Work Packages

Design of field survey

The team will discuss, develop and agree on a survey plan by considering scientific criteria as well as cost and logistic constraints. Chaosheng Zhang and Yunfan Li will conduct a dedicated, preliminary hot and cool spot analysis of existing Tellus soil data with a particular focus on the most likely biologically important elements. 200 sites will be selected for sampling. 

Soil microbial biodiversity

Using the survey design developed an extensive field sampling campaign will be conducted.  Soil samples for microbial analysis will be collected from identified locations using sterile procedures. Following the protocols (including soil storage, transport etc) established by the Earth Microbiome Project, microbial DNA will be extracted from soil samples using Qiagen DNeasy PowerSoil Kit (Qiagen, Hilden, Germany), followed by PCR amplification of bacterial 16S rRNA V3-V4 region, fungal ITS2 region and archaeal 16S rRNA V4-V5 region using specific primers. High-throughput sequencing using the Illumina MiSeq platform and standard bioinformatics will be performed by Novogene Co. Ltd to characterize the diversity and composition of the microbial community. Abundances of bacterial, fungal and archaeal taxa in soil samples will be assessed using the RT-qPCR technique.

Soil faunal biodiversity:

Traditional sampling and morphological identification of all or most soil fauna is not feasible given the ambitious sampling required for mapping purposes versus the labour and time requirements for full soil animal sorting, enumeration and identification. Therefore, a rapid protocol will be used. The protocol will follow selected techniques of the SoilBon Foodweb initiative, a global collaborative network to monitor soil animal communities using rapid, consistent methodologies (Potapov et al., 2022). In brief, a standard 19.6 cm2 area (5 cm diameter) steel corer will be used to extract 10 cm deep soil cores. Soil microfauna (nematodes) and mesofauna (enchytraeid worms) will be extracted from separate cores using Baermann funnels for wet extraction. Soil micro-arthropods (mites and springtails) will be extracted from separate cores using Tullgren/Berlese-type dry extraction with a light bulb. Animal enumeration, identification and body-mass estimation will be done using a novel high-throughput approach based on image analysis of mixed communities. High-resolution images of each sample will be generated by a flatbed scanner and analysed by a computer-vision pipeline based on deep learning algorithms currently under development and peer-review (details Potapov et al., 2022).

Production of hot spot and cool spot maps of soil geochemistry

The existing soil geochemical data from ‘Towards a National Soil Database’ (NSD) project and the Tellus project will be prepared and input in ArcGIS software. The lower density sampling of NSD with 1310 samples provides the national coverage, while the Tellus geochemical data with a higher sampling density currently has a coverage of half of the country. There are 45 geochemical parameters in NSD (including pH values, soil organic carbon, available P, available K, available Mg, as well as major and trace elements) and more than 50 geochemical parameters in the Tellus data. The hot spot analysis tools of local Moran’s I (Local Indicator of Spatial Association, LISA) and Getis-Ord Gi* will be employed to reveal the hot spots and cool spots of these geochemical parameters, with the results shown as maps.

Production of maps showing the spatial association between soil geochemistry and biodiversity

The newly collected biodiversity data in this project will be input into ArcGIS and the spatial distribution maps of soil biodiversity will be produced. Hot and cool spots of soil biodiversity will be revealed using the above-mentioned technologies (LISA and Getis Ord Gi*). The spatial association of hot/cool spots of geochemistry and biodiversity will be explored with overlay in GIS. Areas with high/low values of geochemical parameters and biodiversity indices will thus be identified. The links between soil geochemistry and biodiversity will then be explored based on these maps.

Quantification of community structure and pattern detection

Data will be organised into matrices that will associate biological data to environmental variables and spatially explicit information. Patterns of correlation in the matrices will be quantified with a range of univariate and multivariate analyses. Some key aspects of these analyses will be the quantification of diversity while controlling for differences in animal density or sequencing depth and yields (e.g., rarefaction curves), and an exploration of community dissimilarity such as changes in microbial composition at multiple taxonomic/phylogenetic scales. The phylogenetic approach will be particularly important for microbes, given the molecular nature of the data.  Variance partitioning models will quantify the relative effects of multiple sources of variation with a particular focus on spatial patterns, and community phylogenetic metrics. Null models of community and phylogenetic beta diversity will be tested to quantify patterns of taxa turnover and assess community convergence to (or divergence from) a core community associated with different habitats and locations

Association and ecological networks

Network analysis will be used to associate different groups of biota, and microbial functional genes, and how the environment modulates these associations. The networks will be assembled from co-occurrence matrices. There are important limitations in co-occurrence and correlation patterns as means to quantify interaction networks (Blanchet et al., 2020; Caruso et al., 2012). To overcome these limitations, the analysis will take advantage of methods of network reconstruction from partial information based on maximum entropy and maximum likelihood estimators (Squartini & Garlaschelli 2017). These methods account for uncertainty in the network configuration (e.g. spurious associations, ephemeral or rare associations, sparse matrices) and model the association data as fluctuations around an equilibrium configuration, thereby explicitly modelling the uncertainty on the occurrence of certain links (in this case whether an interaction is actually occurring or not). Network reconstruction will be based on structural constraints derived from the observed matrices. The network ensembles offer a baseline and also a null model to detect divergence of networks from equilibrium configurations (Squartini et al., 2013; Battiston et al., 2016).

Identification of indicators and techniques

There will be two classes of predictors of these metrics: spatial position and environmental variables. Environmental variables will be related to each other in different and complex ways, and will also depend on the spatial position of sampling locations. We will use structural equation modelling (SEM) to link exogenous variables (e.g. the environment) to endogenous variables (biota distribution). Starting from an SEM a priori model, we will fit various model variants to the same data to eventually define a set of parsimonious models that adequately fit the data (Grace, 2006; Shipley, 2016 Burnham & Anderson, 2002). The models will help identify key drivers of soil biodiversity, and the factors to monitor to predict soil biodiversity levels in unsampled sites. This information will offer a solid basis not only to identify hotspots of biodiversity but also to understand the factors that support these hotspots.