Effective prediction of biodiversity in tidal flat habitats using an artificial neural network

摘要: Abstract Accurate predictions of benthic macrofaunal biodiversity greatly benefit the efficient planning and management habitat restoration efforts in tidal flat habitats. Artificial neural network (ANN) prediction models for such were developed tested based on 13 biophysical variables, collected from 50 sites flats along coast Korea during 1991–2006. The model showed high training, cross-validation testing. Besides training testing procedures, an independent dataset a different time period (2007–2010) was used to test robustness practical usage model. High (r = 0.84) validated networks proper learning predictive relationship its generality. Key influential variables identified by follow-up sensitivity analyses related with topographic dimension, environmental heterogeneity, water column properties. Study demonstrates successful application ANN accurate understanding dynamics candidate variables.