CIAT Research Online - Accepted Manuscript Assessing the sensitivity and repeatability of permanganate oxidizable carbon as a soil health metric: An interlab comparison across soils The International Center for Tropical Agriculture (CIAT) believes that open access contributes to its mission of reducing hunger and poverty, and improving human nutrition in the tropics through research aimed at increasing the eco-efficiency of agriculture. CIAT is committed to creating and sharing knowledge and information openly and globally. We do this through collaborative research as well as through the open sharing of our data, tools, and publications. Citation: Wade, J.; Maltais-Landry, G.; Lucas, D. E.; Bongiorno, G.; Bowles, T. M.; Calderón, F. J.; Culman, S. W.; Daughtridge, R.; Ernakovich, J. G.; Fonte, S. J.; Giang, D.; Herman, B. L.; Guan, L.; Jastrow, J. D.; Loh, B. H.H.; Kelly, C.; Mann, M. E.; Matamala, R.; Miernicki, E. A.; Peterson, B.; Pulleman, M. M.; Scow, K. M.; Snapp, S. S.; Thomas, V.; Daoyuan Wang, X. T.; Jelinski, N. A.; Liles, G. C.; Barrios-Masias, F. H.; Rippner, D. A.; Silveira, M. L. & Margenota, A. J. 2020. Assessing the sensitivity and repeatability of permanganate oxidizablecarbon as a soil health metric: An interlab comparison across soils. Geoderma. In press Publisher’s DOI: https://doi.org/10.1016/j.geoderma.2020.114235 Access through CIAT Research Online: https://hdl.handle.net/10568/107131 Terms: © 2019. CIAT has provided you with this accepted manuscript in line with CIAT’s open access policy and in accordance with the Publisher’s policy on self-archiving. This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. You may re-use or share this manuscript as long as you acknowledge the authors by citing the version of the record listed above. You may not change this manuscript in any way or use it commercially. For more information, please contact CIAT Library at CIAT-Library@cgiar.org. Title 1 Assessing the sensitivity and repeatability of permanganate oxidizable carbon as a soil 2 health metric: an interlab comparison across soils 3 Authors 4 Jordon Wade1, Gabriel Maltais-Landry2, Dawn E. Lucas2, Giulia Bongiorno3,4, Timothy 5 M. Bowles5, Francisco J. Calderón6, Steve W. Culman7, Rachel Daughtridge8, Jessica 6 G. Ernakovich9, Steven J. Fonte10, Dinh Giang9, Bethany L. Herman7, Lindsey Guan5, 7 Julie D. Jastrow12, Bryan H.H. Loh9; Courtland Kelly10; Meredith E. Mann7; Roser 8 Matamala12, Elizabeth A. Miernicki1, Brandon Peterson6, Mirjam M. Pulleman3,13, Kate 9 M. Scow11, Sieglinde S. Snapp14,15, Vanessa Thomas14, Xinyi Tu14, Daoyuan Wang11, 10 Nicolas A. Jelinski16, Garrett C. Liles17, Felipe H. Barrios-Masias18, Devin A. Rippner11, 11 Maria L. Silveira2,19, Andrew J. Margenot1 12 13 Affiliations 14 1 Department of Crop Sciences, University of Illinois at Urbana-Champaign 15 2 Soil and Water Sciences Department, University of Florida 16 3 Soil Biology Group, Wageningen University, the Netherlands 17 4 Department of Soil Science, Research Institute of Organic Agriculture (FiBL) 18 5 Department of Environmental Science, Policy, & Management, University of California, 19 Berkeley 20 6 Central Great Plains Research Station, USDA-ARS 21 7 School of Environment & Natural Resources, the Ohio State University 22 8 Department of Natural Resources and Environmental Sciences, University of Illinois at 23 Urbana-Champaign 24 9 Department of Natural Resources and the Environment, University of New Hampshire; 25 10 Department of Soil and Crop Sciences, Colorado State University 26 11 Department of Land, Air and Water Resources, University of California Davis 27 12 Environmental Science Division, Argonne National Laboratory 28 13 International Center for Tropical Agriculture (CIAT), Colombia 29 14 Department of Plant, Soil and Microbial Sciences, Michigan State University 30 15 Center for Global Change and Earth Observations, Michigan State University 31 16 Department of Soil, Water and Climate, University of Minnesota 32 17 College of Agriculture, California State University Chico 33 18 Department of Agriculture, Veterinary and Rangeland Sciences, University of Nevada, 34 Reno 35 19 Range Cattle Research and Education Center, University of Florida 36 37 Abstract 38 Soil organic matter is central to the soil health framework. Therefore, reliable indicators 39 of changes in soil organic matter are essential to inform land management decisions. 40 Permanganate oxidizable carbon (POXC), an emerging soil health indicator, has shown 41 promise for being sensitive to soil management. However, strict standardization is 42 required for widespread implementation in research and commercial contexts. Here, we 43 used 36 soils—three from each of the 12 USDA soil orders—to determine the effects of 44 sieve size and soil mass of analysis on POXC results. Using replicated measurements 45 across 12 labs in the US and the EU (n = 7,951 samples), we quantified the relative 46 importance of 1) variation between labs, 2) variation within labs, 3) effect soil mass, and 47 4) effect of soil sieve size on the repeatability of POXC. We found a wide range of 48 overall variability in POXC values across labs (0.03 to 171.8%; mean = 13.4%), and 49 much of this variability was attributable to within-lab variation (median = 6.5%) 50 independently of soil mass or sieve size. Greater soil mass (2.5 g) decreased absolute 51 POXC values by a mean of 177 mg kg-1 soil and decreased analytical variability by 52 6.5%. For soils with organic carbon (SOC) > 10%, greater soil mass (2.5 g) resulted in 53 more frequent POXC values above the limit of detection whereas the lower soil mass 54 (0.75 g) resulted in POXC values below the limit of detection for SOC contents < 5%. A 55 finer sieve size increased absolute values of POXC by 124 mg kg-1 while decreasing the 56 analytical variability by 1.8%. In general, soils with greater SOC contents had lower 57 analytical variability. These results point to potential standardizations of the POXC 58 protocol that can decrease the variability of the metric. We recommend that the POXC 59 protocol be standardized to use 2.5 g for soils < 10% SOC. Sieve size was a relatively 60 small contributor to analytical variability and therefore we recommend that this decision 61 be tailored to the study purpose. Tradeoffs associated with these standardizations can 62 be mitigated, ultimately providing guidance on how to standardize POXC for routine 63 analysis. 64 Acknowledgements 65 We are grateful for the contributions of Cheryl Mackowiak, Jehangir H. Bhadha, Ashley 66 Smyth, and Yuncong Li in helping source several soils for this study. Additionally, we 67 are grateful to Willeke van Tintelen for technical assistance in the laboratory. Rachel C. 68 Daughtridge’s work was supported in part by the U.S. Department of Energy, Office of 69 Science, Office of Workforce Development for Teachers and Scientists (WDTS) under 70 the Science Undergraduate Laboratory Internships Program (SULI). 71 1. Introduction 72 Soil organic matter is a vital component of ecosystem functioning (Schmidt et al., 2011), 73 as well as crop production (Oldfield et al., 2019; Sanderman et al., 2017). For decades, 74 the characterization of organic matter in soils has been predominantly described 75 through chemical extractions or fractionations (Lehmann and Kleber, 2015). 76 Increasingly, this paradigm is being left behind in favor of a model that integrates 77 chemical composition, physical accessibility, and biological activity to describe organic 78 matter dynamics (Blankinship et al., 2018; Dungait et al., 2012; Lehmann et al., 2008). 79 80 The soil health framework emphasizes the degree to which dynamic properties of a soil 81 can be optimized for multifunctionality. While there are intrinsic properties of each soil, 82 the chemical, physical, and biological components of soil organic matter form the core 83 of soil health (Lal, 2016). Soil health seeks to unify these previously disparate 84 components of soil into a cohesive framework for implementation in agroecosystems 85 (Kibblewhite et al., 2008). To aid in the implementation of this framework, a novel set of 86 metrics are being developed with usability by land managers as one of the central goals 87 (Doran and Zeiss, 2000). As these metrics undergo development, vetting, and 88 calibration, an emerging soil health indicator is the fraction of carbon (C) that is oxidized 89 by potassium permanganate (KMnO4), which we will refer to here as permanganate 90 oxidizable C (POXC) (Culman et al., 2012; Moebius-Clune et al., 2017). While POXC is 91 a chemically-defined fraction of soil C, it is often thought or proposed to be reflective of 92 a biologically active pool (Moebius-Clune et al., 2017; NRCS, 2019) and has been 93 shown to be related to microbial community composition (Ramírez et al., 2019). Recent 94 work has also shown that POXC is positively related to aggregate stability (Fine et al., 95 2017; Wade et al., 2019) and inversely related to dispersible clays (Jensen et al., 2019), 96 suggesting a potential physical component of this measurement as well. This 97 interrelatedness of POXC to multiple components of soil health—as well its sensitivity to 98 changes in management, potential for high throughput, and relatively low equipment 99 costs—have made it an attractive metric for soil health assessments (Bongiorno et al., 100 2019). One unique aspect of the soil health framework is the focus on indicator usability 101 and interpretability. Soil health indicators must provide information that is both reliable 102 and actionable for land managers in their decision making process. 103 104 For decades, permanganate oxidation has been used to describe the portion of soil 105 organic matter that is thought to turnover quickly and have a relatively short residence 106 time (Blair et al., 1995; Matsuda and Schnitzer, 1972; Willard et al., 1956). Oxidation by 107 relatively dilute solutions (<0.5 mol L-1) of permanganate (MnO4-) has been used to 108 describe both C (Lefroy et al., 1993; Loginow et al., 1987) and N dynamics (Bundy and 109 Bremner, 1973; Carski and Sparks, 1987). The concentration of MnO4- used in these 110 evaluations have varied over an order of magnitude, ranging from 20 mmol L-1 to 333 111 mmol L-1 (Loginow et al., 1987; Weil et al., 2003). Higher concentrations with longer 112 shaking times have been found to produce inconsistent results (Tirol-Padre and Ladha, 113 2004) and are less sensitive to changes in management (Lucas and Weil, 2012; Weil et 114 al., 2003), prompting the use of lower concentrations and shorter reaction times. While 115 there is a broad consensus to use 20 mL of 0.02 mol L-1 MnO4-, slight variations in 116 shaking time still exist, with both 12 (Culman et al., 2012; Hurisso et al., 2016; Weil et 117 al., 2003) and 10 minutes (Bongiorno et al., 2019; Moebius-Clune et al., 2017; NRCS, 118 2019) of total reaction time being utilized. However, even these slight variations are 119 indicative of a broader convergence from previous times of up to 24 hours (Tirol-Padre 120 and Ladha, 2004). 121 122 Although there has been convergence on the concentration of the solution and reaction 123 time, other potential methodological considerations have been less studied. Of 124 particular interest for standardization are the mass of soil reacted and the sieve size 125 through which that soil has been passed. These methodological decisions have the 126 potential to influence both absolute values of POXC (i.e. sensitivity), as well as the 127 analytical variability (i.e. repeatability). Balancing these considerations is essential to 128 ensure reliable quantification across edaphic contexts. More broadly, these 129 methodological decisions have implications for the utility of POXC to inform land 130 management decisions. Here, we will examine how soil mass and sieve size influence 131 absolute POXC values and the analytical variability. In order to differentiate between 132 treatment effects and lab or operator effects, soils were sent to twelve labs in the United 133 States and Europe. The objectives of this study were to determine 1) changes in 134 absolute POXC values associated with soil mass and sieve size decisions, 2) the range 135 of potential within-lab variability of POXC, and 3) the relative contributions of soil mass 136 and sieve size to analytical variability of POXC. To evaluate the robustness of these 137 findings, we examined these sources of variability using three soils from each of the 138 twelve soil orders of the USDA classification system (n = 36 soils total). 139 140 We hypothesized that a decrease in soil mass from 2.5 g to 0.75 g would increase 141 absolute POXC values due to the greater ratio of oxidant (MnO4-) to substrate (soil 142 organic C [SOC]). However, we hypothesized that this lower soil mass would result in 143 greater variability from sample to sample (i.e. between analytical reps), increasing the 144 analytical variability. We also hypothesized that decreasing sieve size from < 2.0 mm to 145 < 0.5 mm would expose physically occluded organic matter to oxidation, increasing 146 absolute values of POXC and would produce more consistent values. Finally, we 147 hypothesized that within-lab variability would be the largest contributor to the variability 148 of the metric, i.e. that internal lab practices would outweigh the variability associated 149 with methodological considerations. 150 151 2. Materials and Methods 152 2.1. Soil sampling and characterization 153 Three soils for each of the twelve USDA orders were obtained from a combination of 154 archived collections and field sampling. Surface A horizons were sampled for mineral 155 soils and O horizons were obtained for Histosols and Gelisols. Soils were air-dried and 156 sieved to < 2 mm prior to characterization for chemical and physical properties. Soil pH 157 was determined in water (1:2 m/v) after equilibrating for 30 min (Thomas, 1996). Soil 158 texture was determined by the hydrometer method, using overnight shaking (16 h) in 159 sodium hexametaphosphate to disperse mineral particles (Bouyoucos, 1962). Total 160 organic C was determined by dry combustion chromatography (Nelson and Sommers, 161 1996) and soil organic C (SOC) was estimated for soils after gravimetric determination 162 of potential carbonates using dilute HCl (Harris et al., 2001). 163 164 Soils analyzed in the current study included three soils from each of the twelve USDA 165 soil orders and encompassed a wide range of soil physicochemical properties (Table 1). 166 Soil organic carbon contents ranged from 0.21% to 37.7% by mass (median = 2.3%, 167 mean = 6.4%). Clay contents ranged from 0.0 g kg soil-1 to 716.5 g kg-1 soil (median = 168 226.6 g kg-1 soil, mean = 244.3 g kg-1 soil) and sand contents ranged from 66.5 g kg-1 169 soil to 949.8 g kg-1 soil (median = 404.7 g kg-1 soil, mean = 439.1 g kg-1 soil). Further 170 soil physicochemical properties are summarized in Table 1. Soil series and USDA 171 taxonomic classification information are summarized in Table S1. 172 173 2.2. Sample processing and distribution 174 In order to minimize artifacts due to soil processing, we air-dried, hand sieved the soil 175 by gently pressing soil through < 0.5 mm or < 2.0 mm sieves, and then homogenized 176 the sample before sending to twelve different laboratories in the US and Europe. Sieve 177 sizes of < 0.5 mm and < 2.0 mm were based on the two most common sieve sizes used 178 for high-grade chemical analyses (e.g. synchrotron, mass spectroscopy) and in 179 commercial soil test labs, respectively. Then, each participating laboratory performed 180 the KMnO4 oxidation on five analytical replicates using a mass of 0.75 g (±0.02 g) or 181 2.50 g (±0.05 g). The mass of 0.75 g was empirically derived a priori as the mass for 182 which low SOC soils produced detectable levels of POXC and high SOC soils were 183 within the maximum limit of quantification (e.g. 4800 mg POXC kg-1 for 0.75 g soil). All 184 oxidations were performed by the same operator within each laboratory. 185 186 2.3. Permanganate oxidation and reading by colorimetry 187 We performed each oxidation using the methods of Lucas and Weil (2012), specifically 188 the protocol outlined by Culman et al. (2012) (https://lter.kbs.msu.edu/protocols/133). In 189 brief, soils were weighed into 50 mL polypropylene tubes prior to the oxidation step. We 190 initiated the oxidation reaction by adding 18 mL of deionized water and 2 mL of 0.2 mol 191 L-1 KMnO4 (final reaction concentration = 0.02 mol L-1 MnO4-) to each tube containing 192 pre-weighed soil, shaking for exactly 2 min on a reciprocal shaker and allowing to settle 193 for exactly 10 minutes. After settling for exactly 10 min, we terminated the reaction by 194 transferring 0.5 mL of supernatant into a fresh 50 mL tube with 49.5 mL of deionized 195 water, which we then inverted to mix thoroughly, resulting in a homogenized 1:100 196 dilution. Since oxidation occurs over time and total C oxidization is time sensitive, 197 consistency in the timing of termination between replicates and across batches is 198 essential to reproducible measurements. To minimize variability in reaction termination 199 time, the five analytical replicates were run in sequence. The 1:100 dilution was then 200 transferred to either microcuvettes or a 96-well plate reader and analyzed by UV-Vis 201 spectrophotometry to quantify MnO4- remaining in solution by absorbance at 550 nm. 202 After the reaction has been terminated and the 1:100 dilution completed, the MnO4- in 203 solution should be consistent and the subsequent quantification step is less time 204 sensitive. No difference in analytical variability or absolute values were found between 205 readings from microcuvettes and readings from 96-well plate readers (F1,10 = 1.7, p = 206 0.225, data not shown). Adjusted absorbance was then calculated by subtracting the 207 mean of three blanks from the raw absorbance for each sample. 208 209 We then used the adjusted absorbance to calculate the total POXC using the following 210 equation: 211 212 POXC &mg kg-1 soil0= 230.02 mol L-1- 8a + &b × Absadj0@A × &9000 mg C mol-10 × (0.02 L solution)HMass (kg) (Eq. 1) 213 214 where 0.02 mol L-1 is the initial concentration of the oxidation solution, a is the intercept 215 of the standard curve, b is the slope of the standard curve, Absadj is the adjusted 216 absorbance, 9000 mg C mol-1 is the assumed mass of C oxidized by 1 mol of Mn7+ 217 oxidizing to Mn4+ (Weil et al., 2003), 0.02 L is the volume of solution in the oxidation 218 step, and Mass is the mass of soil (in kg) reacted in the tube. To maximize consistency 219 between labs, we calculated a and b for each batch and then used the resulting 220 equations to simultaneously calculate POXC values from adjusted absorbance values 221 (Absadj). To maximize consistency across labs, we used the same KMnO4 222 concentrations to construct all standard curves: 0.020 mol L-1, 0.015 mol L-1, 0.010 mol 223 L-1, and 0.005 mol L-1. Using the theoretical maximum of MnO4- reduced per unit of soil 224 mass (9000 mg C mol-1 MnO4-), we considered values between 0 and 1440 mg kg-1 soil 225 valid for the 2.5 g soil treatment and 0 to 4800 mg kg-1 soil valid for the 0.75 g soil 226 treatment. Values outside of this range were excluded from consideration in 227 measurements of analytical variability, but were used to calculate detection rates. All 228 values are expressed on an air-dried weight basis. 229 230 2.4. Statistical Analyses 231 All statistical analyses were run in RStudio version 1.2.5001 (RStudio Team, 2019). 232 Absolute values were determined by averaging all values for each combination of soil, 233 mass, and sieve size. Averaging for each group was performed using the group_by( ) 234 and summarise( ) command in the dplyr package (Wickham et al., 2019). 235 236 Given that the sieving was performed in one location, we considered these pre-237 oxidation treatments—mass and sieve size—fixed effects in our statistical model. Thus, 238 we assume that any variation due to sieve size and mass occur independently of any 239 lab-specific variation, i.e. these effects are not nested within each laboratory. Similarly, 240 the variability attributed to each soil was considered fixed and not nested within each 241 laboratory. Thus, our initial linear model to assess sources of variation was: 242 243 𝑌 = 𝐿𝑎𝑏 + 𝑆𝑜𝑖𝑙 × 𝑀𝑎𝑠𝑠TUVWTXVYT × 𝑆𝑖𝑒𝑣𝑒 𝑆𝑖𝑧𝑒TUVWTXVYT + ℇ (Eq. 2) 244 245 where Y is the coefficient of variation (CV; expressed as a %) of the five analytical 246 replicates. We then eliminated interaction terms that did not significantly contribute to 247 analytical variability (p > 0.10) to develop our reduced model. Sieve size was retained 248 due to its interaction with Soil (Table 4). 249 250 To determine whether each soil should be assessed individually or grouped by soil 251 order, we compared a reduced model where Soil referred to each of the 36 unique soils 252 with a reduced model where Soil referred to the USDA Soil Order. We compared these 253 reduced models using the Akaike Information Criteria (AIC) (Akaike, 1974) and the 254 likelihood ratio test (Perneger, 2001), both of which indicated that characterizing each 255 soil uniquely was a more parsimonious model fit. Analytical variability by soil order can 256 be found in Table S2 and Figure S1. 257 258 Our final reduced model was estimated using the lm( ) command. The resulting linear 259 regression was log transformed to meet linear regression assumptions. Statistical 260 significance of each variable was determined with the Anova( ) command in the car 261 package (Fox and Weisberg, 2019). Normality of residuals was ascertained using the 262 shapiro.test( ) command to conduct a Shapiro-Wilk test of normality. Additional 263 verification of normality of residuals, homogeneity of variance, and the leverage of 264 individual observations was performed visually using the plot( ) command. We 265 quantified relative contributions of each variable (i.e. lab, soil, or soil mass/sieve size 266 treatment) to overall analytical variability using the estimated marginal means for each 267 variable. Estimated marginal means (also referred to as least-square means) for each 268 factor or set of factors in the final reduced model were obtained using the emmean( ) 269 command in the emmeans package (Lenth, 2019). Therefore, each estimate of 270 analytical variability—whether by soil, lab, soil ´ sieve size, or otherwise—is an 271 estimated marginal mean, which accounts for other sources of variability in our model 272 (Eq. 2). In order to properly determine effect sizes of each treatment and the relative 273 precision associated with that treatment, we used estimation statistics, rather than 274 significance testing (Ho et al., 2019a). Unlike significance testing, which uses p-values 275 to measure against the null hypothesis of no difference, estimation statistics uses a 276 bootstrapped estimate of effect size and the precision associated with that effect size. 277 This circumvents overreliance on p < 0.05, which can often lead to a lack of 278 reproducibility of effects (Halsey et al., 2015), particularly when using categorical 279 variables, e.g. soil mass and sieve size. Estimation statistics were performed using the 280 dabest( ) command in dabestr package (Ho et al., 2019b). Due to the highly skewed 281 distributions of mass and sieve size effects, robustness to outliers was ensured by using 282 5000 bias-corrected and accelerated bootstrap resamplings (Efron, 1987) and using the 283 median, rather than the mean value. Estimated effect sizes and the associated 95% 284 confidence intervals are quantified and included in each plot. 285 286 3. Results and Discussion 287 3.1. Soil characterization and POXC values 288 Detectable POXC values—defined as 0 to 1440 mg kg-1 soil for the 2.5 g soil treatment 289 and 0 to 4800 mg kg-1 soil for the 0.75 g soil treatment—spanned nearly the entire 290 detection range for both masses. For the 2.5 g mass, POXC values ranged from 4 to 291 1406 mg POXC kg-1 soil (median = 427 mg POXC kg-1 soil, mean = 504 mg POXC kg-1 292 soil). For the 0.75 g mass, POXC values ranged from 0.8 mg to 4790 mg POXC kg-1 soil 293 (median = 651 mg POXC kg-1 soil, mean = 1161 mg POXC kg-1 soil), indicating a 34% 294 increase in the median measured POXC value by decreasing the mass of soil analyzed. 295 Average POXC values and analytical variability for each combination of soil and 296 treatment are listed in Table 2. The median absolute deviation—which is similar to 297 standard deviation, but is more robust for our skewed data—ranged from 18 mg POXC 298 kg-1- soil to 1236 mg POXC kg-1 soil (Table 2). The wide range of absolute differences in 299 POXC values underscores how much the POXC values can vary for the same soil. The 300 95% confidence interval for the median absolute deviation was 126 to 178 mg POXC 301 kg-1 soil (mean=152 mg POXC kg-1 soil, data not shown) indicating that absolute 302 deviations of ~150 mg POXC kg-1 soil are likely to occur. Since greater deviations tend 303 to happen for soils with higher average POXC values, there is a need to relativize error, 304 such as using the coefficient of variation or a percent change. Given the range of soil 305 properties and detected POXC values, these values are likely representative of a broad 306 set of edaphic conditions. 307 308 3.2. Methodological effects on absolute values of POXC 309 One of the most common processing treatments for soils is sieving or grinding of the 310 soils prior to analysis. The resulting sieve size classes are especially salient in the 311 measurement of soil C fractions, due to the physical occlusion of C within soil structures 312 (von Lützow et al., 2008). Accordingly, we hypothesized that the smaller sieve size 313 (<0.5 vs < 2.0 mm) would increase POXC values, a hypothesis that was largely 314 confirmed (Figure 1a). On average, the < 0.5 mm sieve size increased POXC values by 315 124 mg POXC kg-1 soil (median = 56 mg POXC kg-1 soil). This is a further extension of 316 the results of Hurisso et al. (2018b), who found that decreasing sieve size from < 8 mm 317 to < 2 mm increased POXC values an average of 141 mg POXC kg-1 soil across three 318 soils. However, our results show that the size of this effect can vary considerably, from -319 188 mg POXC kg-1 soil (i.e. a decrease in POXC) to 876 mg POXC kg-1 soil. On a 320 relative basis, this is an average of 4.0% greater POXC values at the < 0.5 mm sieve 321 size than at the < 2.0 mm (range = -272.6% to 37.6%). It is expected that smaller sieve 322 size would increase POXC values and that this would vary, but not that POXC values 323 would decrease (6 of 36 soils). While there was no apparent physicochemical 324 characteristic(s) underlying these six soils (soils 11, 17, 22, 29, 30, and 33) that could 325 explain this unexpected result, five of the six decreases were for soils with < 100 mg 326 POXC kg-1 soil or < 20% relative decrease. Therefore, while < 0.5 mm sieve size has an 327 inconsistent direction on POXC values, the resulting analyses are likely to be 328 substantively similar. This finding also suggests that it may not be possible to 329 standardize sieve size in a “one size fits all” approach. 330 331 In addition to decisions regarding sieve size, the mass of soil required for analysis is 332 vital to consider in the standardization of POXC because it can change the amount of 333 SOC present in the sample for oxidation by a fixed amount of MnO4-. Our hypothesis 334 that a decrease in soil mass from 2.5 g to 0.75 g would increase POXC values was 335 mostly confirmed. The 0.75 g mass had POXC values that were 177 mg POXC kg-1 soil 336 greater than the 2.5 g mass (median = 111 mg POXC kg-1 soil), an average increase of 337 32.4% (range = -5.0% to 114.6%). This trend was consistent for 35 of the 36 soils, with 338 the exception of soil 31 for which POXC slightly decreased by 16 mg POXC kg-1 soil. 339 While this ratio of soil (and soil SOC) to 0.02 mol L-1 KMnO4 has not been specifically 340 evaluated for POXC, the effect of soil-to-solution ratios has been shown to influence 341 absolute values for a variety of soil chemical extractions, including extractable organic 342 carbon (Kaiser et al., 2001), inorganic N (Li et al., 2012), phosphate (Fuhrman et al., 343 2005), heavy metals (Yin et al., 2002), and even pH (Hendershot et al., 1993). 344 345 Standardization of a soil mass-to-solution ratio for POXC may provide consistency 346 among studies and should be more thoroughly investigated. Previous work using 347 KMnO4 to measure labile C standardized the analysis on the basis of total SOC, rather 348 than on a soil mass basis (Blair et al., 1995). However, this practice has gone largely 349 unadopted due to the increased labor associated with its use. Although standardizing on 350 a mass basis is scientifically rigorous, it may be less amenable to the high-throughput 351 context of commercial soil test labs, which often employ a volumetric scoop for rapid soil 352 preparation (Hoskins and Ross, 2011; Miller et al., 2013; Mylavarapu and Miller, 2014; 353 Peck, 2015). This further modification of POXC for high throughput analysis (and more 354 generally any soil test) may be a significant source of variation, necessitating lab 355 proficiency testing for appropriate quality assurance/quality control (QA/QC). While 356 QA/QC might establish tolerance limits, it would not address the potential for an unequal 357 distribution of bias across SOC contents. 358 359 3.3. Methodological considerations and detection limits 360 The finite amount of MnO4- in the oxidizing solution (0.4 mmol) results in upper limits of 361 quantification for the measurement of POXC. These upper limits change based on the 362 soil mass, which is reflected in the exclusion of high SOC soils (> 10% SOC) from 363 consideration at 2.5 g (Table 3). For the soils included at each soil mass, nearly all of 364 the samples were within the limits of detection (> 95%). However, the proportion of 365 values falling within the detection limits was not consistent across SOC contents (Figure 366 2). In the 2.5 g soil mass, which were exclusively using soils < 10% SOC, detection 367 rates fell off at SOC contents < 1.0% due to non-detectable levels of POXC in samples 368 from both < 0.5 mm and < 2.0 mm sieve sizes (Figure 2c and 2d). However, detection 369 rates were generally higher for 2.5 g than for 0.75 g (Table 3). This difference is largely 370 attributable to the lower absolute values of POXC measured using 2.5 g relative to 0.75 371 g soil (Figure 1b). The lower POXC values at 2.5 g results in a greater proportion of 372 POXC values within detection limits, increasing the overall detection rate. In the 0.75 g 373 soil mass, we found decreased detection at both high SOC contents (> 10%) and at low 374 SOC contents (< 5%) in the < 0.5 mm sieve size. This combination resulted in values 375 that were simultaneously above and below detection limits. Samples analyzed with 0.75 376 g mass at < 2.0 mm sieve size maintained detection rates at higher SOC contents (> 377 10%). For 0.75 g mass, soils with lower SOC contents (< 5.0%) and <2.0 mm sieve size 378 had a sharper decrease in detection than for < 0.5 mm, presumably due to increased 379 consumption of MnO4- (i.e. higher POXC values) associated with the < 0.5 mm sieve 380 size (Figure 1a). Thus, while 0.75 g may allow for soils with a broader range of SOC 381 contents to be measured for POXC, soils with greater SOC content will require an 382 increased number of replications to ensure detectable values. Similarly, 2.5 g soil 383 masses may provide consistent detection at SOC contents < 10%, but greater 384 replication will be required at very low SOC values (< 1%). These results collectively 385 show that across SOC contents, replication is needed to ensure a sufficient number of 386 detectable values. 387 388 3.4. Overall analytical variability 389 The coefficient of variation of the five analytical replicates for each unique set of soil, 390 sieve size, and soil mass ranged from 0.04 to 171.8% (median= 7.85, mean= 13.41; 391 data not shown). The distribution of the total analytical variability was highly positively 392 skewed, i.e. had several extreme high values. The 95% confidence interval for the 393 overall distribution—after transformation to meet normality assumptions and 394 backtransformation into natural values—was 7.2 to 8.0% with a mean overall variability 395 of 7.6% (data not shown). These values of POXC analytical variability largely agree with 396 the values obtained by Hurisso et al. (2018a) and are comparable to the analytical 397 variability of Mehlich-3 extractable P and K also evaluated therein. However, a clearer 398 understanding of the contributing sources of this variability will help improve the 399 reliability and robustness of this metric across contexts. 400 401 3.5. Inter-lab analytical variability 402 We found salient inter-lab effects on the variability of POXC measurements (Table 4, p 403 < 0.00001), although soil mass also influenced analytical variability (Table 4, p < 404 0.00001). Additionally, each soil had differing levels of analytical variability (Table 4, p < 405 0.00001), an effect which varied by sieve size (Table 4, p = 0.009). Although F-values 406 can often be used as measures of overall effect, the transformations needed to meet 407 assumptions of normality prevent a direct comparison of F-values to be made. 408 Therefore, we will examine the overall contribution (in backtransformed values) in the 409 following sections to quantify the relative contributions of each of these sources to the 410 overall observed analytical variability. 411 412 3.6. Intra-lab analytical variability 413 A significant source of variation of any soil measurement is attributable to lab-specific 414 variability. Here, we found that analytical variability ranged from 2.9 to 15.8% within a 415 given lab (Figure 2; median = 6.5, mean = 7.7). Most labs (25 to 75% quartiles) had 416 intra-lab analytical variability ranging from 5.1 to 10.7%. There were no systematic 417 differences in reliability between labs that used a 96-well plate reader and labs that 418 used microcuvettes (F1,10 = 1.7, p = 0.225, data not shown). Internal practices such as 419 KMnO4 reagent storage conditions, consistency of reaction times, error in the dilution 420 step, or pipetting errors prior to dilution may have contributed to this variability, though 421 these are beyond the scope of the current study. However, the suitability of POXC for 422 high-throughput commercial testing labs (Bongiorno et al., 2019) and the potential for 423 substantial inter-laboratory variability (Table 4; p < 0.00001) suggest that POXC would 424 benefit from lab proficiency testing. 425 426 3.7. Contribution of soil mass to analytical variability 427 Alterations of the mass of soil used in POXC analysis can produce differences in 428 absolute values (Figure 1b), but the implications of soil mass for analytical variability are 429 understudied. We hypothesized that a greater soil mass (2.5 g) would allow for a 430 greater and thus more consistent mass of SOC in a given sample, resulting in lower 431 analytical variability than for a lower soil mass (0.75 g). Confirming this hypothesis, 432 analytical variability was 6.5% greater for POXC values measured with 0.75 g relative to 433 the 2.5 g (Figure 4a). Median analytical variability more than doubled from 5.1% at 2.5 g 434 to 11.6% at 0.75 g. This increase in median CV value is equivalent to the median lab-435 specific variability (Figure 3), underscoring the importance of this standardization for 436 routine, repeatable POXC analysis. 437 438 3.8. Contribution of sieve size to analytical variability 439 The standardization of soil sieve size prior to analysis is a common consideration for 440 soil analytical methods. As soil aggregates are broken and the physically protected soil 441 C is exposed, this organic matter becomes more susceptible to oxidation, regardless of 442 chemical composition (Dungait et al., 2012). Therefore, we hypothesized that a finer 443 sieve size (i.e. < 0.5 mm) would produce a more consistent measurement of POXC than 444 a larger sieve size (i.e. < 2.0 mm), yielding lower analytical variability. While this 445 hypothesis was largely confirmed (Figure 4b), the magnitude of this difference was 446 smaller than expected. A 1.8% decrease in median CV from 8.4% in < 2 mm sieve size 447 to 6.6% in < 0.5 mm represents a modest improvement in analytical variability. This 448 difference is much lower than the ~10% decreases that Hurisso et al. (2018b) reported 449 for soils sieved to < 8.0mm (CV » 20%) to < 2.0 mm (CV » 10%). Thus, both the 450 absolute variability and the relative change in that variability was lower in our study than 451 in Hurisso et al. (2018b). However, the variability at < 2.0 mm was comparable across 452 both studies (CV » 10%). Our results and those of Hurisso et al. (2018b) collectively 453 demonstrate higher analytical variability of POXC values in coarser sieve size 454 treatments. However, sieving to smaller sizes seems to produce diminishing returns in 455 terms of analytical variability. Additionally, manual sieving used in this study, as is 456 commonly practiced by research labs, produced similar analytical variability as samples 457 mechanically flail ground to the same size (< 2.0 mm), as is commonly employed in 458 commercial test lab settings (Hurisso et al., 2018b). 459 460 While sieve size exerted a strong main effect on analytical variability (Table 4, Figure 461 4b), the relationship was not straightforward, as indicated by the soil ´ sieve size 462 interaction (Table 3, p = 0.009). Upon further examination of the interaction term, we 463 found differences in both direction and magnitude of this effect (Figure 5). The 464 magnitude of the sieve size effect on analytical variability—the absolute value of the 465 difference between < 0.5 mm and < 2.0mm sieve sizes—ranged from 0.04% in soil 28 466 to 10.6% in soil 32 (mean D = 1.64%). The magnitude of this difference was inversely 467 related to SOC content (F1, 70 = 4.4, p = 0.039, data not shown). Thus, soils with a lower 468 SOC content expressed larger differences in analytical variability between sieve sizes. 469 In most soils (31 out of 36), this entailed lower analytical variability at < 0.5 mm sieve 470 size than at < 2.0 mm, as hypothesized. However, several soils exhibited increased 471 variability at the < 0.5 mm sieve size, and this did not appear to be explained by 472 physicochemical properties. Of the soils with this inverse relationship, three of the five 473 (soils 11, 27, and 28) had negligible (< 1%) changes in variability, whereas soils 17 and 474 19 had larger changes in variability (3.6% and 5.9%, respectively). Nevertheless, the 475 trend of substantially lower or effectively unchanged analytical variability using the < 0.5 476 mm sieve size was consistent across soils. 477 478 3.9. Soil-specific sources of variability 479 One of the more formidable hurdles to widespread implementation of POXC is the 480 differing degree of soil-specific analytical variability (Table 4; p < 0.00001). Using the 481 standard soil physicochemical properties examined here (Table 1), we found that SOC 482 content had the strongest effect on soil-specific variability. We found similar, negative 483 logarithmic relationships between SOC and soil analytical variability for both < 0.5 mm 484 and < 2.0 mm sieve sizes (Figure 6a and 6b). Thus, lower SOC contents had greater 485 overall analytical variability than higher SOC contents, independent of sieve size. At 486 lower SOC contents, the amount of SOC per sample is likely more susceptible to slight 487 variations between replicates, ultimately increasing analytical variability. 488 489 3.10. Conversion between masses and sieve sizes 490 Differences in methodology—varying the soil mass and sieve size—resulted in changes 491 in absolute POXC values by both sieve size and mass (Figure 1a and 1b, respectively). 492 The differences in absolute values attributable to these methodological differences 493 prevent direct comparison of POXC values across treatments. To facilitate 494 comparisons, we developed equations to convert among POXC values derived using 495 different masses and sieve sizes (Table 5). For these soils, conversions between soil 496 mass and sieve size treatments were generally accurate, with R2 values > 0.90. The 497 root mean square error (RMSE)—a measurement of the expected error of the estimate 498 in mg POXC kg-1 soil—ranged from 55 to 100 mg POXC kg-1 soil for soils < 10% SOC. 499 Conversions within a given sieve size were more accurate than conversions within a 500 mass (R2 values and lower RMSE). Of the two sieve sizes, conversions were most 501 accurate in the < 0.5 mm sieve size (R2 = 0.968, RMSE = 53 mg POXC kg-1 soil). 502 Therefore, we found that approximations can be made across these soil mass and sieve 503 size treatments. 504 505 4. Future Work 506 To date, the most prominent adaptations of the POXC measurement in soils have 507 altered the concentration of MnO4- (Weil et al., 2003) and the mass of soil (Culman et 508 al., 2012) but still accepting the underlying chemistry assumptions. Future work should 509 address two central assumptions to further refine the measurement. First, assumptions 510 about the oxidation-reduction processes should be examined. As Gruver (2015) noted, 511 the assumed relationship of 1 mol Mn7+ oxidizing 9000 mg of C (Eq. 1) assumes that C0 512 ® C4+ and Mn7+ ® Mn4+, a stoichiometric relationship of 0.75 mol C to 1 mol Mn. At 513 circumneutral pH, Mn7+ reduces to Mn4+, but at acidic pH, a nearly complete reduction 514 of Mn7+ to Mn2+ would be expected (Ladbury and Cullis, 1958). The redox state of SOC 515 can vary considerably by texture (Keiluweit et al., 2018, 2017) and can reflect the 516 composition of C inputs (Spokas, 2010). Because the stoichiometry of carbon oxidation 517 and permanganate reduction is a necessary assumption in the calculation of POXC, this 518 value should be empirically determined for soil conditions (e.g., texture, SOC quality) 519 that can impact the conversion of a measure of MnO4- reduction to a SOC 520 concentration. It is possible that “the use of a constant stoichiometric relationship when 521 calculating [POXC] may be more a matter of convenience than accuracy” (Gruver, 522 2015). Secondly, quantification of POXC on a SOC, not soil mass basis (i.e., constant 523 ratio of MnO4-:SOC) as in Lefroy et al. (1993) and Blair et al. (1995) could represent a 524 substantial improvement in the repeatability of the metric by accounting for 525 nonlinearities (Gruver, 2015). Establishment of a consistent ratio of reducing agent (soil 526 C) to oxidizing agent (MnO4-) may be necessary to ensure the reliability of a 527 measurement based on a reduction-oxidation reaction. 528 529 Although methodological considerations have been central in many discussions, the 530 question remains: “What fraction of soil C is the permanganate oxidizing?” Several 531 evaluations of this question have been largely inconclusive (Margenot et al., 2017; 532 Romero et al., 2018). Better understanding the nature of POXC is critical to its use as 533 an indicator of soil health given the mechanistic assumptions of this operationally 534 defined C fraction implicit in its description as “active C” or “microbial food” (NRCS, 535 2019). 536 537 5. Recommendations 538 The refinement of soil methods is essential to providing reliable tools for land 539 management decisions. However, land managers often prefer to use multiple indicators 540 to inform their decisions, particularly within the realm of soil health (Andrews et al., 541 2002). In pursuance of this goal, soil health indicators have often been proposed as 542 heuristic in place of more accurate yet labor-intensive measurements. For the current 543 proposition and application of POXC as an operational metric sensitive to management, 544 sensitivity and/or precision must be weighed carefully against usability and ease of 545 implementation. While we did not exhaustively test all potential sample processing 546 treatments, we have focused on two methodological variations that have (in our 547 experience) proved especially problematic. Accordingly, we developed the following 548 recommendations with two goals of (1) minimizing analytical variability while (2) 549 maximizing the utility of the POXC metric across soil contexts. Given the range of soil 550 characteristics in our current dataset, we believe these recommendations are applicable 551 across nearly all soil contexts. 552 1. In-house quality control practices: Our data demonstrate that one of the least-553 generalizable, yet most significant sources of variation is within-lab variability 554 (Figure 3). However, our data suggests that low variability (< 5%) is easily 555 attainable. In lieu of external lab proficiency testing for POXC, individual labs are 556 recommended to develop in-house quality control practices, such as internal 557 reference soils or technician performance testing, to minimize this significant 558 source of variability. 559 2. Soil mass of 2.5 g: A soil mass of 2.5 g resulted in lower analytical variability 560 (Figure 4a), but was not suitable for soils with SOC contents > 10% by mass due 561 to full consumption of MnO4- (i.e., no quantification possible). This threshold 562 roughly corresponds to the ~12%SOC threshold that distinguishes mineral from 563 organic soils in USDA Soil Taxonomy (Soil Survey Staff, 2014). Therefore, for 564 studies comparing both mineral and organic soils, we recommend a lower mass 565 (0.75 g). Standardization of soil mass within the same study or monitoring 566 program is recommended because mass can markedly affect POXC values, and 567 this magnitude of change is greater than any other source of variability assessed 568 here. 569 3. Sieve size of < 2.0 mm: A finer sieve size decreased analytical variability (Figure 570 4b) for the majority of soils (Figure 5) by 1.8%. However, the additional sieving to 571 < 0.5 mm requires more labor and/or time. Given the negligible decrease in 572 variability at < 0.5 mm sieve size, we believe this additional labor time is an 573 opportunity cost for other in-house quality control metrics (see recommendation 574 1) that contribute a larger amount of variability. 575 4. Replication: Our data shows that POXC, like many other soil metrics, has a 576 substantial degree of analytical variability. Therefore, analytical replication is 577 necessary, although the reasoning for increased replication varies. Soils with 578 lower SOC contents generally have higher variability (Figure 6a and 6b), 579 necessitating additional replicates to ascertain that sample POXC values are an 580 accurate approximation of the population mean. At SOC contents > 10% or <5%, 581 greater replication is needed to ensure that the calculated POXC values are 582 within detection limits (Figure 3), but POXC is most often measured using one to 583 three replicates (NRCS, 2019). The issue of underpowered or uncertain 584 hypothesis testing is ubiquitous in soil science analyses (Ladoni et al., 2015; 585 Welsch et al., 2019). Calculated replication numbers can be found in Table S4. 586 While we do not have values for other standard soil measurements for the 587 current dataset, Hurisso et al. (2018a) found that POXC had comparable 588 analytical variability to organic matter via loss-on-ignition, a common component 589 of agronomic soil tests. 590 591 These recommendations balance a tradeoff between the sensitivity and the reliability of 592 the metric. Specifically, the lower absolute POXC values at 2.5g (Figure 1b) reduces the 593 overall sensitivity of the metric, relative to values based on 0.75g. The decision to use a 594 greater soil mass also decreases the range of SOC contents at which the measurement 595 is viable, potentially complicating analyses. For example, all three Gelisols and two of 596 the three Histosols fully consumed the 0.4 mmol of MnO4- when 2.5 g soil was 597 employed, preventing measurement of POXC. We believe that these considerations are 598 outweighed by the substantial decrease in analytical variability (Figure 4a, D = 6.5%) 599 and consistency with previously published literature (Culman et al., 2012). 600 601 6. Conclusions 602 As with any emerging soil health metric, POXC must be thoroughly evaluated prior to 603 widespread adoption. Thus, standardization of methods and a clearer understanding of 604 relative sources of variability are essential steps along the path towards implementation 605 in commercial soil test labs. Across a wide range of edaphic properties, we found that 606 research labs (n = 12, US and EU) differed in their within-lab variability, which ranged 607 from 3 to 16%. A finer sieve size (< 0.5 mm) increased the absolute values of POXC 608 (mean = 124 mg POXC kg-1 soil; median = 56 mg POXC kg-1 soil) and decreased the 609 analytical variability 1.8%, relative to < 2.0 mm sieve size. Using a greater soil mass 610 (2.5 g) decreased the absolute POXC values (mean = 177 mg POXC kg-1 soil; median = 611 111 mg POXC kg-1 soil) and the analytical variability (D = 6.5%). However, at the greater 612 soil mass the full consumption of MnO4- (i.e., ‘bleaching’) for soils with SOC >10% 613 exceeded the limit of quantification and meant that POXC could not be measured. 614 Conversely, at the lower soil mass (0.75 g), some soils with SOC < 5% could be below 615 the detection limit. Although variability in POXC measurements was in part soil-specific, 616 it was generally inverse to SOC content. Therefore, we recommend that routine POXC 617 analysis of <10% SOC soils (most mineral soils) be conducted using multiple analytical 618 replicates and a soil mass of 2.5 g. For analyses that include high organic matter soils 619 (>10% SOC), we recommend decreasing soil mass to 0.75 g for more appropriate 620 comparisons across soils. While the < 0.5 mm sieve size decreased analytical variability 621 relative to < 2.0 mm (D = -1.8%), the increase in labor associated with the finer sieve 622 size suggests that this additional effort is likely not merited. Given the wide range of 623 edaphic contexts in the current study, we believe that these recommendations are 624 robust across soil and climatic contexts. 625 Tables and Figures 626 627 Table 1. Classification and characteristics of soils used in multilab comparison of permanganate 628 oxidizable C (POXC). Soils are A horizons for mineral soils and O horizons for Histosols and 629 Gelisols. Soil series information, including USDA classification and land use is available for 630 each soil in Table S1. 631 632 Soil ID USDA Order SOC (%) C:N pH Clay (g kg-1) Sand (g kg-1) CECa (meq 100 g-1) Location 1 Oxisol 1.7 12.7 5.0 716 79 9.1 Kisumu, Kenya 2 Oxisol 0.6 9.7 4.8 480 243 12.5 CA, USA 3 Oxisol 1.4 11.7 5.6 694 74 11.6 Vihiga, Kenya 4 Vertisol 1.8 14.7 7.7 446 81 32.0 CA, USA 5 Vertisol 1.6 11.2 6.1 392 269 25.3 CA, USA 6 Vertisol 1.0 12.4 7.0 540 135 31.1 CA, USA 7 Histosol 8.3 14.8 7.7 236 502 31.2 CA, USA 8 Histosol 37.7 25.3 5.3 125 500 2.0 MN, USA 9 Histosol 29.7 16.0 7.7 102 414 55.8 FL, USA 10 Inceptisol 3.3 10.8 4.8 176 395 8.9 RI, USA 11 Inceptisol 1.2 15.5 4.5 281 311 7.0 CA, USA 12 Inceptisol 1.5 11.4 6.6 317 156 21.4 CA, USA 13 Mollisol 3.7 14.6 6.0 275 66 26.9 IL, USA 14 Mollisol 1.1 10.4 8.1 284 330 15.3 MO, USA 15 Mollisol 3.1 12.4 6.1 256 105 14.7 CA, USA 16 Alfisol 1.1 10.1 8.0 228 292 17.4 IL, USA 17 Alfisol 0.9 8.4 6.5 300 200 19.1 CA, USA 18 Alfisol 2.4 10.2 5.7 75 850 26.6 IL, USA 19 Ultisol 0.8 ndb 6.1 236 503 3.0 FL, USA 20 Ultisol 4.1 21.7 5.8 128 668 8.0 CA, USA 21 Ultisol 7.0 22.0 5.4 225 474 20.4 CA, USA 22 Entisol 8.9 68.5 8.1 203 418 92.3 FL, USA 23 Entisol 0.9 8.9 7.3 25 925 21.1 CA, USA 24 Entisol 1.5 13.9 6.4 107 680 11.2 CA, USA 25 Andisol 8.0 19.6 6.0 103 819 9.4 CA, USA 26 Andisol 7.4 19.6 5.8 77 743 11.7 CA, USA 27 Andisol 4.4 21.1 6.3 0 918 10.5 CA, USA 28 Spodosol 6.2 13.2 4.7 0 950 10.0 FL, USA 29 Spodosol 2.3 25.0 4.4 25 900 3.7 FL, USA 30 Spodosol 2.7 11.9 5.7 0 364 24.8 FL, USA 31 Aridisol 1.3 9.9 7.8 0 671 24.9 CA, USA 32 Aridisol 0.7 12.2 7.1 0 874 18.9 NV, USA 33 Aridisol 0.2 nd 8.1 716 79 7.4 NV, USA 34 Gelisol 17.5 19.9 6.9 480 243 25.0 AK, USA 35 Gelisol 20.8 15.8 5.3 694 74 22.5 AK, USA 36 Gelisol 32.8 29.8 5.2 446 81 31.7 AK, USA a CEC = cation exchange capacity; b nd = non-detectable total soil N (<0.05%) 633 634 Table 2. Summary of POXC values for each soil and treatment. Values are mean, 635 median, and median absolute difference (all in mg POXC kg-1 soil). Analytical variability 636 (% coefficient of variation) for each soil is also summarized. Averages (mean and 637 median) are n = 50 for 0.75 g and n = 60 for 2.5 g. CV values are averaged from each 638 of the 12 labs. 639 Soil ID 2.5 g 0.75 g < 0.5 mm < 2.0 mm < 0.5 mm < 2.0 mm Average CV (%) Average CV (%) Average CV (%) Average CV (%) 1 342, 324, 41 8.7 274, 259, 57 11.0 373, 323, 211 23.7 317, 251, 198 21.1 2 287, 278, 52 9.3 253, 230, 41 11.9 361, 271, 102 25.0 250, 202, 141 37.3 3 274, 267, 42 10.0 247, 237, 41 12.1 390, 329, 150 27.7 321, 315, 123 22.9 4 497, 516, 55 5.8 410, 414, 57 5.1 612, 616, 151 15.9 502, 509, 106 16.8 5 557, 573, 32 4.6 476, 511, 48 4.5 643, 655, 156 11.2 611, 593, 163 13.9 6 345, 357, 37 8.5 278, 282, 41 10.3 464, 390, 172 27.6 315, 275, 120 26.3 7 1155, 1166, 224 3.8 917, 929, 166 7.4 1891, 1965, 414 4.6 1435, 1412, 322 8.6 8 -- -- -- -- 4450, 4687, 124 4.6 3660, 3657, 564 6.4 9 -- -- -- -- 4319, 4414, 180 1.7 4040, 4127, 222 2.1 10 427, 385, 60 7.2 347, 317, 72 8.7 514, 457, 207 14.5 525, 392, 162 14.0 11 268, 275, 57 14.4 252, 239, 76 12.7 380, 271, 196 37.9 437, 293, 198 41.4 12 554, 556, 73 4.3 497, 496, 87 6.7 600, 562, 159 12.7 548, 477, 192 15.3 13 625, 627, 45 5.1 576, 582, 40 7.8 774, 750, 163 11.1 731, 749, 111 14.9 14 241, 202, 32 11.3 195, 159, 34 10.5 270, 231, 96 30.8 241, 227, 102 26.6 15 852, 847, 70 4.3 710, 674, 136 6.7 1214, 953, 195 11.7 1104, 814, 130 11.5 16 245, 231, 46 8.5 190, 182, 38 12.0 288, 214, 115 27.9 343, 235, 72 26.5 17 229, 219, 24 11.6 234, 218, 40 12.9 283, 180, 136 47.1 299, 234, 131 23.1 18 638, 630, 57 4.2 553, 549, 77 7.1 905, 768, 160 14.1 813, 721, 124 10.4 19 167, 153, 35 17.4 174, 165, 36 16.9 244, 212, 117 44.7 212, 204, 120 21.0 20 907, 931, 110 2.2 640, 645, 116 7.1 1230, 1199, 289 7.0 865, 856, 227 13.7 21 1171, 1210, 90 2.2 946, 959, 131 4.0 1650, 1648, 295 5.3 1243, 1218, 241 11.6 22 374, 387, 43 5.9 365, 357, 41 9.9 409, 404, 130 18.9 475, 433, 131 26.4 23 447, 493, 63 6.6 389, 410, 37 8.2 531, 547, 123 14.7 519, 496, 141 28.2 24 618, 617, 122 6.9 387, 386, 57 11.9 792, 739, 209 14.6 492, 482, 136 24.9 25 984, 1011, 365 8.3 967, 946, 210 8.0 1590, 1499, 638 12.6 1495, 1404, 324 11.7 26 831, 852, 306 9.7 705, 691, 156 9.6 1231, 1247, 484 16.4 1009, 1073, 245 16.9 27 420, 439, 188 14.7 378, 395, 93 12.8 612, 560, 141 20.4 534, 473, 167 21.8 28 1039, 1033, 234 6.8 1075, 1034, 152 6.7 1719, 1765, 470 9.2 1664, 1641, 291 8.8 29 514, 505, 61 8.7 501, 489, 43 5.7 714, 640, 123 17.9 690, 598, 171 27.5 30 780, 807, 72 4.4 829, 839, 100 5.9 997, 988, 159 10.9 1084, 1048, 201 12.9 31 333, 348, 47 5.1 313, 333, 52 6.2 380, 324, 85 15.8 343, 328, 143 19.4 32 302, 229, 47 9.4 215, 144, 51 20.9 600, 274, 66 25.8 537, 195, 90 26.7 33 102, 90, 24 30.5 149, 70, 18 24.5 115, 111, 53 27.3 96, 90, 56 38.3 34 -- -- -- -- 4486, 4657, 141 1.0 4291, 4416, 153 2.0 35 -- -- -- -- 2953, 2519, 1236 12.4 2429, 2410, 747 10.5 36 -- -- -- -- 4318, 4521, 298 3.1 4094, 4145, 397 3.7 640 Table 3. Soil IDs included in each processing treatment and the proportion of assays 641 that resulted in MnO4- concentration within the range of detection. 642 Sieve Size Mass Soil IDs excluded n Detection limit (mg POXC kg-1 soil) % Detectable < 0.5 mm 0.75 g none 2141 4800 96.0 2.50 g 8, 9, 34-36 1835 1440 99.5 < 2.0 mm 0.75 g none 2142 4800 95.7 2.50 g 8, 9, 34-36 1833 1440 99.1 643 644 Table 4. Final reduced model assessing relative sources of analytical variability (%CV) 645 for all detectable POXC values. All experimental factors were initially included in the 646 model and non-significant effects (p ³ 0.10) were eliminated from the model one at a 647 time. Final model was log transformed to meet assumptions of normality. 648 649 Experimental Factor df F-value p-value Lab 11 62.1 < 0.00001 Mass 1 438.7 < 0.00001 Sieve size 1 47.0 < 0.00001 Soil ID 35 35.5 < 0.00001 Soil ID ´ Sieve size 35 1.7 0.009 650 651 Table 5. Equations to convert between processing treatments for soils with SOC below 652 10%. All calculations are using mg POXC kg-1 soil. Bootstrapped 95% confidence 653 intervals for each equation can be found in Table S3. 654 Conversion Equation RMSEa R2 < 2.0mm: 0.75 g ® 2.5 g POXC2.5g = 0.633*POXC0.75g + 56.40 55.4 0.954 < 0.5mm: 0.75 g ® 2.5 g POXC2.5g = 0.608*POXC0.75g + 85.95 59.4 0.968 0.75 g: < 0.5mm ® < 2.0mmb POXC<2mm = 0.821*POXC<0.5mm + 43.76 100.0 0.938 2.5 g: < 0.5mm ® < 2.0mm POXC<2mm = 0.820*POXC<0.5mm + 27.72 67.7 0.932 a root mean squared error, in mg POXC kg-1 soil; b conversion including soils with SOC above ~ 10%: POXC<2mm = 0.902*POXC<0.5mm – 13.42, RMSE = 135.1, R2 = 0.986 655 656 Figure 1. Absolute differences in POXC values between (a) sieve sizes and (b) mass. 657 Note the differences in scale. Values are estimated marginal means for each soil using 658 Eq. 2 and are in mg POXC kg -1 soil. 659 660 661 Figure 2. Detections rates (%) as a function of SOC for (a) < 2.0 mm and 0.75 g, (b) < 662 0.5 mm and 0.75 g, (c) < 2.0 mm and 2.5 g, and (d) < 0.5 mm and 2.5 g. Dashed lines 663 indicate average detection across all soils for that set of sieve size and soil mass. 664 665 666 Figure 3. Range of analytical variability that is attributable to within-lab sources. Top 667 represents the continuous probability function, while the bottom indicates the specific 668 variation for each of the twelve labs. Values are estimated marginal means from Eq. 2 669 for each lab. 670 671 672 Figure 4. Changes in analytical variability of permanganate oxidizable C (POXC) values 673 due to change in (a) soil mass and (b) sieve size. Values are backtransformed 674 estimated marginal means for each unique combination of soil, soil mass, sieve size, 675 and lab. The 95% confidence intervals for the difference between treatments is based 676 on bias-corrected accelerated bootstrap resampling with 5000 resamplings. 677 678 679 Figure 5. Interactive effects of sieve size and soil on analytical variability on 680 permanganate oxidizable C (POXC) values. Bolded soils indicate that CV of < 0.5 mm 681 was greater than < 2.0 mm. Dashed line indicates mean CV for all soils (CV = 8.86). All 682 values are backtransformed estimated marginal means using Eq. 2. 683 684 685 Figure 6. Analytical variability of permanganate oxidizable C (POXC) for each soil by 686 SOC content at (a) sieve size < 2.0 mm and (b) sieve size < 0.5 mm (mass = 0.75g for 687 both). 688 689 690 References 691 Akaike, H., 1974. A new look at the statistical model identification. IEEE Transactions 692 on Automatic Control 19, 716–723. https://doi.org/10.1109/TAC.1974.1100705 693 Andrews, S.S., Karlen, D.L., Mitchell, J.P., 2002. A comparison of soil quality indexing 694 methods for vegetable production systems in Northern California. Agriculture, 695 ecosystems & environment 90, 25–45. 696 Blair, G.J., Lefroy, R.D., Lisle, L., 1995. Soil carbon fractions based on their degree of 697 oxidation, and the development of a carbon management index for agricultural 698 systems. Australian journal of agricultural research 46, 1459–1466. 699 Blankinship, J.C., Berhe, A.A., Crow, S.E., Druhan, J.L., Heckman, K.A., Keiluweit, M., 700 Lawrence, C.R., Marín-Spiotta, E., Plante, A.F., Rasmussen, C., 2018. Improving 701 understanding of soil organic matter dynamics by triangulating theories, 702 measurements, and models. Biogeochemistry 140, 1–13. 703 Bongiorno, G., Bünemann, E.K., Oguejiofor, C.U., Meier, J., Gort, G., Comans, R., 704 Mäder, P., Brussaard, L., de Goede, R., 2019. Sensitivity of labile carbon 705 fractions to tillage and organic matter management and their potential as 706 comprehensive soil quality indicators across pedoclimatic conditions in Europe. 707 Ecological Indicators 99, 38–50. 708 Bouyoucos, G.J., 1962. Hydrometer Method Improved for Making Particle Size 709 Analyses of Soils. Agronomy journal 54, 464–465. 710 Bundy, L.G., Bremner, J.M., 1973. Determination of ammonium n and nitrate n in acid 711 permanganate solution used to absorb ammonia, nitric oxide, and nitrogen 712 dioxide evolved from soils. Communications in soil science and plant analysis 4, 713 179–184. 714 Carski, T.H., Sparks, D.L., 1987. Differentiation of Soil Nitrogen Fractions Using a 715 Kinetic Approach 1. Soil Science Society of America Journal 51, 314–317. 716 Culman, S.W., Snapp, S.S., Freeman, M.A., Schipanski, M.E., Beniston, J., Lal, R., 717 Drinkwater, L.E., Franzluebbers, A.J., Glover, J.D., Grandy, A.S., Lee, J., Six, J., 718 Maul, J.E., Mirksy, S.B., Spargo, J.T., Wander, M.M., 2012. Permanganate 719 Oxidizable Carbon Reflects a Processed Soil Fraction that is Sensitive to 720 Management. Soil Science Society of America Journal 76, 494–504. 721 https://doi.org/10.2136/sssaj2011.0286 722 Doran, J.W., Zeiss, M.R., 2000. Soil health and sustainability: managing the biotic 723 component of soil quality. Applied soil ecology 15, 3–11. 724 Dungait, J.A.J., Hopkins, D.W., Gregory, A.S., Whitmore, A.P., 2012. Soil organic 725 matter turnover is governed by accessibility not recalcitrance. Global Change 726 Biology 18, 1781–1796. https://doi.org/10.1111/j.1365-2486.2012.02665.x 727 Efron, B., 1987. Better bootstrap confidence intervals. Journal of the American statistical 728 Association 82, 171–185. 729 Fine, A.K., van Es, H.M., Schindelbeck, R.R., 2017. Statistics, scoring functions, and 730 regional analysis of a comprehensive soil health database. Soil Science Society 731 of America Journal 81, 589–601. 732 Fox, J., Weisberg, S., 2019. An {R} Companion to Applied Regression. Sage, Thousand 733 Oaks, CA. 734 Fuhrman, J.K., Zhang, H., Schroder, J.L., Davis, R.L., Payton, M.E., 2005. Water-735 Soluble Phosphorus as Affected by Soil to Extractant Ratios, Extraction Times, 736 and Electrolyte. Communications in soil science and plant analysis 36, 925–935. 737 Gruver, J., 2015. Evaluating the sensitivity and linearity of a permanganate-oxidizable 738 carbon method. Communications in soil science and plant analysis 46, 490–510. 739 Halsey, L.G., Curran-Everett, D., Vowler, S.L., Drummond, G.B., 2015. The fickle P 740 value generates irreproducible results. Nature methods 12, 179. 741 Harris, D., Horwáth, W.R., van Kessel, C., 2001. Acid fumigation of soils to remove 742 carbonates prior to total organic carbon or CARBON-13 isotopic analysis. Soil 743 Science Society of America Journal 65, 1853. 744 https://doi.org/10.2136/sssaj2001.1853 745 Hendershot, W.H., Lalande, H., Duquette, M., 1993. Soil reaction and exchangeable 746 acidity. Soil sampling and methods of analysis 2. 747 Ho, J., Tumkaya, T., Aryal, S., Choi, H., Claridge-Chang, A., 2019a. Moving beyond P 748 values: data analysis with estimation graphics. Nature Methods 1. 749 Ho, J., Tumkaya, T., Aryal, S., Choi, H., Claridge-Chang, A., 2019b. Moving beyond P 750 values: Everyday data analysis with estimation plots. BioRxiv 377978. 751 Hoskins, B., Ross, D., 2011. Soil Sample Preparation and Extraction. The Northeast 752 Coordinating Committee for Soil Testing. 753 Hurisso, T.T., Culman, S.W., Horwath, W.R., Wade, J., Cass, D., Beniston, J.W., 754 Bowles, T.M., Grandy, A.S., Franzluebbers, A.J., Schipanski, M.E., 2016. 755 Comparison of permanganate-oxidizable carbon and mineralizable carbon for 756 assessment of organic matter stabilization and mineralization. Soil Science 757 Society of America Journal 80, 1352–1364. 758 Hurisso, T.T., Culman, S.W., Zhao, K., 2018a. Repeatability and Spatiotemporal 759 Variability of Emerging Soil Health Indicators Relative to Routine Soil Nutrient 760 Tests. Soil Science Society of America Journal 82, 939–948. 761 https://doi.org/10.2136/sssaj2018.03.0098 762 Hurisso, T.T., Culman, S.W., Zone, P., Sharma, S., 2018b. Absolute values and 763 precision of emerging soil health indicators as affected by soil sieve size. 764 Communications in Soil Science and Plant Analysis 1–9. 765 Jensen, J.L., Schjønning, P., Watts, C.W., Christensen, B.T., Peltre, C., Munkholm, L.J., 766 2019. Relating soil C and organic matter fractions to soil structural stability. 767 Geoderma 337, 834–843. 768 Kaiser, K., Kaupenjohann, M., Zech, W., 2001. Sorption of dissolved organic carbon in 769 soils: effects of soil sample storage, soil-to-solution ratio, and temperature. 770 Geoderma 99, 317–328. 771 Keiluweit, M., Gee, K., Denney, A., Fendorf, S., 2018. Anoxic microsites in upland soils 772 dominantly controlled by clay content. Soil Biology and Biochemistry 118, 42–50. 773 Keiluweit, M., Wanzek, T., Kleber, M., Nico, P., Fendorf, S., 2017. Anaerobic microsites 774 have an unaccounted role in soil carbon stabilization. Nature communications 8, 775 1771. 776 Kibblewhite, M., Ritz, K., Swift, M., 2008. Soil health in agricultural systems. 777 Philosophical Transactions of the Royal Society of London B: Biological Sciences 778 363, 685–701. 779 Ladbury, J.W., Cullis, C.F., 1958. Kinetics and mechanism of oxidation by 780 permanganate. Chemical Reviews 58, 403–438. 781 Ladoni, M., Basir, A., Kravchenko, A., 2015. Which soil carbon fraction is the best for 782 assessing management differences? A statistical power perspective. Soil 783 Science Society of America Journal 79, 848–857. 784 Lal, R., 2016. Soil health and carbon management. Food and Energy Security 5, 212–785 222. 786 Lefroy, R.D., Blair, G.J., Strong, W.M., 1993. Changes in soil organic matter with 787 cropping as measured by organic carbon fractions and 13 C natural isotope 788 abundance. Plant and soil 155, 399–402. 789 Lehmann, J., Kleber, M., 2015. The contentious nature of soil organic matter. Nature 790 528, 60. 791 Lehmann, J., Solomon, D., Kinyangi, J., Dathe, L., Wirick, S., Jacobsen, C., 2008. 792 Spatial complexity of soil organic matter forms at nanometre scales. Nature 793 Geoscience 1, 238. 794 Lenth, R., 2019. emmeans: Estimated Marginal Means, aka Least-Squares Means. 795 Li, K., Zhao, Y., Yuan, X., Zhao, H., Wang, Z., Li, S., Malhi, S.S., 2012. Comparison of 796 factors affecting soil nitrate nitrogen and ammonium nitrogen extraction. 797 Communications in soil science and plant analysis 43, 571–588. 798 Loginow, W., Wisniewski, W., Gonet, S.S., Ciescinska, B., 1987. Fractionation of 799 organic carbon based on susceptibility to oxidation. Polish Journal of Soil 800 Science (Poland). 801 Lucas, S.T., Weil, R.R., 2012. Can a Labile Carbon Test be Used to Predict Crop 802 Responses to Improve Soil Organic Matter Management? Agronomy Journal 803 104, 1160–1170. https://doi.org/10.2134/agronj2011.0415 804 Margenot, A.J., Calderón, F.J., Magrini, K.A., Evans, R.J., 2017. Application of DRIFTS, 805 13C NMR, and py-MBMS to characterize the effects of soil science oxidation 806 assays on soil organic matter composition in a Mollic Xerofluvent. Applied 807 spectroscopy 71, 1506–1518. 808 Matsuda, K., Schnitzer, M., 1972. The permanganate oxidation of humic acids extracted 809 from acid soils. Soil Science 114, 185–193. 810 Miller, R.O., Gavlak, R., Horneck, D., 2013. Soil, Plant, and Water Reference Methods 811 for the Western Region, 4th ed. Western Coordinating Committe on Nutrient 812 Management. 813 Moebius-Clune, B.N., Moebius-Clune, D.J., Gugino, B.K., Idowu, O.J., Schindelbeck, 814 R.R., Ristow, A.J., van Es, H.M., Thies, J.E., Shayler, H.A., McBride, M.B., Kurtz, 815 K.S.M., Wolfe, D.W., Abawi, G.S., 2017. Comprehensive Assessment of Soil 816 Health – The Cornell Framework Manual, 3.1. ed. Cornell University, Geneva, 817 NY. 818 Mylavarapu, R., Miller, R., 2014. Soil Preparation, Measurement, and Storage, in: Soil 819 Test Methods From the Southeastern United States. Southern Extension and 820 Research Activity Information Exchange Group. 821 Nelson, D.W., Sommers, L.E., 1996. Total carbon, organic carbon, and organic matter, 822 in: Methods of Soil Analysis Part 3—Chemical Methods, SSSA Book Series. pp. 823 961–1010. 824 NRCS, 2019. Recommended Soil Health Indicators and Associated Laboratory 825 Procedures (Technical Note No. 450– 03). US Department of Agriculture. 826 Oldfield, E.E., Bradford, M.A., Wood, S.A., 2019. Global meta-analysis of the 827 relationship between soil organic matter and crop yields. Soil 5, 15–32. 828 Peck, T.R., 2015. Standard Soil Scoop, in: Recommended Chemical Soil Test 829 Procedures for the North Central Region. North Central Regional Research 830 Publication. 831 Perneger, T.V., 2001. Sifting the evidence: Likelihood ratios are alternatives to P values. 832 BMJ: British Medical Journal 322, 1184. 833 Ramírez, P.B., Fuentes-Alburquenque, S., Díez, B., Vargas, I., Bonilla, C.A., 2019. Soil 834 microbial community responses to labile organic carbon fractions in relation to 835 soil type and land use along a climate gradient. Soil Biology and Biochemistry 836 107692. 837 Romero, C.M., Engel, R.E., D’Andrilli, J., Chen, C., Zabinski, C., Miller, P.R., Wallander, 838 R., 2018. Patterns of change in permanganate oxidizable soil organic matter from 839 semiarid drylands reflected by absorbance spectroscopy and Fourier transform 840 ion cyclotron resonance mass spectrometry. Organic geochemistry 120, 19–30. 841 RStudio Team, 2019. RStudio: Integrated Development for R. RStudio, Inc., Boston, 842 MA. 843 Sanderman, J., Creamer, C., Baisden, W.T., Farrell, M., Fallon, S., 2017. Greater soil 844 carbon stocks and faster turnover rates with increasing agricultural productivity. 845 Soil 3, 1–16. 846 Schmidt, M.W., Torn, M.S., Abiven, S., Dittmar, T., Guggenberger, G., Janssens, I.A., 847 Kleber, M., Kögel-Knabner, I., Lehmann, J., Manning, D.A., 2011. Persistence of 848 soil organic matter as an ecosystem property. Nature 478, 49. 849 Soil Survey Staff, 2014. Keys to Soil Taxonomy, 12th ed. USDA-Natural Resources 850 Conservation Service, Washington, D.C. 851 Spokas, K.A., 2010. Review of the stability of biochar in soils: predictability of O: C 852 molar ratios. Carbon Management 1, 289–303. 853 Thomas, G., 1996. Soil pH and soil acidity. In ‘Methods of soil analysis. Part 3. 854 Chemical methods’.(Ed. DL Sparks) pp. 475–490. Soil Science Society of 855 America: Madison, WI. 856 Tirol-Padre, A., Ladha, J.K., 2004. Assessing the Reliability of Permanganate-857 Oxidizable Carbon as an Index of Soil Labile Carbon. Soil Science Society of 858 America Journal 68, 969. https://doi.org/10.2136/sssaj2004.9690 859 von Lützow, M., Kögel-Knabner, I., Ludwig, B., Matzner, E., Flessa, H., Ekschmitt, K., 860 Guggenberger, G., Marschner, B., Kalbitz, K., 2008. Stabilization mechanisms of 861 organic matter in four temperate soils: Development and application of a 862 conceptual model. Z. Pflanzenernähr. Bodenk. 171, 111–124. 863 https://doi.org/10.1002/jpln.200700047 864 Wade, J., Culman, S.W., Sharma, S., Mann, M., Demyan, M.S., Mercer, K.L., Basta, 865 N.T., 2019. How Does Phosphorus Restriction Impact Soil Health Parameters in 866 Midwestern Corn–Soybean Systems? Agronomy Journal. 867 Weil, R.R., Islam, K.R., Stine, M.A., Gruver, J.B., Samson-Liebig, S.E., 2003. Estimating 868 active carbon for soil quality assessment: A simplified method for laboratory and 869 field use. American Journal of Alternative Agriculture 18, 3–17. 870 https://doi.org/10.1079/AJAA200228 871 Welsch, J., Songling, C., Buckley, H.L., Lehto, N.J., Jones, E.E., Case, B.S., 2019. How 872 many samples? Soil variability affects confidence in the use of common 873 agroecosystem soil indicators. Ecological Indicators 102, 401–409. 874 Wickham, H., François, R., Henry, L., Müller, K., 2019. dplyr: A Grammar of Data 875 Manipulation. 876 Willard, H.H., Furman, N.H., Bricker, C.E., 1956. Oxidations with standard potassium 877 permanganate solutions. Elements of quantitative analysis, theory and practice. 878 D. Van Nostrand Co., Princeton, NJ 217. 879 Yin, Y., Impellitteri, C.A., You, S.-J., Allen, H.E., 2002. The importance of organic matter 880 distribution and extract soil: solution ratio on the desorption of heavy metals from 881 soils. Science of the Total Environment 287, 107–119. 882 883 Supplementary 884 Table S1. Characterization of soils used in interlab comparison. 885 Soil ID Soil Series USDA Classification Textural class 1 NA Typic Kandiudox C 2 NA Eutroustox C 3 NA Rhodic Hapludox C 4 Capay Fine, smectitic, thermic Typic Haploxerert SiC 5 Neerdobe Fine, smectitic, thermic Xeric Duraquert CL 6 Anita Clayey, smectitic, thermic, shallow Xeric Duraquert C 7 Rindge Euic, thermic Typic Haplosaprist SCL 8 Merwin Loamy, mixed, dysic, frigid Terric Haplohemist L 9 Dania Euic, hyperthermic, shallow Lithic Haplosaprist L 10 Canton Coarse-loamy, mixed, superactive, mesic Typic Dystrudept L 11 Baldock Fine-loamy, mixed (calcareous), mesic Typic Haplaquept CL 12 Auburn Loamy, mixed, superactive, thermic Lithic Haploxerept SiCL 13 Flanagan fine, smectitic, mesic Aquic Argiudoll SiCL 14 Wakenda Fine-silty, mixed, superactive, mesic Typic Argiudoll CL 15 Supan Fine-loamy, mixed, superactive, mesic Pachic Argixeroll SiL 16 Menfro Fine-silty, mixed, superactive, mesic Typic Hapludalf L 17 Contra Costa Fine, mixed, superactive, thermic Mollic Haploxeralf SiCL 18 Elliot Fine, illitic, mesic Aquollic Hapludalf LS 19 Orangeburg Fine-loamy, kaolinitic, thermic Typic Kandiudult SCL 20 Aiken Fine, parasesquic, mesic Andic Palehumult SL 21 Narlon Fine, mixed, semiactive, thermic Typic Albaquult L 22 Krome Loamy-skeletal, carbonatic, hyperthermic Lithic Udorthent L 23 Reiff Sandy-skeletal, mixed, frigid Vitrandic Xerorthent S 24 Oceano Mixed, thermic Lamellic Xeropsamment SL 25 McCarthy Medial-skeletal, amorphic, mesic Humic Haploxerand LS 26 Waca Medial-skeletal, amorphic, frigid Humic Vitrixerand SL 27 Asta Medial-skeletal, amorphic, mesic Typic Haploxerand S 28 Smyrna Sandy, siliceous, hyperthermic Aeric Alaquod S 29 Ona Sandy, siliceous, hyperthermic Typic Alaquod S 30 Ona Sandy, siliceous, hyperthermic Typic Alaquod SiL 31 Panoche Fine-loamy, mixed, superactive, thermic Typic Haplocambid SL 32 Verdico Fine, smectitic, mesic Vertic Paleargid S 33 Pizene Fine-loamy, mixed, superactive, mesic Typic Natrargid C 34 Klasi Typic Aquiturbel C 35 Klawasi Typic Historthel C 36 Klasi Typic Aquiturbel SiC 886 887 Table S2. Analytical variability (expressed as a %) for each soil order. Values are 888 estimated means and confidence intervals (CI) that have been backtransformed. 889 Soil Order Mean 95% CI Standard Error Alfisol 10.5 (9.1, 12.0) 1.1 Andisol 9.5 (8.3, 10.9) 1.1 Aridisol 14.2 (12.2, 16.5) 1.1 Entisol 9.1 (7.9, 10.4) 1.1 Gelisol 2.3 (1.8, 2.8) 1.1 Histosol 2.5 (2.1, 3.0) 1.1 Inceptisol 9.3 (8.1, 10.7) 1.1 Mollisol 6.4 (5.5, 7.3) 1.1 Oxisol 11.8 (10.2, 13.7) 1.1 Spodosol 6.8 (5.9, 7.8) 1.1 Ultisol 6.7 (5.8, 7.7) 1.1 Vertisol 7.6 (6.6, 8.7) 1.1 890 891 Table S3. Estimated means and bootstrapped 95% confidence intervals (5000 892 resamplings) for estimated parameters of regression parameters to convert between 893 processing treatments. 894 Conversion Intercept Slope < 2.0mm: 0.75 g ® 2.5 g 56.36 (17.32, 84.70) 0.633 (0.600, 0.698) < 0.5mm: 0.75 g ® 2.5 g 86.03 (45.60, 127.77) 0.608 (0.561, 0.668) 0.75 g: < 0.5mm ® < 2.0mma 43.77 (-23.66, 101.51) 0.821 (0.724, 0.939) 2.5 g: < 0.5mm ® < 2.0mm 27.85 (-24.85, 59.78) 0.820 (0.729, 0.951) a conversion including soils with SOC above ~ 10%: intercept = -13.35 (-66.71, 45.41); slope = 0.902 (0.814, 0.949) 895 896 Table S4. Number of replicates needed for a given confidence interval to be ± 20% of 897 the population mean. 898 Mass 2.5g Sieve Size < 0.5 mm < 2.0 mm Confidence Level 99% 95% 90% 80% 99% 95% 90% 80% SOC Content (%) 0.2 – 2.0 38 22 16 9 30 17 12 7 2.1 – 5.0 49 28 20 12 20 12 8 5 5.1 – 10.0 36 21 15 9 17 10 7 4 Mass 0.75g Sieve Size < 0.5 mm < 2.0 mm Confidence Level 99% 95% 90% 80% 99% 95% 90% 80% SOC Content (%) 0.2 – 2.0 38 22 16 9 30 17 12 7 2.1 – 5.0 49 28 20 12 20 12 8 5 5.1 – 10.0 36 21 15 9 17 10 7 4 899 900 Figure S1. Analytical variability by soil order. Letters indicate significant difference by 901 Tukey’s HSD (a = 0.05). 902 903 View publication stats