基于主成分-时间序列模型的地下水位预测

Groundwater level forecast based on principal component analysis and multivariate time series model

地下水位预测是区域水资源管理的重要依据。针对地下水位在时间序列上表现出高度的随机性和滞后性,建立了基于主成分分析与多变量时间序列CAR(Controlled Auto-Regressive)模型耦合的地下水位预报模型,将该模型应用于济南市陡沟灌区地下水位预测,结果显示,模型模拟值与实测值的决定系数R^2和Nash-Suttcliffe系数Ens均达到0.90以上;以2011年为基准年,当降水量减少10%~20%,蒸发量和生活用水量增加10%~20%,调入27.39万~137.0万m^3地表水用于农业灌溉时,到2030年灌区地下水位将维持在30.99~31.29 m,较基准年上升0.12~0.42 m。在区域水资源紧缺的背景下,适当引入地表水灌溉,减少地下水的开采,灌区地下水位将逐步回升,对于灌区的可持续发展和区域水资源的合理利用具有重要意义。

Predication of groundwater level is an important basis for the management of regional water resources. Based on the high randomness and hysteresis characteristics of groundwater in time series, a groundwater level prediction model that is based on principal component analysis and muhivariable time series CAR model is built and used for the predica- tion of groundwater level at Dongou irrigation area of Ji＇nan. According to the results, the determination coefficient R2 and the Nash-Suttcliffe coefficient E,~ of the simulated value and the measured value all reached 0. 90 and the above. By taking 2011 as the base year, when precipitation reduces 10%--20%, evaporation and domestic water consumption in- creases 10%--20% and 273 900--1 370 000 m3 surface water is diverted for agricultural irrigation, the groundwater lev- el at the irrigation area will be maintained at 30. 99--31.29 m in 2030, increasing 0. 12--0. 42 m than that of the base year. Under the background of regional water resources shortage, proper diverting surface water for irrigation and reduc- ing groundwater exploitation can gradually increase the groundwater level at irrigation area and have great significance for the sustainable development of irrigation area and the reasonable utilization of regional water resources.