Statistical Analysis Report

Main samples (field 1-14)

Customer	Innovation Centre for Organic Farming, Tove Mariegaard Pedersen
Customer ID	DA00204-22
Project	The microbial community of the field, 2. year (2022).
Sample Type	Soil
Number of samples	48 samples
Type of data	ITS2 region

Introduction to the biostatistical analysis

The Project

The current report describes microbiome profiles of 48 samples collected in the second year of the project, year 2022, across 48 fields. For each field, one sample was collected to represent the field, corresponding to the ‘main’ samples collected for each filed from 2021. These samples were taken for each field based on 16 subsamples taken in a w-pattern throughout the field.

In 2023, the scope will be expanded to include a total of approx. 100 fields (both conventional and organic fields). In this report we aalyse the samples from 2022 alone. Later we will perform joined analysis to both identify robust patterns arcross the years and evaluate if any patterns appear year-specific indicating varying conditions such as whether.
The aim is to evaluate how the microbiome of the fields associate with other field parameters of both agricultural practices and soil indicators of nutrients, type and structure.
Only when the full data set for 2021-2023 is ready, a final analysis can be conducted.

We separate the evaluations into two R3 reports:

• The first report presents results for the analysis of microbial taxonomy for samples form 2022, where the samples’ microbiome profiles are related to information on cultivation practices / cultivation conditions.
• The second report presents results for the analysis of microbial functional features for samples form 2022, where the samples’ microbiome profiles are related to information on cultivation practices / cultivation conditions.

Analysis

In “Report 3”, biostatistical analyses are performed and the results presented, building on the data generated and evaluated in the 2 prior reports (Report 1: Sequencing and data processing report, Report 2: Microbiome profiling report).
Through biostatistical analysis we relate the microbiome profiles to the key variables selected for year 2022. The focus here is to evaluate how and to what extent the variables shape and relate to the soil microbiome composition and diversity. We therefore focus on the overall structure of the microbiome also called the microbiome composition and the diversity.

The key variables assessed in this report are summarized with summary statistics across the 48 samples in the below table.

Summary Statistics

Variable	N	Mean	Std. Dev.	Min	Pctl. 25	Pctl. 75	Max
JB_value	48
… 1	19	39.6%
… 2	5	10.4%
… 5	1	2.1%
… 6	20	41.7%
… 7	3	6.2%
Earthworm_status	48
… 0	11	22.9%
… 1	37	77.1%
Cold_soil	48
… 0	33	68.8%
… 1	15	31.2%
Compact_soil	48
… 0	40	83.3%
… 1	8	16.7%
field_well_drained	48
… 0	7	14.6%
… 1	41	85.4%
Mulching_of_straw	48
… 0	25	52.1%
… 1	23	47.9%
Clovergrass_within_3_years	48
… 0	36	75%
… 1	12	25%
No_plough	48
… 0	36	75%
… 1	12	25%
ConservationAgriculture	48
… 0	45	93.8%
… 1	3	6.2%
Years_since_plowing	48	3.417	2.988	1	1	4.25	11
Rt	48	6.487	0.489	5.7	6.2	6.75	7.6
Phosphorus	48	3.081	1.301	0.7	2.175	3.925	6
Potassium	48	10.473	6.247	1.5	7.325	14	41
Magnesium	48	6.89	2.577	1.9	5.5	8.15	16
Cobber	48	2.481	0.885	1	1.8	3.1	5.1
Organic_material_perc	48	3.025	1.363	1.16	2.173	3.372	8.74
Clay_perc	48	9.258	4.727	2.4	4.675	12.85	20
Nitrogen_perc	48	0.144	0.052	0.07	0.1	0.17	0.28
Organic_farm	48
… 0	24	50%
… 1	24	50%
Years_since_turning_organic	48	4.521	4.758	1	1	7.25	15
Livestock	48
… 0	20	41.7%
… 1	28	58.3%
Livestock_manure	48
… 0	16	33.3%
… 1	32	66.7%
Commercial.fertilizer	48
… 0	24	50%
… 1	24	50%
Vinasse	48
… 0	46	95.8%
… 1	2	4.2%
Cast	48
… 0	46	95.8%
… 1	2	4.2%
Degassed.fertilizer	48
… 0	35	72.9%
… 1	13	27.1%
Chalked	47
… 0	38	80.9%
… 1	9	19.1%

Table 1: Summary statistics of the key variables selected for evaluation in relation to the fields microbiome profiles in year 2022.

Evaluation of overall microbiome profiles

We initiate the evaluation of the 10 samples (1 per field) with a stacked barplot of the microbiome profiles in each sample. This allows us to make a first evaluation of the extent of difference in the taxonomic profiles between the fields.

Note that in order to show the organisms with a color scheme that is interpretable, it is necessary to filter the profiles and select a subset of the most abundant clades to be included in the plots. The filtering used is specified in the axis labels of each plot (e.g. >2% in the relative abundance plots mean that a clade must have a relative abundance across samples of more than 2% in order to be included in the plot).

Stacked barplots

The stacked barplots allow us to visually access the stability of the taxonomic profile across the fields, and get a feeling of the level to which individual clades are found across field or more sporadic. Compared to the bacterial part of the microbiome, the fungi show a large deviation between fields, with both large variation in some, and others that are dominated by a few clades. And we see how the dominating clade is also different between many fields.

Phylum

---

Figure 1: Visualization of the fungal community in the samples. Stacked barplots of taxonomic clades in each of the evaluated samples. Clade abundance was transformed to relative abundance to sum to 100% in each sample.

---

Class

Figure 2: Visualization of the fungal community in the samples. Stacked barplots of taxonomic clades in each of the evaluated samples. Clade abundance was transformed to relative abundance to sum to 100% in each sample.

---

Order

Figure 3: Visualization of the fungal community in the samples. Stacked barplots of taxonomic clades in each of the evaluated samples. Clade abundance was transformed to relative abundance to sum to 100% in each sample.

---

Overall microbiome communities related to factor variables

We use the overall microbiome profiles presented in the stacked barplots above, to calculate a measure of difference in the microbiome composition between samples (beta-diversity). The calculated beta-diversity measures are used for visual inspection of the relationship between the microbiome profiles in the so called ordination plots (see below), and in a statistical model named ADONIS (or PERMANOVA, see details below) to evaluate if the overall microbiome composition associates with the selected variables. For these evaluations we focus on the variables that can be viewed as binary or generating groups (e.g. JB value, Earthworm status, Crops, Mulching of straw and Years since plowing). These variables are “grouping variables” that allow us to group samples into subgroups. The remaining variables constitute continuous values (concentrations or percentages) and therefore, we use a different set of tools to evaluate how they relate to the overall microbiome composition (see below).

Note that “years since plowing” and “Years since turning organic” are analyzed as integers with increasing values (0,1,2, etc) and could be analyzed both with ADONIS and the model used for continuous variables (ENVFIT, see below). However, we find it interesting to visualize the shift in the composition according to “years” in an ordination plot, and the variable is therefore included here.

Visualization by ordination (beta-diversity)

As described in Report 2, beta-diversity is a measure of how similar or dissimilar the fungal community is between each pair of samples. The measures are useful for statistical analysis and visualization of the overall microbiome community. In ordination plots, each sample is a point and the distance between the points increases with increasing dissimilarity in the microbiome communities.

Here we evaluate the microbiome communities using the Bray-Curtis, Aitchison and Jaccard beta-diversity measures. If not all plots are shown in this report, you can find them in the project folder. We use a combination of beta-diversity measures as each measure highlights different properties of the microbiome. See more details in Report 2.

We use the different measures in combination with different microbiome profiles (taxonomic levels and normalization) as follows:

• Bray Curtis and Jaccard are computed from the relative abundance data, at the the genus level
• Aitchison is computed from the total abundance data transformed with central-log-ratio (CLR), at the genus level

The Aitchison distance is a simple euclidean distance calculated using CLR transformed microbiome profiles. An analysis of CLR transformed data will reveal how the organisms behave relative to the per-sample average microbiome. Values for a microbe can therefore be negative after CLR transformation - meaning that it makes up a smaller amount of the microbiome than the average abundant microbe. This is a very different way to view the microbiome than Bray-curtis and Jaccard that uses the data as relative proportions (i.e. how big a proportion of the sample’s microbiome does the individual microbe comprise). This might appear unnecessarily mathematical and unrelated to agrobiology but the CLR transformation has proved to be able to pinpoint patterns in microbiomes that are driven by environmental factors such as nutrient content or treatment applied to the samples. We therefore evaluate structures in the dataset using all three measures.

Permutational Multivariate Analysis of Variance

To evaluate if the compositional differences evaluated below using ordination plots explain a notable amount of the variation in the microbial composition, and if the amount of explained variation is statistically significant, we perform an analysis named Permutational Multivariate Analysis of Variance (ADONIS). ADONIS uses sums of squares of a multivariate dataset and is analogous to MANOVA (multivariate analysis of variance) using beta-diversity measures. It uses distance matrices among sources of variation and fits linear models to the distance matrices using a permutation test with pseudo-F ratios and can therefore be considered as a “permutational manova”.

For the analysis we use Bray-Curtis, Jaccard and Aitchison beta-diversity measures and perform the analysis at the phylum level down to the ASV level. The latter is used in amplicon sequencing in which a group of exact sequences is referred to as an amplicon sequence variant (ASV).

Each table shows results from evaluation of the effect of one variable and there is thus one table per variable.

In each plot, samples are colored by the variable being assessed.

---

JB value

Figure 4: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.1335	0.131	0.0559	0.805	0.0763	0.585
Class	0.115	0.186	0.0715	0.772	0.092	0.237
Order	0.1077	0.182	0.0922	0.245	0.0984	0.114
Family	0.0994	0.222	0.0946	0.166	0.0994	0.068
Genus	0.1109	0.071	0.101	0.061	0.1046	0.024
ASV	0.1107	0.028	0.0983	0.003	0.1034	0.03

Table 2: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

Earthworm status

Figure 5: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.0581	0.046	0.0169	0.54	0.0252	0.254
Class	0.0288	0.235	0.0258	0.234	0.0233	0.289
Order	0.034	0.106	0.0168	0.765	0.0208	0.481
Family	0.035	0.05	0.0214	0.42	0.0219	0.389
Genus	0.0332	0.047	0.0223	0.326	0.0219	0.394
ASV	0.0303	0.043	0.0218	0.323	0.0217	0.384

Table 3: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

Cold_soil

Figure 6: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.0104	0.632	0.0233	0.346	0.0275	0.208
Class	0.0079	0.901	0.0262	0.219	0.0254	0.187
Order	0.0092	0.927	0.0234	0.288	0.0251	0.151
Family	0.0186	0.594	0.0228	0.3	0.025	0.121
Genus	0.0181	0.652	0.0246	0.136	0.0256	0.077
ASV	0.0246	0.196	0.0229	0.113	0.023	0.19

Table 4: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

Compact soil

Figure 7: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.0044	0.882	0.0091	0.812	0.0098	0.948
Class	0.0241	0.324	0.017	0.684	0.0194	0.579
Order	0.0248	0.293	0.0176	0.728	0.0179	0.776
Family	0.0234	0.322	0.0157	0.923	0.0166	0.923
Genus	0.0226	0.362	0.0178	0.845	0.0182	0.815
ASV	0.0259	0.138	0.0197	0.878	0.0194	0.735

Table 5: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

Field well drained

Figure 8: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.1124	0.008	0.0219	0.386	0.0328	0.119
Class	0.0468	0.043	0.0172	0.667	0.0262	0.18
Order	0.0397	0.046	0.0239	0.274	0.0282	0.074
Family	0.0386	0.037	0.0265	0.119	0.026	0.118
Genus	0.039	0.024	0.0254	0.13	0.0246	0.168
ASV	0.0268	0.095	0.0222	0.244	0.0233	0.219

Table 6: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

Mulching of straw

Figure 9: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.0093	0.693	0.0209	0.41	0.017	0.64
Class	0.0636	0.009	0.0236	0.308	0.0263	0.16
Order	0.0576	0.006	0.0209	0.471	0.024	0.218
Family	0.0426	0.021	0.0191	0.629	0.0265	0.071
Genus	0.0364	0.035	0.0225	0.295	0.027	0.046
ASV	0.0369	0.008	0.0251	0.014	0.0264	0.01

Table 7: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

Clovergrass within 3 years

Figure 10: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.0039	0.886	0.0294	0.224	0.0203	0.438
Class	0.0408	0.085	0.0364	0.046	0.0321	0.03
Order	0.0334	0.118	0.0349	0.027	0.0346	0.007
Family	0.0373	0.046	0.0274	0.078	0.0362	0.006
Genus	0.0347	0.053	0.026	0.094	0.0324	0.008
ASV	0.0363	0.012	0.0259	0.007	0.0308	0.002

Table 8: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

No plough

Figure 11: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.0098	0.662	0.0174	0.522	0.0119	0.877
Class	0.0239	0.323	0.0369	0.05	0.0187	0.668
Order	0.0312	0.159	0.0344	0.032	0.0199	0.615
Family	0.0292	0.146	0.0249	0.174	0.0212	0.477
Genus	0.0268	0.173	0.0232	0.24	0.0219	0.365
ASV	0.0248	0.179	0.0207	0.606	0.0201	0.679

Table 9: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

Conservation Agriculture

Figure 12: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.0353	0.143	0.1048	0.004	0.0395	0.078
Class	0.0497	0.046	0.0706	0.001	0.0388	0.018
Order	0.067	0.005	0.0563	0.001	0.0304	0.074
Family	0.0592	0.003	0.0375	0.021	0.0237	0.283
Genus	0.0423	0.022	0.0324	0.032	0.0208	0.511
ASV	0.0318	0.051	0.0243	0.077	0.0142	0.983

Table 10: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

Years since plowing

Figure 13: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.0076	0.744	0.0243	0.322	0.0177	0.581
Class	0.0166	0.559	0.0362	0.055	0.0236	0.296
Order	0.0261	0.254	0.0328	0.045	0.0212	0.467
Family	0.0302	0.12	0.0242	0.226	0.022	0.391
Genus	0.0277	0.158	0.0214	0.414	0.0218	0.386
ASV	0.0259	0.15	0.021	0.511	0.0205	0.623

Table 11: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

Organic farm

Figure 14: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.0147	0.518	0.045	0.076	0.0213	0.441
Class	0.0187	0.47	0.0393	0.034	0.0245	0.258
Order	0.0201	0.455	0.0305	0.071	0.0279	0.074
Family	0.0276	0.18	0.0243	0.21	0.0277	0.047
Genus	0.0319	0.093	0.0246	0.156	0.0267	0.058
ASV	0.0308	0.05	0.0223	0.208	0.0237	0.097

Table 12: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

Years since turning organic

Figure 15: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.0092	0.671	0.0571	0.031	0.0467	0.016
Class	0.0172	0.534	0.0373	0.046	0.0324	0.043
Order	0.0195	0.473	0.0336	0.025	0.0344	0.005
Family	0.0268	0.207	0.0295	0.054	0.0295	0.021
Genus	0.0275	0.144	0.0283	0.049	0.0281	0.033
ASV	0.0307	0.049	0.0238	0.055	0.0265	0.028

Table 13: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

Livestock

Figure 16: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.0161	0.527	8e-04	0.971	0.0152	0.731
Class	0.0171	0.574	0.0244	0.278	0.0275	0.111
Order	0.0169	0.605	0.0228	0.335	0.0267	0.093
Family	0.0287	0.171	0.0205	0.506	0.0254	0.117
Genus	0.0291	0.126	0.0225	0.275	0.0265	0.073
ASV	0.0301	0.041	0.0235	0.074	0.0253	0.048

Table 14: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

Livestock manure

Figure 17: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.0101	0.654	0.0069	0.879	0.0193	0.533
Class	0.0326	0.147	0.0164	0.7	0.0211	0.469
Order	0.0308	0.145	0.0183	0.672	0.0244	0.206
Family	0.0247	0.248	0.023	0.297	0.0255	0.132
Genus	0.0234	0.299	0.0231	0.236	0.0229	0.282
ASV	0.0235	0.272	0.0209	0.59	0.0228	0.242

Table 15: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

Commercial fertilizer

Figure 18: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.0147	0.518	0.045	0.076	0.0213	0.441
Class	0.0187	0.47	0.0393	0.034	0.0245	0.258
Order	0.0201	0.455	0.0305	0.071	0.0279	0.074
Family	0.0276	0.18	0.0243	0.21	0.0277	0.047
Genus	0.0319	0.093	0.0246	0.156	0.0267	0.058
ASV	0.0308	0.05	0.0223	0.208	0.0237	0.097

Table 16: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

Vinasse

Figure 19: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.0039	0.898	0.0255	0.296	0.0202	0.394
Class	0.018	0.447	0.0181	0.546	0.0196	0.485
Order	0.016	0.63	0.0231	0.298	0.0218	0.36
Family	0.0221	0.351	0.0171	0.731	0.0201	0.514
Genus	0.0242	0.257	0.0178	0.723	0.0196	0.575
ASV	0.0252	0.211	0.0195	0.761	0.0201	0.528

Table 17: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

Cast

Figure 20: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.0032	0.922	0.0238	0.275	0.0139	0.705
Class	0.016	0.521	0.0216	0.367	0.0159	0.794
Order	0.0145	0.704	0.0206	0.447	0.0155	0.846
Family	0.0117	0.897	0.0178	0.621	0.0156	0.891
Genus	0.0131	0.898	0.0187	0.591	0.0173	0.799
ASV	0.0143	0.935	0.0196	0.764	0.0192	0.614

Table 18: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

Degassed fertilizer

Figure 21: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.0271	0.271	0.0241	0.324	0.0158	0.672
Class	0.0211	0.395	0.0275	0.171	0.0211	0.437
Order	0.0175	0.57	0.0294	0.084	0.025	0.182
Family	0.0209	0.437	0.0228	0.28	0.0209	0.491
Genus	0.0209	0.446	0.0219	0.347	0.0208	0.525
ASV	0.0227	0.312	0.0212	0.46	0.0231	0.231

Table 19: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

Chalked

Figure 22: Visualization of structure of the fungal community between the samples. Ordination plots using different beta-diversity measures and data transformations as stated in the plot titles. Dots are colored by the variable of interest as seen to the right of the figure panels.

---

	Bray-Curtis		Jaccard		Aitchison
Taxa level	R2	p	R2	p	R2	p
Phylum	0.0629	0.039	0.0263	0.311	0.028	0.207
Class	0.0475	0.043	0.0279	0.208	0.0274	0.144
Order	0.0491	0.022	0.0275	0.149	0.0292	0.051
Family	0.0598	0.003	0.0255	0.162	0.0272	0.075
Genus	0.0536	0.004	0.0273	0.071	0.0285	0.037
ASV	0.0428	0.005	0.0252	0.026	0.0271	0.03

Table 20: Results from ADONIS analysis. The table shows results from ADONIS analyses including samples from all fields. The analysis was performed using 999 permutations to robustly calculate significance. The table shows the obtained R-squared values that indicate the percentage of variation that the variable could explain and the corresponding p-values.

Overall microbiome communities related to continuous variables

We now turn to the continuous variables. We continue to use the beta-diversity measure from the microbiome, but use a different analysis model named ENVFIT. We again display the samples in ordination plots but add to these plots arrows representing the association of the variables with the microbiome.

ENVFIT

The function fits environmental variables onto an ordination using multiple regression and thus looks for linear relationships between the axis that drive the main variation in the microbiome and the environmental variables.

The output of the analysis gives the direction of the association in terms of coordinates in the ordination plot (the arrow head). In the plots, the arrows are scaled by their correlation (square root of R2) so that “weak” variables have shorter arrows than “strong” variables. The arrows point in the direction where they have maximal correlations with the ordination axis. The lengths of arrows for fitted vectors are automatically adjusted for the physical size of the plot, and the arrow lengths can thus not be compared across plots, only the arrows within a plot can be compared.

Note that the ENVFIT analysis uses the ordination exes and therefore is restricted to relate the environmental factors to the part of the microbiome that is represented by these axes (and thus not the full microbiome variation). This fact is also relevant when comparing R2 (the squared correlation coefficient) from ADONIS and envfit: ADONIS decomposes the entire dissimilarity matrix into “variance” explained by each covariate. In NMDS (used by ENVFIT), you have reduced the entire dissimilarity matrix to three dimensions and then look for correlations in those three dimensions with the covariates. The “discrepancy” is therefore due to the fact that different amounts of variation are assessed by the two models. Therefore, if ADONIS gives higher R2 values, this suggests that the effect of the variable is on the parts of the dissimilarity matrix not represented well by the 3-d NMDS solution.

• R2 - variation explained by the model of multiple regression; the square-root of this value is used to scale lengths of vectors (arrows) in the ordination diagrams (variables with higher squared (R2) are represented by longer arrows).
• Pr(>r) - the significance of the multiple regression, calculated by permutation test. Indicates whether the variable is related to ordination axes more than if it is a randomly generated one.

Figure 23: Visualization of the ENVFIT analysis. The association of the evaluated environmental variables with the ordination axis is displayed by arrows, with longer arrows indicating a stronger association.

---

.	Bray		Jaccard		Euclidean
Variable	R2	p	R2	p	R2	p
Rt	0.017	0.708	0.098	0.098	0.081	0.174
Phosphorus	0.102	0.082	0.044	0.367	0	0.991
Potassium	0.012	0.712	0.16	0.043	0.043	0.348
Magnesium	0.007	0.849	0.195	0.021	0.033	0.451
Cobber	0.082	0.143	0.054	0.27	0.016	0.707
Organic material (%)	0.11	0.093	0.054	0.229	0.228	0.005
Clay (%)	0.054	0.281	0.234	0.001	0.212	0.004
Nitrogen (%)	0.088	0.146	0.116	0.069	0.105	0.093

Table 21: Results from ENVFIT analysis across all fields. The table shows results from ENVFIT analyses including samples from all fields. The table shows the obtained R-squared values and the corresponding p-values from the ENVFIT analysis, for each of the assessed variables in each of the three beta-diversity ordinations.

---

Non-linear relationships for continuous environmental variables

The ENVFIT model looks for linear relationships between the main variation of the microbiome (the separation of the samples along the first ordination axes) and the environmental variables. It can also be relevant to look for non-linear relationships - so called smooth surfaces - where the environmental variable for example is highest in the centrally located samples on the ordination plot and lower in samples located towards the edges. We use a mode called Ordisurf to look for such non-linear patterns. Ordisurf internally uses generalized additive models with integrated smoothness estimation, so-called GAMs, to fit the variable. Below is a table of the results from the analysis looking for smooth surfaces for the selected variables and further down are the surfaces illustrated on ordination plots.

Euclidean
	F	P
Rt	0.223	1.49e-01
Phosphorus	0.27	2.14e-01
Potassium	0.042	3.67e-01
Magnesium	0	5.77e-01
Cobber	1.101	2.41e-02
Organic_material_perc	2.138	5.75e-04
Clay_perc	1.213	3.37e-03
Nitrogen_perc	0.633	5.57e-02

Table 22: Results from non-linear analysis. The table shows results from the non-linear analyses performed using Ordisurf. The table shows the obtained F values and the corresponding p-values, for each of the assessed variables based on the Aitchison beta-diversity.