Akademik Veri Yönetim Sistemi
4th International Civil Engineering & Architecture Conference , Trabzon, Türkiye, 17 - 19 Mayıs 2025, cilt.1, ss.1014-1019, (Tam Metin Bildiri)
Water
resources planning and management policies are developed based on flow
forecasts over a certain period. The use of machine learning methods in the
construction of flow forecasting models is preferred for reasons such as saving
time, the quality and quantity of the data to be needed is less and has a lower
processing volume. Within the scope of this study, forecasting models for the
monthly inflows of Altınkaya Dam were constructed using the MARS method, which
is one of the regression-based methods. Monthly total precipitation and mean
temperature values covering the period 1987-2022 of the stations considered to
represent the dam basin were used as input parameters and monthly dam inflow
values were used as output parameters. Considering the lagged states of the
determined parameters, 6 different feature selection methods were used.
Correlation coefficient (R), root mean square error (RMSE), mean absolute error
(MAE), Nash Sutcliffe (NSE) efficiency coefficient were used to determine the
most successful prediction model. The most successful model among the different
model combinations was the M2 model created by backward elimination method. It
was concluded that the MARS method was sufficiently successful in predicting
monthly dam inflow. The prediction of dam inflow based on climatic factors such
as temperature and precipitation play a critical role not only for making
correct operational decisions but also for water resources management in the
long term.