Implementing integrated measurements of Essential Biodiversity Variables at a national scale

Funding: the Strategic Science Investment Funding for Crown Research Institutes from the Ministry of Business, Innovation and Employment.


INTRODUCTION
Biodiversity is being lost globally at an increasing rate (Tittensor et al., 2014), caused by habitat destruction, land-use intensification, pollution, overharvesting, climate change, and biological invasions (Dornelas et al., 2014;Gossner et al., 2016). Most evidence of biodiversity trends at national and global scales is aggregated from local data sources, often subjectively placed (McGill, Dornelas, Gotelli, & Magurran, 2015) or otherwise biased (Fournier, White, & Heard, 2019;Geijzendorffer et al., 2016). Different components of biodiversity are often measured at different sample points and seldom simultaneously, which limits understanding of interrelationships among them (Dornelas et al., 2019, Pereira & Cooper 2006. The lack of a harmonised observation system delivering regular, timely data (Pereira et al., 2013) hampers understanding of how biodiversity responds to pressures (Pecl et al., 2017).
Despite repeated calls for widespread, objective biodiversity data (e.g. Jackson et al., 2016), there are few examples (e.g. Alberta Biodiversity Monitoring Institute, 2015) and only one that we know of at a national scale (in Mexico; Garcia-Alaniz et al., 2017).
Systematic national biodiversity monitoring requires a range of methods, from those assessing ecosystem structure (e.g. remote sensing; Pereira et al., 2013) to those assessing species populations.
Repeated measures of permanent sample points allow assessment of trends in species populations and community composition (Pereira et al., 2013), enabling global meta-analyses (e.g. Dornelas et al., 2014;Vellend et al., 2013). For example repeated measures of permanent vegetation plots reveal changes in plant populations and community composition at large scales (e.g. Mayor, Cahill, He, & Boutin, 2015).
Systematic national monitoring allows evaluation of regional or local variation, because trends in biodiversity are scale dependent (McGill et al., 2015). It can provide baselines for ecosystem-based management, without which poor decision-making and environmental policies can result (Seargeant, Moynahan, & Johnson, 2012;Yaffee, 1997).
The United Nations Convention on Biological Diversity (CBD) provides a global imperative for national biodiversity monitoring. Pereira et al. (2013) formulated Essential Biodiversity Variables (EBVs) suitable for determining progress towards the Aichi Targets of the CBD, set for 2020. The EBV classes include species populations and community composition, which include individual EBVs well suited to repeated measurements at point locations. New Zealand is a global biodiversity hotspot (Myers, Mittermeier, Mittermeier, da Fonseca, & Kent, 2000), with very high endemism. As a signatory to the CBD, New Zealand needs to report progress towards Aichi Targets and it has passed legis-  Schmeller et al., 2018). Species populations and community composition of some non-native terrestrial mammals are also measured because they exert strong influences on native flora and fauna (Walker & Bellingham 2011). New Zealand already reports EBVs based on national-scale remote sensing (in the ecosystem structure EBV class; e.g. Cieraad, Walker, Price, & Barringer, 2015) and the data from point locations sampled systematically can supplement and add value to these EBVs by validating interpretation of imagery (Table 1; Pereira et al., 2013). National measurements of EBVs at point locations recognise interrelationships between native and non-native biodiversity ( Figure. 1) and are all quantified within weeks of each other at the same permanent sample points. We demonstrate these interrelationships at a national scale, which was impossible before the programme began.

Measuring and reporting EBVs across New Zealand
The main islands of New Zealand (North Island, South Island, Stewart Island, and immediately adjacent islands) extend between 34 • 23′ S and 47 • 17′ S and comprise 266,256 kmš. Axial mountain ranges extend about 80% of the country's length (up to 3,724 m elevation in the South Island), while volcanic peaks (up to 2,797 m) feature prominently in the North Island. The climate is oceanic temperate (McGlone, Buitenwerf, & Richardson, 2016). Before human settlement, most land below treeline was forested, with alpine grasslands above treeline. Māori settlement (from about 1280) caused deforestation of drier regions, east of the axial ranges (Perry, Wilmshurst, & McGlone, 2014). European settlement from about 1830 deforested the wetter regions and lowlands, and altered the biota by introducing European agricultural grasses, crops, and livestock such that little native vegetation cover now remains in some regions (Walker & Bellingham 2011). Many introduced plants and animals are invasive and have caused substantial, rapid change in native biodiversity. For example non-native predatory TA B L E 1 Essential Biodiversity Variables (EBVs) for which New Zealand's national biodiversity monitoring programme contributes data specifically (X, or other qualifiers) and to which it could potentially contribute data (P; in situ data could supplement EBVs measured primarily by remote sensing), within EBV classes (Pereira et al., 2013). Dashes represent EBV classes or individual EBVs to which no contribution is made. Individual EBVs and their descriptions are from https://geobon.org/ebvs/what-are-ebvs/

Species populations
Species distribution X X X Population abundance X X X Population structure by age/size class Trees only --

Community composition
Taxonomic diversity X X Partial Species interactions P P P

Ecosystem structure
Ecosystem extent and fragmentation P, in situ P, in situ P, in situ Ecosystem composition by functional types P, in situ P, in situ P, in situ  (Table   S1). Most (62.8%) of the 11.67 million ha in native vegetation cover is administered by DOC; the remainder is mostly privately owned.

Species populations and community composition as EBV classes measured nationally
The EBVs implemented across New Zealand provide data for two of six EBV classes, that is species populations and community composition (Pereira et al., 2013), and provide comprehensive quantification of three individual EBVs for vegetation, birds, and non-native mammals, and of another EBV (population structure) for forest trees (Table 1).
Vegetation, birds, and non-native mammals have been of enduring public interest and are major foci for management and restoration (Allen, Bellingham, & Wiser, 2003;Norton, 2009). Effects of biological invasions are a particular focus. Individual EBVs employ widely used methods (Supporting Information S1) but never before combined simultaneously at the same sample points; each have therefore required iterative refinement (e.g. Forsyth et al., 2018a;Gormley et al., 2015). Methods provide data on 15 of 32 non-native mammals, particularly herbivores ( Figure 1) and two omnivores (brushtail possums

Some non-native mammals are almost mutually exclusive in their distributions
Non-native mammals (brushtail possums, brown hares (Lepus europaeus), European rabbits (Oryctolagus cuniculus), and ungulates) occurred on 88% of 823 sample points across public land ( Figure 5).
The areas invaded by brushtail possums and those by brown hares were almost mutually exclusive (co-occurrence at only 12% of sample points; χ 2 1 = 19.5, p < .001). Brushtail possums occurred throughout forests and shrublands, across all latitudes, more often below 1,000 m (79% of sample points) than between 1,000 and 1,500 m (64% of sample points; χ 2 1 = 5.78, p = .016; Figure 5a). They occurred even less frequently in non-woody ecosystems (44% of sample points), where they were 74% less abundant than in forests and shrublands (mean trap catch index = 1.7% ± 0.28 (SEM) % vs. 6.5% ± 0.42%; two-sample t 811 = 9.43, p < .001). Brown hares occurred mostly in the eastern South Island, seldom in the North Island (12 of 191 sample points there; Figure 5b) and, in contrast to possums, scarcely occurred in forests and shrublands (only 7% of sample points below 1,500 m).
Although brown hares occurred in some non-woody ecosystems below 500 m (14% of those sample points), they were much more frequent in them above 500 m (60% of those sample points), including TA B L E 2 Number of plots where each method was applied each year between 2001 and 2018, partitioned between public and private land, and land cover assigned by dominance by native or non-native vegetation (according to Cieraad et al., 2015)

The deforested eastern South Island is heavily invaded by non-native plants, birds, and mammals
The severity of biological invasions by each of non-native plants, birds, and mammals across public land was assessed using a binary classification based on thresholds defined by either management targets (e.g. Warburton & Livingstone 2015) or expert opinion (Bellingham, Cieraad, Gormley, & Richardson, 2015). Thresholds of a high degree of invasion for non-native plants was >25% of cumulative cover; for non-native birds, their species richness exceeded that of native birds; and for non-native mammals when at least one species (or

DISCUSSION
The approach taken by New Zealand's national biodiversity monitoring programme of measuring multiple EBVs simultaneously has resulted in defensible, objective reporting of the national state of biodiversity and sets in place an infrastructure that can be measured repeatedly to determine trends and interrelationships among EBVs. Seargeant et al. (2012) identified attributes needed to implement longterm biodiversity monitoring. First is leveraging off existing infrastructure. Reporting EBVs throughout New Zealand built on existing F I G U R E 4 Box-and-whisker plots of cumulative counts of all individual native and non-native bird species for 1340 sample points, grouped by (a) forests and shrublands versus non-woody in three elevational classes, and (b) major forest physiognomic groups and shrublands (from Wiser et al., 2011). Counts sharing the same letter are not significantly different at α = .05 (Tukey HSD test,). Boxes indicate the 25th and 75th percentiles and the horizontal line indicates the median. The whiskers cover data points no more than 1.5 times the interquartile range from the box infrastructure for national carbon reporting, using the same permanent plots in natural forests and shrublands (Holdaway et al., 2017) to measure vegetation EBVs (plant species distributions, population abundances, and taxonomic diversity; Table 1)  porting Information S1). Third, they recommended regular reporting, and the programme has contributed to DOC's annual reports from its inception (e.g. Bellingham et al., 2015), to the most recent state of environment reports (Ministry for the Environment andStatistics New Zealand, 2015, 2018), and the recent national CBD report. Data from public land are obtainable on request from the New Zealand Department of Conservation and Ministry for the Environment, while data from private land require additional permission from landowners and regional councils.

Implementing a national programme
New Zealand's national programme provides a baseline of measurements at temporal and spatial scales that relate directly to objectives at local management scales, as recommended by Seargeant et al. (2012), and it provides evidence to show local or regional conformity or departure from widespread patterns. For example, control of red deer (Cervus elaphus scoticus) was conducted in one forested watershed to determine effects on seedling recruitment, and the local estimates of ungulate density were the same as the median nationally, supporting a view that the effects of control were generalizable (Bellingham et al., 2016). Local management of deer and other ungulates in New Zealand forests has often been predicated on low representation of small stems outside long-established but subjectively placed fenced exclosures compared with those inside. The systematic assessment showed that forests nationally are similar to those 'inside' exclosures (Peltzer et al., 2014), indicating that locally intense effects of ungulate browsing are not generalizable. National data can also provide a rational basis for assigning resources for management, preventing inappropriate diversion of resources to some local sites.

Integration of EBVs
The integration of bird and vegetation EBVs from the same sample points demonstrates the importance of forests as habitat for New F I G U R E 5 Relative abundance of four non-native mammals across 823 sample points by elevation and latitude, ranging from 0 (smallest points) to maximum abundance (largest circles) in woody (green) and non-woody (blue) ecosystems. For possums, circle sizes relate to the mean trap-catch index (possums per 100 trap nights) recorded across four transects. For ungulates, rabbits, and hares, circle sizes relate to the mean pellet count across the four transects. Circle sizes of abundance are not comparable among the four mammal groups Zealand native birds, and that research is needed to determine why non-native birds sometimes dominate in non-woody ecosystems.
Existing national data sources, such as bird species' occurrence in 10 × 10 km cells throughout New Zealand (Robertson, Hyvönen, Fraser, & Pickard, 2007), could not provide this information because most cells contain more than one vegetation type (Dymond, Shepherd, Newsome, & Belliss, 2017), and are therefore unsuitable for linking bird community composition and vegetation habitat. Similarly, the limited co-occurrence of brushtail possums and brown hares ( Figure 5) could not be deduced from existing distribution maps (King 2005).
National plot-based data will improve precision in species distribution models, enabling them to accommodate the fine-scale differences in climate and habitat essential for forecasting interacting effects of global change.

Benefits of national measurement of EBVs
Systematic national sampling is suitable for reporting trends of the common and dominant species that exert the strongest influence on ecosystem processes and underpin provisioning of many ecosystem services (Avolio et al., 2019). Some can become rare rapidly, for example the near-total loss of once-dominant American chestnut (Castanea dentata) from eastern North American forests in the early 20th century caused by the non-native pathogen Cryphonectria parasitica (Paillet, 2002). Others show gradual, but significant declines, for example 19 abundant North American land bird species each experienced population reductions of >50 million birds over 48 years (Rosenberg et al., 2019), and common plants showed the greatest declines in occurrence over 20 years in Germany (Jansen, Bonn, Bowler, Bruelheide, & Eichenberg, 2020). Repeated national sampling of common species will provide early warning signals to prompt action (Schmeller et al., 2018;Wintle, Runge, & Bekessy, 2010 (Lindenmayer, Piggott, & Wintle, 2013) to prevent mohua reaching its current 'nationally vulnerable' status (Robertson et al., 2013).  2015).  bemoaned the lack of questions to guide that Program, but we disagree with this criticism. Because even the best ecologists are poor at anticipating the behaviour of extremely well-studied systems (Doak et al., 2008), questions based on deterministic models can constrain data collection and sampling design (Wintle et al., 2010). Too little is known about many environmental drivers of New Zealand's biodiversity in space and time. For example, distributional limits of many endemic trees and their relative abundances are poorly understood (Lee, 1998) and the distributions of many birds defy simple explanations in terms of habitat suitability or predation pressure (e.g. the patchy distribution of South Island robin (Petroica australis); Powlesland, 2013). More generally, we disagree that measuring change in biodiversity requires questions to guide it or hypotheses to test. Governments routinely report trends in GDP, health, crime, and education metrics without hypotheses or questions (McGlone, 2014): these data generate, rather than respond to, questions. Long-term collection of climate data, atmospheric CO 2 concentrations, and precipitation and stream chemistry was driven not by questions, but because the data were seen as fundamental (Lovett et al., 2007;McGlone, 2014): the trends from these data spawned hypotheses and formed the convincing evidence base for global and national policies (Lovett et al., 2007). We believe long-term, systematic collection of biodiversity data is equally fundamental, and that it will similarly generate questions and hypotheses, and form the evidence base needed to set policies to maintain biodiversity (Kidd, Bekessy, & Garrard, 2019), and evaluate their effectiveness (Visconti et al., 2019).
Such an approach allows documenting and learning from ecological "surprises" (sensu Doak et al., 2008, e.g. unprecedented interactions between climate change and pests and pathogens). The longer the monitoring is maintained, the greater the chance of revealing unprecedented community dynamics (Lindenmayer, Likens, Krebs, & Hobbs, 2010) and detecting effects of infrequent, but ecologically important, events. For example since New Zealand's programme began, the first tropical cyclone in 46 years to disturb western South Island forests occurred in 2014 (Macara, 2015), a M w 7.8 earthquake in 2016 generated thousands of landslides across the north-eastern South Island (Hamling et al., 2017), and a novel pathogen, myrtle rust (Austropuccinia psidii), reached many regions in 2017, infecting some native Myrtaceae species. The national network of sample points allows in situ quantification of the disturbance regime EBV (Table 1), and the data from the other EBVs allow unprecedented quantification of biotic interactions in response to perturbations (e.g. whether myrtle rust effects on fleshy fruited and nectar-bearing Myrtaceae also affect bird species that feed on them).

Limitations and relationships to other monitoring initiatives
Systematic national sampling is unsuitable to assess status and trends in uncommon ecosystems (sensu Williams, Wiser, Clarkson, & Stanley, 2007), which will be sampled by few replicates, if at all (Gardner, 2010;. Uncommon ecosystems, and the species restricted to them, therefore require their own monitoring systems that encompass their full environmental range. Likewise, local networks can target taxa that are too demanding of expertise to measure nationally (e.g. of invertebrates; Watts, Stringer, Innes, & Monks, 2017), and species with interannual changes in abundance so great that five-yearly measurements are inappropriate (e.g. rodents; Ruscoe, Wilson, McElrea, McElrea, & Richardson, 2004). Local networks that provide intensive sampling in space and time (e.g. Elliott et al., 2010) add interpretive value to national programmes (Jetz et al., 2019), as well as assessing the effectiveness of local management (e.g. Forsyth, Ramsey, Perry, McKay, & Wright, 2018b). The growing contribution of citizen science can also be integrated alongside the national programme (e.g. enhanced prediction of species occupancy; Jetz et al., 2019).

CONCLUSION
New Zealand's national programme is now providing primary data on species populations and community composition throughout the country, which have been hitherto lacking compared with well-surveyed, densely populated countries (e.g. in Europe). The programme's systematic sampling sets it apart from post hoc assembling of data from multiple subjectively placed samples to determine biodiversity status and trend (e.g. Butchart et al., 2010) or from unrepresentative sites (e.g. national monitoring networks that only sample nature reserves, as in China; Xu et al., 2017).
The sustainability of programmes such as New Zealand's is uncertain, in part because of high start-up costs and the lag between implementation and realising benefits, which are most apparent once trends are demonstrated (Watson & Novelly 2004). The national programme currently assesses biodiversity mostly on public land; more emphasis is now needed in private land, which can be achieved through greater regional government participation. The more agencies that are involved (central and regional government, and scientific research agencies), the greater its chances of sustainability (cf. Jackson et al., 2016). This poses challenges in terms of maintenance and sharing of infrastructure (e.g. databases) and coordination of field efforts and capacity building. For now, New Zealand's national reporting of EBVs is a step towards it becoming normalised, in the same manner as its national and international reporting of human health and education statistics. Bellingham, Richardson, Gormley, Allen, Forsyth, McKay, MacLeod, van Dam-Bates, and Wright conceived the ideas and designed methodology. Cook, Crisp, McKay, and Wright directed collection of the data.

AUTHORS' CONTRIBUTIONS
Richardson and Gormley analysed the data. Bellingham and Richardson led the writing of the manuscript. All authors contributed critically to the drafts and gave final approval for publication.