Iterative Forecasting Improves Near-Term Predictions of Methane Ebullition Rates

Near-term, ecological forecasting with iterative model refitting and uncertainty partitioning has great promise for improving our understanding of processes the predictive skill models, but to date been infrequently applied predict biogeochemical fluxes. Bubble fluxes methane (CH 4 ) from aquatic sediments atmosphere (ebullition) dominate freshwater greenhouse gas emissions, it remains unknown how best make robust near-term CH ebullition predictions using models. Near-term workflows have potential address several current challenges in predicting rates, including: development models that can be across time horizons ecosystems, identification timescales which provide useful information, quantification predictions. To assess capacity near-term, improve rate predictions, we developed tested a workflow rates small eutrophic reservoir throughout one open-water period. The included repeated updating forecast over newly-collected data via refitting. We compared forecasts both alternative generated without persistence null model. Our estimated up 2 weeks into future [RMSE at 1-week ahead = 0.53 0.48 log e (mg m −2 d −1 2-week horizons]. Forecasts outperformed 1- horizons. Driver process contributed most total uncertainty, suggesting improvements should focus on improved mechanistic drivers. Altogether, study suggests improves week-to-week provides insight predictability future, identifies sources are important contributors