ABSTRACT Accurate and meaningful prediction of saltwater intrusion in coastal aquifers requires appropriate prediction tools. Artificial intelligence-based prediction models and their ensembles have been a better choice for mimicking the complex and nonlinear seawater intrusion progressions in coastal aquifers. This study utilizes a weighted-average ensemble of ‘heterogeneous’ prediction models to predict the saltwater intrusion progression in a coastal aquifer study area. The Dempster-Shafer theory of evidence is employed to calculate the weights of five different prediction model algorithms. Corresponding weights for individual prediction models are utilized in developing the ensemble prediction. Ensemble prediction performance for salinity intrusion in coastal aquifers in this effort is evaluated using several descriptive metrics. The values of the descriptive metrics suggest that the ensemble model performs in the same way as the best model in the ensemble. The methodology is evaluated for an illustrative coastal aquifer study area exposed to pumping-induced saltwater intrusion.

中文翻译：

---

利用基于 Dempster-Shafer 证据理论的预测模型的加权平均异质集合对沿海含水层的咸水入侵进行预测

摘要 对沿海含水层海水入侵的准确和有意义的预测需要适当的预测工具。基于人工智能的预测模型及其集成是模拟沿海含水层中复杂和非线性海水入侵进程的更好选择。本研究利用“异质”预测模型的加权平均集合来预测沿海含水层研究区的咸水入侵进展。Dempster-Shafer 证据理论用于计算五种不同预测模型算法的权重。单个预测模型的相应权重用于开发集成预测。在这项工作中，对沿海含水层盐度入侵的集合预测性能是使用几个描述性指标进行评估的。描述性指标的值表明集成模型的性能与集成中的最佳模型相同。该方法针对暴露于抽水引起的盐水入侵的示例性沿海含水层研究区进行了评估。