Refining Marine Net Primary Production Estimates: Advanced Uncertainty Quantification through Probability Prediction Models
Abstract. In marine ecosystems, Net Primary Production (NPP) is pivotal, not merely as a critical indicator of ecosystem health, but also as an integral component in the global carbon cycling process. This study introduces an advanced probability prediction model to refine the precision of NPP estimation and to deepen our comprehension of its inherent uncertainties. A comprehensive comparative analysis is undertaken, juxtaposing a Bayesian probability prediction model, predicated on empirical distribution, with a probability prediction model anchored in deep learning. The objective is to meticulously quantify the uncertainty associated with NPP. The findings underscore the applicability of probability prediction in investigating the uncertainty of marine NPP. Both models proficiently delineate the dynamic trends and inherent uncertainties in NPP, with the neural network model exhibiting superior accuracy and dependability. Additionally, these probability prediction models are adeptly applied to prognosticate NPP in specific marine regions, efficaciously elucidating the interannual trends in NPP variation. This research contributes not only a more precise method for quantifying NPP uncertainty but also bolsters scientific support for the stewardship of marine ecosystems and the preservation of environmental integrity.