Abstract:

To investigate the dynamic characteristics of groundwater table depth （GTD） in different hydrological regions of arid oasis areas under the development of irrigated agriculture， the oasis area in the central part of Sangong River Basin in Xinjiang was selected as the target study region， and this oasis area was divided into three hydraulic units from south to north， i.e.， the alluvial fan oasis area （ADFO）， upper alluvial plain oasis area （APOU） and lower alluvial plain oasis area （APOL）. Using the GTD data of 9 long-term monitoring wells as well as irrigation agriculture development， hydrometeorological and regional socio-economic information from 1995 to 2016， the variation characteristics and influencing factors of GTD were analyzed based on a variety of powerful methods such as ensemble empirical mode decomposition， wavelet analysis and grey correlation degree， and a BP neural network model was developed to predict the change of GTD in the studied region under the future changing environment. We note that the interannual variations of GTD fluctuated greatly in the oasis area of Sangong River Basin， with a continuous downward trend during the past 22 years， especially in ADFO area with an average annual decline rate of 1.03 m. The change points of GTD for all wells were found to have occurred during 2006-2010， which represents the transition period of agricultural irrigation schemes from traditional flood irrigation to water-saving irrigation， and the GTD during the water-saving irrigation period （after 2012） was deeper than that the traditional flood irrigation period （before 2006） in each hydrological region， with an increase of ADFO （12.25-15.59 m） > APOU （5.30-8.23 m） > APOL （1.03-1.71 m）. The main influencing factors of GTD change in the basin are the cultivated land area， groundwater pumping and mountain annual runoff. The simulation and validation results indicate that the BP neural network model coupled with groundwater table depth in different hydrological regions has good modelling accuracy， and under the implementation of the policy of reducing water consumption by returning farmland from 2017 to 2036， the GTD in the ADFO， APOU and APOL will rise by 6.74 m， 2.55 m and 0.35 m， respectively. This study would provide directives for maintaining the sustainability of groundwater in oasis-desert systems in other similar endorheic watersheds.

Key words: groundwater table depth, different hydrological regions, dynamic change characteristics, BP neural network, Sangong River Basin