Estimation of Spatial Distribution of Snow Water Equivalent and Snow Density of the West Azarbaijan Province's Basins

A significant portion of the annual precipitation of the mountainous West Azarbaijan (WA) of the I. R. of Iran occurs as snowfalls. Therefore، an assessment of the water equivalent of the snowpack and its density is of utmost importance in estimating the water yield of the WA basins. AS snowfall gradually increase with latitude and elevation، and since establishing snow courses and regular snow sampling in extensive areas is both costly and difficult، it is prudent to use a limited amount of data، suitable algorithms، and modeling to predict these 2 parameters for inaccessible places in WA. In this study، using the artificial neural network، variable snow density and snow water equivalent in the basins were estimated. In designing the artificial neural network three independent variables of longitude، latitude and altitude of the snow survey are used as the inputs. Artificial neural network design consisted of three neurons in input layer، two neurons in output layer، 6 to 18 neurons in one or more of the middle layers. The results showed that the structure of 2-6-3 to 2-6-6-3 and 2-6-6-6-3، yielded more logical answers. Thus، in the 2-6-3 structure، network design were converged after 270 iterations. Using the Levenberg-Merquate algorithm and the results obtained from the 2-6-3 structure، the neural network successfully estimated 92% of variations in snow density and snow water equivalent in WA. Furthermore، the stability results indicated that the 2-6-3 structure when used for analyzing the data from 35 snow measuring stations could satisfactorily predict 91% of variations in snow density and snow water equivalent for the 2009-2010 period. As the neural network uses fewer، easily accessible inputs as compared with the models used in similar studies، it has a better merit in snow-related hydrological studies