Gavin McNicol, a postdoc in the lab (now at Stanford), has just published his work on coastal forest carbon. Coastal rainforests are the most biomass-carbon dense forest biome in the world, and Gavin focused on building a high spatial resolution/broad extent map of the soil/non-biomass portion of that - which is actually abut 60% of the total. He did a great job, working at 90m resolution across 10 degrees of latitude, a truly difficult task.
In total, we’re looking at approximately 4.5 Pg C in the top 1m, mostly driven by precipitation and topography.
One of the most interesting aspects of the project was Gavin’s comparison to global maps, which are used in global C budgets and models. Our work here is highly detailed, and works with an extensive soil observation network that was put together specifically for this task. So it’s useful to compare and validate global maps, since the scales are at least comparable (e.g., we’re not comparing one observation to a global map, but a huge region/part of the globe).
The global maps are generally pretty poor - 4 fold lower than SoilGrids250m, the finest resolution global map available and significantly more accurate against observations. That was true for other global maps as well.
This type of modeling, at this scale, is quite useful for evaluating global products that are so important for our global change research - they are produced at the right scale for comparison and thus far better for validation than just a series of points that may skew accuracy spatially. I think it’s fairly clear we need more of these types of studies and evaluations.
McNicol G, Bulmer C, D'Amore DV, Sanborn P, Saunders S, Giesbrecht I, Gonzalez-Arriola S, Bidlack AL, Butman D, Buma B. Large, climate-sensitive soil carbon stocks mapped with pedology-informed machine learning in the North Pacific coastal temperate rainforest. Environmental Research Letters. In press.