An Artificial Neural Network Model for Damoghan Reservoir Inflow Forecasting Using Snow Cover Data Derived from NOAA

The information of discharges in future monthes are useful for better reservoir
management. This study aims to forecast monthly inflow to Damoghan reservoir, in
Semnan province, by dynamic artificial neural network (DANN) models. Input data
of the models include monthly flow discharge, precipitation, mean temperature and
snow cover area of previous monthes. In the research record of 22 years of the data
are used. Determination snow cover area was done using thresholds in histograms of
snow in visible and thermal channels. Dynamic artificial networks were determined
with one hidden layer, Levenberg-Marquardt as training function and sigmoid as
transfer function. Moreover five models were run with different input data and the
results were compared. Root mean square error to observed flow discharge
( RMSE Qobs ), mean bias error (MBE), mean absolute relative error (MARE),
maximum relative error (REmax) and R2 (correlation coefficient) were the criteria
used for models evaluation. The best result was gained with three inputs including
inflow discharge, precipitation and snow cover area. Comparing cumulative relative
errors of models, the selected model is capable of forecasting next six months.
Moreover snow cover area improves the result of the model with flow discharge and
precipitation as input data by decreasing 10% and 26% in mean absolute relative
error and maximum relative error, respectively.

An Artificial Neural Network Model for Damoghan Reservoir Inflow Forecasting Using Snow Cover Data Derived from NOAA Keywords:

Damoghan reservoir , dynamic artificial neural network , inflow forecasting , NOAA-AVHRR , snow cover area