Search

• Manifold distance metric better reflects similarities between ecological states in Empirical Dynamic Model. • Prediction accuracy improved by incorporating manifold distance into simplex projection and S-map. • Multivariate embedding and S-map amplify the advantage of manifold distance for prediction accuracy. • Effectiveness of manifold distance metric demonstrated by both synthetic and empirical data. Empirical Dynamic Modeling (EDM) has been a powerful tool for complex ecosystem prediction by providing an equation-free modelling framework. Theoretically, it allows future ecosystem behavior to be predicted by connecting current state to the similar, adjacent and future state on the attractor manifold which is reconstructed by single or multiple time series observed from natural systems. However, the Euclidean distance metric used in these algorithms could bias the true distance on the attractor manifold and consequently decrease the prediction performance. This could become worse if the dimension of the ecosystem is much higher and the system behavior is much complicated so that the reconstructed attractor manifold is more intricate. Therefore, manifold distance metric for both Simplex Projection and S-map was proposed. Our results clearly showed that the prediction accuracy of EDM had a general improvement after manifold distance metric was adopted. Experiments conducted on both synthetic and empirical data proved this advancement. Interestingly, these improvements were unequal for different implementations and the number of variables for embedding. Analysis demonstrated that S-map under multivariate embedding achieved the best prediction performance when manifold distance metric was applied. This suggested that the proposed manifold distance metric can work particularly well for predicting high dimensional ecosystem with complex behaviors. The main contribution of this research is that a new ecological indicator has been developed to more accurately estimate the similarity between ecological states in a reconstructed manifold and therefore provide higher prediction accuracy for EDM framework. [ABSTRACT FROM AUTHOR]

Data deficiency prevents the development of reliable machine learning models for many agroecosystems, especially those characterized by a dearth of knowledge derived from field data. However, other similar agroecosystems with extensive data resources can be of use. We propose a new predictive modeling approach based upon the concept of transfer learning to solve the problem of data deficiency in predicting productivity of agroecosystems, where productivity is a nonlinear function of various interacting biotic and abiotic factors. We describe the process of building metamodels (machine learning models built and trained on simulation data) from simulations built for one agroecosystem (US wild blueberry) as the source domain, where the data resource is abundant. Metamodels are evaluated and the best metamodel representing the system dynamics is selected. The best metamodel is re-parameterized and calibrated to another agroecosystem (Northeast China bog blueberry) as the target domain where field collected data are lacking. Experimental results showed that our metamodel developed for wild blueberry achieved 78% accuracy in fruit-set prediction for bog blueberry. To demonstrate its usefulness, we applied this calibrated metamodel to investigate the response of bog blueberry to various weather conditions. We found that an 8% reduction in fruit-set of bog blueberry is likely to happen if weather becomes warmer and wetter as predicted by climate models. In addition, southern and eastern production regions will suffer more severe fruit-set decline than the other growing regions. Predictions also suggest that increasing commercially available honeybee densities to 18 bees/m2/min, or bumble bee densities to 0.6 bees/m2/min, is a viable way to compensate for the predicted 8% climate induced fruit-set decline in the future. [ABSTRACT FROM AUTHOR]