80th Bownocker Distinguished Lecturer

 

It has long been an academic question as to why rates of silicate-water reactions appear slower in the field than in the laboratory. Today, this question has become increasingly important to policy-makers who are interested in attacking the problem of global warming by grinding basalt and dispersing it on farm fields to enhance weathering-driven removal of CO2 from the atmosphere. The enthusiasm for this approach to carbon dioxide removal, and the many ongoing experiments worldwide, beg the question, why is it so hard to predict accurate rates of the long-studied process of mineral weathering in field systems?

From flask reactors to soils to hillsides to watersheds, rate estimates become increasingly slower. Most of the reasons explaining the lab-field discrepancy are related to surface area, hydrology, heterogeneities, and biota. In addition, some issues are system‐level effects. The use of improved reactive transport models, sometimes coupled to hydrologic models or driven by climate data, can decrease the lab-field discrepancy. But, from a general point of view, we cannot predict rates of CO2 drawdown for soil amendments a priori because i) weathering is a complex spatially-nested system, ii) our best physics-based models are not reducible, and iii) our best data-driven models are hampered by data sparcity. To make accurate predictions of weathering, we will need to develop hybrid models that use physics laws incorporated into machine learning approaches that we can improve with data assimilation from many more field-level weathering experiments.