The North American Regional Carbon Flux Model: Predicting Carbon Flux Using Inversely Estimated Target Data With a Neural Network
Can we create a carbon flux model that uses inverse model estimates of carbon flux to train a neural network? Instead of using scaled-up carbon flux data, we can use inverse model estimates that capture large-scale carbon flux features because they are directly linked to the atmospheric CO2 data. This differs from other models because instead of using site-level data for carbon flux, our model is using region-wide averages to provide a better idea of the activity of carbon flux over broader areas. But unlike site-level estimates, inverse fluxes lack some of the detailed information like the temporal resolution. By also using regional environmental variables, we can make a model for CO2 flux for North America by utilizing the inverse model method to create data-driven carbon flux estimates for a large region.