SPUN works with local researchers to develop sampling campaigns that help characterize mycorrhizal biodiversity in diverse underground ecosystems. We do this by combining geo-located mycorrhizal DNA sequencing data with ecological variables to generate spatial predictions of mycorrhizal diversity through a machine learning approach. This approach allows us to predict values of mycorrhizal diversity across the diverse ecosystems and regions of the globe. Next, we quantify the uncertainty of these model predictions and identify under-sampled ecoregions to help guide future mycorrhizal research. In order to improve the accuracy of our models, we ground-truth our predictions by working with local researchers to physically sample the soil in the ecosystems with the highest uncertainty values, as identified by our models, and then test them against our predictions. With each new set of samples, our models get more accurate and our uncertainty values go down.

On the ground, SPUN works with local scientists and institutions to identify sampling sites from which soil samples will be taken. The soil samples are collected according to a standardized protocol that ensures consistency in our analyses. You can read about our sampling protocols here. These samples are then processed in a lab, where DNA is extracted from the soil samples followed by the specific amplification of arbuscular and ectomycorrhizal fungal DNA. The amplified region of mycorrhizal DNA is then sent to a sequencing lab for the identification of the exact mycorrhizal species and lineages in the sample. These data are fed back into our models, which improve our predictions. We aim to generate data that is useful to governments, policy makers, NGOs and others to help diversify conservation agendas.

Learn more about our expeditions by checking out the stories below.