Soil health in context: An evaluation of soil health indicator sensitivity, variability, and data synthesis in agricultural systems

The concept of soil health has captured the attention of stakeholders across the agricultural supply chain, providing an approachable metaphor for the dynamic, multi-functional nature of soil systems. A wide range of indicators, including soil physical, chemical, and biological properties, have been proposed to operationalize soil health. This work consists of three studies that address specific challenges to the practical implementation of soil health assessment: 1) a lack of soil health indicator data representing the combination of agronomic practices and soil conditions relevant to Minnesota, 2) the need to communicate with non-experts about soil variability, especially as it relates to sampling for soil health assessment, interpreting soil health tests, and extrapolating existing research to similar soils, and 3) the need for rigorous and transparent quantitative syntheses of soil health research to support evidence-based decision-making.

To address the first challenge, we conducted a statewide survey of 15 soil health indicators on 27 farms located across the state of Minnesota. Most soil health indicators were sensitive to differences in regional soil conditions and hillslope position, but wet aggregate stability was the only indicator consistently sensitive to management. To address the second challenge, we used an unsupervised learning algorithm, k-means, to aggregate publicly available soil data representing >6,500 map units and 1 million hectares into eight clusters based on similarities in topsoil properties. We found that grouping soils by cluster explained a similar amount of variation in biological and physical soil health indicators as grouping by Major Land Resource Area and explained more variation than soil taxonomic sub-group. To address the third challenge, we conducted an umbrella review, or review of reviews, to evaluate the scope, methods, and reporting rigor of n=204 meta-analyses addressing the effects of agricultural management on soil health indicators. We identified common methodological weaknesses in these studies, finding that the vast majority omitted key steps related to critical assessment of primary studies and mitigating bias during the review process. Taken together, these results provide new insights into key sources of variation in soil health indicators for Minnesota soils, a novel framework for grouping similar soils in Minnesota to aid communication and sampling for soil health, and a critical evaluation for how data synthesis methods can be improved to better support soil health research and decision-making.