Evaluation of artificial neural network models and time series in the estimation of hydrostatic level (Case study: South Khorasan Province-Birjand Aquifer)

Groundwater has always been one of the major sources of drinking and agriculture especially in arid and semi-arid agriculture regions. birjandaquifer in south khorasan province,due to its location in the arid region,considers the use of groundwater as the most important and at the same time sole source of freshwater. modeling and prediction of wells' hydrostatic level is one of the basic tasks to achieve the optimal management of water resources.the current article mainly aims to evaluate the effectiveness of artificial neural network technique and time series in the prediction of hydrostatic level of groundwater. for this purpose,the data on hydrostatic level of 13 piezometer wells present in birjand plain as one of the sub-basins of lut desert was utilized in the 16-year statistical period of 1997-2013 on a monthly basis.in the current study,the discharge parameters including(amount of water extraction per cubic meter from drinking water wells,industry and agriculture),the water entering each polygon in terms of cubic meters (due to precipitationin the area) and water surface level (m) per piezometers in the previous time step were used,and the model output waswater level at the current time step.wells' hydrostatic level was simulated separately by neural network technique and time series (sarima),and finally rmse,mae and r2 were used to determine prediction accuracy of each of two methods. the study results indicated the high precision of both techniques of neural network and time series in the prediction of the hydrostatic level of the wells in the region.