As people, animals and materials are transported across increasingly large distances in a globalized world, threats to our biosecurity and food security are rising. Aotearoa New Zealand is an island nation with many endemic species, a strong local agricultural industry, and a need to protect these from pest threats, as well as the economy from fraudulent commodities. Mitigation of such threats is much more effective if their origins and pathways for entry are understood. We propose that this may be addressed in Aotearoa using strontium isotope analysis of both pests and products. Bioavailable radiogenic isotopes of strontium are ubiquitous markers of provenance that are increasingly used to trace the origin of animals and plants as well as products, but currently a baseline map across Aotearoa is lacking, preventing use of this technique. Here, we have improved an existing methodology to develop a regional bioavailable strontium isoscape using the best available geospatial datasets for Aotearoa. The isoscape explains 53% of the variation (R2 = 0.53 and RMSE = 0.00098) across the region, for which the primary drivers are the underlying geology, soil pH, and aerosol deposition (dust and sea salt). We tested the potential of this model to determine the origin of cow milk produced across Aotearoa. Predictions for cow milk (n = 33) highlighted all potential origin locations that share similar 87Sr/86Sr values, with the closest predictions averaging 7.05 km away from their true place of origin. These results demonstrate that this bioavailable strontium isoscape is effective for tracing locally produced agricultural products in Aotearoa. Accordingly, it could be used to certify the origin of Aotearoa’s products, while also helping to determine if new pest detections were of locally breeding populations or not, or to raise awareness of imported illegal agricultural products.
Citation: Kramer RT, Kinaston RL, Holder PW, Armstrong KF, King CL, Sipple WDK, et al. (2022) A bioavailable strontium (87Sr/86Sr) isoscape for Aotearoa New Zealand: Implications for food forensics and biosecurity. PLoS ONE 17(3): e0264458. https://doi.org/10.1371/journal.pone.0264458
Editor: Lee W. Cooper, University of Maryland Center for Environmental Science, UNITED STATES
Received: October 18, 2021; Accepted: February 10, 2022; Published: March 16, 2022
Copyright: © 2022 Kramer et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: CLK received funding from the Royal Society of New Zealand Marsden Fund Fast-Start Grant (Award #: 17-UOO-149): https://www.royalsociety.org.nz/. KFA and PWH received funding from Better Border Biosecurity through a contract from The New Zealand Institute for Plant and Food Research Limited (Contract #: 37373): https://www.b3nz.org.nz/ and from the Plant Biosecurity Cooperative Research Centre, Australia (Contract #: PBCRC2111): http://legacy.pbcrc.com.au/. PWH received funding from the Tertiary Education Commission Centre of Research Excellence PhD funding to the Bio-Protection Research Centre: www.tec.govt.nz/funding RTK received funding from the following organizations: Centre for Global Migrations Postgraduate Research 2019 and 2020 Grants: https://www.otago.ac.nz/global-migrations/ Australasian Society for Human Biology Studentship: https://www.australasianhumanbiology.com/ Royal Society Te Apārangi Skinner Fund: https://www.royalsociety.org.nz/ University of Otago Doctoral Scholarship: https://www.otago.ac.nz/graduate-research/scholarships/phd/ Origin Inference from Geospatial Isotope Networks Grant (NSF Award ABI-1565128): https://www.nsf.gov/awardsearch/showAward?AWD_ID=1565128 All funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Aotearoa New Zealand’s (hereafter Aotearoa) local and export economy as well as its natural environment have been jeopardized by the introduction of foreign pests and associated diseases [1–5] as well as the selling of impure and fraudulent food products marketed as Aotearoa-made [6–9]. Specifically, Aotearoa’s food industry is vulnerable to tampered, counterfeited, adulterated, and simulated food products [6,8]. Ensuring the purity and integrity of Aotearoa products protects a multi-billion-dollar food industry, with dairy ($16.6 billion NZD), wine ($1.75 billion NZD), and mānuka honey ($314 million NZD) being key agricultural products that are the most common fraudulent goods on the national and international markets . Fraudulent and cheaper varieties of each product result in significant financial losses for the Aotearoa economy . For mānuka honey, it is estimated that the annual recorded profit of $314 million NZD is four times inferior to what would be expected without counterfeit products diluting the market . Consequently, the Treasury of the New Zealand Government estimates that the Ministry of Primary Industries (MPI) spends approximately $400 million NZD each year on biosecurity programs that assess, contain, and prevent foreign pests and fraudulent foods from entering and exiting the country . Understanding the origins and pathways of entry is a key piece of information towards this effort but is often very difficult or impossible to provide.
Provenancing biological products and organisms in Aotearoa using biogeochemical analyses has been investigated for both food forensics and biosecurity. To date it has been restricted to use of the isotopes of light elements (typically δ2H, δ18O, δ15N, δ13C, and some others) and trace element concentrations, including for the verification of products such as milk powders [10,11], wine , fetal bovine serum , or for insect pest biosecurity breaches . Currently in Aotearoa, private company Oritain offers origin verification for food and related products . Crown Research Institutes GNS Science , Bio-Protection Research Centre at Lincoln University  and the Centre for Trace Element Analysis at the University of Otago  also perform provenancing consultancy but focus more on research and conservation. Oritain utilizes light element stable isotopes (δ2H, δ18O, δ15N, δ13C) and trace elements (Na, K, Zn, Fe, +35 others) to authenticate and provenance materials within the country and globally. These geochemical methods for provenancing are proven and successful predictors of geographic origin for a variety of materials [18–22], but they have limitations.
Of the light element stable isotopes, δ2H and δ18O found in biological tissues are primarily derived from regional precipitation, drinking water, and atmospheric diatomic oxygen [23,24]. Ehleringer et al.  found that δ2H and δ18O present in organic tissues displays “spatially explicit patterns” that vary predictably with geography including altitude, latitude, temperature, and continentality [19,25]. However, their interpretation is complex as these isotopes fractionate during biological and physical processes as they circulate through ecosystems and tissues [23,26]. Therefore, while specific tissues usually preserve the spatial patterns observed in precipitation, predicting them requires the impractical development of tissue-specific conversion equations to account for isotopic fractionation [27–29]. In addition, δ2H and δ18O are sensitive to evaporative losses and to exchange, which complicates sample collection, storage, and application [19,25,30–32]. Importantly, δ2H and δ18O variations on the landscape vary continuously and at low spatial resolution, as well as seasonally, making the values redundant regionally [33–35]. Although, in rare situations, the spatial and seasonal distributions of δ2H and δ18O can be useful to differentiate between materials originating from the northern and southern hemispheres . Generally, however, all else being equal it can be challenging to isotopically distinguish between ecosystems that are geologically and climatically similar using only δ2H and δ18O [34,36] because the variation and patterning of atmospheric isotopes may not be distinct enough. This issue is most important in Aotearoa and the greater Pacific where islands and small land masses do not experience the same continentality effects that lead to distinct isotopic variation in other regions of the world.
The other light isotopes, δ13C and δ15N, are primarily used to look at variations in diet for humans and animals and farming practices [22,37–42]. Singularly, these isotope tracers have limited application to geographical origin prediction but combining them with other isotopes can assist with provenancing efforts. This is because δ13C and δ15N vary predictably with the climate, dietary resources, and long-term land-use effects as they differentiate between C3 and C4 plants, protein consumption in the diet, and identifying the use of conventional and organic fertilizers in farming systems [37,40,43]. Lastly, all light isotopes are susceptible to variation introduced through the global supermarket, where nonlocal products may be incorporated into modern human diets and blur the geochemical signature within the sample [19,37,44,45].
In addition to the light stable isotopes used for provenancing, trace elements (chemical elements found in low concentrations of less than 100 parts per million (ppm) in the mineral matrix) are also not used to predict origins but can be used for samples of known origin to determine if they have similar or distinguishable chemical profiles from other, potentially fraudulent, samples [46,47]. Chemical profile comparisons can be based on just a few key elements or on a whole suite of elements depending on the sample and reference materials available . For example, trace elements could identify fraudulent food products, like wine or tea, by comparing the chemical fingerprint of the suspicious sample to the authentic product [20,21]. The caveat is that this requires a priori knowledge about the chemical profile for all potential regions of origin to determine if the sample of interest classifies into a particular region or group . While feasible, this would require an extensive database to store the chemical fingerprints for every type of reference material that may need to be provenanced. Furthermore, Koffman et al.  found that trace elements from Aotearoa sediments lack the regional variability observed when using lead, neodymium, and strontium isotopes, but trace elements have successfully been used to demonstrate regional variability in Aotearoa soils at multiple scales [49,50].
The analysis of the radiogenic strontium isotope ratio, 87Sr/86Sr, is an alternative approach that has the potential to be broadly applicable without the need for situation-specific research and development. Internationally, this isotope system has been a key investigative method used to predict the region-of-origin for plants, animals, insects, and other biological materials [18,51–56]. Ecosphere 87Sr/86Sr variation primarily reflects the underlying geology but also other processes including chemical weathering, alluvial and fluvial erosion, soil processes, and aerosol deposition (sea salt, volcanic ash, dust, loess) [51,55,57]. Strontium in the geosphere enters the tissues of biological organisms through their uptake of water and ingestion of dietary resources. This integrated 87Sr/86Sr fraction is referred to as biologically available or “bioavailable”. Any minor isotopic fractionation that occurs in the environment is corrected accordingly during analysis [51,52,55,58,59]. Therefore, biological tissues usually retain the “isotopic fingerprint” of the local ecosystem from which dietary resources were obtained . Different biological tissues (plant leaves, hair, nails, teeth, and bone) form and remodel at different rates and the integrated 87Sr/86Sr values in these tissues should reflect where organisms lived at different time frames of their lives [60–62].
Strontium isotope analysis as a tool in forensic tracing has received limited attention to date in Aotearoa mostly due to the cost of analysis when scaled to very large numbers of samples, and the lack of a bioavailable 87Sr/86Sr baseline. Recent efforts have gone some way to tackling these issues, with technical advances to address the former  and with Duxfield et al.  aggregating published geological data to create a 87Sr/86Sr baseline. The recent baseline  summarized the variation in geological 87Sr/86Sr for Aotearoa and did not predict the bioavailable 87Sr/86Sr produced through the complex environmental system that 87Sr/86Sr cycles through. Duxfield et al.  used a case study to compare plant bioavailable 87Sr/86Sr values to their geologic baseline and found that the values fell within the expected ranges based on the underlying lithology but demonstrated smaller 87Sr/86Sr value ranges than the geological unit they were associated with. These results are most likely due to the geological 87Sr/86Sr baseline not considering the exogenous sources of 87Sr/86Sr in the biosphere that are influenced by climatic, atmospheric, and environmental conditions. Also, some of the published geological data they used are based on 87Sr/86Sr compositions for individual mineral phases rather than the bulk rock, which homogenizes mineral signatures.
Similar to baselines, isoscapes are models that predict the spatial distribution of isotopes by incorporating various sources that contribute to the bioavailable isotope “pool” of a region [65–72]. Strontium isoscapes have been produced at the continental scale for North America , Central America , Africa , and Europe , at regional scales for Australia , the Caribbean , China , the Netherlands , France , southwest Sweden , the Modena , and South Korea , as well as for island locales such as mainland UK  and Ireland . For strontium isotopes, the leading approach to create regional and global isoscapes [51,56,87–89] uses bioavailable 87Sr/86Sr values of georeferenced samples (soil, plants, and small local organisms) and a machine-learning framework that considers contributions from the ecosphere to predict bioavailable 87Sr/86Sr values across the region of interest.
In this paper, we introduce the first bioavailable 87Sr/86Sr isoscape model for Aotearoa constructed using a machine-learning approach [51,88] and the best available dataset of geospatial predictors. Specifically, the isoscape construction uses a random forest (RF) model framework that considers a variety of climatic, atmospheric, and environmental variables that may contribute to the spatial distribution of bioavailable 87Sr/86Sr throughout Aotearoa. The RF model also uses bioavailable 87Sr/86Sr data from plants, soils, and animals living in Aotearoa to calibrate the model and ensure that the predicted 87Sr/86Sr model values reflect the actual 87Sr/86Sr values obtained from the real-world samples.
Bataille et al.  discuss the optimal substrate to sample (plants and soils versus small local animals) when constructing 87Sr/86Sr isoscapes and concluded that while local animals are preferred because they integrate multiple sources of bioavailable 87Sr/86Sr, plants and soils are acceptable alternatives. It is, however, important to consider that plants of different rooting depths take up varying 87Sr/86Sr values depending on the exchangeable soil fraction (the fraction of topsoil sample that is extractable when leached using an ammonium nitrate solution) [51,90,91]. To compensate for this, the sampling strategy of this study targeted plants of varying root depths (shallow, medium, and deep) at each sampling location to ensure that any intra-site 87Sr/86Sr variability was captured. Shallow roots are defined as slender, branched, fibrous or creeping roots that grow close to the surface. Medium roots include larger plants whose roots penetrate one to two meters below the surface soil. Deep-rooted plants have tap roots that consist of a primary root that penetrates deep into the soil at depths greater than three meters. Using the RF-based methodology allowed us to explain the main environmental influences on the 87Sr/86Sr isoscape and we then validated the use of for provenancing using cow milk samples collected from farms across Aotearoa.
Materials and methods
Bioavailable 87Sr/86Sr sample distribution and descriptive statistics
Plant sample collection sites were chosen based on a strict set of inclusion criteria to circumvent potential anthropogenic and natural Sr contaminants. These criteria avoided farms, pastures, drainage ditches, and other human-made earthworks, and being restricted to public lands with road-access. Riparian areas, land in flood areas and along watercourses, were avoided to avoid any mixing of exogenous Sr sources . Also, the topographic relief and elevation of the regions limited where and when samples could be taken. These factors greatly limited the distribution of the samples (Fig 1). Potential sample stops were planned from an ArcGIS shapefile layer with a highway road feature class overlaid on a geological map to mark where there was a geological change within each area along the route. Leaves were collected from plants of each root-type (shallow, medium, and deep) located within a 5 m diameter of one another at each sample site (when available) to capture a higher resolution of 87Sr/86Sr variation for the single locality. Samples were stored in labeled, plastic sample bags and dehydrated later the same day in a clean SunBeam food dehydrator before being stored in new clean and labeled, plastic sample bags. Plant type, root-depths, and other metadata are provided in the Supplementary Information (S3 File).
Fig 1 developed in ArcGIS Pro and includes regional boundaries and coastlines feature layers (sourced from Natural Earth) and hillshade feature layer (created using GNS DEM 8m shapefile). Map projected to the New Zealand Transverse Mercator (NZTM 2000).
Topsoil samples (0–20 cm) were collected by GNS Science in the Nelson, Otago, and Southland regions (Fig 1). Samples were collected by hand auger and the sub-2-mm portion was retained after sieving and drying at 40°C [49,93]. Additional soil and plant 87Sr/86Sr data were provided by PWH and KFA for the Christchurch, Bay of Plenty, Auckland, and Northland regions. To date, very few modern local human and animal bioavailable 87Sr/86Sr data have been published for Aotearoa and only eight (two sheep and six humans) were available to include in this study (Fig 1).
Multiple analytical methods were utilized to analyze the plant and soil samples and generate the 87Sr/86Sr data. GNS topsoil (n = 71) and collected plant samples (n = 185) were prepared under the supervision of the lab technician in the Centre for Trace Element Analysis, University of Otago, Dunedin. GNS topsoil subsamples of 1g were leached in 2.5 mL of 1 M ammonium nitrate (NH4NO3) solution and agitated overnight to extract the bioavailable strontium fraction [53,81,92,94]. Then, the topsoil samples underwent Sr separation column procedures as described previously [76,95,96], while the collected plant samples utilized an automated ion-exchange chromatography method using a 3-ml column (ESI, Part number CF-MC-SrCa-3000) filled with DGA (diglycolamide) resin (TrisKem International, Bruz, France) as described by Wijenayake .
Strontium isotope measurement of the collected plant and GNS topsoil samples was conducted at the Centre for Trace Element Analysis, University of Otago using a Nu Plasma-HR MC ICP-MS (Nu Instruments Ltd., UK) following previously reported protocols [76,95–99]. Repeated measurement of the NIST SRM 987 strontium isotope reference material (sourced from the National Institute of Standards and Technology (NIST), USA) and the HPS in-house strontium isotope standard, that bracketed every six samples, were used to monitor the accuracy and external precision of the measurements. We obtained average values of 0.71025 ± 0.00002 (2 SD, n = 70) for NIST SRM 987 and 0.70761 ± 0.00009 (2 SD, n = 58) for HPS in very good agreement with previously reported values. Any instrumental mass fractionation present was corrected for using the exponential mass fractionation lay by normalization to 86Sr/88Sr = 0.1194. Procedural blanks were run with each batch of 6 samples and all yielded negligible Sr levels of < 250 pg. The additional soil and plant samples (n = 126) utilized Sr separation column procedures as described by Pin & Bassin . Sr isotope measurement of the soil samples from  was conducted using a Nu Plasma-HR MC ICP-MS (Nu Instruments Ltd., UK) at the Centre for Trace Element Analysis, University of Otago, Dunedin (standard = NIST SRM 987, n = 19, 87Sr/86Sr 0.710274 ±0.000023 (2 SD)); and the additional soil and plant samples  measured using Isotopx Phoenix TIMS (thermal ionization mass spectrometry) at the University of Adelaide (NIST SRM 987, n = 7, 87Sr/86Sr = 0.710245 ± 0.000008 (2 SD).
Further details regarding preparation and analysis for all plant and soil samples are available in the Supplementary Information (S1.1–5 in S1 File). Altogether, 414 bioavailable 87Sr/86Sr data (Fig 1) were used to construct the strontium isoscape, specifically comprising data from 314 plants (182 unique locations), 92 topsoils (84 unique locations), and eight mammals (two unique locations).
Bioavailable 87Sr/86Sr isoscape
This study follows the methodology established by Bataille and colleagues [51,93] using a RF model framework. Model construction used bioavailable, geo-referenced 87Sr/86Sr data gathered for Aotearoa, many geomatic auxiliary variables in the form of global rasters (gridded matrix of cells) used as covariates, and an existing process-based bedrock 87Sr/86Sr model [87,88] to produce the final predicted 87Sr/86Sr isoscape model. The RF R-script provided by Bataille et al. [51,88] optimizes the regression model with the root mean square error (RMSE) and uses five 10-fold repeated cross-validations. Additionally, the R-script includes using the Variable Selection Under Random Forest (VSURF) package  to identify relevant and highly predictive variables [51,87–89]. The relationships between the variables selected by the VSURF function and the bioavailable 87Sr/86Sr variability were assessed using variable importance purity measure and partial dependence plots . The auxiliary variables were obtained from various sources summarized in S2 Table 1 in S2 File and represent geological, climatic, and environmental variables that may influence bioavailable 87Sr/86Sr variability. Most variables were available as global rasters that were trimmed to an Aotearoa extent and projected to the New Zealand Transverse Mercator 2000 coordinate system.
To assess the uncertainty of the isoscape, we used quantile RF regression model to generate quartile-1 and quartile-3 models using the log-transformed 87Sr/86Sr values. We use a log-transformation because the 87Sr/86Sr data have a positively skewed distribution. Then, the quartile-1 model was subtracted from the quartile-3 model (i.e., Q3-Q1) to generate a final interquartile range raster at a resolution of 1 km.
Cow milk samples
Dairy farms throughout the country that contributed cow milk samples agreed to a controlled feeding regime where all cattle were expected to be pasture-fed on-site and were not provided with supplementary feed options . Therefore, the assumption was that the 87Sr/86Sr values of the cow milk would reflect the underlying geology and local atmospheric conditions with no contamination from exogenous 87Sr/86Sr sources. Details regarding the isotopic preparation and analytical methods for the cow milk samples are provided in the Supplementary Information (S1.5 in S1 File).
To assess the validity of the bioavailable 87Sr/86Sr isoscape developed, we predict the region-of-origin for cow milk samples and then assess the accuracy of that assignment relative to their actual origin. Firstly, origin predictions are computed using assignR , which operates in a semi-parametric Bayesian framework to calculate the posterior probability of the sample belonging to each cell within the isoscape raster. Statistically, assignR uses a maximum likelihood assignment model and employs Bayes Theorem to calculate the probability that a sample originates from a geographic location given the isotopic signature of the sample . The statistical methodology of the assignR code  is adapted from Wunder  and Vander Zanden et al. . To facilitate visualization, we created maps displaying the top 33% by area of the predicted surface for each sample (SI4) with the top 33% areas coded as 1 and other areas coded as 0. Once the top 33% probability maps are produced, they are brought into ArcGIS Pro. Then, we calculated the accuracy of the nearest predicted geographic origin (part of the top 33% probability) for each sample in terms of how close they were in km to the actual place of origin for the cow, using the Measure Distance tool in ArcGIS Pro. We also measured the distance from the nearest predicted cell for the top 20% and top 10% probability surfaces to assess the trade-offs between accuracy and precision for these arbitrary thresholds. We do not provide the maps for the top 20% and 10% predictions, but they can be provided upon request.
87Sr/86Sr isotopes and Aotearoa bioavailable 87Sr/86Sr isoscape
The bioavailable 87Sr/86Sr data for the plant and soil samples are available in the Supplementary Information (S3 File). The combined values for plants, soils, and local mammals demonstrate comparable distributions, with Quartile 1 = 0.70738 ± 0.00002, median = 0.70832 ± 0.00003, and Quartile 3 = 0.70897 ± 0.00108 (Fig 2). The animal substrate samples display a tighter range of 87Sr/86Sr values compared to the plants and topsoils, with soil displaying the largest variability in 87Sr/86Sr, but 90% of all data fall within values of 0.70500 and 0.71250 (Fig 2).
The animal substrate (n = 8) samples in pink display a tighter range of 87Sr/86Sr values compared to the plants in green (n = 314) and topsoils in blue (n = 92) samples that are all used to construct and calibrate the bioavailable 87Sr/86Sr isoscape model.
The final 87Sr/86Sr isoscape for Aotearoa uses the best performing RF model that considered all auxiliary variables (S2 Table 1 in S2 File) and 414 bioavailable 87Sr/86Sr values (S3 File). The RF regression model produces a map demonstrating the mean 87Sr/86Sr prediction (R2 = 0.53, RMSE = 0.00098) ranging from 0.70567 to 0.71118, for the entire country including the Chatham Islands (Fig 3A). The accompanying interquartile range raster (Fig 3B) demonstrates a standard error ranging from 0.0001 to 0.002 for the country.
Bioavailable 87Sr/86Sr isoscape (A) and interquartile range raster (B) for Aotearoa created using random forest regression. The bioavailable 87Sr/86Sr isoscape (A) (R2 = 0.53, RMSE = 0.00098) demonstrates the predicted 87Sr/86Sr values, ranging from 0.70567 to 0.71118, for the entire country including the Chatham Islands. Log-transformed 87Sr/86Sr values of the bioavailable data were used to construct quartile-1 and quartile-3 regression models that were then subtracted from one another (Q3-Q1) to create the final interquartile range raster (B) with values ranging from 0.0001 to 0.002. Figure developed in ArcGIS Pro with a coastlines feature layer (sourced from Natural Earth) and projected to NZTM 2000.
Assessment of Aotearoa 87Sr/86Sr model performance
The VSURF package selected 11 predictive variables (Fig 4A). These include dust and sea salt aerosol deposition (r.dust, r.ssa, r.ssaw), geological attributes (r.age, r.toprock, r.GNSagemax), temperature (r.mat), elevation (r.elevation), and soil characteristics (r.ph, r.clay, r.nitrogen). Initially, the VSURF selected 10 features, but when r.nitrogen was forcibly added to the RF function, the amount of variation explained increased by ~2%. Therefore, r.nitrogen was included in the final model regression. The dominant predictive variables are the rock type classification (r.toprock) and soil pH in H₂O solution (r.ph) based on their Percent Increased Mean Squared Error (%IncMSE) values (Fig 4A). The higher the %IncMSE value, the more important the predictor variable . The n-fold cross validation explains 53% of the variation in the bioavailable 87Sr/86Sr model, with an RMSE of 0.00098 over the dataset (Fig 4B). The uncertainty appears uniform across the prediction range (Fig 4B).
Plots demonstrate (A) model variable importance plot and (B) n-fold cross validation with best fit line in red.
Partial dependence plots are used to examine the association between the predictors and bioavailable 87Sr/86Sr (Fig 5). These illustrate that 87Sr/86Sr values increase with increasing soil pH (r.ph) and elevation (r.elevation). On the other hand, inverse relationships are apparent where 87Sr/86Sr values increase with decreasing mean annual temperature (r.mat), principal surface lithology type (r.toprock), rock age (r.age, r.GNSagemax), and clay content (r.clay). The “toprock” raster (Fig 5) consists of 67 different lithological classifications (detailed in S2 Table 2 in S2 File) where the first few categories have the highest 87Sr/86Sr values: 1 = floodplain alluvium; 2 = volcanic ashes older than Taupō pumice; 3 = Loess; 4 = Sandstone; 5 = Greywacke; etc. Because this variable is categorical, its correlation with 87Sr/86Sr is most likely coincidental. Both rock age variables, r.GNSagemax and r.age, (Fig 5) demonstrate a threshold-based relationship with bioavailable 87Sr/86Sr values showing little influence of age above 300 million years. This is consistent with previous findings demonstrating that the age of rocks is a major contributor to bioavailable 87Sr/86Sr values [73,87,88]. The remaining predictive variables, dust and sea salt aerosol deposition (r.dust, r.ssa, r.ssaw) and nitrogen content (r.nitrogen), display non-linear relationships with the 87Sr/86Sr values.
Units of measurement are provided on the x-axis for each variable. For r.toprock, the x-axis represents the numbered categories of rock type (S2 Table 2 in S2 File and S2 Fig 1 in S2 File) obtained from the New Zealand Land Resource Information System online portal. Refer to the Supplementary Information, S2 Table 1 in S2 File, for descriptions of all variables used in the RF model and their sources.
Model validation using cow milk
Generally, we found that cow milk with 87Sr/86Sr values near 0.70800 were difficult to assign because most of the isoscape values across Aotearoa fall in a narrow band between the range of 0.70750 and 0.70930. When a sample is within this interval, the probability surfaces show broad regions with high probabilities of origin corresponding to low assignment precision. Conversely, the model performs well when predicting origin for samples that are below 0.70750 or above 0.70930 because these values are much less common on the Aotearoa isoscape allowing for better precision. We produced prediction maps showing the top 33% of isoscape raster cells with the highest posterior probability for each cow milk sample (n = 33) available in the Supplementary Information (S4 File). The cow milk 87Sr/86Sr and the predicted 87Sr/86Sr from the isoscape show a good correlation (R2 = 0.52) when compared with one another (Fig 6). When error bars are included, all cow samples fall within the 95% confidence interval of the linear regression line on the actual versus predicted bioavailable 87Sr/86Sr value plot (Fig 6), except sample Cow 7.
Error bar values were extracted from the uncertainty raster for each cow milk sample location. The red line represents the line of best fit and the shaded region represents the 95% confidence interval (CI). The black line shows the 1:1 line between predicted and observed 87Sr/86Sr values. Sample points are color-coded by region (regions depicted in Fig 1). All cow milk samples fall within the 95% CI, except sample Cow 7 labeled on plot.
The maximum likelihood assignment model was quite accurate and all but one milk sample had high probability cells (top 33%) located within proximity (average of 7.05 kilometers) to the actual dairy farm of origin for each represented region (Tables 1 and 2). We also measured the distance between the nearest probability cell and the dairy farm of actual origin for the top 20% and top 10% probability surfaces to assess the trade-off between precision and accuracy when different probability thresholds are used (Tables 1 and 2). While the accuracy of most origin predictions is high using the top 33% probability threshold, the precision is lacking for most samples. Conversely, the top 20% and top 10% probability thresholds produce precise origin predictions (i.e., they predict fewer regions of potential origin), but the accuracy (distance to known origin) decreases (Tables 1 and 2).
Cow milk samples originating from the Canterbury region produce the most accurate predictions (Tables 1 and 2). For all Canterbury cow milk samples, the prediction maps have high probability cells within 0 to 1.56 km of the actual location. The Waikato region milk samples produce highly and moderately accurate predictions with a 6.97 km average distance away from the nearest predicted cell (Table 2). Samples Cow 4, 5, 8, and 9 are highly accurate with predictions 0–0.64 km from their actual locations. The remaining samples in the Waikato region have predictions 6.42 to 16.84 km away from their known locations (Table 2). For the Southland region, all cow milk samples originated near the city of Invercargill and display variable prediction accuracies. Samples Cow 19–23 and 26–27 are highly accurate (less than 5 km away); Cow 24, 25, and 28 are moderately accurate (6 to 20 km away); and Cow 29 is the least accurate being 86.24 km away from the nearest predicted cell (Table 2).
Bioavailable strontium isoscape for Aotearoa
Understanding the geology and climate allows improved modeling of Sr isotope cycles from the geosphere into the biosphere for provenancing materials. Accordingly, the covariates used in the RF model here account for geological and atmospheric conditions in Aotearoa (S2 Table 1 in S2 File). Seven of the eleven highest performing covariates selected by the VSURF package consisted of geological variables (Figs 4 and 5), while the remaining four were atmospheric variables showing the contribution of dust and sea salt aerosols and mean annual temperature to bioavailable 87Sr/86Sr values.
Of the geological variables, there are several features that can explain the observed range of 87Sr/86Sr values (0.70560–0.71120 ± 0.002, Fig 3) across Aotearoa. These values reflect the variable composition of bedrock for the country, which has a rich geological history  that has resulted in areas with distinct mineralogy and rock ages, both factors which strongly affect strontium and consequently bioavailable 87Sr/86Sr . Aotearoa has a compositionally diverse basement geology with formation ages between c. 500–100 million years ago (Ma) (Fig 7). The modern-day distribution of geology reflects terrane accretion and c. 23 Ma of tectonism along the strike-slip system, the Alpine Fault, which bisects the country [108,110,111]. The Alpine Fault juxtaposes terranes which were once contiguous for hundreds of kilometers, resulting in rocks of terrane affinity being present at either end of the country. Fig 7 shows the clear relationship between the geological terranes, Alpine Fault, and bioavailable 87Sr/86Sr, especially in the South Island. In particular, the model predicts the lowest bioavailable 87Sr/86Sr values (0.70570–0.70750) for the northwestern and southwestern portions of the South Island and Rakiura (Stewart Island). These rocks share a similar geological origin, belonging to the Paleozoic Buller and Takaka terranes of the Indo-Australian plate. The rocks of these terranes are broadly comprised of metasediments and intruding plutonic igneous rocks, called batholiths, that formed along the previously active margin of Gondwana between 500–100 Ma [112,113]. Many of these rocks are of igneous origin or derivation, and therefore tend to have very low 87Sr/86Sr values despite their older age [114,115], though some plutonic igneous rocks are known to have higher 87Sr/86Sr values owing to their derivation from melting ancient crust .
Diagrams showing simplified (A) basement terrane geology and (B) bioavailable 87Sr/86Sr isoscape highlighting regions with low, intermediate, and high 87Sr/86Sr values. The geology in (A) follows Edbrooke et al. . The Alpine Fault is shown as a solid black line; the division between Eastern Province and Western Province basement rocks is shown by a black dashed line; and the Taupo Volcanic Zone (TVZ) outline is shown by a dotted red line. Low, intermediate, and high Sr value ranges for simplified isoscape (B) were determined using the “Equal Intervals” symbology function in ArcGIS Pro. Black denotes regions of low 87Sr/86Sr values (0.70570–0.70750), dark grey denotes intermediate 87Sr/86Sr values (0.70750–0.70930), and light grey denotes regions with high 87Sr/86Sr values (0.70930–0.71120). Figure developed in ArcGIS Pro using a coastlines feature layer (sourced from Natural Earth) and projected to NZTM 2000.
The lowest predicted bioavailable 87Sr/86Sr isoscape values in the North Island also partially correspond with basement geology–some areas of low predicted 87Sr/86Sr isoscape values correspond with the Dun Mountain-Maitai and the Murihiku terranes (Fig 7). The Dun Mountain-Maitai Terrane is composed of ultramafic rocks overlaid by plutonic and volcanic sequences, that are covered by sedimentary deposits, while the Murihiku Terrane consists largely of sedimentary and volcanic materials . However, in the central North Island bioavailable strontium ratios also appear to be heavily influenced by Quaternary volcanic deposits relating to eruptives from the Taupō Volcanic Zone (S2 Fig 1 in S2 File). These eruptive rocks include ignimbrites and volcanic ashes that date volcanism from around 1.6 Ma to present day . Therefore, these deposits and the ashes that predate them are young and are expected to display low 87Sr/86Sr values. The area immediately surrounding Mount Taranaki in the westernmost region of the central North Island (Fig 7) has been active since c. 130 thousand years ago (Ka)  and is also comprised of volcanic rocks that display the low 87Sr/86Sr values characteristic of younger igneous formations . The partial dependence plots (Fig 5) demonstrate this relationship between 87Sr/86Sr values and age as well.
Conversely, the highest 87Sr/86Sr values (0.70930–0.71120) of Aotearoa are predicted for areas comprised of the Mesozoic Caples, Rakaia, and Pahau terranes which stretch from the lower South Island along the eastern portion of the North Island (Fig 7). These terranes have mudstone/sandstone protoliths that are variably metamorphosed, but generally increase in metamorphic grade the closer they are to the Alpine fault [121,122]. The Caples Terrane consists of volcaniclastic feldspathic metasedimentary rocks . The Rakaia and Pahau terranes contain variably metamorphosed quartzofeldspathic sedimentary rocks . These rocks are overlain by aerially expansive alluvium and loess deposits in the central east portion of the South Island, around the Canterbury Plains (Fig 7 and S2 Fig 1 in S2 File). The elevated 87Sr/86Sr values for these regions (Fig 7) most likely result from the more felsic (rich in silica) content of the rocks .
The juxtaposition of differing geologies along the Alpine Fault also has clear effects on the distribution of bioavailable 87Sr/86Sr values in Aotearoa. This fault is primarily a strike-slip system, however uplift on the southeastern side of the fault has resulted in the formation of the Southern Alps. This mountain chain runs along the Alpine Fault and marks the distinction between high and intermediate 87Sr/86Sr values in the model (Fig 7). The elevation partial dependence plot (Fig 5) shows that 87Sr/86Sr values increase with increasing elevation. In addition, erosional forces associated with uplifted areas have generated massive amounts of gravel and dust that are older and have high 87Sr/86Sr values. This gravel and dust have been transported by glaciers, glacier-fed streams, and prevailing westerly winds into the Canterbury plains to the east of the Alps [48,108,125], and therefore affecting 87Sr/86Sr values in this region. Koffman et a.  note that the glacial activity and associated glacial outwash expanded the Canterbury Plains by approximately 30,000 km2 by depositing sediments with 87Sr/86Sr values ranging from 0.70950 to 0.71650. Glaciers to the west of the Southern Alps did not create similar outwash plains because they terminated near the edge of the continental shelf creating a distinction between high 87Sr/86Sr values to the East and intermediate 87Sr/86Sr values (0.70750–0.70930) in the West .
Although geological 87Sr/86Sr affects bioavailable 87Sr/86Sr, atmospheric variables clearly interact with geology to produce a complex isoscape. Despite its small size, Aotearoa’s climate varies vastly with a warm subtropical climate in the Northland region, a cool temperate climate throughout most of the country, and alpine weather in the mountainous West Coast and Southland regions of the South Island . The mean annual temperature plot (Fig 5) demonstrates that temperature has a significant effect on bioavailable 87Sr/86Sr in Aotearoa where higher 87Sr/86Sr values occur in areas with lower temperatures. This pattern correlates with geological and latitudinal differences between the North and South Islands. The former experiences warmer weather and has more volcanic and low-Sr lithologies (Fig 7), while the latter experiences colder weather and the highest 87Sr/86Sr values occur along the Southern Alps and the Canterbury plains. Precipitation is another atmospheric variable that plays a key role in bioavailable 87Sr/86Sr distributions. The Southern Alps experience some of the highest recordings for mean annual precipitation in Aotearoa, measuring up to 6,000 mm/year, while the land to their east records the lowest rainfall at 250–1,000 mm/year [125,126]. Increased precipitation on this coastal setting probably facilitates a high rate of marine aerosol deposition that is characterized by lower 87Sr/86Sr signatures, as well as historically high deposition of volcanic ash from surrounding volcanic centers particularly in the northwestern South Island and the North Island . This is reflected in the partial dependence plot for the sea salt aerosol deposition where higher rates of deposition are associated with lower 87Sr/86Sr values (Fig 5). However, the lack of bioavailable 87Sr/86Sr samples from these regions does limit the accuracy of the isoscape here (Fig 7). No cow milk samples were obtained from these regions.
Testing the bioavailable isoscape using cow milk samples
Using the first bioavailable 87Sr/86Sr isoscape developed here for Aotearoa, this study sought to validate the use of strontium isoscapes for provenancing research, such as food product verification, by predicting the origin for cow milk sampled from around the country. Fig 6 illustrates that all cow milk samples, with one exception (Cow 7), fall within the 95% confidence interval of the line of best fit. This indicates that their measured 87Sr/86Sr values reflect the predicted values obtained from the bioavailable 87Sr/86Sr isoscape, illustrating that the bioavailable 87Sr/86Sr isoscape performs well and has the potential to be utilized in provenancing applications.
The relatively high regression residual for sample Cow 7 might be due to the addition of non-local feed to the cow’s diet including health supplements, supplementary feed from other Aotearoa regions including ryegrasses (with high calcium content), imported palm kernel expeller (PKE), or even transportation of the cow between regions in Aotearoa that could introduce foreign 87Sr/86Sr values. However, dairy farms that contributed cow milk samples  had enforced a controlled feeding regime where all cattle were expected to be pasture-fed on-site and were not provided with supplementary feed options. In Aotearoa, livestock and raw cow’s milk are commonly transported inter-regionally. Prior to going through the pasteurization and production processes, and depending on processing plant availability or maintenance, milk may be transported some distance, between the west and east coasts, particularly in the South Island. The outlier sample, Cow 7, may therefore represent a cow that 1) was not kept on a strict diet or 2) had been transported from another region in Aotearoa.
The accuracy and precision of predictions was measured as the distance from their known place of origin to the nearest predicted area of potential origin using different probability quantile thresholds (10%, 20%, and 33%). As expected, the top 33% threshold produced the most accurate (i.e., closest to known origin) predictions, but those predictions covered more potential areas and were less precise (Tables 1 and 2 and S4 File). On the other hand, the top 20% and top 10% thresholds provided more constrained potential region-of-origin predictions, making them more precise, but their accuracy decreased (Tables 1 and 2). Specifically, the average distance away from the place of known origin was 7.05 km for the top 33% threshold, 16.99 km for the top 20%, and 30.73 km for the top 10% threshold (Table 2). This showed that there was a trade-off between precision and accuracy for prediction outputs. In the Supplementary Information (S4 File), we provide prediction maps of the top 33% probability quantile for each cow milk sample.
In total, 73% (24 out of 33) of milk samples fell into the intermediate 87Sr/86Sr range with values between 0.70750 and 0.70930 (Figs 6 and 7). Milk that fell into this range originated from the Waikato, Canterbury, Southland, Northland, Manawatu-Wanganui, and Nelson regions (Figs 1 and 6). As indicated by the isoscape, there is high redundancy for bioavailable 87Sr/86Sr across Aotearoa. When samples had 87Sr/86Sr values that fell outside of this intermediate range, region-of-origin predictions were more likely to be very accurate and precise. For example, samples Cow 4 and Cow 5 had the lowest 87Sr/86Sr values (< 0.70750) and produced highly accurate and precise predictions for all probability quantile thresholds (Table 1, Fig 6 and S4 File). Furthermore, samples Cow 11–14, 16, 17, and 30 had 87Sr/86Sr values greater than 0.70930 and all predicted regions of potential origin less than 3 km away from their known place of origin for the top 33% and 20% thresholds (Table 1, Fig 6, and S4 File). When 87Sr/86Sr values did not fall into the low or high ranges, the accuracy of origin predictions decreased, meaning that the distance away from the known place of origin tended to be larger because the assignment model predicted many potential regions of origin with similar bioavailable 87Sr/86Sr values. In these cases, utilizing a second isotope system, such as δ2H and δ18O, may help to constrain the predicted region-of-origin [31,37].
In terms of accuracy, Canterbury region cow milk samples (n = 8) produced the most accurate origin predictions across all probability thresholds with the smallest average distance between the known origin and the nearest predicted cell (Table 2 and S4 File). This strongly implies that the Canterbury cows have a local diet and do not have foreign feed introducing exogenous 87Sr/86Sr values. Conversely, Southland region milk samples (n = 11) produced the least accurate origin predictions across all thresholds (Tables 1 and 2). Fig 6 shows that the majority of the Southland cow milk samples have observed values below the 1:1 black line, implying that their actual 87Sr/86Sr values are higher than the predicted 87Sr/86Sr isoscape values. As mentioned previously, much of the South Island was glaciated about ~26,000 to ~18,000 BP and glacially fed rivers, specifically the Clutha and Mataura rivers and their tributaries, carried sediment from the Wakatipu Valley in the far western portion of the Otago region into the Southland and eastern Otago regions . The fluvial transportation of sediments from the Wakatipu Valley may be a major contributor to the elevated 87Sr/86Sr values of the Southland cow milk samples. Furthermore, the low accuracy may result from the increased geologic diversity in Southland versus Canterbury, where the former is comprised of several terranes, ranging from mafic to felsic, and the former is dominated by the Rakaia Terrane (Fig 7).
Even though the Southland cow milk samples fell below the 1:1 line (Fig 6), many of the samples (Cow 19–23, 26, and 27) produced accurate predictions less than 5 km away from their known origin and three other Southland samples (Cow 24, 25, and 28) were only 13 to 19 km off their targets using the top 33% threshold (Table 1). One sample, Cow 29, had the highest 87Sr/86Sr value of the Southland group and was the least accurate where the nearest predicted cell was 86–123 km away from the cow’s known place of origin (Table 1). Since the other milk samples from the Southland region produce accurate to moderately accurate predictions, we do not suspect that the elevated 87Sr/86Sr value of sample Cow 29 was due to its proximity to the coast (i.e., sea salt or dust aerosol deposition) or the introduction of foreign feed. Instead, it is likely that Cow 29 had been transported down to the Southland region from the Canterbury region which is more consistent with its elevated 87Sr/86Sr values. Milk samples were obtained through farms on a voluntary basis, so determining whether this sample had actually been transported from Canterbury was not possible.
Further applications for provenancing materials in Aotearoa
Apart from food science applications, the bioavailable 87Sr/86Sr isoscape can assist with identifying the origin of new pests, unidentified materials, illegal agricultural products, and origin mislabeling. Specifically, 87Sr/86Sr can help biosecurity organizations within Aotearoa determine if a pest encountered post-border represents a newly introduced pest or an established population bearing local 87Sr/86Sr values [1,99,100,127,128]. For the key pastoral industry, Ferguson et al.  estimated that invasive invertebrates alone cost the sector between $1.7 to $2.3 billion NZD annually. New pests and diseases are introduced to Aotearoa through shipping and air traffic pathways [128–131]. One of the highest risk pests for Aotearoa is the exotic brown marmorated stink bug (Halymorpha halys) because it feeds on over 300 species of plants and infestation can ruin entire crops . Recent attempts to define the origins of this invasive stink bug used δ2H and δ18O isotopes from their wings to determine whether a bug detected post-border represented a recently introduced foreign bug or an established population . Holder et al.  concluded that the distinct δ2H and δ18O isotope signature of their wings suggested a cooler climate origin, supporting evidence that the specimen was not from a locally breeding, southern hemisphere summer population. Currently, there are no rapid response tools to assess the provenance of these pests when they are introduced which limits the ability to develop effective measures to prevent their arrival in Aotearoa. This leaves agencies poorly informed as to the actual risk, with difficult decisions about embarking on expensive responses that may otherwise have to assume the presence of a locally established population.
Recently, Murphy et al.  developed a new mass spectrometric method to analyze 87Sr/86Sr in low-Sr samples, such as a single insect, in less time than traditional methods. This new analytical method enables this technology to be used within the limits for biosecurity, commanding very short time frames and availability of very little biological material. Now, coupled with this 87Sr/86Sr isoscape, well supported decisions based on probabilities can be made as to the likely Aotearoa or offshore origin of foreign pests, which is key to determine the most efficient response to prevent their establishment . Prior to this point this had not been possible, anywhere. The identification and removal of foreign pests protects crops from foreign diseases and safeguards the future of endemic flora and fauna species throughout the country.
This 87Sr/86Sr-focused approach can be applied globally to determine the local or foreign nature of pests on borders and ports of entry. However, this would require that areas of interest construct bioavailable 87Sr/86Sr isoscapes to compare their samples against. The Aotearoa 87Sr/86Sr isoscape produced by this study could be used alongside other lines of evidence (e.g., isotopic systems, chemical fingerprints), and strontium baseline models from other regions of the world to test whether “Made in New Zealand” food products display expected 87Sr/86Sr signatures or if they are fraudulent products.
Furthermore, law enforcement agencies could use the isoscape to predict the region of ‘growing areas’ or determine country of origin using bioavailable 87Sr/86Sr from confiscated illicit drugs, like marijuana or heroin [132,133]. Additionally, although Aotearoa does not encounter many unidentified human remains, if they were recovered and traditional identification methods failed to produce a positive identification, 87Sr/86Sr values from the unidentified person’s teeth could help predict their region or country of origin [19,25,26,37,65,74,77,134,135].
The method of provenancing used in this study is useful because it provides a visual representation of the predicted probability surface that allows for ease of communication and comprehension. Though the predicted maps need to be generated by a specialist, once created, a non-specialist can interpret the prediction map and make decisions about where an unknown material may originate from within Aotearoa using their own criteria. The main limitation with applying the isoscape to provenancing investigations is that the predictions highlight many potential regions-of-origin that have similar 87Sr/86Sr values when using the top 33% threshold. The threshold can be set to 20% or 10% but, as demonstrated here, these thresholds reduce the accuracy of origin predictions. Instead, predicted areas could be further constrained by combining the predictive strength of 87Sr/86Sr with other isotope systems, such as δ2H and δ18O [37,74,89]. Isoscapes for δ2H and δ18O already exist for Aotearoa and can easily be combined in a dual-isotope approach to predict the region-of-origin for unknown materials using the built-in assignment features of the assignR package [104,136–138].
As people, animals, and materials are transported across increasingly large distances in a globalized world, issues with biosecurity and food security are rising. The global pandemic of COVID-19 has highlighted the difficulties we face in tracing and controlling the origins of animals and products that can transport viral loads. In Aotearoa, there are many endemic species and a strong local agricultural industry that must be protected from biosecurity threats. There is an urgent need to have tools which enable the provenancing of pests and agricultural products arriving in and being transported around the islands of Aotearoa, as well as confirm origins of products advertised as ‘New Zealand made’ that enter into overseas markets to protect valuable commodities from fraud. Isotopes are ubiquitous markers of provenance that are increasingly used to trace the origin of food or animals. In this study, we introduced the first bioavailable 87Sr/86Sr isoscape for Aotearoa and demonstrated how the isoscape can be used to certify the origin of agricultural products. We improved upon existing methodology to develop a bioavailable 87Sr/86Sr isoscape using the best available geospatial datasets to tune the regional isoscape. As anticipated, the primary drivers of bioavailable 87Sr/86Sr variability are the underlying geology, soil pH, and aeolian (dust and sea salt) deposits. We then tested and proved that there is potential for utilizing 87Sr/86Sr isotopes to determine the origin of cow milk and other agricultural products in Aotearoa.
Currently, there is little to no geo-referenced bioavailable 87Sr/86Sr data derived from plants, animals, or soil in Aotearoa, beyond this study. As more data become available, the bioavailable 87Sr/86Sr isoscape model can be recalibrated and further improved, though we do acknowledge that sampling to fill current geographic gaps in the model is a costly undertaking. With the availability of this baseline bioavailable 87Sr/86Sr isoscape, our hope is to promote further research using 87Sr/86Sr isotopes in Aotearoa and abroad.
Current provenancing methods in Aotearoa rely heavily on trace elements and light stable isotopes (δ2H, δ15N, δ18O, δ13C) to predict region-of-origin. The information provided by light isotopes and trace elements are useful for excluding potential regions of origin from consideration, but predictions can be further constrained if combined with other geologically derived isotope systems, like 87Sr/86Sr [37,89]. Future provenancing projects in Aotearoa should implement a multi-isotope approach that uses 87Sr/86Sr alongside other light stable isotopes to produce both accurate and precise region-of-origin predictions.
S1 File. Sample preparation and analysis.
S1.1. Aotearoa Plant Sample Preparation. S1.2. Aotearoa Topsoil Sample Preparation. S1.3. Collected Plant and Topsoil MC-ICP-MS Analysis. S1.4. Additional Plant and Topsoil Preparation and Analysis. S1.5. Cow Milk Sample Preparation and Analysis.
S2 File. Random forest model and auxiliary variables.
S2 Table 1. Geological and climatic variables used in random forest regression. D = discrete variable; C = continuous variable. S2 Table 2. “Toprock” Category (#1–67) Descriptions. S2 Fig 1. The top performing covariate “Toprock” selected by VSURF package. The 67 toprock principal surface lithology types have been simplified but all types are detailed in S2 Table 2. Data sourced from LRIS (https://lris.scinfo.org.nz/layer/48065-nzlri-rock/; 2010) [S5.18] and GNS (fault lines shapefile) [S5.20]. The “toprock” shapefile does not include data for Rakiura (Stewart Island) or Wharekauri (Chatham Islands).
S3 File. Bioavailable 87Sr/86Sr Isoscape–Color Version.
S3 Fig 1. Bioavailable 87Sr/86Sr isoscape (color version). The highest 87Sr/86Sr ratios are depicted in blue and the lowest are shown in purple. The bioavailable 87Sr/86Sr isoscape (R2 = 0.53, RMSE = 0.00098) demonstrates the predicted 87Sr/86Sr values, ranging from 0.70567 to 0.71118, for the entire country including the Chatham Islands. Figure developed in ArcGIS Pro using a coastlines feature layer (sourced from Natural Earth) and projected to NZTM 2000.
S4 File. Aggregated 87Sr/86Sr data.
Excel = “S4_BioSrData”.
S5 File. Cow milk sample predictions.
PDF = “S5_CowMilkSamplePredictions”. Cow Milk Sample Predictions—Top 33% Posterior Probabilities. Regions highlighted in blue represent the top 33% of areas predicted as potential regions-of-origin using bioavailable 87Sr/86Sr where values “1” = Likely Origin and “2” = Unlikely Origin. The box in top left shows the area surrounding the cow milk sample’s actual place-of-origin marked by yellow circle. The hillshade layer was created using a GNS DEM 8m shapefile.
A special thanks to the Centre for Trace Elements team at the University of Otago for the countless hours of lab guidance and supervision. Additional thanks to AsureQuality for assistance collecting B3 samples and to David Murphy, Queensland University of Technology for managing analysis of the B3 samples. We acknowledge the SPATIAL (Spatio-temporal Isotope Analytics Lab, University of Utah) Short Course for providing the environment to design this work and the funds to carry it out to RTK.
- 1. Holder PW, Armstrong K, Van Hale R, Millet MA, Frew R, Clough TJ, et al. Isotopes and trace elements as natal origin markers of Helicoverpa armigera–an experimental model for biosecurity pests. PLOS ONE. 2014;9: e92384. pmid:24664236
- 2. Holder PW, Van Hale RJ, Frew R, George S, Armstrong KF. Natal origin of the invasive biosecurity pest, brown marmorated stink bug (Halyomorpha halys: Penatomidae), determined by dual‐element stable isotope‐ratio mass spectrometry. Pest Manag Sci. 2020;76: 1456–63. pmid:31659828
- 3. Hulme PE. Plant invasions in New Zealand: global lessons in prevention, eradication and control. Biol Invasions. 2020;22: 1539–1562.
- 4. Russell JC, Innes JG, Brown PH, Byrom AE. Predator-free New Zealand: conservation country. BioScience. 2015;65: 520–525. pmid:26955079
- 5. Veitch CR, Clout MN. Human dimensions in the management of invasive species in New Zealand. In: McNeely JA, editor. The great reshuffling: Human dimensions of invasive alien species. IUCN; 2001. pp. 63–71.
- 6. Cozzolino D, Cynkar WU, Shah N, Smith PA. Can spectroscopy geographically classify Sauvignon Blanc wines from Australia and New Zealand? Food Chem. 2011;126: 673–8.
- 7. Biosecurity Pests and diseases we want to keep out of New Zealand. Ministry of Primary Industries. 2017 [cited 3 September 2021]. Available from: https://www.mpi.govt.nz/biosecuritypests-and-diseases-we-want-to-keep-out-of-new-zealand/alphabetical-list-of-priority-pests-and-diseases/#Plants.
- 8. Quirke S, Ball R. Counterfeiting in the Primary Industry Sector and the threat to New Zealand’s economy. Nat Security J. 2019;1: 61–74.
- 9. The Estimates of Appropriations for the Government of New Zealand–Primary Sector. Treasury. 2020 [cited 8 June 2021]. Available from: https://www.treasury.govt.nz/sites/default/files/2020-05/est20-v9-intro.pdf.
- 10. Ehtesham E, Hayman AR, McComb KA, Van Hale R, Frew RD. Correlation of geographical location with stable isotope values of hydrogen and carbon of fatty acids from New Zealand milk and bulk milk powder. J Agric Food Chem. 2013;61: 8914–23. pmid:23957873
- 11. McLeod RJ, Prosser CG, Wakefield JW. Identification of goat milk powder by manufacturer using multiple chemical parameters. J Dairy Sci. 2016;99: 982–93. pmid:26709188
- 12. Angus NS, O’keeffe TJ, Stuart KR, Miskelly GM. Regional classification of New Zealand red wines using inductively‐coupled plasma‐mass spectrometry (ICP‐MS). Aust J Grape Wine Res. 2006;12: 170–6.
- 13. Versteegen R, Shatova O, Lind S, Linterman K. Testing for geographic origin of fetal bovine serum. BioProcess J. 2019;18: 1–8.
- 14. Oritain. Available from: https://www.oritain.com.
- 15. Crown Research Institutes GNS Science. Available from: https://www.gns.cri.nz.
- 16. Bio-Protection Research Centre. Lincoln University. Available from: https://www.research.lincoln.ac.nz/our-research/faculties-research-centres/bio-protection-research-centre.
- 17. Centre for Trace Element Analysis. University of Otago. Available from: https://www.otago.ac.nz/geology/facilities/geochemical-analysis/trace-element-analysis.
- 18. Baffi C, Trincherini PR. Food traceability using the 87Sr/86Sr isotopic ratio mass spectrometry. Euro Food Res Technol. 2016;242: 1411–1439.
- 19. Chesson LA, Barnette JE, Bowen GJ, Brooks JR, Casale JF, Cerling TE, et al. Applying the principles of isotope analysis in plant and animal ecology to forensic science in the Americas. Oecologia 2018;187: 1077–94. pmid:29955984
- 20. Granato D, Putnik P, Kovačević DB, Santos JS, Calado V, Rocha RS, et al. Trends in chemometrics: Food authentication, microbiology, and effects of processing. Compr Rev Food Sci Food Saf. 2018;17: 663–677. pmid:33350122
- 21. Liu Z, Zhang W, Zhang Y, Chen T, Shao S, Zhou L, et al. Assuring food safety and traceability of polished rice from different production regions in China and Southeast Asia using chemometric models. Food Control. 2019;99: 1–10.
- 22. West JB, Bowen GJ, Cerling TE, Ehleringer JR. Stable isotopes as one of nature’s ecological recorders. Trends Ecol Evol. 2006;21: 408–414. pmid:16753238
- 23. Bentley RA, Knipper C. Geographical patterns in biologically available strontium, carbon and oxygen isotope signatures in prehistoric Southwest Germany. Archaeometry. 2005;47: 629–644.
- 24. Ehleringer JR, Bowen GJ, Chesson LA, West AG, Podlesak DW, Cerling TE. Hydrogen and oxygen isotope ratios in human hair are related to geography. P Nat Acad Sci. 2008;105: 2788–2793. pmid:18299562
- 25. Cerling TE, Barnette JE, Bowen GJ, Chesson LA, Ehleringer JR, Remien CH, et al. Forensic stable isotope biogeochemistry. Annu Rev Earth Planet Sci. 2016;44: 175–206.
- 26. Ehleringer JR, Thompson AH, Podlesak DW, Bowen GJ, Chesson LA, Cerling TE, et al. A framework for the incorporation of isotopes and isoscapes in geospatial forensic investigations. In: West JB, Bowen GJ, Dawson TE, Tu KP, editors. Isoscapes: Understanding Movement, Pattern, and Process on Earth through Isotope Mapping. Dordrecht: Springer; 2010. pp. 357–387. https://doi.org/10.1007/978-90-481-3354-3_17.
- 27. Chenery CA, Pashley V, Lamb AL, Sloane HJ, Evans JA. 2012. The oxygen isotope relationship between the phosphate and structural carbonate fractions of human bioapatite. Rapid Commun Mass Sp. 2012;26: 309–19. pmid:22223318
- 28. Daux V, Lécuyer C, Héran MA, Amiot R, Simon L, Fourel F, et al. Oxygen isotope fractionation between human phosphate and water revisited. J Hum Evol. 2008;55: 1138–47. pmid:18721999
- 29. Iacumin P, Bocherens H, Mariotti A, Longinelli A. Oxygen isotope analyses of co-existing carbonate and phosphate in biogenic apatite: a way to monitor diagenetic alteration of bone phosphate? Earth Planet Sci Lett. 1996;142: 1–6.
- 30. Benettin P, Nehemy MF, Cernusak LA, Kahmen A, McDonnell JJ. On the use of leaf water to determine plant water source: A proof of concept. Hydrol Process. 2021;35: e14073.
- 31. Magozzi S, Bataille CP, Hobson KA, Wunder MB, Howa JD, Contina A, et al. Calibration chain transformation improves the comparability of organic hydrogen and oxygen stable isotope data. Methods Ecol Evol. 2021;12: 732–747.
- 32. Terzer‐Wassmuth S, Wassenaar LI. Performance of low‐cost stainless‐steel beverage kegs for long‐term storage integrity and easy dispensing of water isotope (δ1⁸O, δ2H) reference materials. Rapid Commun Mass Sp. 2021;35: e9164. pmid:34270817
- 33. Aggarwal PK, Araguás-Araguás LJ, Groening M, Kulkarni KM, Kurttas T, Newman BD, et al. Global hydrological isotope data and data networks. In: West JB, Bowen GJ, Dawson TE, Tu KP, editors. Isoscapes: Understanding Movement, Pattern, and Process on Earth through Isotope Mapping. Dordrecht: Springer; 2010. pp. 33–50. https://doi.org/10.1007/978-90-481-3354-3_2
- 34. Bowen GJ, Cai Z, Fiorella RP, Putman AL. Isotopes in the water cycle: regional-to global-scale patterns and applications. Annu Rev Earth Planet Sci. 2019;47: 453–479.
- 35. Gat JR. Oxygen and hydrogen isotopes in the hydrologic cycle. Annu Rev Earth Planet Sci. 1996;24: 225–262.
- 36. Fry B. Stable Isotope Ecology. New York: Springer. 2006. https://doi.org/10.1007/0-387-33745-8
- 37. Bartelink EJ, Chesson LA. Recent applications of isotope analysis to forensic anthropology. Forensic Sci Res. 2019;4: 29–44. pmid:30915415
- 38. DeNiro MJ, Epstein S. Influence of diet on the distribution of carbon isotopes in animals. Geochim Cosmochim Acta. 1978;42: 495–506.
- 39. DeNiro MJ, Epstein S. Influence of diet on the distribution of nitrogen isotopes in animals. Geochim Cosmochim Acta. 1981;45: 341–51.
- 40. Lee‐Thorp JA. On isotopes and old bones. Archaeometry. 2008;50: 925–950.
- 41. Rogers KM, Turnbull RE, Martin AP, Baisden WT, Rattenbury MS. Stable isotopes reveal human influences on southern New Zealand soils. Appl Geochem. 2017;82: 15–24.
- 42. Yuan Y, Zhang W, Zhang Y, Liu Z, Shao S, Zhou L, et al. Differentiating organically farmed rice from conventional and green rice harvested from an experimental field trial using stable isotopes and multi-element chemometrics. J Agricul Food Chem. 2018;66: 2607–2615. pmid:29419296
- 43. Rogers KM. Nitrogen isotopes as a screening tool to determine the growing regimen of some organic and nonorganic supermarket produce from New Zealand. J Agricul Food Chem. 2008;56: 4078–4083. pmid:18489112
- 44. Bird MI, Crabtree SA, Haig J, Ulm S, Wurster CM. A global carbon and nitrogen isotope perspective on modern and ancient human diet. P Nat Acad Sci. 2021;118: e2024642118. pmid:33941703
- 45. Nardoto GB, Silva S, Kendall C, Ehleringer JR, Chesson LA, Ferraz ES, et al. Geographical patterns of human diet derived from stable‐isotope analysis of fingernails. Ame J Phys Anthropol. 2006;131: 137–46. pmid:16552735
- 46. Berry JE. Trace metal analysis by laser ablation-inductively coupled plasma-mass spectrometry and x-ray K-edge densitometry of forensic samples. Doctoral Thesis, Iowa State University. 2015. Available from: https://www.proquest.com/dissertations-theses/trace-metal-analysis-laser-ablation-inductively/docview/1724044945/se-2?accountid=14700.
- 47. Isotope Pye K. and trace element analysis of human teeth and bones for forensic purposes. Geol Soc Spec Publ. 2004;232: 215–236.
- 48. Koffman BG, Goldstein SL, Winckler G, Borunda A, Kaplan MR, Bolge L, et al. New Zealand as a source of mineral dust to the atmosphere and ocean. Quat Sci Rev. 2021;251: 106659.
- 49. Martin AP, Turnbull RE, Rattenbury MS, Cohen DR, Hoogewerff JA, Rogers KM, et al. The regional geochemical baseline soil survey of southern New Zealand: Design and initial interpretation. J Geochem Explor. 2016;167: 70–82.
- 50. Turnbull RE, Rogers KM, Martin AP, Rattenbury MS, Morgan R. Human impacts recorded in chemical and isotopic fingerprints of soils from Dunedin City, New Zealand. Sci Total Environ. 2019;673: 455–469. pmid:30991335
- 51. Bataille CP, Crowley BE, Wooller MJ, Bowen GJ. Advances in global bioavailable strontium isoscapes. Palaeogeogr Palaeoclimatol Palaeoecol. 2020;555: 109849.
- 52. Bentley RA. Strontium isotopes from the earth to the archaeological skeleton: a review. J Archaeol Method Theory. 2006;13: 135–187.
- 53. Capo RC, Stewart BW, Chadwick OA. Strontium isotopes as tracers of ecosystem processes: theory and methods. Geoderma 1998;82: 197–225.
- 54. Crowley BE, Miller JH, Bataille CP. Strontium isotopes (87Sr/86Sr) in terrestrial ecological and palaeoecological research: empirical efforts and recent advances in continental‐scale models. Biol Rev. 2015;92: 43–59. pmid:26392144
- 55. Graustein WC. 87Sr/86Sr Ratios Measure the Sources and Flow of Strontium in Terrestrial Ecosystems. In: Rundel PW, Ehleringer JR, Nagy KA, editors. Stable Isotopes in Ecological Research. Springer: New York; 1989. pp. 491–512. https://doi.org/10.1007/978-1-4612-3498-2_28 pmid:2669387
- 56. Holt E, Evans JE, Madgwick R. Strontium (87Sr/86Sr) mapping: A critical review of methods and approaches. Earth-Sci Rev. 2021;216: 103593.
- 57. Serna A, Prates L, Mange E, Salazar-García DC, Bataille CP. Implications for paleomobility studies of the effects of quaternary volcanism on bioavailable strontium: A test case in North Patagonia (Argentina). J Archaeol Sci. 2020;121: 105198.
- 58. Flockhart DT, Kyser TK, Chipley D, Miller NG, Norris DR. Experimental evidence shows no fractionation of strontium isotopes (87Sr/86Sr) among soil, plants, and herbivores: implications for tracking wildlife and forensic science. Isot Environ Health Stud. 2015;51: 372–381. pmid:25789981
- 59. Lewis J, Pike AW, Coath CD, Evershed RP. Strontium concentration, radiogenic (87Sr/86Sr) and stable (δ88Sr) strontium isotope systematics in a controlled feeding study. Sci Technol Archaeol Res. 2017;3: 45–57.
- 60. Fahy GE, Deter C, Pitfield R, Miszkiewicz JJ, Mahoney P. Bone deep: variation in stable isotope ratios and histomorphometric measurements of bone remodelling within adult humans. J Archaeol Sci. 2017;87: 10–16.
- 61. Hedges RE, Clement JG, Thomas CD, O’connell TC. Collagen turnover in the adult femoral mid‐shaft: Modeled from anthropogenic radiocarbon tracer measurements. Am J Phys Anthropol. 2007;133: 808–816. pmid:17405135
- 62. Matsubayashi J, Tayasu I. Collagen turnover and isotopic records in cortical bone. J Archaeol Sci. 2019;106: 37–44.
- 63. Murphy DT, Allen CM, Ghidan O, Dickson A, Hu WP, Briggs E, et al. Analysing Sr isotopes in low‐Sr samples such as single insects with inductively coupled plasma tandem mass spectrometry using N₂O as a reaction gas for in‐line Rb separation. Rapid Commun Mass Sp. 2020;34: e8604. pmid:31756774
- 64. Duxfield T, King C, Buckley H, Petchey P, Nowell GM, Kinaston RL. Establishing a strontium isotope baseline in New Zealand for future archaeological migration studies: A case study. J Archaeol Sci Rep. 2020;32: 102412.
- 65. Beard BL, Johnson CM. Strontium isotope composition of skeletal material can determine the birthplace and geographic mobility of humans and animals. J Forensic Sci. 2000;45: 1049–1061. pmid:11005180
- 66. Bowen GJ. 2010. Statistical and geostatistical mapping of precipitation water isotope ratios. West JB, Bowen GJ, Dawson TE, Tu KP, editors. Isoscapes: Understanding Movement, Pattern, and Process on Earth through Isotope Mapping. Dordrecht: Springer; 2010. pp. 139–160. https://doi.org/0.1007/978-90-481-3354-3_7
- 67. Bowen GJ, Liu Z, Vander Zanden HB, Zhao L, Takahashi G. Geographic assignment with stable isotopes in IsoMAP. Methods Ecol Evol. 2014;5: 201–206.
- 68. West JB, Sobek A, Ehleringer JR. A simplified GIS approach to modeling global leaf water isoscapes. PLOS ONE. 2008;3: e2447. pmid:18560592
- 69. West JB, Bowen GJ, Dawson TE, Tu KP, editors. Isoscapes: Understanding Movement, Pattern, and Process on Earth through Isotope Mapping. Dordrecht: Springer; 2010. https://doi.org/10.1007/978-90-481-3354-3
- 70. West JB, Kreuzer HW, Ehleringer JR. Approaches to plant hydrogen and oxygen isoscapes generation. In: West JB, Bowen GJ, Dawson TE, Tu KP, editors. Isoscapes: Understanding Movement, Pattern, and Process on Earth through Isotope Mapping. Dordrecht: Springer; 2010. pp. 161–178. https://doi.org/10.1007/978-90-481-3354-3_8
- 71. Wunder MB. Using isoscapes to model probability surfaces for determining geographic origins. In: West JB, Bowen GJ, Dawson TE, Tu KP, editors. Isoscapes: Understanding Movement, Pattern, and Process on Earth through Isotope Mapping. Dordrecht: Springer; 2010. pp. 251–270. https://doi.org/10.1007/978-90-481-3354-3_12
- 72. Wunder MB. Determining geographic patterns of migration and dispersal using stable isotopes in keratins. J Mammal. 2012;93: 360–367.
- 73. Bataille CP, Bowen GJ. Mapping 87Sr/86Sr variations in bedrock and water for large scale provenance studies. Chem Geol. 2012;304: 39–52.
- 74. Kramer RT, Bartelink EJ, Herrmann NP, Spradley MK, Bataille CP. Application of stable isotopes and geostatistics to infer region of geographic origin for deceased unidentified Latin American migrants. In: Parra R, Zapico S, Ubelaker DH, editors. Humanitarian Forensic Science: Interacting with the dead and the living. New Jersey: Wiley & Sons Ltd; 2020. pp. 425–440. https://doi.org/10.1002/9781119482062.ch27
- 75. Janzen A, Bataille CP, Copeland SR, Quinn RL, Ambrose SH, Reed D, Hamilton M, Grimes V, Richards MP, le Roux P, Roberts P. Spatial variation in bioavailable strontium isotope ratios (87Sr/86Sr) in Kenya and northern Tanzania: Implications for ecology, paleoanthropology, and archaeology. Palaeogeogr Palaeoclimatol Palaeoecol. 2020;560: 109957.
- 76. Hoogewerff JA, Reimann C, Ueckermann H, Frei R, Frei KM, Van Aswegen T, et al. Bioavailable 87Sr/86Sr in European soils: A baseline for provenancing studies. Sci Total Environ. 2019;672: 1033–1044. pmid:30999220
- 77. Adams S, Grün R, McGahan D, Zhao JX, Feng Y, Nguyen A, et al. A strontium isoscape of north-east Australia for human provenance and repatriation. Geoarchaeology. 2019;34: 231–251.
- 78. Bataille CP, Laffoon J, Bowen GJ. Mapping multiple source effects on the strontium isotopic signatures of ecosystems from the circum‐Caribbean region. Ecosphere. 2012;3: 1–24.
- 79. Wang X, Tang Z. The first large-scale bioavailable Sr isotope map of China and its implication for provenance studies. Earth-Sci Rev. 2020;201: 103353.
- 80. Kootker LM, van Lanen RJ, Kars H, Davies GR. Strontium isoscapes in The Netherlands. Spatial variations in 87Sr/86Sr as a proxy for palaeomobility. J Archaeol Sci Rep. 2016;6: 1–13.
- 81. Willmes M, Bataille CP, James HF, Moffat I, McMorrow L, Kinsley L, et al. Mapping of bioavailable strontium isotope ratios in France for archaeological provenance studies. Appl Geochem. 2018;90: 75–86.
- 82. Blank M, Sjögren KG, Knipper C, Frei KM, Storå J. Isotope values of the bioavailable strontium in inland southwestern Sweden—A baseline for mobility studies. PLOS ONE 2018;13: e0204649. pmid:30286158
- 83. Durante C, Bertacchini L, Cocchi M, Manzini D, Marchetti A, Rossi MC, et al. Development of 87Sr/86Sr maps as targeted strategy to support wine quality. Food Chem. 2018;255: 139–146. pmid:29571459
- 84. Song BY, Ryu JS, Shin HS, Lee KS. Determination of the source of bioavailable Sr using 87Sr/86Sr tracers: a case study of hot pepper and rice. J Agricul Food Chem. 2014;62: 9232–9238. pmid:25186083
- 85. Snoeck C, Pouncett J, Claeys P, Goderis S, Mattielli N, Pearson MP, et al. Strontium isotope analysis on cremated human remains from Stonehenge support links with west Wales. Sci Rep. 2018;8: 1–8. pmid:29311619
- 86. Snoeck C, Ryan S, Pouncett J, Pellegrini M, Claeys P, Wainwright AN, et al. Towards a biologically available strontium isotope baseline for Ireland. Sci Total Environ. 2020;712: 136248. pmid:31945525
- 87. Bataille CP, Brennan SR, Hartmann J, Moosdorf N, Wooller MJ, Bowen GJ. A geostatistical framework for predicting variations in strontium concentrations and isotope ratios in Alaskan rivers. Chem Geol. 2014;389: 1–5.
- 88. Bataille CP, Von Holstein IC, Laffoon JE, Willmes M, Liu XM, Davies GR. A bioavailable strontium isoscape for Western Europe: A machine learning approach. PLOS ONE. 2018;13: e0197386. pmid:29847595
- 89. Bataille CP, Jaouen K, Milano S, Trost M, Steinbrenner S, Crubézy É, et al. Triple sulfur-oxygen-strontium isotopes probabilistic geographic assignment of archaeological remains using a novel sulfur isoscape of western Europe. PLOS ONE. 2021;16: e0250383. pmid:33951062
- 90. Hartman G, Richards M. Mapping and defining sources of variability in bioavailable strontium isotope ratios in the Eastern Mediterranean. Geochim Cosmochim Acta. 2014;126: 250–264.
- 91. Poszwa A, Ferry B, Dambrine E, Pollier B, Wickman T, Loubet M, et al. Variations of bioavailable Sr concentration and 87Sr/86Sr ratio in boreal forest ecosystems: Role of biocycling, mineral weathering and depth of root uptake. Biogeochem. 2004;67: 1–20.
- 92. Sillen A, Hall G, Richardson S, Armstrong R. 87Sr/86Sr ratios in modern and fossil food-webs of the Sterkfontein Valley: implications for early hominid habitat preference. Geochim Cosmochim Acta. 1998;62: 2463–2473.
- 93. Rattenbury MS, Martin AP, Baisden WT, Turnbull RE, Rogers KM. Geochemical baseline soil surveys for understanding element and isotope variation across New Zealand. New Zealand J Agric Res. 2018;61: 347–357.
- 94. DIN ISO 19730. Soil Quality—Extraction of Trace Elements From Soil Using Ammonium Nitrate Solution. Beuth: Berlin; 2008.
- 95. Scholes RC, Hageman KJ, Closs GP, Stirling CH, Reid MR, Gabrielsson R, et al. Predictors of pesticide concentrations in freshwater trout–The role of life history. Environ Pollut. 2016;219: 253–261. pmid:27814542
- 96. Shaw B, Buckley H, Summerhayes G, Stirling C, Reid M. Prehistoric migration at Nebira, South Coast of Papua New Guinea: New insights into interaction using isotope and trace element concentration analyses. J Anthropol Archaeol. 2011;30: 344–58.
- 97. Wijenayake K. 2021. Origin Authentication of Bovine Milk. Doctoral Thesis, University of Otago. 2021.
- 98. Cocker K, Shane P, Stirling CH, Reid MR. Constraints on assembly of Tongariro and Ruapehu andesite magmas based on Sr-isotope compositions of plagioclase and groundmass. Lithos. 2021;400: 106406.
- 99. Gabrielsson RM, Kim J, Reid MR, Stirling CH, Numata M, Closs GP. Does the trace element composition of brown trout Salmo trutta eggs remain unchanged in spawning redds? J Fish Biol. 2012;81: 1871–1879. pmid:23130688
- 100. Pin C, Bassin C. Evaluation of a strontium-specific extraction chromatographic method for isotopic analysis in geological materials. Anal Chim Acta. 1992;269: 249–255.
- 101. Holder PW. Isotopes and trace elements as geographic origin markers for biosecurity pests. Doctoral Thesis, Lincoln University. 2012. Available from: https://researcharchive.lincoln.ac.nz/bitstream/handle/10182/5249/Holder_phd.pdf?sequence=3&isAllowed=y.
- 102. Armstrong KF, Holder PW, Murphy DT. Decision intelligence: determining pest natal origin. Final report for contract PBCRC2111 Plant Biosecurity Cooperative Research Centre 37. 2017. Available from: https://nla.gov.au/nla.obj-698245580/view.
- 103. Genuer R, Poggi JM, Tuleau-Malot C. VSURF Function. R package version 1.1.0; 2019 [cited 1 March 2021]. Available from: https://www.rdocumentation.org/packages/VSURF/versions/1.1.0/.
- 104. Chao M, Bowen GJ. assignR: Infer Geographic Origin from Isotopic Data. R package version 1.1.3; 2019 [cited 15 July 2020]. Available from: https://CRAN.R-project.org/package=assignR.
- 105. Wunder MB, Kester CL, Knopf FL, Rye RO. A test of geographic assignment using isotope tracers in feathers of known origin. Oecologia. 2005;144: 607–617. pmid:15891836
- 106. Vander Zanden HB, Wunder MB, Hobson KA, Wilgenburg SL, Wassenaar LI, Welker JM, et al. Contrasting assignment of migratory organisms to geographic origins using long-term versus year-specific precipitation isotope maps. Methods Ecol Evol. 2014;5: 891–900.
- 107. Breiman L. Random forests. Mach Learn, 2001;45: 5–32.
- 108. Ballance PF. 2017. New Zealand Geology: an illustrated guide. Auckland: Geoscience Society of New Zealand Miscellaneous Publications; 2017.
- 109. Faure G, Powell JL. Strontium isotope geology. Berlin: Springer-Verlag; 1972. https://doi.org/10.1007/978-3-642-65367-4
- 110. Norris RJ, Cooper AF. Late Quaternary slip rates and slip partitioning on the Alpine Fault, New Zealand. J Struct Geol. 2001;23: 507–520.
- 111. Wellman HW, Willett RW. The geology of the West Coast from Abut Head to Milford Sound—Part 1. Trans Proc R Soc NZ. 1942;71: 282–306.
- 112. Mortimer N, Tulloch AJ, Spark RN, Walker NW, Ladley E, Allibone A, et al. Overview of the Median Batholith, New Zealand: a new interpretation of the geology of the Median Tectonic Zone and adjacent rocks. J Afr Earth Sci. 1999;29: 257–68.
- 113. Mortimer N. New Zealand’s geological foundations. Gondwana Res. 2004;1: 261–272.
- 114. Muir RJ, Ireland TR, Weaver SD, Bradshaw JD, Evans JA, Eby GN, et al. Geochronology and geochemistry of a Mesozoic magmatic arc system, Fiordland, New Zealand. J Geol Soc London. 1998;155: 1037–1053.
- 115. Münker C. The isotope and trace element budget of the Cambrian Devil River Arc System, New Zealand: identification of four source components. J Petrol. 2000;41: 759–788.
- 116. Tulloch AJ, Ramezani J, Kimbrough DL, Faure K, Allibone AH. U-Pb geochronology of mid-Paleozoic plutonism in western New Zealand: Implications for S-type granite generation and growth of the east Gondwana margin. GSA Bulletin. 2009;121: 1236–1261.
- 117. Edbrooke SW, Heron DW, Forsyth PJ, Jongens R. Geological map of New Zealand 1:1000000. GNS Science Geological Map 2. 2 sheets. Lower Hutt: GNS Science; 2015.
- 118. Campbell HJ, Mortimer N, Turnbull IM. Murihiku Supergroup, New Zealand: redefined. J Roy Soc NZ. 2003;33: 85–95.
- 119. Grant PJ. Major periods of erosion and alluvial sedimentation in New Zealand during the Late Holocene. J Roy Soc NZ. 1985;15: 67–121.
- 120. Alloway B, McComb P, Neall V, Vucetich C, Gibb J, Sherburn S, et al. Stratigraphy, age, and correlation of voluminous debris‐avalanche events from an ancestral Egmont Volcano: implications for coastal plain construction and regional hazard assessment. J Roy Soc NZ. 2005;35: 229–67.
- 121. Cox SC, Barrell DJA (compilers). Geology of the Aoraki area: scale 1:250,000. Institute of Geological & Nuclear Sciences 1:250,000 geological map 15, 71 p. + 1 folded map. Lower Hutt: GNS Science; 2007.
- 122. Grindley GW. Structure of the alpine schists of South Westland, Southern Alps, New Zealand. New Zeal J Geol Geop. 1963;6: 872–930.
- 123. Mortimer N, McLaren S, Dunlap WJ. Ar-Ar dating of K-feldspar in low grade metamorphic rocks: Example of an exhumed Mesozoic accretionary wedge and forearc, South Island, New Zealand. Tectonics. 2012;31: TC3020.
- 124. Roser BP, Korsch RH. Geochemical characterization, evolution and source of a Mesozoic accretionary wedge: the Torlesse terrane, New Zealand. Geol Mag. 1999;136: 493–512.
- 125. Mackintosh L. Overview of New Zealand’s Climate. 2001 [cited 11 June 2021]. In: NIWA Taihoro Nukurangi [Internet]. Auckland: NIWA 2001. Available from: https://niwa.co.nz/education-and-training/schools/resources/climate/overview.
- 126. Macara GR. The climate and weather of West Coast. 2016 [cited 11 June 2021]. In: NIWA Taihoro Nukurangi [Internet]. Auckland: NIWA 2016. Available from: https://niwa.co.nz/sites/niwa.co.nz/files/West_Coast_Climatology_NIWA_web.pdf.
- 127. Ferguson CM, Barratt BI, Bell N, Goldson SL, Hardwick S, Jackson M, et al. Quantifying the economic cost of invertebrate pests to New Zealand’s pastoral industry. New Zealand J Agric Res. 2019;62: 255–315.
- 128. Faulkner KT, Hulme PE, Pagad S, Wilson JR, Robertson MP. Classifying the introduction pathways of alien species: are we moving in the right direction? NeoBiota 2020;62: 143–159.
- 129. Brenton-Rule EC, Barbieri RF, Lester PJ. Corruption, development and governance indicators predict invasive species risk from trade. P Roy Soc B-Biol Sci. 2016;283: 20160901. pmid:27306055
- 130. Sikes BA, Bufford JL, Hulme PE, Cooper JA, Johnston PR, Duncan RP. Import volumes and biosecurity interventions shape the arrival rate of fungal pathogens. PLOS Biol. 2018;16: e2006025. pmid:29851948
- 131. Toy SJ, Newfield MJ. The accidental introduction of invasive animals as hitchhikers through inanimate pathways: a New Zealand perspective. Rev Sci Tech. 2010;29: 123–33. pmid:20617653
- 132. West JB, Hurley JM, Dudás FÖ, Ehleringer JR. The stable isotope ratios of marijuana. II. Strontium isotopes relate to geographic origin. J Forensic Sci. 2009;54: 1261–1269. pmid:19804525
- 133. DeBord J, Pourmand A, Jantzi SC, Panicker S, Almirall J. Profiling of heroin and assignment of provenance by 87Sr/86Sr isotope ratio analysis. Inorganica Chim Acta. 2017;468: 294–299.
- 134. Bartelink EJ, Chesson LA, Tipple BJ, Hall S, Kramer RT. Multi‐isotope approaches for region‐of‐origin predictions of undocumented border crossers from the US–Mexico border: Biocultural perspectives on diet and travel history. In: Parra R, Zapico S, Ubelaker DH, editors. Humanitarian Forensic Science: Interacting with the dead and the living. New Jersey: Wiley & Sons Ltd; 2020. pp. 369–384. https://doi.org/10.1002/9781119482062.ch24
- 135. Laffoon JE, Sonnemann TF, Shafie T, Hofman CL, Brandes U, Davies GR. Investigating human geographic origins using dual-isotope (87Sr/86Sr, δ18O) assignment approaches. PLOS ONE. 2017;12: e0172562. pmid:28222163
- 136. Baisden WT, Keller ED, Van Hale R, Frew RD, Wassenaar LI. Precipitation isoscapes for New Zealand: enhanced temporal detail using precipitation-weighted daily climatology. Isot Environ Health Stud. 2016;52: 343–52. pmid:27007914
- 137. McComb KA, Sarker S, Hoogewerff JA, Hayman AR, Frew RD. A δ2H Isoscape of blackberry as an example application for determining the geographic origins of plant materials in New Zealand. PLOS ONE. 2019;14: e0226152. pmid:31815966
- 138. Rogers KM, Wassenaar LI, Soto DX, Bartle JA. A feather‐precipitation hydrogen isoscape model for New Zealand: implications for eco‐forensics. Ecosphere. 2012;3: 1–3.