New publication evaluating modeling approaches for estimating net ecosystem exchange in Canadian peatlands

A new paper published in Scientific Reports compiled in-situ carbon dioxide data for 21 peatland sites across Canada, spanning seven ecoregions. With the aim of evaluating the performance of different models in predicting carbon dioxide flux, these measurements were coupled with 32 remotely sensed and gridded hydroclimatic variables. The final model inputs after reducing multicollinearity and balancing model error were evapotranspiration, burn area index, normalized difference water index, and modeled soil moisture. The model that achieved the highest performance was the Light-Gradient Boosting Machine model, which iteratively corrects errors from previous attempts. This model was found to be generalizable to other ecoregions, provided the range of fluxes within that region falls within that of the training data. These findings offer a framework for regional scaling to improve Canada’s spatially explicit carbon dioxide emission estimates.

Read the full paper here: https://doi.org/10.1038/s41598-026-67275-6