基于神经网络的福田红树林生态系统健康评估

Ecosystem Health Assessment of Mangrove Wetland in Futian Based on Neural Network

摘要: 本研究构建了基于神经网络模型的环境因素与生态系统健康指标间的映射关系以及基于健康度(I)和偏离度(b)的红树林生态系统健康评价方法，并以底栖动物多样性和黑脸琵鹭数量为指标对深圳福田红树林生态系统健康状况进行了评估与预测。结果表明：人类活动指标(噪声和人口数量)与黑脸琵鹭数量，以及水质指标(溶解氧、总磷、可溶性磷、硝态氮、氨态氮、盐度、pH、总氮和电导率)与底栖动物多样性间的灰色关联度均大于0.9；基于神经网络的鸟类数量和底栖动物多样性的模拟结果与野外监测的数值和变化趋势一致。黑脸琵鹭的数量及其I值预计在未来几年会呈现下降的趋势，而底栖动物多样性及其I值将整体呈上升趋势，说明该区水环境健康状况有较大改善，但整个湿地生态系统仍处于不健康状态。偏离度(b)随时间的波动较大，说明该区生态系统健康状况的稳定性较弱，在未来仍需加强红树林的管理和保护工作。

Abstract: This study constructs a mapping relationship between environmental factors and ecosystem health indicators based on neural network models and a mangrove ecosystem health assessment method based on health degree (I) and deviation degree (b). Furthermore, health status of the Futian mangrove ecosystem in Shenzhen was assessed and predicted based on the diversity of benthic animals and the number of black-faced spoonbills (Platalea minor). The results showed that the value of grey correlation degree is greater than 0.9 between human activity indicators (noise and population) and the number of black-faced spoonbills, as well as water quality indicators (dissolved oxygen, total phosphorus, soluble phosphorus, nitrate nitrogen, ammonia nitrogen, salinity, pH, total nitrogen, and conductivity) and the diversity of benthic animals. The simulation results of the bird population and benthic diversity based on the neural network model are consistent with the values and trends of field monitoring indicators. The number of black-faced spoonbills and their I values is expected to decline in the next few years, while the diversity of benthic animals and their I values will show an overall upward trend, indicating that the water environment health status of the area has been greatly improved, while the entire wetland ecosystem is still unhealthy. Deviation degree (b) fluctuations over time indicate that the stability of the ecosystem health in the area is weak, and mangrove management and protection need to be strengthened in the future.