Living within the Earth’s soil are millions of insects, worms and other invertebrates, which help keep the ground healthy and fertile. There is a growing concern that changing land-use habits, such as agriculture and urban development, are causing these populations of invertebrates to decline. However, to what extent different types of land use negatively impact soil invertebrates is not clear.

Healthy habitats often have a greater variety of species. This biodiversity can be measured in a number of ways, ranging from counting the number of species, to more complex approaches that calculate a species’ role in an ecosystem or how close it is to extinction. Finding a way to sensitively measure the biodiversity of soil invertebrates could further researcher’s understanding of how different types of land use are affecting these communities.

A new method known as DNA metabarcoding has made it easier to distinguish between different species and calculate the biodiversity of entire populations. Now, Dopheide et al. have used this technique to study invertebrate communities from 75 sites across New Zealand which have been impacted by different land-use habits. This revealed that the most reliable and consistent way to uncover how land use affects soil invertebrates was to measure the rarity of species (i.e. the number of unique species present at each site).

Dopheide et al. found that agriculture negatively affected soil invertebrates and that most types of invertebrates responded in a similar way. Horticulture – such as orchards and vineyards – had the most severe impact, with the lowest variety of species compared to grassland or forest.

Other measurements of biodiversity, such as the number of different species, may underestimate the negative impact agriculture is having on invertebrate communities. The findings of Dopheide et al. highlight why developing strategies to preserve and restore these communities is so important. However, more work is needed to understand what specifically is causing biodiversity to decline and how this effect can be reversed.

Land-use changes through deforestation, agricultural development, and urbanisation have caused worldwide impacts on the biodiversity of terrestrial communities and ecosystems (Dirzo et al., 2014; Newbold et al., 2015). Invertebrates are the most diverse and abundant component of animal biodiversity worldwide and are major contributors of terrestrial ecosystem services such as pollination, soil formation, and nutrient cycling (Lavelle et al., 2006; Wagg et al., 2014; Yang and Gratton, 2014). Long-term declines in the richness and biomass of insects and other terrestrial invertebrates are predicted to have major impacts on food webs and ecosystem functions (Eisenhauer et al., 2019; Hallmann et al., 2017; Potts et al., 2010). Despite this, most invertebrate species remain undescribed, and there is an incomplete understanding of land-use effects on invertebrate biodiversity, particularly for those that reside in soils (Cameron et al., 2018; Eisenhauer et al., 2019).

Biodiversity loss is typically measured as reductions in species richness (i.e., total number of species; e.g. Forister et al., 2010; George et al., 2019; Newbold et al., 2015). Despite widespread concern about biodiversity loss, evidence for impacts of anthropogenic land use on terrestrial invertebrate species richness is mixed, with studies often detecting richness declines for some taxa or groups but not others (Allan et al., 2014; Attwood et al., 2008; Blaum et al., 2009). Among the few studies that have examined land-use impacts on below-ground invertebrate communities, one detected negative impacts of long-term disturbance on soil invertebrate richness (Callaham et al., 2006), another detected increasing alpha diversity and homogenisation of soil invertebrates with increasing grassland intensification (Gossner et al., 2016); while others detected inconsistent richness patterns among different soil invertebrate taxa across land uses (George et al., 2019; Tsiafouli et al., 2015; Wood et al., 2017). These inconsistent patterns make it difficult to draw general conclusions about the impacts of land use on soil invertebrate biodiversity (Allan et al., 2014), and make the use of individual taxa as bioindicators problematic (Gerlach et al., 2013).

Inconsistent patterns in biodiversity measurement may reflect limitations of the diversity index used. In particular, species richness provides no indication of the distribution, taxonomy or function of species or communities (Fleishman et al., 2006; Hillebrand et al., 2018), potentially overlooking the nature and extent of land-use impacts on soil invertebrate communities. In contrast, rarity (sometimes termed ‘endemism richness’; Kier et al., 2009) measures the extent to which species are widely distributed generalists or limited to particular sites or land-use types. Rarity may thus indicate homogenising effects of land use on communities (McKinney and Lockwood, 1999; Smart et al., 2006), and the conservation value of sites (Kier and Barthlott, 2001). Furthermore, rare species can contribute disproportionately to ecosystem functioning (Dee et al., 2019; Leitão et al., 2016; Lyons et al., 2005; Mouillot et al., 2013). Rarity may therefore more accurately reflect the impacts of land use on soil invertebrate communities than species richness.

Rarity and other diversity metrics can also be placed in a phylogenetic context. Phylogenetic diversity reflects the evolutionary history and taxonomic range of communities and associated traits and functions (Faith, 1992), thus providing robust information for conservation assessment purposes (Faith, 1992; Forest et al., 2007; González-Orozco et al., 2015; Mishler et al., 2014). Phylogenetic diversity can also act as a proxy for functional diversity, albeit imperfectly (Mazel et al., 2018; Srivastava et al., 2012; Winter et al., 2013). Phylogenetic rarity, calculated as the portion of a phylogeny that is unique to a region or habitat (Mishler et al., 2014; Rosauer et al., 2009), combines elements of both rarity and phylogenetic diversity; high phylogenetic rarity implies that a community contains a taxonomically distinct assemblage of species and associated ecosystem functions. Mean pairwise distance, meanwhile, measures the phylogenetic relatedness of species within a community, which may reflect land-use driven filtering or competitive exclusion processes (Webb et al., 2002). The additional information represented by rarity and phylogenetic biodiversity metrics suggests that land-use related patterns based on these values may be clearer and more consistent among soil invertebrate taxa than those based on species richness and other non-phylogenetic diversity measures. Furthermore, rarity and phylogenetic rarity may be more sensitive indicators of land-use impacts on soil invertebrate communities than richness or phylogenetic diversity, because the former metrics reflect the distribution of species and lineages whereas the latter do not. These possibilities remain untested.

Here we present a comprehensive analysis of soil invertebrate biodiversity across different land-use types at a national spatial scale. We use modern DNA metabarcoding methods to measure invertebrate responses, as this enables the rapid and detailed identification of large numbers of invertebrate specimens from multiple taxonomic groups simultaneously (Drummond et al., 2015; George et al., 2019; Wood et al., 2017; Yu et al., 2012) and allows more efficient calculation of biodiversity metrics than previously possible. We analysed the invertebrate faunas in soil samples collected from 75 sites distributed across five different major land-use categories (natural forest, planted forest, low-producing and high-producing grassland, and perennial cropland) throughout New Zealand. Based on these data, we calculated six different biodiversity metrics: species richness, effective species numbers, rarity, phylogenetic diversity, phylogenetic rarity, and mean pairwise distance; as well as standardised effect size (SES) values for the latter phylogenetic metrics. We used these metrics to assess the impacts of land use on a comprehensive range of soil invertebrate taxa. We tested the following hypotheses: 1) all soil invertebrate taxa show the same biodiversity trends across the five land-use types; 2) patterns of soil invertebrate rarity, phylogenetic diversity, and phylogenetic rarity across the five land-use types are more consistent among taxa than species richness or non-phylogenetic diversity; 3) rarity and phylogenetic rarity of soil invertebrates are more sensitive to land use than richness, diversity, or phylogenetic diversity.

This research provides clear evidence of adverse impacts of agricultural land use upon soil invertebrate communities. Effects of land use on biological communities are usually measured as shifts in species richness. However, rarity metrics were much more sensitive to land use and more consistent among taxa than richness or effective species numbers in our study, suggesting that the latter metrics may underestimate land-use impacts on biodiversity. Rarity is a function of the number of species with limited distributions or narrow habitat specificity. These rare species can have important roles in ecosystem processes (Dee et al., 2019; Leitão et al., 2016; Lyons et al., 2005), and are inherently more vulnerable to extinction. Overlooking species rarity, as richness does, therefore obscures the effects of different land uses on communities, with potential detrimental consequences for the function and resilience of ecosystems. Our results suggest that efficient DNA-based measurement of plot-level rarity improves our understanding of rare species occurrence and provides an effective basis for incorporating soil invertebrates into conservation planning.

Rare species include not only habitat specialists, but also transient and conditionally rare taxa. It is possible that OTUs that were rare in this study may be more common in locations not sampled. Nonetheless, our observation that patterns of rarity among land-use categories were the most consistent among different taxa suggests that rarity is an ecologically meaningful measure of ecosystem biodiversity. This is supported by prior studies suggesting that rare species are particularly sensitive to ecosystem change. For example, rare plant and fungal species appear to be particularly sensitive to changes in environment (Avis et al., 2008; Dickie et al., 2009; Dickie and Reich, 2005; McIntyre and Lavorel, 1994), and pollinating insect losses are concentrated among rare species (Powney et al., 2019). Furthermore, most of the terrestrial invertebrate species currently considered to be at risk or threat of extinction in New Zealand are naturally uncommon (Stringer and Hitchmough, 2012).

Phylogenetic diversity – and especially phylogenetic rarity – explained larger proportions of land-use variance across taxa than their non-phylogenetic counterparts, and phylogenetic rarity was overall the most sensitive metric to land-use differences. Phylogenetic metrics incorporate evolutionary and functional aspects of biodiversity (Faith, 1992; Faith, 2015; Mazel et al., 2018). New Zealand has a long history of geographic isolation and glaciation, reflected by the presence of many deeply divergent invertebrate lineages (Buckley et al., 2015; Trewick et al., 2011). The high levels of invertebrate phylogenetic rarity in natural forest sites likely reflects assemblages of long-present soil invertebrates that are highly adapted to these habitats, but ill-suited to the modified land-use types included in the study. These trends might differ in regions with greater connectivity, longer-term agriculture, and different geological history. Phylogenetic diversity SES and mean pairwise distance SES values showed different evidence of land-use effects compared with their non-SES counterparts, suggesting, for example, that Lepidoptera, mite and Rotifera communities are less dispersed, suggesting loss of lineages, in agricultural sites compared with forest habitats. In contrast, Malacostraca communities appear to be under-dispersed in natural forest sites, and to gain lineages due to anthropogenic land use. Phylogenetic rarity SES values further support the finding of consistently reduced rarity in agricultural sites, independent of species richness effects. Together, these observations indicate that phylogenetic information provides additional insights into soil invertebrate biodiversity patterns, as has been observed for other groups (González-Orozco et al., 2015; Mishler et al., 2014).

Sequence data, metadata, processed data, bioinformatic processing and analysis code used to generate the results in the manuscript (with one exception, detailed below) are deposited in the Manaaki Whenua-Landcare Research DataStore, accessible at: https://doi.org/10.7931/w3j3-5v40. Our sample sites include many Māori and/or privately-owned locations. We have have removed site location details from our metadata out of respect for concerns of Māori and other landowners. The removal of site location details precludes recreation of the map of sample site locations (Figure 6) and analyses of latitude effects, but otherwise has no impact on our results.

1. 1. Dopheide A
   2. Makiola A
   3. Orwin KH
   4. Holdaway R
   5. Wood JR
   6. Dickie I

   (2019) Manaaki Whenua - Landcare Research DataStore

   Land use impacts on soil invertebrate biodiversity.

   https://doi.org/10.7931/w3j3-5v40

1. 1. Allan E
   2. Bossdorf O
   3. Dormann CF
   4. Prati D
   5. Gossner MM
   6. Tscharntke T
   7. Blüthgen N
   8. Bellach M
   9. Birkhofer K
   10. Boch S
   11. Böhm S
   12. Börschig C
   13. Chatzinotas A
   14. Christ S
   15. Daniel R
   16. Diekötter T
   17. Fischer C
   18. Friedl T
   19. Glaser K
   20. Hallmann C
   21. Hodac L
   22. Hölzel N
   23. Jung K
   24. Klein AM
   25. Klaus VH
   26. Kleinebecker T
   27. Krauss J
   28. Lange M
   29. Morris EK
   30. Müller J
   31. Nacke H
   32. Pasalic E
   33. Rillig MC
   34. Rothenwöhrer C
   35. Schall P
   36. Scherber C
   37. Schulze W
   38. Socher SA
   39. Steckel J
   40. Steffan-Dewenter I
   41. Türke M
   42. Weiner CN
   43. Werner M
   44. Westphal C
   45. Wolters V
   46. Wubet T
   47. Gockel S
   48. Gorke M
   49. Hemp A
   50. Renner SC
   51. Schöning I
   52. Pfeiffer S
   53. König-Ries B
   54. Buscot F
   55. Linsenmair KE
   56. Schulze ED
   57. Weisser WW
   58. Fischer M

   (2014) Interannual variation in land-use intensity enhances grassland multidiversity

   PNAS 111:308–313.

   https://doi.org/10.1073/pnas.1312213111

   • PubMed
   • Google Scholar
2. 1. Andelman SJ
   2. Fagan WF

   (2000) Umbrellas and flagships: efficient conservation surrogates or expensive mistakes?

   PNAS 97:5954–5959.

   https://doi.org/10.1073/pnas.100126797

   • PubMed
   • Google Scholar
3. 1. Attwood SJ
   2. Maron M
   3. House APN
   4. Zammit C

   (2008) Do arthropod assemblages display globally consistent responses to intensified agricultural land use and management?

   Global Ecology and Biogeography 17:585–599.

   https://doi.org/10.1111/j.1466-8238.2008.00399.x

   • Google Scholar
4. 1. Avis PG
   2. Mueller GM
   3. Lussenhop J

   (2008) Ectomycorrhizal fungal communities in two north american oak forests respond to nitrogen addition

   New Phytologist 179:472–483.

   https://doi.org/10.1111/j.1469-8137.2008.02491.x

   • PubMed
   • Google Scholar
5. 1. Balvanera P
   2. Pfisterer AB
   3. Buchmann N
   4. He JS
   5. Nakashizuka T
   6. Raffaelli D
   7. Schmid B

   (2006) Quantifying the evidence for biodiversity effects on ecosystem functioning and services

   Ecology Letters 9:1146–1156.

   https://doi.org/10.1111/j.1461-0248.2006.00963.x

   • PubMed
   • Google Scholar
6. 1. Blaum N
   2. Seymour C
   3. Rossmanith E
   4. Schwager M
   5. Jeltsch F

   (2009) Changes in arthropod diversity along a land use driven gradient of shrub cover in savanna rangelands: identification of suitable indicators

   Biodiversity and Conservation 18:1187–1199.

   https://doi.org/10.1007/s10531-008-9498-x

   • Google Scholar
7. 1. Börschig C
   2. Klein A-M
   3. von Wehrden H
   4. Krauss J

   (2013) Traits of butterfly communities change from specialist to generalist characteristics with increasing land-use intensity

   Basic and Applied Ecology 14:547–554.

   https://doi.org/10.1016/j.baae.2013.09.002

   • Google Scholar
8. 1. Buckley TR
   2. Krosch M
   3. Leschen RAB

   (2015) Evolution of New Zealand insects: summary and prospectus for future research

   Austral Entomology 54:1–27.

   https://doi.org/10.1111/aen.12116

   • Google Scholar
9. 1. Callaham MA
   2. Richter DD
   3. Coleman DC
   4. Hofmockel M

   (2006) Long-term land-use effects on soil invertebrate communities in Southern Piedmont soils, USA

   European Journal of Soil Biology 42:S150–S156.

   https://doi.org/10.1016/j.ejsobi.2006.06.001

   • Google Scholar
10. 1. Cameron EK
    2. Martins IS
    3. Lavelle P
    4. Mathieu J
    5. Tedersoo L
    6. Gottschall F
    7. Guerra CA
    8. Hines J
    9. Patoine G
    10. Siebert J
    11. Winter M
    12. Cesarz S
    13. Delgado-Baquerizo M
    14. Ferlian O
    15. Fierer N
    16. Kreft H
    17. Lovejoy TE
    18. Montanarella L
    19. Orgiazzi A
    20. Pereira HM
    21. Phillips HRP
    22. Settele J
    23. Wall DH
    24. Eisenhauer N

    (2018) Global gaps in soil biodiversity data

    Nature Ecology & Evolution 2:1042–1043.

    https://doi.org/10.1038/s41559-018-0573-8

    • PubMed
    • Google Scholar
11. 1. Campos PF
    2. Gilbert TM

    (2012) DNA extraction from keratin and chitin

    Methods in Molecular Biology 840:43–49.

    https://doi.org/10.1007/978-1-61779-516-9_6

    • PubMed
    • Google Scholar
12. Software

    1. Charney N

    (2012) vegetarian: jost diversity measures for community data

    CRAN.

    https://cran.r-project.org/package=vegetarian
13. 1. Clavel J
    2. Julliard R
    3. Devictor V

    (2011) Worldwide decline of specialist species: toward a global functional homogenization?

    Frontiers in Ecology and the Environment 9:222–228.

    https://doi.org/10.1890/080216

    • Google Scholar
14. 1. Dee LE
    2. Cowles J
    3. Isbell F
    4. Pau S
    5. Gaines SD
    6. Reich PB

    (2019) When do ecosystem services depend on rare species?

    Trends in Ecology & Evolution 34:746–758.

    https://doi.org/10.1016/j.tree.2019.03.010

    • PubMed
    • Google Scholar
15. 1. Dickie IA
    2. Richardson SJ
    3. Wiser SK

    (2009) Ectomycorrhizal fungal communities and soil chemistry in harvested and unharvested temperate Nothofagus rainforests

    Canadian Journal of Forest Research 39:1069–1079.

    https://doi.org/10.1139/X09-036

    • Google Scholar
16. 1. Dickie IAN
    2. Reich PB

    (2005) Ectomycorrhizal fungal communities at forest edges

    Journal of Ecology 93:244–255.

    https://doi.org/10.1111/j.1365-2745.2005.00977.x

    • Google Scholar
17. 1. Dirzo R
    2. Young HS
    3. Galetti M
    4. Ceballos G
    5. Isaac NJ
    6. Collen B

    (2014) Defaunation in the anthropocene

    Science 345:401–406.

    https://doi.org/10.1126/science.1251817

    • PubMed
    • Google Scholar
18. Conference

    1. Drummond AJ
    2. Newcomb RD
    3. Buckley TR
    4. Xie D
    5. Dopheide A
    6. Potter BCM
    7. Heled J
    8. Ross HA
    9. Tooman L
    10. Grosser S
    11. Park D
    12. Demetras NJ
    13. Stevens MI
    14. Russell JC
    15. Anderson SH
    16. Carter A
    17. Nelson N

    (2015) Evaluating a multigene environmental DNA approach for biodiversity assessment

    GigaScience. pp. 4–46.

    https://doi.org/10.1186/s13742-015-0086-1

    • Google Scholar
19. 1. Edgar RC

    (2013) UPARSE: highly accurate OTU sequences from microbial amplicon reads

    Nature Methods 10:996–998.

    https://doi.org/10.1038/nmeth.2604

    • PubMed
    • Google Scholar
20. 1. Eisenhauer N
    2. Bonn A
    3. A. Guerra C

    (2019) Recognizing the quiet extinction of invertebrates

    Nature Communications 10:50.

    https://doi.org/10.1038/s41467-018-07916-1

    • Google Scholar
21. 1. Faith DP

    (1992) Conservation evaluation and phylogenetic diversity

    Biological Conservation 61:1–10.

    https://doi.org/10.1016/0006-3207(92)91201-3

    • Google Scholar
22. 1. Faith DP

    (2015) Phylogenetic diversity, functional trait diversity and extinction: avoiding tipping points and worst-case losses

    Philosophical Transactions of the Royal Society B: Biological Sciences 370:20140011.

    https://doi.org/10.1098/rstb.2014.0011

    • PubMed
    • Google Scholar
23. 1. Fleishman E
    2. Noss R
    3. Noon B

    (2006) Utility and limitations of species richness metrics for conservation planning

    Ecological Indicators 6:543–553.

    https://doi.org/10.1016/j.ecolind.2005.07.005

    • Google Scholar
24. 1. Folmer O
    2. Black M
    3. Hoeh W
    4. Lutz R
    5. Vrijenhoek R

    (1994)

    DNA primers for amplification of mitochondrial cytochrome c oxidase subunit I from diverse metazoan invertebrates

    Molecular Marine Biology and Biotechnology 3:294–299.

    • PubMed
    • Google Scholar
25. 1. Forest F
    2. Grenyer R
    3. Rouget M
    4. Davies TJ
    5. Cowling RM
    6. Faith DP
    7. Balmford A
    8. Manning JC
    9. Procheş S
    10. van der Bank M
    11. Reeves G
    12. Hedderson TA
    13. Savolainen V

    (2007) Preserving the evolutionary potential of floras in biodiversity hotspots

    Nature 445:757–760.

    https://doi.org/10.1038/nature05587

    • PubMed
    • Google Scholar
26. 1. Forister ML
    2. McCall AC
    3. Sanders NJ
    4. Fordyce JA
    5. Thorne JH
    6. O'Brien J
    7. Waetjen DP
    8. Shapiro AM

    (2010) Compounded effects of climate change and habitat alteration shift patterns of butterfly diversity

    PNAS 107:2088–2092.

    https://doi.org/10.1073/pnas.0909686107

    • PubMed
    • Google Scholar
27. 1. Gámez-Virués S
    2. Perović DJ
    3. Gossner MM
    4. Börschig C
    5. Blüthgen N
    6. de Jong H
    7. Simons NK
    8. Klein A-M
    9. Krauss J
    10. Maier G
    11. Scherber C
    12. Steckel J
    13. Rothenwöhrer C
    14. Steffan-Dewenter I
    15. Weiner CN
    16. Weisser W
    17. Werner M
    18. Tscharntke T
    19. Westphal C

    (2015) Landscape simplification filters species traits and drives biotic homogenization

    Nature Communications 6:8568.

    https://doi.org/10.1038/ncomms9568

    • Google Scholar
28. 1. George PBL
    2. Lallias D
    3. Creer S
    4. Seaton FM
    5. Kenny JG
    6. Eccles RM
    7. Griffiths RI
    8. Lebron I
    9. Emmett BA
    10. Robinson DA
    11. Jones DL

    (2019) Divergent national-scale trends of microbial and animal biodiversity revealed across diverse temperate soil ecosystems

    Nature Communications 10:1107.

    https://doi.org/10.1038/s41467-019-09031-1

    • PubMed
    • Google Scholar
29. 1. Gerlach J
    2. Samways M
    3. Pryke J

    (2013) Terrestrial invertebrates as bioindicators: an overview of available taxonomic groups

    Journal of Insect Conservation 17:831–850.

    https://doi.org/10.1007/s10841-013-9565-9

    • Google Scholar
30. 1. González-Orozco CE
    2. Mishler BD
    3. Miller JT
    4. Laffan SW
    5. Knerr N
    6. Unmack P
    7. Georges A
    8. Thornhill AH
    9. Rosauer DF
    10. Gruber B

    (2015) Assessing biodiversity and endemism using phylogenetic methods across multiple taxonomic groups

    Ecology and Evolution 5:5177–5192.

    https://doi.org/10.1002/ece3.1747

    • PubMed
    • Google Scholar
31. 1. Gossner MM
    2. Lewinsohn TM
    3. Kahl T
    4. Grassein F
    5. Boch S
    6. Prati D
    7. Birkhofer K
    8. Renner SC
    9. Sikorski J
    10. Wubet T
    11. Arndt H
    12. Baumgartner V
    13. Blaser S
    14. Blüthgen N
    15. Börschig C
    16. Buscot F
    17. Diekötter T
    18. Jorge LR
    19. Jung K
    20. Keyel AC
    21. Klein A-M
    22. Klemmer S
    23. Krauss J
    24. Lange M
    25. Müller J
    26. Overmann J
    27. Pašalić E
    28. Penone C
    29. Perović DJ
    30. Purschke O
    31. Schall P
    32. Socher SA
    33. Sonnemann I
    34. Tschapka M
    35. Tscharntke T
    36. Türke M
    37. Venter PC
    38. Weiner CN
    39. Werner M
    40. Wolters V
    41. Wurst S
    42. Westphal C
    43. Fischer M
    44. Weisser WW
    45. Allan E

    (2016) Land-use intensification causes multitrophic homogenization of grassland communities

    Nature 540:266–269.

    https://doi.org/10.1038/nature20575

    • Google Scholar
32. 1. Hallmann CA
    2. Sorg M
    3. Jongejans E
    4. Siepel H
    5. Hofland N
    6. Schwan H
    7. Stenmans W
    8. Müller A
    9. Sumser H
    10. Hörren T
    11. Goulson D
    12. de Kroon H

    (2017) More than 75 percent decline over 27 years in total flying insect biomass in protected Areas

    PLOS ONE 12:e0185809.

    https://doi.org/10.1371/journal.pone.0185809

    • PubMed
    • Google Scholar
33. 1. Hendrickx F
    2. Maelfait J-P
    3. Van Wingerden W
    4. Schweiger O
    5. Speelmans M
    6. Aviron S
    7. Augenstein I
    8. Billeter R
    9. Bailey D
    10. Bukacek R
    11. Burel F
    12. Diekötter T
    13. Dirksen J
    14. Herzog F
    15. Liira J
    16. Roubalova M
    17. Vandomme V
    18. Bugter R

    (2007) How landscape structure, land-use intensity and habitat diversity affect components of total arthropod diversity in agricultural landscapes

    Journal of Applied Ecology 44:340–351.

    https://doi.org/10.1111/j.1365-2664.2006.01270.x

    • Google Scholar
34. 1. Hillebrand H
    2. Blasius B
    3. Borer ET
    4. Chase JM
    5. Downing JA
    6. Eriksson BK
    7. Filstrup CT
    8. Harpole WS
    9. Hodapp D
    10. Larsen S
    11. Lewandowska AM
    12. Seabloom EW
    13. Van de Waal DB
    14. Ryabov AB

    (2018) Biodiversity change is uncoupled from species richness trends: consequences for conservation and monitoring

    Journal of Applied Ecology 55:169–184.

    https://doi.org/10.1111/1365-2664.12959

    • Google Scholar
35. 1. Hobbs RJ
    2. Arico S
    3. Aronson J
    4. Baron JS
    5. Bridgewater P
    6. Cramer VA
    7. Epstein PR
    8. Ewel JJ
    9. Klink CA
    10. Lugo AE
    11. Norton D
    12. Ojima D
    13. Richardson DM
    14. Sanderson EW
    15. Valladares F
    16. Vila M
    17. Zamora R
    18. Zobel M

    (2006) Novel ecosystems: theoretical and management aspects of the new ecological world order

    Global Ecology and Biogeography 15:1–7.

    https://doi.org/10.1111/j.1466-822X.2006.00212.x

    • Google Scholar
36. Book

    1. Hurst JM
    2. Allen RB

    (2007)

    A Permanent Plot Method for Monitoring Indigenous Forests: Field Protocols

    Manaaki Whenua-Landcare Research.

    • Google Scholar
37. 1. Jonsson M
    2. Yeates GW
    3. Wardle DA

    (2009) Patterns of invertebrate density and taxonomic richness across gradients of area, isolation, and vegetation diversity in a lake-island system

    Ecography 32:963–972.

    https://doi.org/10.1111/j.1600-0587.2009.05784.x

    • Google Scholar
38. 1. Katoh K
    2. Standley DM

    (2013) MAFFT multiple sequence alignment software version 7: improvements in performance and usability

    Molecular Biology and Evolution 30:772–780.

    https://doi.org/10.1093/molbev/mst010

    • PubMed
    • Google Scholar
39. 1. Kembel SW
    2. Cowan PD
    3. Helmus MR
    4. Cornwell WK
    5. Morlon H
    6. Ackerly DD
    7. Blomberg SP
    8. Webb CO

    (2010) Picante: r tools for integrating phylogenies and ecology

    Bioinformatics 26:1463–1464.

    https://doi.org/10.1093/bioinformatics/btq166

    • PubMed
    • Google Scholar
40. 1. Kier G
    2. Kreft H
    3. Lee TM
    4. Jetz W
    5. Ibisch PL
    6. Nowicki C
    7. Mutke J
    8. Barthlott W

    (2009) A global assessment of endemism and species richness across island and mainland regions

    PNAS 106:9322–9327.

    https://doi.org/10.1073/pnas.0810306106

    • PubMed
    • Google Scholar
41. 1. Kier G
    2. Barthlott W

    (2001) Measuring and mapping endemism and species richness: a new methodological approach and its application on the flora of africa

    Biodiversity and Conservation 10:1513–1529.

    https://doi.org/10.1023/A:1011812528849

    • Google Scholar
42. 1. Lavelle P
    2. Decaëns T
    3. Aubert M
    4. Barot S
    5. Blouin M
    6. Bureau F
    7. Margerie P
    8. Mora P
    9. Rossi J-P

    (2006) Soil invertebrates and ecosystem services

    European Journal of Soil Biology 42:S3–S15.

    https://doi.org/10.1016/j.ejsobi.2006.10.002

    • Google Scholar
43. 1. Leandro C
    2. Jay-Robert P
    3. Vergnes A

    (2017) Bias and perspectives in insect conservation: a european scale analysis

    Biological Conservation 215:213–224.

    https://doi.org/10.1016/j.biocon.2017.07.033

    • Google Scholar
44. 1. Leitão RP
    2. Zuanon J
    3. Villéger S
    4. Williams SE
    5. Baraloto C
    6. Fortunel C
    7. Mendonça FP
    8. Mouillot D

    (2016) Rare species contribute disproportionately to the functional structure of species assemblages

    Proceedings of the Royal Society B: Biological Sciences 283:20160084.

    https://doi.org/10.1098/rspb.2016.0084

    • Google Scholar
45. 1. Leray M
    2. Yang JY
    3. Meyer CP
    4. Mills SC
    5. Agudelo N
    6. Ranwez V
    7. Boehm JT
    8. Machida RJ

    (2013) A new versatile primer set targeting a short fragment of the mitochondrial COI region for metabarcoding metazoan diversity: application for characterizing coral reef fish gut contents

    Frontiers in Zoology 10:34.

    https://doi.org/10.1186/1742-9994-10-34

    • PubMed
    • Google Scholar
46. 1. Lozupone C
    2. Knight R

    (2005) UniFrac: a new phylogenetic method for comparing microbial communities

    Applied and Environmental Microbiology 71:8228–8235.

    https://doi.org/10.1128/AEM.71.12.8228-8235.2005

    • PubMed
    • Google Scholar
47. 1. Lyons KG
    2. Brigham CA
    3. Traut BH
    4. Schwartz MW

    (2005) Rare species and ecosystem functioning

    Conservation Biology 19:1019–1024.

    https://doi.org/10.1111/j.1523-1739.2005.00106.x

    • Google Scholar
48. 1. Makiola A
    2. Dickie IA
    3. Holdaway RJ
    4. Wood JR
    5. Orwin KH
    6. Glare TR

    (2019) Land use is a determinant of plant pathogen alpha- but not beta-diversity

    Molecular Ecology 28:3786–3798.

    https://doi.org/10.1111/mec.15177

    • PubMed
    • Google Scholar
49. Report

    1. Manktelow D
    2. Stevens P
    3. Walker J
    4. Gurnsey S
    5. Park N
    6. Zabkiewicz J
    7. Teulon D
    8. Rahman A

    (2005) Trends in Pesticide Use in New Zealand: 2004

    Report to the Ministry for the Environment HortResearch Client Report No 17962.

    https://dioxinnz.com/Spray-NZ-Hist/PDF/nz-pesticide-trends.pdf

    • Google Scholar
50. 1. Martin M

    (2011) Cutadapt removes adapter sequences from high-throughput sequencing reads

    EMBnet.journal 17:10–12.

    https://doi.org/10.14806/ej.17.1.200

    • Google Scholar
51. 1. Mazel F
    2. Pennell MW
    3. Cadotte MW
    4. Diaz S
    5. Dalla Riva GV
    6. Grenyer R
    7. Leprieur F
    8. Mooers AO
    9. Mouillot D
    10. Tucker CM
    11. Pearse WD

    (2018) Prioritizing phylogenetic diversity captures functional diversity unreliably

    Nature Communications 9:2888.

    https://doi.org/10.1038/s41467-018-05126-3

    • Google Scholar
52. 1. McIntyre S
    2. Lavorel S

    (1994) Predicting richness of native, rare, and exotic plants in response to habitat and disturbance variables across a variegated landscape

    Conservation Biology 8:521–531.

    https://doi.org/10.1046/j.1523-1739.1994.08020521.x

    • Google Scholar
53. 1. McKinney ML
    2. Lockwood JL

    (1999) Biotic homogenization: a few winners replacing many losers in the next mass extinction

    Trends in Ecology & Evolution 14:450–453.

    https://doi.org/10.1016/S0169-5347(99)01679-1

    • PubMed
    • Google Scholar
54. 1. McMurdie PJ
    2. Holmes S

    (2013) phyloseq: An R Package for Reproducible Interactive Analysis and Graphics of Microbiome Census Data

    PLOS ONE 8:e61217.

    https://doi.org/10.1371/journal.pone.0061217

    • Google Scholar
55. 1. Miller ET
    2. Farine DR
    3. Trisos CH

    (2017) Phylogenetic community structure metrics and null models: a review with new methods and software

    Ecography 40:461–477.

    https://doi.org/10.1111/ecog.02070

    • Google Scholar
56. 1. Mishler BD
    2. Knerr N
    3. González-Orozco CE
    4. Thornhill AH
    5. Laffan SW
    6. Miller JT

    (2014) Phylogenetic measures of biodiversity and neo- and paleo-endemism in australian Acacia

    Nature Communications 5:4473.

    https://doi.org/10.1038/ncomms5473

    • PubMed
    • Google Scholar
57. 1. Mouillot D
    2. Bellwood DR
    3. Baraloto C
    4. Chave J
    5. Galzin R
    6. Harmelin-Vivien M
    7. Kulbicki M
    8. Lavergne S
    9. Lavorel S
    10. Mouquet N
    11. Paine CE
    12. Renaud J
    13. Thuiller W

    (2013) Rare species support vulnerable functions in high-diversity ecosystems

    PLOS Biology 11:e1001569.

    https://doi.org/10.1371/journal.pbio.1001569

    • PubMed
    • Google Scholar
58. 1. Newbold T
    2. Hudson LN
    3. Hill SL
    4. Contu S
    5. Lysenko I
    6. Senior RA
    7. Börger L
    8. Bennett DJ
    9. Choimes A
    10. Collen B
    11. Day J
    12. De Palma A
    13. Díaz S
    14. Echeverria-Londoño S
    15. Edgar MJ
    16. Feldman A
    17. Garon M
    18. Harrison ML
    19. Alhusseini T
    20. Ingram DJ
    21. Itescu Y
    22. Kattge J
    23. Kemp V
    24. Kirkpatrick L
    25. Kleyer M
    26. Correia DL
    27. Martin CD
    28. Meiri S
    29. Novosolov M
    30. Pan Y
    31. Phillips HR
    32. Purves DW
    33. Robinson A
    34. Simpson J
    35. Tuck SL
    36. Weiher E
    37. White HJ
    38. Ewers RM
    39. Mace GM
    40. Scharlemann JP
    41. Purvis A

    (2015) Global effects of land use on local terrestrial biodiversity

    Nature 520:45–50.

    https://doi.org/10.1038/nature14324

    • PubMed
    • Google Scholar
59. Software

    1. Niperess D

    (2010) Phylo.endemism: an R function for calculating phylogenetic endemism of ecological samples

    GitHub.

    https://rdrr.io/github/davidnipperess/PDcalc/man/phyloendemism.html
60. 1. Oberprieler SK
    2. Andersen AN
    3. Gillespie GR
    4. Einoder LD

    (2019) Vertebrates are poor umbrellas for invertebrates: cross‐taxon congruence in an australian tropical savanna

    Ecosphere 10:e02755.

    https://doi.org/10.1002/ecs2.2755

    • Google Scholar
61. Software

    1. Oksanen J
    2. Blanchet F
    3. Guillaume Friendly M
    4. Kindt R
    5. Legendre P
    6. McGlinn D
    7. Minchin PR
    8. O’Hara RB
    9. Simpson GL
    10. Solymos P
    11. Stevens MH
    12. Szoecs H
    13. Wagner H

    (2017) vegan: community ecology package

    CRAN.

    http://cran.r-project.org/package=vegan
62. 1. Orwin KH
    2. Dickie IA
    3. Wood JR
    4. Bonner KI
    5. Holdaway RJ

    (2016) Soil microbial community structure explains the resistance of respiration to a dry–rewet cycle, but not soil functioning under static conditions

    Functional Ecology 30:1430–1439.

    https://doi.org/10.1111/1365-2435.12610

    • Google Scholar
63. Book

    1. Pawson SM
    2. Brockerhoff EG
    3. Meenken ED
    4. Didham RK

    (2009) Non-native plantation forests as alternative habitat for native forest beetles in a heavily modified landscape

    In: Brockerhoff E. G, Jactel H, Parrotta J. A, Quine C. P, Sayer J, Hawksworth D. L, editors. Plantation Forests and Biodiversity: Oxymoron or Opportunity?. Springer. pp. 203–224.

    https://doi.org/10.1007/s10531-008-9380-x

    • Google Scholar
64. 1. Pawson SM
    2. Ecroyd CE
    3. Seaton R
    4. Shaw WB
    5. Brockerhoff EG

    (2010)

    New Zealand’s exotic plantation forests as habitats for threatened indigenous species

    New Zealand Journal of Ecology 34:342–355.

    • Google Scholar
65. 1. Porter TM
    2. Hajibabaei M

    (2018) Automated high throughput animal CO1 metabarcode classification

    Scientific Reports 8:4226.

    https://doi.org/10.1038/s41598-018-22505-4

    • PubMed
    • Google Scholar
66. 1. Potts SG
    2. Biesmeijer JC
    3. Kremen C
    4. Neumann P
    5. Schweiger O
    6. Kunin WE

    (2010) Global pollinator declines: trends, impacts and drivers

    Trends in Ecology & Evolution 25:345–353.

    https://doi.org/10.1016/j.tree.2010.01.007

    • PubMed
    • Google Scholar
67. 1. Powney GD
    2. Carvell C
    3. Edwards M
    4. Morris RKA
    5. Roy HE
    6. Woodcock BA
    7. Isaac NJB

    (2019) Widespread losses of pollinating insects in Britain

    Nature Communications 10:1018.

    https://doi.org/10.1038/s41467-019-08974-9

    • PubMed
    • Google Scholar
68. 1. Price MN
    2. Dehal PS
    3. Arkin AP

    (2010) FastTree 2--approximately maximum-likelihood trees for large alignments

    PLOS ONE 5:e9490.

    https://doi.org/10.1371/journal.pone.0009490

    • PubMed
    • Google Scholar
69. Software

    1. R Development Core Team

    (2016) R: a language and environment for statistical computing

    R Foundation for Statistical Computing, Vienna, Austria.

    https://www.r-project.org/
70. 1. Rognes T
    2. Flouri T
    3. Nichols B
    4. Quince C
    5. Mahé F

    (2016) VSEARCH: a versatile open source tool for metagenomics

    PeerJ 4:e2584.

    https://doi.org/10.7717/peerj.2584

    • PubMed
    • Google Scholar
71. 1. Rosauer D
    2. Laffan SW
    3. Crisp MD
    4. Donnellan SC
    5. Cook LG

    (2009) Phylogenetic endemism: a new approach for identifying geographical concentrations of evolutionary history

    Molecular Ecology 18:4061–4072.

    https://doi.org/10.1111/j.1365-294X.2009.04311.x

    • PubMed
    • Google Scholar
72. Software

    1. RStudio team

    (2015) RStudio: Integrated Development Environment for R

     RStudio, Inc.

    http://www.rstudio.org
73. 1. Schnell IB
    2. Bohmann K
    3. Gilbert MT

    (2015) Tag jumps illuminated--reducing sequence-to-sample misidentifications in metabarcoding studies

    Molecular Ecology Resources 15:1289–1303.

    https://doi.org/10.1111/1755-0998.12402

    • PubMed
    • Google Scholar
74. 1. Schuldt A
    2. Assmann T

    (2010) Invertebrate diversity and national responsibility for species conservation across Europe – A multi-taxon approach

    Biological Conservation 143:2747–2756.

    https://doi.org/10.1016/j.biocon.2010.07.022

    • Google Scholar
75. 1. Smart SM
    2. Thompson K
    3. Marrs RH
    4. Le Duc MG
    5. Maskell LC
    6. Firbank LG

    (2006) Biotic homogenization and changes in species diversity across human-modified ecosystems

    Proceedings of the Royal Society B: Biological Sciences 273:2659–2665.

    https://doi.org/10.1098/rspb.2006.3630

    • Google Scholar
76. 1. Srivastava DS
    2. Cadotte MW
    3. MacDonald AA
    4. Marushia RG
    5. Mirotchnick N

    (2012) Phylogenetic diversity and the functioning of ecosystems

    Ecology Letters 15:637–648.

    https://doi.org/10.1111/j.1461-0248.2012.01795.x

    • PubMed
    • Google Scholar
77. 1. Stamatakis A

    (2014) RAxML version 8: a tool for phylogenetic analysis and post-analysis of large phylogenies

    Bioinformatics 30:1312–1313.

    https://doi.org/10.1093/bioinformatics/btu033

    • PubMed
    • Google Scholar
78. 1. Stein A
    2. Gerstner K
    3. Kreft H

    (2014) Environmental heterogeneity as a universal driver of species richness across taxa, biomes and spatial scales

    Ecology Letters 17:866–880.

    https://doi.org/10.1111/ele.12277

    • PubMed
    • Google Scholar
79. 1. Stringer IAN
    2. Hitchmough RA

    (2012) Assessing the conservation status of New Zealand's native terrestrial invertebrates

    New Zealand Entomologist 3535:77–84.

    https://doi.org/10.1080/00779962.2012.686309

    • Google Scholar
80. 1. Trewick SA
    2. Wallis GP
    3. Morgan-Richards M

    (2011) The invertebrate life of New Zealand: a phylogeographic approach

    Insects 2:297–325.

    https://doi.org/10.3390/insects2030297

    • PubMed
    • Google Scholar
81. 1. Tscharntke T
    2. Klein AM
    3. Kruess A
    4. Steffan-Dewenter I
    5. Thies C

    (2005) Landscape perspectives on agricultural intensification and biodiversity - ecosystem service management

    Ecology Letters 8:857–874.

    https://doi.org/10.1111/j.1461-0248.2005.00782.x

    • Google Scholar
82. 1. Tsiafouli MA
    2. Thébault E
    3. Sgardelis SP
    4. de Ruiter PC
    5. van der Putten WH
    6. Birkhofer K
    7. Hemerik L
    8. de Vries FT
    9. Bardgett RD
    10. Brady MV
    11. Bjornlund L
    12. Jørgensen HB
    13. Christensen S
    14. Hertefeldt TD
    15. Hotes S
    16. Gera Hol WH
    17. Frouz J
    18. Liiri M
    19. Mortimer SR
    20. Setälä H
    21. Tzanopoulos J
    22. Uteseny K
    23. Pižl V
    24. Stary J
    25. Wolters V
    26. Hedlund K

    (2015) Intensive agriculture reduces soil biodiversity across Europe

    Global Change Biology 21:973–985.

    https://doi.org/10.1111/gcb.12752

    • PubMed
    • Google Scholar
83. 1. Wagg C
    2. Bender SF
    3. Widmer F
    4. van der Heijden MG

    (2014) Soil biodiversity and soil community composition determine ecosystem multifunctionality

    PNAS 111:5266–5270.

    https://doi.org/10.1073/pnas.1320054111

    • PubMed
    • Google Scholar
84. 1. Wang Q
    2. Garrity GM
    3. Tiedje JM
    4. Cole JR

    (2007) Naive bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy

    Applied and Environmental Microbiology 73:5261–5267.

    https://doi.org/10.1128/AEM.00062-07

    • PubMed
    • Google Scholar
85. 1. Webb CO
    2. Ackerly DD
    3. McPeek MA
    4. Donoghue MJ

    (2002) Phylogenies and community ecology

    Annual Review of Ecology and Systematics 33:475–505.

    https://doi.org/10.1146/annurev.ecolsys.33.010802.150448

    • Google Scholar
86. 1. Winter M
    2. Devictor V
    3. Schweiger O

    (2013) Phylogenetic diversity and nature conservation: where are we?

    Trends in Ecology & Evolution 28:199–204.

    https://doi.org/10.1016/j.tree.2012.10.015

    • Google Scholar
87. 1. Wood JR
    2. Holdaway RJ
    3. Orwin KH
    4. Morse C
    5. Bonner KI
    6. Davis C
    7. Bolstridge N
    8. Dickie IA

    (2017) No single driver of biodiversity: divergent responses of multiple taxa across land use types

    Ecosphere 8:e01997.

    https://doi.org/10.1002/ecs2.1997

    • Google Scholar
88. 1. Yachi S
    2. Loreau M

    (1999) Biodiversity and ecosystem productivity in a fluctuating environment: the insurance hypothesis

    PNAS 96:1463–1468.

    https://doi.org/10.1073/pnas.96.4.1463

    • PubMed
    • Google Scholar
89. 1. Yang LH
    2. Gratton C

    (2014) Insects as drivers of ecosystem processes

    Current Opinion in Insect Science 2:26–32.

    https://doi.org/10.1016/j.cois.2014.06.004

    • Google Scholar
90. 1. Yu DW
    2. Ji Y
    3. Emerson BC
    4. Wang X
    5. Ye C
    6. Yang C
    7. Ding Z

    (2012) Biodiversity soup: metabarcoding of arthropods for rapid biodiversity assessment and biomonitoring

    Methods in Ecology and Evolution 3:613–623.

    https://doi.org/10.1111/j.2041-210X.2012.00198.x

    • Google Scholar

Acceptance summary:

Five categories of land-use that could be ranked according to increasing human impact showed strong linear declines in the majority of 17 invertebrate soil animal groups. These negative impacts were best detected when phylogenetic rarity was used as a measure whereas other diversity measures were less responsive and may therefore be less useful indicators. The comprehensive assessment of the soil invertebrate communities was made possible by the use of DNA metabarcoding.

Decision letter after peer review:

Thank you for submitting your article "Endemism is a better indicator of soil invertebrate biodiversity loss with land use change than richness or diversity" for consideration by eLife. Your article has been reviewed by three peer reviewers, including Bernhard Schmid as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Ian Baldwin as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Marc Cadotte (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

This manuscript represents a major effort to use DNA barcoding in analyzing soil communities at 75 sites across New Zealand. More than 11'000 taxa could be identified and assigned to different groups of invertebrates and microbes. The hypothesis is that diversity decreases along a gradient of land-use intensity represented by five land-use categories natural forest -> planted forest -> low-producing grassland -> high-producing grassland -> perennial cropland. This hypothesis is confirmed, especially if species are weighted by the reciprocal of the number of sites at which they occur, suggesting that they are habitat specialists and therefore respond most strongly to land use.

Essential revisions:

The manuscript should be improved in three respects:

First, there are several terms that are not well defined in the manuscript. "Land use" is sometimes considered as a state of an ecosystem and then called land-use category (please always use the hyphen when "land-use" is an adjectival noun) but sometimes considered as land-use change or intensification etc. Please use consistent terms throughout and make sure you define them well. In this context and related to the second main point I suggest that you put the five land-use categories explicitly into a linear sequence from one to five and call it land-use intensity.

Endemism is defined in the Materials and methods, but all reviewers would prefer a more generic term such as rare species because you cannot extrapolate from 75 sites to the entire area of your country, which would be necessary to know if a species that only occurred at 1 or 2 sites was geographically restricted. You should also mention the caveat that some of your rare species could just be "sink" or "passenger" species not adapted to the site conditions, while your hypothesis assumes that they are habitat specialists. One reviewer suggests in this context that you could actually do a sensitivity analysis without the rarest species that only occur at one site.

Second, you should get much more from your data if you expand your analyses. In principle, you selected your five categories such that they can be ordered along a gradient. Best would be if you could calculate some value for land-use intensity for each category or even each site, but even in the absence of this you are allowed to simply assign land-use intensity values from 1-5 and use this linear contrast in all ANOVA (the easiest way is to code land-use categories from 1-5 and use it as continuous variable LUI and factor LUC: then fit LUI+LUC in this order and you get a stronger test of your main hypothesis, namely the linear contrast LUI, and a test for deviations from the linear contrast, which is measured by LUC when fitted after LUI). Please replace all box-plots with means +/- standard errors. In addition, you could show regression lines for the LUI-gradient wherever it is significant. There would be the more sophisticated method of isotonic regression (Gaines, Steven D., and William R. Rice. 1990. Analysis of Biological Data When there are Ordered Expectations. The American Naturalist 135: 310-317.), but clearly the simpler method indicated above is also valid.

The above will remove the need for the excessive use of pairwise comparisons which are statistically not independent and ill-suited to test ordered hypotheses. A further reduction of unnecessary repeats and more comprehensive analysis is to compare the responses of the different taxa to LUI(+LUC) in a single ANOVA by putting the metrics for the different taxa into a single column and add a column with the taxa identities (or even two, one for phyla and one for classes/orders). You can then fit the following ANOVA-model:

Metric ~ LUI+LUC+phylum+order+LUI:phylum+LUC:phylum+LUI:order+LUC:order, or more simply without LUC terms and only using phylum or order. The interactions indicated by ":" test if the different taxonomic groups respond differently. Using the above you can exchange Figure 3—source data 1B and omit Supplementary Table 4. There is still the problem of comparing the R² values (or better just use %SS, which are increments in multiple R²) mentioned by one reviewer and we ask you to consider their suggestions. It also seems you analyzed R² values as (meta-)data; this requires explanation and justification in the Materials and methods, because R² values will have a special distribution that deviates from normal.

Furthermore, as noted by one reviewer, from your Figure 6 it appears that the land-use categories are not well "randomized" across space and thus covariates could explain some of your results. You could include some of these first in the fitting sequence of ANOVAs, but we accept that you don't want to put in too many covariates because there are not so many sites to test more complicated models and sometimes correction for covariates can be unjustified if they are not a "true cause" of an observed effect.

One reviewer also suggests that you should try phylogenetic metrics that are not mathematically related to richness.

Your suggestion that the results imply homogenization needs to be tested statistically, which in fact can be done quite easily e.g. by calculating β diversities. You also use a term "heterogeneity", which obviously is taken from output of an R function. However, you cannot expect readers to check this up but rather need to give a clear definition of heterogeneity and how it is calculated.

Third, the Introduction and Discussion sections contain many repetitive statements and statements that lead the reader away from the main story line. The Introduction should focus on the effects of land-use intensity on soil organisms that here have been studied with DNA barcoding, offering a new level of resolution. Your focus and hypothesis on differences in responses between common and rare taxa is really secondary and less "a priori" than the main hypothesis that land-use intensity reduces soil biodiversity. The Discussion contains many speculations, in part related to this secondary hypothesis, that weaken the stronger results regarding the main hypothesis. Generally, we find it very difficult to say rare taxa can "better" indicate biodiversity responses to land-use intensity, because obviously you don't know what the "true biodiversity" is, which you implicitly use as reference.

Reviewer #1:

See above summary.

Reviewer #2:

In this paper the authors use meta barcoding to assess the diversity of soil invertebrates and relate diversity to land use. They find that more intensively managed habitats contain a lower diversity of soil fauna and, in particular, a lower diversity of rare (narrowly distributed) species. The diversity declines are consistent across groups: all groups decline or don't respond to land use, none increase. I think the paper is valuable in looking at soil fauna, which are poorly studied, and the results are interesting in showing that land use change has large effects on these taxa. It is also interesting that it is the rare species that decline with land use, as has been found in other organism groups. However, I have some reservations about the analysis and the framing of the study in terms of endemism.

1) The study aims to show the effect of land use change (conversion of native forests into grassland or perennial cropping) on soil fauna diversity. However, there is no attempt to correct for confounding factors. How were the sites chosen? Was there an attempt to ensure no confounding between land use and environmental factors such as altitude, soil type etc.? All that is said in the Materials and methods is that the sites were distributed across New Zealand. The Discussion mentions some possible confounding, as it is said that native forests are in more "rugged and less accessible areas". In addition, from Figure 6 it appears that many of the native forests are on the west coast of the South Island and there may therefore be climatic differences from the other sites? Currently there is no attempt to correct for confounding factors in the analysis: land use type is the only factor included in the models. I think it would be worth trying to fit some covariates in these models: climate variables and altitude would make sense and some soil variables such as soil type or pH would be good to consider too. This would make the analysis much more robust in showing that land use is the driver of soil fauna diversity, after correcting for soil and climate. It would also add interesting information on the other drivers of soil fauna diversity.

2) I found the term "endemism" confusing in the context of this study. There are no actual data on whether any of these taxa are endemic to a given area, as these data do not exist for soil fauna. Instead endemism is defined as the number of sites occupied by an OTU. Taxa occurring on few sites are therefore considered "endemic" but there is no information on whether they really are endemic or whether they occur elsewhere, e.g. outside New Zealand. Also, if a taxa is found in only two sites it would be considered endemic here but if those two sites were far apart, e.g. it was found in a northern and southern sample, it might not have a restricted distribution, it might simply be rare throughout the range. I therefore think that it would be clearer to talk about rare species and mention that here rarity is based on site occupancy. The situation may be even more problematic for phylogenetic endemism as in the original paper Rosauer et al., 2009, state that phylogenetic endemism values will be biased if closely related species occurring in other areas are missing from the sample (which is highly likely to be the case in this study).

2b) Related to this, I think it is important to mention in the Discussion that rarity is defined here relative to the sites sampled. Some of the species considered rare in this analysis might not be rare at all if they are common in habitats or areas not sampled in this study. There is nothing the authors can do about it this and I think the approach they have used is completely reasonable, but this limitation ought to be acknowledged in the Discussion.

3) How much is the measure of "endemic richness" affected by OTUs occurring in a single site? I could imagine that the distribution of these singleton taxa is quite stochastic and I wonder if it would be worth recalculating the index without them to check that they are not driving the whole pattern.

4) I am not really convinced that the analysis of R² values (Figure 4) is valid. The R² values for the different groups are not independent of each other as diversities of different taxa are likely to be correlated due to shared environmental responses or interactions between the groups. I am not sure that this analysis is really needed anyway, it is clear from Figure 3A that rare and phylogenetically distinct species respond more strongly to land use. However, if it is retained the authors should justify it and test the robustness of the analysis, perhaps with bootstrapping?

5) It would be useful to report the correlations between the diversities of different groups, in the supplementary information, to show which respond similarly.

6) The evidence that phylogenetic diversity responds more strongly than species richness is very weak. The R² values and effect size of land use do not seem to differ between taxonomic and phylogenetic richness or endemism and phylogenetic endemism. The points in the fourth paragraph of the Discussion are therefore not supported by the analysis. If the authors want to check whether phylogenetic diversity does respond more strongly to land use then they should calculate measures of phylogenetic diversity uncorrelated with richness. This could be done in two ways: 1) by randomizing species between sites, whilst maintaining site richness values, and calculating expected phylogenetic diversity for the random communities. A standardized effect size would show if the observed phylogenetic diversity values are greater or lower than expected by chance (Webb et al., 2002). 2) the authors could extract residuals for phylogenetic diversity after correcting for taxonomic diversity and analyse the residuals. I imagine these approaches could also be used to correct phylogenetic endemism values. Also see Winter et al., 2013.

7) How well does the meta barcoding approach work to recover the species present in the sample? Are there data showing that it accurately recovers the species present? Did you check that species present in the Tullgren extracts were also recovered by the meta barcoding? It would be reassuring if there were some data on the robustness of the approach.

8) It is interesting that the authors find such a large effect of land use change on the soil fauna. Other studies in Europe have found that soil fauna are quite insensitive to land use intensification (Gossner et al., 2016). I imagine that this is because they looked at land use intensification within a habitat (i.e. within grasslands) while this study considers land conversion from forests to grassland and cropland. I think it would be worth commenting on this in the Discussion.

9) I feel it would be better to show plots of model predictions and standard errors, rather than the raw data boxplots currently used. If the models include covariates to correct for environmental confounders then this would allow the authors to plot the effect of land use after correcting for other factors.

10) Discussion, fifth paragraph: Collembola also seem resistant to land use change.

11) Table 1 should not be in the main text and could go to the supplement. If you want to use these post hoc tests then you could use letters in the figures to distinguish the land use types that differ from each other.

12) In Figure 3—source data 1B, it would be good to show effects (coefficients) from the models, rather than only showing the SS and F values.

Reviewer #3:

The paper by Dopheide and colleagues examines the effect of land use on soil faunal diversity. Overall, it is a nice analysis and examines a particularly under-studied community type, and I think it will be of value to the broader community. I do have some reservations about the presentation and to my mind, the most important concern of mine is #3 below.

1) There is quite a bit about homogenization in the inference in the Discussion, but this isn't actually tested for. You could do an analysis of β diversity measures (taxonomic and phylogenetic) to determine if communities are in fact more homogeneous in agricultural settings than in forests – see Swenson 2011 PloS ONE 6: e21264 and Jin et al. 2015, J. Ecol. 103: 742-749.

2) The metrics analyzed are good for comparing amongst one another, but I think another needs to be assessed. PD will be highly correlated with richness, and phylogenetic-endemism will pick up on phylogenetic distinctiveness as well as range size -two different possible responses to land use. I would recommend assessing MPD as well to see how land use influences the relatedness of species in communities independent of range size.

3) Phylogenetic-endemism needs to be clarified. How was endemism estimated? Was it across all 75 samples or within each land use type? Further, and this needs to be commented on in the manuscript, estimating endemism from 75 samples is dubious at best and better reflects habitat specialization than endemism per se. Forest specialists can be extremely widespread and not at all endemic. So, you need to be cautious about language and inference. I would recommend getting rid of 'endemism' altogether, except for the methods when describing the metrics, and instead refer to 'habitat specialists'.

4) Further, the comparison between individual species phylogenetic-endemism and community level aggregation can provide more insight as well. You could look at the distribution of species values and how these scale up to the community e.g., see Cadotte and Davies 2010 Div and Dist 16: 376-385 and R function in Pearse et al. 2015 Bioinformatics btv277. Though this latter suggestion is a recommendation and not a requirement since I have a vested interest, so feel free to ignore.

Detailed thoughts:

Abstract:

– There's no real scientific setup or overarching problem statement, just a statement that the authors want to compare something.

– It is not clear what 'community change' refers to. Externally driven or natural fluctuation or succession? I see Land-use later on, the first we see this is with the Results sentence. Needs a better set up. The first sentence of the Abstract should be about land-use driving diversity change and the need to adequately assess community sensitivity to land use.

Introduction:

Does a nice job of setting up the importance of the study, however, paragraph #1 is long and seems to make a number of different points. I prefer short and concise paragraphs that have single subjects and communication goals.

Results:

Subsection “Overall community composition”, last paragraph – it is not clear what heterogeneity refers to.

Figure 2: this is a difficult figure to distil meaningful information. Is there another way to show how lands alters abundance?

Figure 3B and C: can be moved to supplementary material if space is an issue.

Table 1: not needed, significant differences can be indicated on Figure 3A.

Overall, the results are strong and clearly support the major inferences.

Discussion:

Overall good and clear. Again 'heterogeneity' is unclear, please set this up better since it appears to be an important result. Further, there is little biology in the Discussion. Much of it is about the metrics and it would benefit from more discussion of the biology of the systems.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your revised article "Rarity is a more reliable indicator of land-use impacts on soil invertebrate biodiversity than richness or diversity" for consideration by eLife. Your revision and response letter has been assessed by Bernhard Schmid as Reviewing Editor and Ian Baldwin as the Senior Editor.

The agreement was that we cannot yet proceed with the revision until you have more fully incorporated the major suggestions from the first round of reviews. You did provide extensive responses in your response letter, but did not add important items to you revision. These items are, again:

1) Make a linear contrast for the a priori hypothesis that is so pervasive throughout your paper, namely that there is a sequence in which you can put your five land-use categories (you even use this sequence in figures). There is nothing circular and statistically this is much more justified than multiple comparison. As mentioned before, making such a contrast is as valid as any other contrasts, and we can assure you that it will make a much stronger message. The way you can present it, is that you have a term with the five categories and four degrees of freedom and then the alternative of the linear contrast and the remainder with 1 and 3 degrees of freedom. The total of the SS of the two latter will equal the SS of the former but the MS for the linear contrast will be large. We don't mind if you only put this into the supplementary material, but we do want to see it.

Once your a-priori hypothesis that there is an effect of a continuous increase along the five categories, in spite of the inability to measure this continuity, has been tested highly significant, and the remainder, which tests for deviation from the a-priori hypothesis may even be far from significant, you will have a very strong message as you formulated it before but without statistical backing. To fit the linear contrast, just make an explanatory variable LUI with values 1, 2, 3, 4 and 5 for the land-use categories (LUC). Then fit LUI+LUC sequentially. Compare this with the fit of LUC without the linear contrast.

2) Add the analysis with rarest species excluded to the supplement.

3) We disagree with your statement "We think that this test overlooks important aspects of the results, namely the direction of biodiversity differences(as opposed to their consistency among taxa), and which taxa do or do not respond consistently. It is these aspects that the pairwise comparisons are intended to demonstrate." We believe the contrary is true. But it is your paper and we only would like to see the alternative analysis ((Metric ~ LUI+LUC+phylum+order+LUI:phylum+LUC:phylum+LUI:order+LUC:order) in the supplementary material. Note that the LUI-by-taxa interactions are exactly testing differences in direction of biodiversity responses to land-use categories!

4) Add an analysis with environmental covariates to the supplement.

Once we can see these items, we will progress with reviewing.

Essential revisions:

The manuscript should be improved in three respects:

First, there are several terms that are not well defined in the manuscript. "Land use" is sometimes considered as a state of an ecosystem and then called land-use category (please always use the hyphen when "land-use" is an adjectival noun) but sometimes considered as land-use change or intensification etc. Please use consistent terms throughout and make sure you define them well.

We have defined and reworded land-use terms for clarity and consistency throughout the manuscript, as suggested.

In this context and related to the second main point I suggest that you put the five land-use categories explicitly into a linear sequence from one to five and call it land-use intensity.

We thank the reviewers for this suggestion, which we have considered and explored in depth. Unfortunately, we encountered some fundamental issues that meant we were unable to implement it. This is partly because we cover forest, grassland and horticulture, which are all inherently managed very differently. This meant that we were unable to come up with a meaningful way of defining land use intensity that could apply across all land use categories. We tried ranking the importance of key management practices across the land uses (e.g. via pesticide and fertiliser use, disturbance frequency, vegetation complexity), but this failed to distinguish between plantation forestry and low-producing grassland, and between high producing grassland and perennial cropland. We considered using abiotic data as a way of quantifying intensity, but this effectively just tests whether that particular component of land-use intensity influenced diversity metrics. We also considered using the observed biotic data to infer land-use intensity rankings as suggested by the reviewers, but felt that this would create a new set of biases as it involves circular reasoning i.e. it effectively tests whether the gradient defined by biota influences biota. As we are unable to robustly define a land-use intensity gradient, we have more carefully worded the manuscript so that it is clearer that we are not looking directly at a defined intensity gradient.

Endemism is defined in the Materials and methods, but all reviewers would prefer a more generic term such as rare species because you cannot extrapolate from 75 sites to the entire area of your country, which would be necessary to know if a species that only occurred at 1 or 2 sites was geographically restricted.

We have replaced “endemism” with “rarity” (and “phylogenetic endemism” with “phylogenetic rarity”) throughout the paper.

You should also mention the caveat that some of your rare species could just be "sink" or "passenger" species not adapted to the site conditions, while your hypothesis assumes that they are habitat specialists.

We have added this point to the Discussion.

One reviewer suggests in this context that you could actually do a sensitivity analysis without the rarest species that only occur at one site.

We recalculated all biodiversity estimates with species that occur only in a single site excluded, as suggested. The resulting trends were largely consistent with those observed with all species included, although land-use differences were statistically somewhat weaker. We have noted this in the Materials and methods and Results. For example, mean overall rarity values per land use with OTUs that occur in just one site included were ranked natural forest > low-producing grassland > planted forest > high-producing grassland > perennial cropland, with significantly higher rarity in natural forest compared to all four other land uses (Figure 3). Mean overall rarity values per land use with these OTUs excluded were ranked in the same order, but the differences between land uses were less pronounced, with significantly higher mean rarity in natural forest compared to perennial cropland, but no significant differences between mean rarity values in natural forest compared to planted forest or grassland sites. Therefore, our results are partly, but not entirely, driven by species occurring in a single site.

Second, you should get much more from your data if you expand your analyses. In principle, you selected your five categories such that they can be ordered along a gradient. Best would be if you could calculate some value for land-use intensity for each category or even each site, but even in the absence of this you are allowed to simply assign land-use intensity values from 1-5 and use this linear contrast in all ANOVA (the easiest way is to code land-use categories from 1-5 and use it as continuous variable LUI and factor LUC: then fit LUI+LUC in this order and you get a stronger test of your main hypothesis, namely the linear contrast LUI, and a test for deviations from the linear contrast, which is measured by LUC when fitted after LUI).

We thank the reviewers for this suggestion. We have considered this point at great length. Unfortunately, as discussed above, we think this analysis approach is problematic. We are unaware of any applicable method for determining land-use intensity of our sites that applies across all land uses, and there is no clear a priori reason to rank all five land-use categories in a linear sequence. Any such ranking would be inherently circular (using our results to generate a ranking, then using that ranking as a predictor of our results). Therefore, we don’t feel we can carry out the ANOVA test as suggested.

Please replace all box-plots with means +/- standard errors.

We have replaced all box-plots with means +/- standard errors, as suggested.

In addition, you could show regression lines for the LUI-gradient wherever it is significant.

As we cannot determine a land-use intensity gradient nor order our land-use categories in a linear sequence, we are unable to carry out regression on this.

There would be the more sophisticated method of isotonic regression (Gaines, Steven D., and William R. Rice. 1990. Analysis of Biological Data When there are Ordered Expectations. The American Naturalist 135: 310-317.), but clearly the simpler method indicated above is also valid.

The above will remove the need for the excessive use of pairwise comparisons which are statistically not independent and ill-suited to test ordered hypotheses. A further reduction of unnecessary repeats and more comprehensive analysis is to compare the responses of the different taxa to LUI(+LUC) in a single ANOVA by putting the metrics for the different taxa into a single column and add a column with the taxa identities (or even two, one for phyla and one for classes/orders). You can then fit the following ANOVA-model:

Metric ~ LUI+LUC+phylum+order+LUI:phylum+LUC:phylum+LUI:order+LUC:order, or more simply without LUC terms and only using phylum or order. The interactions indicated by ":" test if the different taxonomic groups respond differently. Using the above you can exchange Figure 3—source data 1B and omit Supplementary Table 4.

Again, we thank the reviewers for this suggestion. As discussed above, we are unable to determine land-use intensity values for our sites or land-use categories, so cannot use the suggested ANOVA model. Given this constraint, we have instead added a variant of the suggested ANOVA (Metric ~ LUC * taxonomic group) to our analysis, to test for consistent responses of different taxa to land use.

We think that this test overlooks important aspects of the results, namely the direction of biodiversity differences (as opposed to their consistency among taxa), and which taxa do or do not respond consistently. It is these aspects that the pairwise comparisons are intended to demonstrate. The point about pairwise tests being ill-suited to testing ordered hypotheses is not applicable, because our land-use categories could not be ordered a priori, and we have been unable to identify a better statistical method for analysing the patterns of land use responses among different taxa. Consequently, we have retained the pairwise tests, but we have revised the Results section so that less emphasis is placed on these tests. This includes removing the tables of pairwise test results, and indicating significant differences directly on the figures, as suggested by reviewer 2 (below).

There is still the problem of comparing the R² values (or better just use %SS, which are increments in multiple R²) mentioned by one reviewer and we ask you to consider their suggestions. It also seems you analyzed R² values as (meta-)data; this requires explanation and justification in the Materials and methods, because R² values will have a special distribution that deviates from normal.

We have revised and replaced the analysis of R² values according to the suggestions above, and those of reviewer 2 (below), as follows. We used non-parametric bootstrapping stratified by taxonomic group to estimate % Sum of Squares explained by land use for each biodiversity metric and plotted these results as a histogram (Figure 5). We compared these values between metrics using a non-parametric statistical test. This has been explained in the Materials and methods.

Furthermore, as noted by one reviewer, from your Figure 6 it appears that the land-use categories are not well "randomized" across space and thus covariates could explain some of your results. You could include some of these first in the fitting sequence of ANOVAs, but we accept that you don't want to put in too many covariates because there are not so many sites to test more complicated models and sometimes correction for covariates can be unjustified if they are not a "true cause" of an observed effect.

The spatial randomisation of sites is constrained by existing patterns of land use. We agree that covariates could explain some of our results, but this is beyond the main intent of this analysis, which is to compare the performance of different biodiversity metrics for detecting impacts of land use on soil invertebrates. In practice, land use is a surrogate for many other variables (e.g. geology, vegetation, and soil chemistry), which may be the true drivers of invertebrate biodiversity metrics.

One reviewer also suggests that you should try phylogenetic metrics that are not mathematically related to richness.

We have added comparisons of mean pairwise distance, and standardised effect sizes for each of the phylogenetic metrics, between land-uses (Figures 3 and 4).

Your suggestion that the results imply homogenization needs to be tested statistically, which in fact can be done quite easily e.g. by calculating β diversities. You also use a term "heterogeneity", which obviously is taken from output of an R function. However, you cannot expect readers to check this up but rather need to give a clear definition of heterogeneity and how it is calculated.

We tested for multivariate homogeneity of sample dispersions between land-use categories using the “betadisper” function in the R package vegan. This test determines the dispersion of groups as the average distance of group members to the group centroid in multivariate space, and uses ANOVA to test for dispersion differences among groups. Our statements about homogenisation were based on this test. We have also added tests of β diversity and phylogenetic β diversity between land-use categories, as suggested, to further support our statements about homogenisation. We have clarified this in the Materials and methods and Results.

Third, the Introduction and Discussion sections contain many repetitive statements and statements that lead the reader away from the main story line. The Introduction should focus on the effects of land-use intensity on soil organisms that here have been studied with DNA barcoding, offering a new level of resolution. Your focus and hypothesis on differences in responses between common and rare taxa is really secondary and less "a priori" than the main hypothesis that land-use intensity reduces soil biodiversity. The Discussion contains many speculations, in part related to this secondary hypothesis, that weaken the stronger results regarding the main hypothesis.

We have revised the Introduction and Discussion according to these suggestions. For example, we have placed more emphasis on the impacts of land use in the Introduction and Discussion and have edited or removed various speculative and repetitive statements.

Generally, we find it very difficult to say rare taxa can "better" indicate biodiversity responses to land-use intensity, because obviously you don't know what the "true biodiversity" is, which you implicitly use as reference.

We agree that we don’t know what the “true biodiversity” is, and we have therefore replaced any statements about rarity being “better” with more precise terms. For example, the title now states that “rarity is a more reliable indicator” and the Abstract that “rarity provides more consistent and clearer evidence” of land-use impacts, rather than rarity is a “better” indicator.

Reviewer #1:

See above summary.

Reviewer #2:

[…] 1) The study aims to show the effect of land use change (conversion of native forests into grassland or perennial cropping) on soil fauna diversity. However, there is no attempt to correct for confounding factors. How were the sites chosen? Was there an attempt to ensure no confounding between land use and environmental factors such as altitude, soil type etc.? All that is said in the Materials and methods is that the sites were distributed across New Zealand. The Discussion mentions some possible confounding, as it is said that native forests are in more "rugged and less accessible areas". In addition, from Figure 6 it appears that many of the native forests are on the west coast of the South Island and there may therefore be climatic differences from the other sites? Currently there is no attempt to correct for confounding factors in the analysis: land use type is the only factor included in the models. I think it would be worth trying to fit some covariates in these models: climate variables and altitude would make sense and some soil variables such as soil type or pH would be good to consider too. This would make the analysis much more robust in showing that land use is the driver of soil fauna diversity, after correcting for soil and climate. It would also add interesting information on the other drivers of soil fauna diversity.

As discussed above, the randomisation of sites is constrained by land use patterns throughout New Zealand. We agree that covariates would explain some of our results, but feel that this is beyond the intent of this paper, which is to compare the performance of different biodiversity metrics for detecting impacts of land use on soil invertebrates, rather than to understand the specific drivers of soil fauna diversity. We view land use in this context as a way of summarising a wide range of different drivers, including those described by the reviewer.

2) I found the term "endemism" confusing in the context of this study. There are no actual data on whether any of these taxa are endemic to a given area, as these data do not exist for soil fauna. Instead endemism is defined as the number of sites occupied by an OTU. Taxa occurring on few sites are therefore considered "endemic" but there is no information on whether they really are endemic or whether they occur elsewhere, e.g. outside New Zealand. Also, if a taxa is found in only two sites it would be considered endemic here but if those two sites were far apart, e.g. it was found in a northern and southern sample, it might not have a restricted distribution, it might simply be rare throughout the range. I therefore think that it would be clearer to talk about rare species and mention that here rarity is based on site occupancy. The situation may be even more problematic for phylogenetic endemism as in the original paper Rosauer et al., 2009, state that phylogenetic endemism values will be biased if closely related species occurring in other areas are missing from the sample (which is highly likely to be the case in this study).

These are good points, and we thank the reviewer for pointing them out. As mentioned above, we have replaced “endemism” with “rarity” throughout the manuscript.

2b) Related to this, I think it is important to mention in the Discussion that rarity is defined here relative to the sites sampled. Some of the species considered rare in this analysis might not be rare at all if they are common in habitats or areas not sampled in this study. There is nothing the authors can do about it this and I think the approach they have used is completely reasonable, but this limitation ought to be acknowledged in the Discussion.

We have acknowledged this point in the Discussion.

3) How much is the measure of "endemic richness" affected by OTUs occurring in a single site? I could imagine that the distribution of these singleton taxa is quite stochastic and I wonder if it would be worth recalculating the index without them to check that they are not driving the whole pattern.

As discussed above, we recalculated biodiversity estimates with these OTUs excluded, resulting in similar endemism richness (now called rarity) trends, but statistically weaker differences among land uses, indicating that these OTUs partially but not entirely drive the observed patterns. We have noted this in the paper (Results and Materials and methods).

4) I am not really convinced that the analysis of R² values (Figure 4) is valid. The R² values for the different groups are not independent of each other as diversities of different taxa are likely to be correlated due to shared environmental responses or interactions between the groups. I am not sure that this analysis is really needed anyway, it is clear from Figure 3A that rare and phylogenetically distinct species respond more strongly to land use. However, if it is retained the authors should justify it and test the robustness of the analysis, perhaps with bootstrapping?

As mentioned above, we have replaced the previous analysis of R² values with a non-parametric bootstrapping analysis followed by a non-parametric statistical test, as suggested (subsections “Biodiversity differences among invertebrate taxa”, and “Biodiversity analyses and statistics”, Figure 5).

5) It would be useful to report the correlations between the diversities of different groups, in the supplementary information, to show which respond similarly.

We have added this information to the Appendix (Appendix—figures 7-12).

6) The evidence that phylogenetic diversity responds more strongly than species richness is very weak. The R² values and effect size of land use do not seem to differ between taxonomic and phylogenetic richness or endemism and phylogenetic endemism. The points in the fourth paragraph of the Discussion are therefore not supported by the analysis. If the authors want to check whether phylogenetic diversity does respond more strongly to land use then they should calculate measures of phylogenetic diversity uncorrelated with richness. This could be done in two ways: 1) by randomizing species between sites, whilst maintaining site richness values, and calculating expected phylogenetic diversity for the random communities. A standardized effect size would show if the observed phylogenetic diversity values are greater or lower than expected by chance (Webb et al., 2002). 2) the authors could extract residuals for phylogenetic diversity after correcting for taxonomic diversity and analyse the residuals. I imagine these approaches could also be used to correct phylogenetic endemism values.

Also see Winter et al., 2013.

We have calculated standardised effect sizes for each of the tested phylogenetic biodiversity metrics, as suggested (Figure 4, Figure 4—figure supplements 1 and 2). Phylogenetic diversity SES resulted in less consistent and different land use patterns among taxa compared to non-SES values. Phylogenetic endemism SES, in contrast, resulted in somewhat increased consistency of land use patterns among taxa compared to non-SES values, strongly suggesting that rarity declines from natural forest to agricultural land use. Together with the bootstrapping analysis described above, we think this does show that phylogenetic metrics do provide additional value.

7) How well does the meta barcoding approach work to recover the species present in the sample? Are there data showing that it accurately recovers the species present? Did you check that species present in the Tullgren extracts were also recovered by the meta barcoding? It would be reassuring if there were some data on the robustness of the approach.

Recovery of species using DNA metabarcoding is imperfect due to factors such as PCR primer mismatches causing PCR amplification biases to varying extents. However, previous work has shown that the approach tends to recover a majority of species in a sample, and enables the identification of many more species and a broader range of taxa than is typically feasible using traditional methods. We are therefore confident that our method provides a robust (albeit imperfect) assessment of soil invertebrate biodiversity.

8) It is interesting that the authors find such a large effect of land use change on the soil fauna. Other studies in Europe have found that soil fauna are quite insensitive to land use intensification (Gossner et al., 2016). I imagine that this is because they looked at land use intensification within a habitat (i.e. within grasslands) while this study considers land conversion from forests to grassland and cropland. I think it would be worth commenting on this in the Discussion.

This is an interesting point. We have added this to the Discussion, as suggested.

9) I feel it would be better to show plots of model predictions and standard errors, rather than the raw data boxplots currently used. If the models include covariates to correct for environmental confounders then this would allow the authors to plot the effect of land use after correcting for other factors.

The plots now show mean values and standard errors. As mentioned above, we think that including confounding factors is beyond the intention of this research.

10) Discussion, fifth paragraph: Collembola also seem resistant to land use change.

This is true, and we have updated the manuscript accordingly (Discussion).

11) Table 1 should not be in the main text and could go to the supplement. If you want to use these post hoc tests then you could use letters in the figures to distinguish the land use types that differ from each other.

We have removed Table 1, and added letters to figures to show differences between land uses, as suggested (Figures 3 and 4 and their figure supplements).

12) In Figure 3—source data 1B, it would be good to show effects (coefficients) from the models, rather than only showing the SS and F values.

The coefficients of a factorial ANOVA are simply the differences in the means compared to the first treatment level (intercept). We have added mean values to the figures, and therefore there is little additional information that would be gained by adding coefficients to the result tables. Rather, we think inclusion of coefficients in the results tables would make them overly complex.

Reviewer #3:

The paper by Dopheide and colleagues examines the effect of land use on soil faunal diversity. Overall, it is a nice analysis and examines a particularly under-studied community type, and I think it will be of value to the broader community. I do have some reservations about the presentation and to my mind, the most important concern of mine is #3 below.

1) There is quite a bit about homogenization in the inference in the Discussion, but this isn't actually tested for. You could do an analysis of β diversity measures (taxonomic and phylogenetic) to determine if communities are in fact more homogeneous in agricultural settings than in forests – see Swenson 2011 PloS ONE 6: e21264 and Jin et al. 2015, J. Ecol. 103: 742-749.

As discussed above, this was based on tests for heterogeneity/homogeneity of multivariate sample dispersions between land uses. We have now added tests of β diversity and phylogenetic β diversity among land uses. Together, these results provide evidence of biotic homogenisation.

2) The metrics analyzed are good for comparing amongst one another, but I think another needs to be assessed. PD will be highly correlated with richness, and phylogenetic-endemism will pick up on phylogenetic distinctiveness as well as range size -two different possible responses to land use. I would recommend assessing MPD as well to see how land use influences the relatedness of species in communities independent of range size.

We have added analyses of MPD to the manuscript, as suggested. In contrast to the other metrics, MPD does not differ among land uses for most taxa.

3) Phylogenetic-endemism needs to be clarified. How was endemism estimated? Was it across all 75 samples or within each land use type? Further, and this needs to be commented on in the manuscript, estimating endemism from 75 samples is dubious at best and better reflects habitat specialization than endemism per se. Forest specialists can be extremely widespread and not at all endemic. So, you need to be cautious about language and inference. I would recommend getting rid of 'endemism' altogether, except for the methods when describing the metrics, and instead refer to 'habitat specialists'.

Endemism and phylogenetic endemism were calculated across all 75 sites. We have clarified this in the Materials and methods. Furthermore, as discussed above, we have replaced “endemism” with “rarity” throughout the manuscript. We prefer to use “rarity” over “habitat specialisation” because the latter is only one of three axes of rarity and can’t be easily determined from our data.

4) Further, the comparison between individual species phylogenetic-endemism and community level aggregation can provide more insight as well. You could look at the distribution of species values and how these scale up to the community e.g., see Cadotte and Davies 2010 Div and Dist 16: 376-385 and R function in Pearse et al. 2015 Bioinformatics btv277. Though this latter suggestion is a recommendation and not a requirement since I have a vested interest, so feel free to ignore.

This is an interesting suggestion, but we think this may be getting beyond the aims of the manuscript and is likely to over-complicate it.

Detailed thoughts:

Abstract:

– There's no real scientific setup or overarching problem statement, just a statement that the authors want to compare something.

We have edited the Abstract such that it now begins with a statement of the problem.

– It is not clear what 'community change' refers to. Externally driven or natural fluctuation or succession? I see Land-use later on, the first we see this is with the results sentence. Needs a better set up. The first sentence of the Abstract should be about land-use driving diversity change and the need to adequately assess community sensitivity to land use.

We have revised the Abstract according to these suggestions.

Introduction:

Does a nice job of setting up the importance of the study, however, paragraph #1 is long and seems to make a number of different points. I prefer short and concise paragraphs that have single subjects and communication goals.

We have revised paragraph 1 as suggested.

Results:

Subsection “Overall community composition”, last paragraph – it is not clear what heterogeneity refers to.

As discussed above, heterogeneity referred to differences in multivariate community sample dispersion among land-use categories. We have clarified this in the manuscript, and added comparisons of β diversity to strengthen our results.

Figure 2: this is a difficult figure to distil meaningful information. Is there another way to show how lands alters abundance?

We respectfully disagree. The heatmap is intended to show the distribution of species across samples along with their relative abundances. There would surely be other ways to show changes in abundance in relation to land use, but we doubt they would be any more clear or concise, given the need to represent a large number of species.

Figure 3B and C: can be moved to supplementary material if space is an issue.

Figures 3B and C are now supplementary figures linked to Figure 3A (now simply Figure 3).

Table 1: not needed, significant differences can be indicated on Figure 3A.

As discussed above, Table 1 has been removed, and we have indicated significant differences on Figure 3, as suggested.

Overall, the results are strong and clearly support the major inferences.

Thanks!

Discussion:

Overall good and clear. Again 'heterogeneity' is unclear, please set this up better since it appears to be an important result. Further, there is little biology in the Discussion. Much of it is about the metrics and it would benefit from more discussion of the biology of the systems.

We have clarified the meaning of heterogeneity throughout the paper, as suggested, and have added more discussion of land use impacts on soil communities.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

1) Make a linear contrast for the a-priori hypothesis that is so pervasive throughout your paper, namely that there is a sequence in which you can put your five land-use categories (you even use this sequence in figures). There is nothing circular and statistically this is much more justified than multiple comparison. As mentioned before, making such a contrast is as valid as any other contrasts, and we can assure you that it will make a much stronger message. The way you can present it, is that you have a term with the five categories and four degrees of freedom and then the alternative of the linear contrast and the remainder with 1 and 3 degrees of freedom. The total of the SS of the two latter will equal the SS of the former but the MS for the linear contrast will be large. We don't mind if you only put this into the supplementary material, but we do want to see it.

Once your a priori hypothesis that there is an effect of a continuous increase along the five categories, in spite of the inability to measure this continuity, has been tested highly significant, and the remainder, which tests for deviation from the a-priori hypothesis may even be far from significant, you will have a very strong message as you formulated it before but without statistical backing. To fit the linear contrast, just make an explanatory variable LUI with values 1, 2, 3, 4 and 5 for the land-use categories (LUC). Then fit LUI+LUC sequentially. Compare this with the fit of LUC without the linear contrast.

As mentioned previously, we have some concerns with this approach. Our first concern is that after several years of research on land-use impacts on biodiversity, we have come to the realisation that the term “land-use intensity” is problematic. Trying to encompass all the axes of variation across diverse land-use categories into a single metric is probably impossible. Our second, related, concern is with the idea of deriving a ranking from the data and then using that ranking as a predictor of the data, which we view as circularity. Evidently the reviewers do not see this as a problem. Nonetheless, we feel that readers might misunderstand anything termed “land-use intensity” as an a priori, rather than derived variable.

We believe that these concerns can be managed through careful presentation and terminology. Rather than referring to “land-use intensity” we have used the term “derived land-use rank” or DLUR at relevant points in the manuscript. The advantage of this term is that it does not imply anything beyond how the term was quantified (i.e., it is simply a rank score, and it is derived). We have explained this approach in the Materials and methods section.

We carried out ANOVA tests for significant DLUR trends for each biodiversity metric and taxonomic group. This provided very similar conclusions to our tests with land use as a categorical factor. The results of these tests are found in Appendix (Appendix—tables 1 and 5).

2) Add the analysis with rarest species excluded to the supplement.

We have added the results of recalculated biodiversity estimates with the rarest species excluded (those that occur only in a single site) to the Appendix (Appendix—figures 3-5). These trends were largely consistent with those observed with all species included, although land-use differences were statistically somewhat weaker. For example, mean overall rarity values per land use were ranked in the same order with the rarest species included and excluded, but statistical differences between land uses were less pronounced in the latter case compared to the former.

3) We disagree with your statement "We think that this test overlooks important aspects of the results, namely the direction of biodiversity differences(as opposed to their consistency among taxa), and which taxa do or do not respond consistently. It is these aspects that the pairwise comparisons are intended to demonstrate." We believe the contrary is true. But it is your paper and we only would like to see the alternative analysis ((Metric ~ LUI+LUC+phylum+order+LUI:phylum+LUC:phylum+LUI:order+LUC:order) in the supplementary material. Note that the LUI-by-taxa interactions are exactly testing differences in direction of biodiversity responses to land-use categories!

We have added the suggested ANOVA to the paper, using DLUR instead of LUI. The ANOVA model used was as follows:

Metric ~ DLUR + LUC + group + DLUR:group + LUC:group

in which “group” is 17 taxonomic groups encompassing all the terrestrial invertebrate taxa detected.

This analysis has been explained in the Materials and methods section, with a summary of the outcomes in the Results, and full results in the Appendix (Appendix—table 2).

4) Add an analysis with environmental covariates to the supplement.

We have added analyses of environmental covariate effects on biodiversity to the paper. We did this in two ways. First, we carried out ANOVA tests for effects of spatial attributes (latitude and altitude) plus land-use category on overall biodiversity metrics. Second, we generated a PCA based on spatial and soil chemistry variables, then carried out ANOVA tests for effects of the major PCA components plus land-use category on overall biodiversity metrics. These analyses are described in the Materials and methods and Results, with the full details in the Appendix (Appendix—figure 6; Appendix—tables 3 and 4)

Author details

1. Andrew Dopheide

   Manaaki Whenua - Landcare Research, Auckland, New Zealand

   Contribution

   Conceptualization, Formal analysis, Software, Bioinformatics, Data curation, Visualization, Writing - original draft, Writing - reviewing and editing

   For correspondence

   dopheidea@landcareresearch.co.nz

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-1554-9832

2. Andreas Makiola

   Bio-Protection Research Centre, Lincoln University, Lincoln, New Zealand

   Contribution

   Investigation, Fieldwork, Writing - reviewing and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-9611-9238

3. Kate H Orwin

   Manaaki Whenua – Landcare Research, Lincoln, New Zealand

   Contribution

   Conceptualization, Funding acquisition, Writing - reviewing and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-0906-728X

4. Robert J Holdaway

   Manaaki Whenua – Landcare Research, Lincoln, New Zealand

   Contribution

   Conceptualization, Funding acquisition, Investigation, Fieldwork, Writing - reviewing and editing

   Competing interests

   No competing interests declared

5. Jamie R Wood

   Manaaki Whenua – Landcare Research, Lincoln, New Zealand

   Contribution

   Funding acquisition, Investigation, Molecular analysis, Writing - reviewing and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-8008-6083

6. Ian A Dickie

   Bio-Protection Research Centre, School of Biological Sciences, University of Canterbury, Christchurch, New Zealand

   Contribution

   Conceptualization, Funding acquisition, Investigation, Fieldwork, Formal analysis, Software, Bioinformatics, Writing - reviewing and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-2740-2128

• 1,900

  Page views

• 297

  Downloads

• 14

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.