Research on inversion of ecosystem dynamics model parameters based on improved Neural Network algorithm

摘要

Bayesian probability reversal and Markov chain Monte Carlo (MCMC) techniques were used in the regional environmental model to evaluate carbon (C) conversion coefficient and simulate carbon pool size uncertainty. Six data sets of C were used on soil respiration, tree biology, bio-leaves, litter, bed layer C, mineral soil measured under CO 2 (350 ppm) in these two contexts, and six data levels and liter height. Go 2 (550 ppm) curve. Increasing and better understanding of carbon dioxide levels in the Earth's atmosphere and the global carbon cycle. Web-based time-series models are based on the environment and climate parameters and create deep learning. To avoid a large part of the research and hard data downloads, the Google Earth Engine platform, based on this model can generate, and all input data will be published on the Internet. Variation method of numerical simulation, the global ocean ecosystem dynamics is used. Data Degradation Cycle Current measurement is a useful tool for extracting quantitative information from environmental information. However, to estimate the parameter value in terms of time inversion, it has been reduced that the sample factor can control the amount of spiral data as in the conventional reverse probe model. This research aims to increase the number of data degradation and increase the number of parameters.