10:45 AM - 12:15 PM
[MGI25-P07] A Novel Groundwater Withdrawal Estimation Framework by Coupling Time-Frequency Analysis Method with Spa-Temporal Pattern Analysis Method – A Case of Central Taiwan
Keywords:Groundwater Wtihdrawal Estimation, Singal Decomposition Method, Empirical Orthogonal Function
Central Taiwan has been suffering from land subsidence for decades due to over-pumping. It causes the safety concerns to passengers of High Speed Railway which pass through the severe subsiding area. However, groundwater withdrawal amount and pumping location haven’t been fully understood as the result of plenty of uncensored wells which makes it more difficult to develop proper groundwater management strategy. However, central Taiwan has relative high resolution of groundwater level observed along space and time. Therefore, this research proposes a data-based method to estimate the spa-temporal distribution of groundwater withdrawal.
The groundwater withdrawal was estimated by two steps, (1)pumping pattern extraction and (2) pumping pattern quantification. As we know, Groundwater level(GWL) is the combination of the signals from different sources e.g. rainfall, pumping…, which have a certain period and spatial pattern. These stimuli could be recognized by analyzing the period and frequency of components hidden in GWL. Therefore, this study try to extract pumping pattern from GWL by a nonparametric and nonstationary signal time-series decomposition method - Ensemble Empirical Mode Decomposition(EEMD) and a sta-temporal pattern analysis method - Empirical Orthogonal Function(EOF). The Hilbert Spectrum Analysis (HSA) also be used to reveal instantaneously frequency of the components decomposed from EEMD. Since the estimation result could be affected by rainfall, the rainfall component was detected and removed from GWL before extracting the pumping pattern. Finally, the groundwater pumping pattern was quantified into pumping amounts by the storage coefficient from a calibrated MODFLOW model.
The method was conducted in central Taiwan which has severe over-pumping issue to exhibit the feasibility. The estimation result shows that the groundwater withdrawal is 1.63 billion tons per year on average from 2008 to 2019. It is similar to previous research which is in the range of 0.9 ~2.2 billion tons per year.
The groundwater withdrawal was estimated by two steps, (1)pumping pattern extraction and (2) pumping pattern quantification. As we know, Groundwater level(GWL) is the combination of the signals from different sources e.g. rainfall, pumping…, which have a certain period and spatial pattern. These stimuli could be recognized by analyzing the period and frequency of components hidden in GWL. Therefore, this study try to extract pumping pattern from GWL by a nonparametric and nonstationary signal time-series decomposition method - Ensemble Empirical Mode Decomposition(EEMD) and a sta-temporal pattern analysis method - Empirical Orthogonal Function(EOF). The Hilbert Spectrum Analysis (HSA) also be used to reveal instantaneously frequency of the components decomposed from EEMD. Since the estimation result could be affected by rainfall, the rainfall component was detected and removed from GWL before extracting the pumping pattern. Finally, the groundwater pumping pattern was quantified into pumping amounts by the storage coefficient from a calibrated MODFLOW model.
The method was conducted in central Taiwan which has severe over-pumping issue to exhibit the feasibility. The estimation result shows that the groundwater withdrawal is 1.63 billion tons per year on average from 2008 to 2019. It is similar to previous research which is in the range of 0.9 ~2.2 billion tons per year.