These methods, instrumentation and/or protocols apply to all data in this dataset:| Methods and protocols used in the collection of this data package |
|---|
| Description: |
Leaf litter decomposition was analyzed for two years from March 2018-March 2020. Experimental design was a reciprocal transplant design, where leaf litter was collected from both a fertilized and reference watershed, and transplanted in the watershed of origin and opposite watershed. One watershed, +N WS3, received 35 kg N ha-1 yr-1 from 1989-2019 in the form of ammonium sulfate ((NH4)2SO4). An adjacent, similarly aged (last clear cut in c. 1969) watershed, Ref WS7, serves as a reference to +N WS3.
We collected freshly fallen leaf litter of the four dominant species in October of 2017 from a single site in each watershed. Leaf litter from each watershed was thoroughly mixed, sorted by species and watershed of origin, then dried at 65 °C for > 48 hours. For two species (black cherry and sweet birch) sourced from Ref WS7, insufficient litter mass was collected, so we used dried and archived leaf litter collected in 2015 (<8% of total leaf litter used in this study) to supplement the 2017 freshly fallen leaf litter.
We measured rates of leaf litter decomposition using 1-mm mesh fiberglass litterbags (~ 20 cm x 10 cm) filled with 2 g (+/- 0.25 g) of dried leaf litter of a single species and from a single source watershed. In March 2018, five replicate litterbags for each combination of tree species and source watershed were randomly assigned to each plot and placed flat on the surface of the mineral soil horizon after removing the litter layer. All litterbags in a plot were arranged in a 1-m x 1-m square, covered with coarse plastic mesh to prevent disturbance, and the litter layer was replaced on top of the coarse mesh. Litterbags were collected four times between deployment (March 2018) and the end of the study (March 2020). One litterbag of each species and watershed of origin was collected from each plot after 3, 6, and 12 months. After 24 months, two litterbags of each species and watershed of origin were collected from each plot for the final collection. Following collection, litter in each bag was gently brushed to remove soil, and roots and invertebrates were removed as best as possible without losing leaf litter material. Litter was then dried at 65 °C for > 48 hours and weighed.
Overall, the experimental design consisted of 2 watersheds of origin x 2 watersheds of transplant x 10 plots per watershed x 4 species x 5 time points for a total of 40 litterbags per plot and 800 litterbags in total. The reciprocal design of this experiment allowed us to assess whether any detectable differences in decomposition rates between watersheds were due to differences in litter chemistry between source watersheds or differences in the soil environment into which litter bags were transplanted.
Initial Litter Chemsitry
To determine initial litter quality, three subsamples of freshly fallen litter collected for each species and watershed of origin were dried, ground using a mechanical mill through a size 20 mesh, and analyzed for C and N content using Dumas combustion in an elemental analyzer (NA 1500 Series 2, Carlo Erba Instruments, Milan, Italy). Dried, partially decomposed leaf litter from all 800 litterbags collected during the two-year experiment were similarly ground and analyzed for C and N content.
Initial lignin and cellulose content of leaf litter from each species and watershed of origin (n=4-5 replicates) were determined using an acid detergent digest method (Goering & Van Soest, 1970; Van Soest, 1967; Tappi, 1981; Holtzapple, 2003). In summary, 4-5 subsamples of dried, ground leaf litter were digested in an acid-detergent fiber digest solution to isolate cellulose, lignin, and ash. This residue was dried at 65 °C for > 48 hours and weighed. To remove and estimate cellulose in the residue, the samples were then soaked in 75% sulfuric acid, rinsed with DI water, dried at 65 °C for > 48 hours, then weighed. Final residue was then heated in a muffle furnace at 525 °C for 2 hours to determine ash-free dry weight. For the purposes of this study, we consider the ash-free mass remaining after the acid detergent and strong acid digests to be “lignin.” We similarly assessed lignin and cellulose content of final decomposed litter (after 24 months) by randomly selecting a subset of three litter samples for each species, source watershed, and watershed of transplant category (48 total).
We estimated the lignocellulose index (LCI) as lignin content/(lignin content + cellulose content) (Melillo et al., 1989). We also calculated the lignin:N ratio of initial and final leaf litter. Final leaf litter lignin:N was calculated for the subset of samples that were analyzed for lignin and % N (48 samples total). Because our initial litter chemistry analysis used a different number of subsample replicates for determination of % N (n=3) and % lignin (n=4 or 5), we paired every % N measurement with every % lignin measurement for a given species and source watershed category to determine the range and statistics of initial leaf litter lignin:N values.
Soil chemistry
The total C and N content of the mineral soil in each of the litter decay plots were measured on samples collected in October 2018. From each plot, we collected three 2.5-cm diameter soil cores to a depth of 5 cm. The three soil cores from each plot were combined, sieved (to pass a 2-mm mesh), dried at 65 °C for > 48 hours, and ground with mortar and pestle prior to analysis of C and N content by Dumas combustion.
Calculations
Percent mass remaining was calculated for each litterbag. Despite our efforts to clean the decomposed litter of soil, some soil could not be removed without potentially losing leaf tissue. To correct for soil contamination, we assumed that the C concentration of the leaf litter remains constant over decay. Thus, any decomposed leaf litter with a C concentration lower than the initial value was considered to be due to mineral soil contamination (Blair and Crossley, 1988; Janzen et al., 2002; Midgley et al., 2015). The following mixing model was used to determine the fraction of final mass that was litter:
fLitter=(Cd-Cs)/(Ci-Cs)
where fLitter = the fraction of the total litterbag sample mass that is actually litter; Cd = the decomposed litter C concentration; Cs = the mineral soil (0-5 cm) C concentration; Ci = the initial leaf litter C concentration. The mass of the decomposed litter sample was then multiplied by fLitter to correct for soil contamination.
After correcting litter mass values for soil contamination, we calculated the decomposition rates of leaf litter using a single-pool negative exponential model,
M_t=〖exp〗^(-kt)
where Mt is the proportion of initial mass remaining at a given timepoint, k is the decomposition rate (year-1) and t is the decomposition time (years) (Jenny et al., 1949; Olsen, 1963). To estimate the decomposition rate (k), an exponential model was fit to the proportion of mass remaining over time (years) for each combination of species, source watershed, and watershed of transplant. In this analysis, plots were the replicates (n=10), and 160 models were fit to 160 sets of litterbags (4 species x 2 WS origin x 2 WS transplant x 10 plots). We also used a model structure with the intercept set to zero to avoid bias in single-pool decay models (Adair et al., 2010). R2 values were > 0.80 for > 80% of model fits, and given the relatively short duration of this study, the single-pool exponential model is thought to best capture early-stage decomposition dynamics (Harmon et al., 2009).
| | Instrument(s): | Dumas combustion in an elemental analyzer (NA 1500 Series 2, Carlo Erba Instruments, Milan, Italy) |
| | Description: |
Statistical analysis for litter chemical properties and decomposition rates:
For the leaf litter decomposition study, we tested for differences in initial litter chemistry between source watersheds and among species with a two-way ANOVA in which species and source watershed were fixed effects and litter chemical properties were the dependent variables (%C, %N, C:N ratio, %cellulose, %lignin, LCI, lignin:N ratio).
To test for differences in final litter chemistry and decomposition rate among litter species, source watershed, and watershed of transplant, we conducted a 3-way ANOVA with litter species, source watershed, and watershed of transplant as fixed effects; and final litter chemical properties and decomposition rate as the dependent variables (%N, C:N ratio, %cellulose, %lignin, LCI, lignin:N ratio, k). Comparisons among means were analyzed with Tukey-Kramer HSD post hoc test. Normal distribution of residuals was tested using the Shapiro-Wilks test, and homogeneity of variance was tested using Levene’s test. Of the decay rates (k), 5 of 160 values greater than 1.3 were statistical outliers (Rosner’s test; Rosner, 1983) that caused violation in the ANOVA assumption of normally distributed residuals. However, we chose not to remove these outliers and continue with the ANOVA test, as doing so did not change the statistical results nor interpretation of the ANOVA test.
Robust two- and three-way ANOVAs (using the R package “rfit”; Kloke and McKean, 2012) were performed to compare initial % lignin, lignin:N ratio and LCI, and final % N and C:N ratio, respectively, as the assumption of a normal distribution of residuals were not met by these dependent variables. The robust ANOVA uses Wilcoxon rank-based estimators based on reduction of dispersion (RD), instead of traditional least squares (LS) estimators, for testing main effects and interaction (Hocking, 1985; Kloke and McKean, 2014). Comparison among means were analyzed using Wilcoxon rank sum test with continuity correction.
To evaluate relationships between decomposition rates and soil chemistry, we performed a regression analysis. Specifically, we regressed k values for each litter species on soil properties (C, %N, and C:N ratio) measured for the watershed of transplant.
|
| | Description: |
Soil density fraction analysis:
Soil sampling
To assess how elevated N inputs may influence the fate of plant inputs, we measured SOM fractions in both study watersheds, using soil collected from four 5-cm diameter soil cores of the 0-15 cm of mineral soil in each plot in October 2018. These were the same plots used for the litter decomposition study. Soils were stored less than six weeks at 4 °C before being sieved (2 mm) to remove plant and rock material, homogenized, and dried at 65 °C for > 48 hours.
Fractionation procedure
We evaluated the physicochemical nature of organic matter in the mineral soil in each plot using a three-pool soil density fractionation framework described by Lavallee et al. (2019). Briefly, SOM was separated into three pools—light POM, heavy POM, and MAOM—based on their densities and sizes, which is thought to represent the degree of organic matter degradation (Gregorich et al., 2006). Two steps were used to isolate the light POM fraction. First, soil samples were shaken for 15 minutes in deionized (DI) water at ~100 oscillations per minute, centrifuged at 1874 g for 15 minutes, and then the supernatant was filtered through a 20 µm nylon filter. Particulate matter on the filter was considered part of the light POM, which we define as plant-like residue with a density < 1.85 g cm-3 ¬because it is minimally bound to soil minerals. Second, we isolated the rest of the light POM by shaking soils in a dense liquid (1.85 g cm-3 of sodium polytungstate, SPT) for 18 hours to disperse soil macroaggregates, centrifuging at 1874 g for 30 minutes, and aspirating the light POM that floated out of the dense liquid onto a 20 µm nylon filter. This light POM was rinsed with DI, combined with the light POM from DI water suspension, and dried at 65 °C for > 48 hours.
In contrast to the light POM fraction, the heavy POM—still plant-like, chemically—has some mineral association or microbial biproducts that increases its density and may protect the SOM in soil aggregates. Thus, following the removal of light POM with a density >1.85 g cm-3, the centrifuged soil pellets containing the heavy fractions (the denser microbial biproducts and soil minerals) were thoroughly rinsed and centrifuged with DI water at least three times to remove excess SPT. The heavy POM and MAOM that remained in the soil pellet were then separated by size by suspending the pellet in DI water and sieving through a 53 µm sieve. The material remaining on the sieve was considered the heavy POM and sand (>1.85 g cm-3 density and > 53 µm in size), while the matter that passed through the sieve was considered the MAOM, silt and clay fraction (< 53 µm).
All soil fractions were dried at 65 °C, then ground for C and N analysis. If 100% (+/- 5%) of initial soil sample mass was not recovered in all fractions, then the fractionation procedure was repeated for that sample; this occurred for 7 of the 80 samples that were fractionated.
| | Instrument(s): | Dumas combustion in an elemental analyzer (NA 1500 Series 2, Carlo Erba Instruments, Milan, Italy) |
| | Description: |
Statistical analysis for soil density fractionation analysis
For the soil density fractionation study, we tested for differences in the chemistry of bulk soil between the watersheds with a one-way, nested ANOVA, in which watershed was a fixed effect, plot was a random nested effect (within WS), and bulk soil chemical properties were the dependent variables (%C, %N, and C:N ratio). To test for differences in fraction of bulk soil in each density fraction and chemistry of individual fractions between watersheds, we conducted a one-way, nested ANVOA with watershed as the fixed effect, plot as the random nested effect (within WS), and fraction of total mass, total C, and total N, and chemical properties of each fraction as the dependent variables (fraction of total mass, C and N for light POM, heavy POM, and MAOM; %C, %N, and the C:N ratios of light POM, heavy POM, and MAOM). Comparisons among means were analyzed with Tukey-Kramer HSD post hoc tests. Normal distribution of residuals was tested using the Shapiro-Wilks test, and homogeneity of variance was tested using Levene’s test. Variables that did not meet these assumptions were transformed using the natural logarithm prior to statistical analysis. All means and errors presented in results are arithmetic means of original, untransformed data. To test our hypothesis that a greater proportion of light POM may contribute to a greater C:N ratio in the bulk soil, we regressed the bulk soil C:N ratio against the fraction of total mass in the light POM.
|
|
|
|
