A method based on neural network for risk prediction of the typical moraine-dammed lake outburst in the Himalayan region

摘要

In this research, based on neural network, thirty typical moraine-dammed lakes were selected as the training set. In accordance with the rules defined, ten evaluation indexes were made dimensionless and used to train the model. Then the research could get the applicable model that evaluated the probabilities of moraine-dammed lake outburst in the Himalayas region of Tibet, China. Then the probability of outburst was predicted for the Laqu Lake based on the developed model, and the predictive value was 0.538. In terms of risk level standards divided, the Laqu Lake was high-risk, which is consistent with the field survey. It well demonstrated the applicability that using the neural network to assess the probabilities of moraine-dammed lake outburst.