Jalil-Masir Hamed, Fattahi Rohollah, Ghanbari-Adivi Elham, Asadi Aghbolaghi Mahdi, Ehteram Mohammad, Ahmed Ali Najah, El-Shafie Ahmed
Department of Water Science Engineering, Shahrekord University, Shahrekord, Iran.
Department of Water Engineering and Hydraulic Structures, Faculty of Civil Engineering, Semnan University, Semnan, Iran.
Environ Sci Pollut Res Int. 2022 Sep;29(44):67180-67213. doi: 10.1007/s11356-022-20472-y. Epub 2022 May 6.
Predicting sediment transport rate (STR) in the presence of flexible vegetation is a critical task for modelers. Sediment transport modeling methods in the coastal region is equally challenging due to the nonlinearity of the STR-vegetation interaction. In the present study, the kernel extreme learning model (KELM) was integrated with the seagull optimization algorithm (SEOA), the crow optimization algorithm (COA), the firefly algorithm (FFA), and particle swarm optimization (PSO) to estimate the STR in the presence of vegetation cover. The rigidity index, D/wave height, Newton number, drag coefficient, and cover density were used as inputs to the models. The root mean square error (RMSE), the mean absolute error (MAE), and percentage of bias (PBIAS) were used to evaluate the capability of models. This study applied the novel ensemble model, and the inclusive multiple model (IMM), to assemble the outputs of the KELM models. In addition, the innovations of this study were the introduction of a new IMM model, and the use of new hybrid KELM models for predicting STR and investigating the effects of various parameters on the STR. At the testing level, the MAE of the IMM model was 22, 60, 68, 73, and 76% lower than those of the KELM-SEOA, KELM-COA, KELM-PSO, and KELM models, respectively. The IMM had a PBIAS of 5, whereas the KELM-SEOA, KELM-COA, KELM-PSOA, and KELM had PBIAS of 9, 12, 14, 18, and 21%, respectively. The results indicated that the increasing drag coefficient and D/wave height had decreased the STR. From the findings, it was revealed that the IMM and KELM-SEOA had higher predictive ability for STR. Since the sediment is one of the most important sources of environmental pollution, therefore, this study is useful for monitoring and controlling environmental pollution.
预测存在柔性植被情况下的沉积物输移率（STR）对建模者而言是一项关键任务。由于STR与植被相互作用的非线性，沿海地区的沉积物输移建模方法同样具有挑战性。在本研究中，将核极限学习模型（KELM）与海鸥优化算法（SEOA）、乌鸦优化算法（COA）、萤火虫算法（FFA）和粒子群优化算法（PSO）相结合，以估算存在植被覆盖时的STR。刚度指数、D/波高、牛顿数、阻力系数和覆盖密度被用作模型的输入。均方根误差（RMSE）、平均绝对误差（MAE）和偏差百分比（PBIAS）用于评估模型的能力。本研究应用了新型集成模型和包容性多模型（IMM）来汇总KELM模型的输出。此外，本研究的创新之处在于引入了新的IMM模型，以及使用新的混合KELM模型来预测STR并研究各种参数对STR的影响。在测试阶段，IMM模型的MAE分别比KELM-SEOA、KELM-COA、KELM-PSO和KELM模型低22%、60%、68%、73%和76%。IMM的PBIAS为5%，而KELM-SEOA、KELM-COA、KELM-PSOA和KELM的PBIAS分别为9%、12%、14%、18%和21%。结果表明，阻力系数和D/波高的增加会降低STR。从研究结果可知，IMM和KELM-SEOA对STR具有更高的预测能力。由于沉积物是最重要的环境污染源之一，因此，本研究对于监测和控制环境污染具有重要意义。
