Objective The eco−geological environment is closely related to the development of human society. Scientific evaluation of the eco−geological environment can deepen our understanding of environmental conditions and promote social and economic development. However, traditional evaluation methods fail to fully capture the key characteristics of the eco−geological environment such as nonlinearity and spatial autocorrelation, which easily lead to problems such as discontinuous evaluation results, abrupt grade jumps and lack of gradual transition and biased identification of extreme environments. This study aims to construct a deep learning evaluation method suitable for complex eco−geological systems to improve evaluation rationality and spatial continuity.

Methods The subject region under investigation is Anji County, Zhejiang Province. Seventeen evaluation indicators including vegetation coverage, land use and slope were selected to construct a three−subsystem evaluation system of ecological, geological and livable environments. Evaluation label maps were produced by expert scoring, and a Convolutional Neural Network (CNN) model was built for training and evaluation. Finally, the evaluation results were compared with the comprehensive index method and random forest method for verification.

Results The excellent and good areas of eco−geological environment in Anji County account for 95.57%. The ecologically excellent areas are concentrated in forest land, the geologically weak areas are high−prone areas of disasters, and the livable excellent areas are mainly plain gentle zones.

Conclusions The method proposed in this paper can reasonably identify extreme eco-geological environments and effectively eliminate evaluation faults and incoherence of adjacent plots. Compared with conventional evaluation methods such as the comprehensive index method and random forest, the CNN model can better characterize the nonlinear relationships and spatial autocorrelation features of eco-geological systems. The findings of this study provide scientific support for the sustainable development in Anji County, and also offer a methodological reference for eco-geological environment assessment in analogous regions.