis with the same humidity, the thermoscale autosampler was modified and purged with the same air stream as the thermoscale furnace. All samples were analysed in triplicate. Exemplary 90 records are reported in Supporting information, Figure S1.
The obtained dependences of mass loss on temperature were averaged and TMLs were obtained, i.e. in total 93 mass losses in 10°C intervals for each soil sample. In this study, the TMLs are reported with upper temperature limit as a subscript. For example, T M L 100 refers to a thermal mass loss obtained between 90 and 100°C. Mass loss in larger temperature ranges are reported with the whole temperature interval in the subscript, e.g. T M L 200−300 indicating thermal mass loss between 200 and 95 300°C.

Determination of chemical and MB properties of soils
Soil organic carbon (SOC) and total N (TN) were determined by dry combustion using a 28 Series LECO analyser.
MB analyses were carried out on field-moist fresh, sieved samples (<2mm). Soil microbial biomass C and N (C bio , N bio ) was determined by means of the fumigation-extraction method according to (ISO 14240-2, 1997). Soil (10 g on dry basis) 100 was weighed in three replicates and preincubated for one day at 25°C. If the natural water content was lower than 30% of water holding capacity (WHC), water was added to bring the sample to 60 % of WHC. Soil was fumigated with ethanol-free chloroform for 24 h. Fumigated and unfumigated soils were extracted with 40 mL 0.5 M K 2 SO 4 . Estimation of oxidizable C in the extracts was performed photometrically by the dichromate-oxidation method (Yakovchenko and Sikora, 1998). The content of organic C of the non-fumigated soil (C ext ) was considered the labile fraction of soil C. Nitrogen in the extracts was 105 oxidized to nitrate using the alkaline persulfate oxidation (Cabrera and Beare, 1993) and nitrate was measured photometrically at 210 nm (Kandeler, 1993). Coefficients kC=0.38 and kN=0.45 were used to calculate microbial biomass C and N (Joergensen, 1995).
Oxidizable C (C OX ) was estimated using wet digestion of 1 g soil sample with 5 ml 0.27 M K 2 Cr 2 O 7 solution and 7.5 ml concentrated H 2 SO 4 according to (ISO 14235, 1998).

110
Samples were moistened to 60% WHC for R B and to 40% WHC for measurement of R S and respiration curves if the natural water content was lower. This allowed thorough mixing of soil samples with the substrate containing 8.42 g of glucose, 1.37 g of ammonium sulfate and 0.21 g of potassium dihydrogen phosphate. The added amount was 10 mg g −1 for estimation of R S and for measurement of the respiration curves. Samples were incubated 96 h (R B , respiration curves) and 10 h (R S ) in flasks (100 mL for R B , R S ; 500 mL for respiration curves) and the decrease in pressure was regularly recorded, while released CO 2 130 was trapped into a solution of 2.5 M NaOH. Consumed O 2 was calculated using the equation of state of ideal gas (ISO 16072, 2002). The rates of R B and R S were calculated using linear regression, and the specific growth rate µ by means of non-linear regression from the relationship between cumulative O 2 consumption and time (Pell et al., 2005). t peakmax denotes the time elapsed between substrate application and the maximum rate of respiration.
Soil samples were preincubated two days at 25°C prior to estimation of a short-term nitrification activity (SNA) according 135 to (ISO 15685, 2012). Thereafter 60 mL of a medium (pH 7.2) containing potassium phosphate buffer (1 mM), sodium chlorate (15 mM) and ammonium sulfate (3.78 mM) was added, the suspension was shaken and 5 ml of supernatant were taken after two and six hours. To stop nitrification, 5 ml of 4 M KCl was added. The suspension was centrifuged and nitrite was analysed in the supernatant using the method with sulfanilamide and N-(1-naphthyl-ethylenediamine) dihydrochloride (Forster, 1995).

Statistical data treatment 140
A two-step process was used in connecting TML/LTML's with measured soil properties.
In the first step, 93 TMLs were obtained for each soil using TG. Then, the Pearson correlation between the soil parameters (i.e. URE) and TMLs obtained in a specific temperature range (i.e. T M L 40−50 ) were searched using the training set (either grassland (11 samples) or arable soils (21 samples)). Then the search continued for other TMLs in the temperature range 30-600°C. If the p-value (probability) was not p ≤ 0.05 then two or more TMLs were used to find the best correlation.

145
In the second step multilinear regression models were developed based on the significant correlations found from step 1.
The general equation of the regression model could be express as Where Y is the dependent variable, i.e. soil property, A 0 is intercept, A 1 -A n are regression coefficients and X 1 -X n are independent variables i.e. mass losses. The statistical criteria used for selection or removal of variables from regressions were 150 based on either the significance (probability) of the F value, or the F value itself. The parameters, which did not correlate with a TML at significance level < 5% were subjected to further analysis, in which two or more TMLs were involved.
The developed equations were verified using verification set consisting of 5 grassland and 10 arable soils. The criterion for testing the significance of the regression model was correlation coefficient r and significance level p ≤ 0.05.
All the correlations were carried out in Excel and Statistica using a 95% confidence interval. Based on the correlations, the 155 regression equations were developed and verified using soils in verification sets.

165
It can be seen that the individual TMLs are parts of usually wider temperature interval in which the correlation coefficients are high. Comparison of results in both figures suggests slightly higher correlations between MB parameters and TML in arable soils. This is also confirmed in Table 1, which reports the multilinear regressions between TML and MB with the highest correlation coefficients and significance level ≤ 5% (Table 1). As aforementioned, the parameters, which did not correlate with a TML at significance level < 5% were subjected to further analysis, in which two or more TMLs were involved (Equation 1).

170
The results are also reported in Table 1. and soil properties. Also in this case can be seen that some LTML correlate with MB soil properties.

The correlations between mass losses and soil parameters
As it can be seen the training sample set provided correlation between soil microbiological parameters and TG data in both soil sets. The closeness of correlation for TMLs and LTMLs is similar for arable and grass land soils with few exceptions. A 190 very significant difference was observed for TN, which can be explained based on work of (Trasar-Cepeda et al., 1997), who demonstrated that relatively undisturbed soils are characterized by an equilibrium between TN and various biological activity characteristics such as microbial biomass C, N-mineralization capacity, and the activities of phosphatase, β-glucosidase and urease. The equilibrium can be disrupted by various chemical stresses (contamination, pH alteration) and physical disturbances (tillage, wet-dry or freeze-thaw cycles) Chaer et al. (2009) and agricultural activity (Miguéns et al., 2007). As a result, TN is 195 sensitive to many management practices as fertilization, tillage and others with (perhaps temporary) reduction of correlation N BIO were closely related to SOM content and quality (Chaer et al., 2009). Shifts in composition of SOM and microbial community could reduce their predictability in arable soils by disturbances in regulation processes by same way as for carbon and nitrogen.
The correlation between respiration and TML has already been demonstrated for soils exposed to 76% RH prior to TG analysis (Kučerík et al., 2013;Siewert et al., 2012). The correlation coefficients increased with microbial respiration measurements, 205 which enables prediction of microbiological activity using T M L 100 or T M L 300 (Kučerík and Siewert, 2014). The results obtained in the current study confirm the earlier conclusion about the influence of agricultural practices on this relationship (Siewert et al., 2012).
Substrate-induced respiration (R s ) identifies metabolically active components of the soil microbial community. Our results show a close correlation with thermal mass losses (TML).
210 Table 3. The equations obtained from correlation between MB and LTMLs with the highest Pearson correlation coefficients as reported in Table 3. The anaerobic ammonification (AMO) reflects microbial mineralisation of organic nitrogenous compounds. If chemolithotrophs perform nitrification without using organic matter for growth, it is difficult to explain the close correlation between TML and nitrification activity (SNA) for both grassland and arable soils. We can only speculate about the indirect impact of SOM e.g.
via binding of ammonium as a substrate for nitrifiers or its exchange from clays.
Urease activity (URE) is known for its sensitivity to soil management, cropping history (causing its decrease) and man-215 agement practice in general, organic matter content (increases upon organic fertilization), soil depth, heavy metals, and environmental factors such as temperature and pH (Martinez-Salgado et al., 2010). Interestingly, URE showed slightly better correlations for arable soils and at different TML than grassland soils. Somewhat unclear are the contradictory results of correlation between TML and URE (mainly on grassland but also arable soils), which may be explained by site-and scale-dependent effect of putative soil changes on URE. We presume this may be caused (among others) by random sampling incoherence sim-220 ilarly as referred by (Corstanje et al., 2007), who showed that urease activity and SOC were found to be uncorrelated at shorter spatial scales (≤ 1 m) but significantly positively correlated at longer scales of (≥15 m).

The verification of results
As it could be observed in Table 4, the validation confirmed mostly the relationships for arable soils, which properties modelled using TG correlated similarly as the training set. Better results were observed for TML -based equations than for LTML. For 225 the grass land soils, the verification confirmed validity only of TML-based equations for C OX and respiration, and surprisingly also for N BIO . LTML-based equations, good results were observed only for C OX .
In our previous works, we analysed >300 untouched soils of various origin and composition sampled all over the word e.g. (Kučerík et al., 2018;Kučerík and Siewert, 2014;Siewert and Kučerík, 2015). We have demonstrated that the TML and LTML are useful for the determination of SOC, TN and soil organic matter fractions. In all cases, the temperatures of TMLs 230 correlating with SOC and TN were consistent across soil types and locations.
In the current work, we observed some significant correlations in the training soil sample set, but the validation of the results was not always successful. This may be related to several reasons.
1. MB activity has rather the long-term than short-term effect on amount (mass loss) of soil organic matter. In other words, the MB activity responses quickly to soil conditions by composition of soil enzymes, but the effect on SOM content is 235 slower. Second issue is related to the soil sampling. The grass land soils can be considered as very stable and protected against physical perturbation (Jensen et al., 2019), and the correlations between TML's and soil properties are always stronger . However, although the grassland soils were sampled at the same time as arable soils, the sampling was carried out under permanent vegetation. This implies that the inputs of rhizosphere was more significant part of grassland soils than arable soils, which influenced the validation negatively. In comparison to bulk soil, rhizosoil 240 is richer in soil microorganisms, loosen separated plant cells and by roots exudates such as, organic acids, proteins and sugars (Hütsch et al., 2002). These rhizosphere inputs significantly influenced the results of a conventional analysis of soil microbiological properties.