基于多变量时间序列模型的地下水埋深预测——以渭库绿洲为例

Groundwater Depth Prediction Based on Multivariate Time Series Model——A Case Study on Weigan and Kuqa Rivers Delta Oasis

为能够更好地预测干旱区渭库绿洲的地下水埋深及合理规划、开发地下水资源。以1995-2014年降雨量、蒸发量、农业用水量和地下水埋深等数据作为数据源展开研究。利用主成分分析法提取贡献率较大的因素,将其应用于多变量时间序列CAR(Controlled Auto-regressive)模型,从而建立地下水埋深的预测模型,并对模型进行验证。研究结果表明:渭库绿洲的地下水埋深受蒸发量、农业用水量的影响大,分别为0.533、0.466;模型预测值与实测值相对误差的平均值仅为1.483%,表明该模型在干旱区绿洲区域预测精度高、效果佳。预测趋势显示:若研究区蒸发量持续增加25%、降雨量减少15%、农业用水量减少15%时,地下水埋深将达到4.21 m,预测地下水埋深对渭库绿洲的生态环境和农业发展提供了有力的参考。

In order to better predict groundwater depth of Weigan and Kuqa rivers delta oasis, rationally plan and develop groundwater resources, the forecasting model of groundwater depth was established based on multivariate time series CAR model according to the data of annual rainfall from 1995 to 2014, the data of evaporation, water, DEM and groundwater depth and other data according to the contribution rate by using principal component analysis. The model was validated and applied to forecast the groundwater depth. The results indicate that the groundwater depth in Weigan and Kuqa rivers delta oasis is tremendously influenced by evaporation and water consumption, respectively, up to 0.533 and 0.466; the average value of relative error between model predictive value and measured value is only 1.483% and the multivariate time series CAR model is effective in prediction; The prediction schemes show that when the evaporation increases by 25%, the rainfall reduces by 15% and the water consumption reduces by 15%, and then the groundwater depth will be 4.21 m. Therefore, the research on of groundwater depth prediction in this paper can provide references for the ecological environment and agricultural development of Weigan and Kuqa rivers delta oasis.