نوع مقاله : مقاله پژوهشی
نویسندگان
1 فارغ التحصیل کارشناسی ارشد، گروه منابع آب، دانشکده علوم آب، دانشگاه شهید چمران اهواز، اهواز، ایران.
2 استاد، گروه منابع آب، دانشکده مهندسی آب و محیط زیست، دانشگاه شهید چمران اهواز، اهواز، ایران.
3 دانشجوی دکتری مهندسی عمران- مهندسی و مدیریت منابع آب، دانشکده عمران و معماری، دانشگاه شهید چمران اهواز، اهواز، ایران.
چکیده
کلیدواژه‌ها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
In recent years, the water crisis, the decline in groundwater levels, and the reduced sustainability of agriculture have underscored the importance of accurately estimating water infiltration into soil. In this study, water infiltration was simulated using data from 17 infiltration tests and five models: Green-Ampt, Kostiakov, Horton, SCS (Soil Conservation Service), and an artificial neural network (ANN). Statistical evaluation indicated that the SCS model had the best performance due to higher accuracy and lower error compared to the other models, followed by the neural network model in second place.
To assess model uncertainty, the fuzzy α-cut method was applied. Results showed that the neural network model had the lowest uncertainty (11% to 21%) in estimating cumulative infiltration, while the SCS model had the highest uncertainty (88% to 61%). The uncertainty analysis revealed that although the SCS model demonstrated higher accuracy, it was more unstable and sensitive to input parameters.
This finding suggests a general principle: in the comparison of engineering models, it is essential to consider not only statistical accuracy but also uncertainty and model reliability. For instance, in this study, the SCS method might have been the preferred choice without uncertainty analysis; however, considering both the accuracy and reliability of the neural network model, it can be regarded as the better option. These results also confirm the superiority of data-driven models in the optimal management of water resources.
کلیدواژه‌ها [English]
)