The combined use of dynamic factor analysis and wavelet analysis to evaluate latent factors controlling complex groundwater level fluctuations in a riverside alluvial aquifer

Yun Yeong Oh, Seong Taek Yun, Soonyoung Yu, Se Yeong Hamm

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

To identify and quantitatively evaluate complex latent factors controlling groundwater level (GWL) fluctuations in a riverside alluvial aquifer influenced by barrage construction, we developed the combined use of dynamic factor analysis (DFA) and wavelet analysis (WA). Time series data of GWL, river water level and precipitation were collected for 3 years (July 2012 to June 2015) from an alluvial aquifer underneath an agricultural area of the Nakdong river basin, South Korea. Based on the wavelet coefficients of the final approximation, the GWL data was clustered into three groups (WCG1 to WCG3). Two dynamic factors (DFs) were then extracted using DFA for each group; thus, six major factors were extracted. Next, the time–frequency variability of the extracted DFs was examined using multiresolution cross-correlation analysis (MRCCA) with the following steps: 1) major driving forces and their scales in GWL fluctuations were identified by comparing maximum correlation coefficients (rmax) between DFs and the GWL time series and 2) the results were supplemented using the wavelet transformed coherence (WTC) analysis between DFs and the hydrological time series. Finally, relative contributions of six major DFs to the GWL fluctuations could be quantitatively assessed by calculating the effective dynamic efficiency (Def). The characteristics and relevant process of the identified six DFs are: 1) WCG1DF4,1 as an indicative of seasonal agricultural pumping (scales = 64–128 days; rmax = 0.68–0.89; Def ≤ 23.1%); 2) WCG1DF4,4 representing the cycle of regional groundwater recharge (scales = 64–128 days; rmax = 0.98–1.00; Def ≤ 11.1%); 3) WCG2DF4,1 indicating the complex interaction between the episodes of precipitation and direct runoff (scales = 2–8 days; rmax = 0.82–0.91; Def ≤ 35.3%) and seasonal GW-RW interaction (scales = 64–128 days; rmax = 0.76–0.91; Def ≤ 14.2%); 4) WCG2DF4,4 reflecting the complex effects of seasonal pervasive pumping and the local recharge cycle (scales = 64–128 days; rmax = 0.86–0.94; Def ≤ 16.4%); 5) WCG3DF4,2 as the result of temporal pumping (scales = 2–8 days; rmax = 0.98–0.99; Def ≤ 7.7%); and 6) WCG3DF4,4 indicating the local recharge cycle (scales = 64–128 days; rmax = 0.76–0.91; Def ≤ 34.2 %). This study shows that major driving forces controlling GWL time series data in a complex hydrological setting can be identified and quantitatively evaluated by the combined use of DFA and WA and applying MRCCA and WTC.

Original languageEnglish
Pages (from-to)938-955
Number of pages18
JournalJournal of Hydrology
Volume555
DOIs
Publication statusPublished - 2017 Dec 1

Fingerprint

wavelet analysis
factor analysis
aquifer
groundwater
time series
wavelet
pumping
recharge
barrage
dynamic analysis
river water
water level
agricultural land
river basin
runoff

Keywords

  • Dynamic factor analysis
  • Effective dynamic efficiency
  • Groundwater level fluctuation
  • Multiresolution cross-correlation
  • Wavelet analysis
  • Wavelet based clustering

ASJC Scopus subject areas

  • Water Science and Technology

Cite this

@article{0fa937c266e042fd8694402f98d74012,
title = "The combined use of dynamic factor analysis and wavelet analysis to evaluate latent factors controlling complex groundwater level fluctuations in a riverside alluvial aquifer",
abstract = "To identify and quantitatively evaluate complex latent factors controlling groundwater level (GWL) fluctuations in a riverside alluvial aquifer influenced by barrage construction, we developed the combined use of dynamic factor analysis (DFA) and wavelet analysis (WA). Time series data of GWL, river water level and precipitation were collected for 3 years (July 2012 to June 2015) from an alluvial aquifer underneath an agricultural area of the Nakdong river basin, South Korea. Based on the wavelet coefficients of the final approximation, the GWL data was clustered into three groups (WCG1 to WCG3). Two dynamic factors (DFs) were then extracted using DFA for each group; thus, six major factors were extracted. Next, the time–frequency variability of the extracted DFs was examined using multiresolution cross-correlation analysis (MRCCA) with the following steps: 1) major driving forces and their scales in GWL fluctuations were identified by comparing maximum correlation coefficients (rmax) between DFs and the GWL time series and 2) the results were supplemented using the wavelet transformed coherence (WTC) analysis between DFs and the hydrological time series. Finally, relative contributions of six major DFs to the GWL fluctuations could be quantitatively assessed by calculating the effective dynamic efficiency (Def). The characteristics and relevant process of the identified six DFs are: 1) WCG1DF4,1 as an indicative of seasonal agricultural pumping (scales = 64–128 days; rmax = 0.68–0.89; Def ≤ 23.1{\%}); 2) WCG1DF4,4 representing the cycle of regional groundwater recharge (scales = 64–128 days; rmax = 0.98–1.00; Def ≤ 11.1{\%}); 3) WCG2DF4,1 indicating the complex interaction between the episodes of precipitation and direct runoff (scales = 2–8 days; rmax = 0.82–0.91; Def ≤ 35.3{\%}) and seasonal GW-RW interaction (scales = 64–128 days; rmax = 0.76–0.91; Def ≤ 14.2{\%}); 4) WCG2DF4,4 reflecting the complex effects of seasonal pervasive pumping and the local recharge cycle (scales = 64–128 days; rmax = 0.86–0.94; Def ≤ 16.4{\%}); 5) WCG3DF4,2 as the result of temporal pumping (scales = 2–8 days; rmax = 0.98–0.99; Def ≤ 7.7{\%}); and 6) WCG3DF4,4 indicating the local recharge cycle (scales = 64–128 days; rmax = 0.76–0.91; Def ≤ 34.2 {\%}). This study shows that major driving forces controlling GWL time series data in a complex hydrological setting can be identified and quantitatively evaluated by the combined use of DFA and WA and applying MRCCA and WTC.",
keywords = "Dynamic factor analysis, Effective dynamic efficiency, Groundwater level fluctuation, Multiresolution cross-correlation, Wavelet analysis, Wavelet based clustering",
author = "Oh, {Yun Yeong} and Yun, {Seong Taek} and Soonyoung Yu and Hamm, {Se Yeong}",
year = "2017",
month = "12",
day = "1",
doi = "10.1016/j.jhydrol.2017.10.070",
language = "English",
volume = "555",
pages = "938--955",
journal = "Journal of Hydrology",
issn = "0022-1694",
publisher = "Elsevier",

}

TY - JOUR

T1 - The combined use of dynamic factor analysis and wavelet analysis to evaluate latent factors controlling complex groundwater level fluctuations in a riverside alluvial aquifer

AU - Oh, Yun Yeong

AU - Yun, Seong Taek

AU - Yu, Soonyoung

AU - Hamm, Se Yeong

PY - 2017/12/1

Y1 - 2017/12/1

N2 - To identify and quantitatively evaluate complex latent factors controlling groundwater level (GWL) fluctuations in a riverside alluvial aquifer influenced by barrage construction, we developed the combined use of dynamic factor analysis (DFA) and wavelet analysis (WA). Time series data of GWL, river water level and precipitation were collected for 3 years (July 2012 to June 2015) from an alluvial aquifer underneath an agricultural area of the Nakdong river basin, South Korea. Based on the wavelet coefficients of the final approximation, the GWL data was clustered into three groups (WCG1 to WCG3). Two dynamic factors (DFs) were then extracted using DFA for each group; thus, six major factors were extracted. Next, the time–frequency variability of the extracted DFs was examined using multiresolution cross-correlation analysis (MRCCA) with the following steps: 1) major driving forces and their scales in GWL fluctuations were identified by comparing maximum correlation coefficients (rmax) between DFs and the GWL time series and 2) the results were supplemented using the wavelet transformed coherence (WTC) analysis between DFs and the hydrological time series. Finally, relative contributions of six major DFs to the GWL fluctuations could be quantitatively assessed by calculating the effective dynamic efficiency (Def). The characteristics and relevant process of the identified six DFs are: 1) WCG1DF4,1 as an indicative of seasonal agricultural pumping (scales = 64–128 days; rmax = 0.68–0.89; Def ≤ 23.1%); 2) WCG1DF4,4 representing the cycle of regional groundwater recharge (scales = 64–128 days; rmax = 0.98–1.00; Def ≤ 11.1%); 3) WCG2DF4,1 indicating the complex interaction between the episodes of precipitation and direct runoff (scales = 2–8 days; rmax = 0.82–0.91; Def ≤ 35.3%) and seasonal GW-RW interaction (scales = 64–128 days; rmax = 0.76–0.91; Def ≤ 14.2%); 4) WCG2DF4,4 reflecting the complex effects of seasonal pervasive pumping and the local recharge cycle (scales = 64–128 days; rmax = 0.86–0.94; Def ≤ 16.4%); 5) WCG3DF4,2 as the result of temporal pumping (scales = 2–8 days; rmax = 0.98–0.99; Def ≤ 7.7%); and 6) WCG3DF4,4 indicating the local recharge cycle (scales = 64–128 days; rmax = 0.76–0.91; Def ≤ 34.2 %). This study shows that major driving forces controlling GWL time series data in a complex hydrological setting can be identified and quantitatively evaluated by the combined use of DFA and WA and applying MRCCA and WTC.

AB - To identify and quantitatively evaluate complex latent factors controlling groundwater level (GWL) fluctuations in a riverside alluvial aquifer influenced by barrage construction, we developed the combined use of dynamic factor analysis (DFA) and wavelet analysis (WA). Time series data of GWL, river water level and precipitation were collected for 3 years (July 2012 to June 2015) from an alluvial aquifer underneath an agricultural area of the Nakdong river basin, South Korea. Based on the wavelet coefficients of the final approximation, the GWL data was clustered into three groups (WCG1 to WCG3). Two dynamic factors (DFs) were then extracted using DFA for each group; thus, six major factors were extracted. Next, the time–frequency variability of the extracted DFs was examined using multiresolution cross-correlation analysis (MRCCA) with the following steps: 1) major driving forces and their scales in GWL fluctuations were identified by comparing maximum correlation coefficients (rmax) between DFs and the GWL time series and 2) the results were supplemented using the wavelet transformed coherence (WTC) analysis between DFs and the hydrological time series. Finally, relative contributions of six major DFs to the GWL fluctuations could be quantitatively assessed by calculating the effective dynamic efficiency (Def). The characteristics and relevant process of the identified six DFs are: 1) WCG1DF4,1 as an indicative of seasonal agricultural pumping (scales = 64–128 days; rmax = 0.68–0.89; Def ≤ 23.1%); 2) WCG1DF4,4 representing the cycle of regional groundwater recharge (scales = 64–128 days; rmax = 0.98–1.00; Def ≤ 11.1%); 3) WCG2DF4,1 indicating the complex interaction between the episodes of precipitation and direct runoff (scales = 2–8 days; rmax = 0.82–0.91; Def ≤ 35.3%) and seasonal GW-RW interaction (scales = 64–128 days; rmax = 0.76–0.91; Def ≤ 14.2%); 4) WCG2DF4,4 reflecting the complex effects of seasonal pervasive pumping and the local recharge cycle (scales = 64–128 days; rmax = 0.86–0.94; Def ≤ 16.4%); 5) WCG3DF4,2 as the result of temporal pumping (scales = 2–8 days; rmax = 0.98–0.99; Def ≤ 7.7%); and 6) WCG3DF4,4 indicating the local recharge cycle (scales = 64–128 days; rmax = 0.76–0.91; Def ≤ 34.2 %). This study shows that major driving forces controlling GWL time series data in a complex hydrological setting can be identified and quantitatively evaluated by the combined use of DFA and WA and applying MRCCA and WTC.

KW - Dynamic factor analysis

KW - Effective dynamic efficiency

KW - Groundwater level fluctuation

KW - Multiresolution cross-correlation

KW - Wavelet analysis

KW - Wavelet based clustering

UR - http://www.scopus.com/inward/record.url?scp=85032933447&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85032933447&partnerID=8YFLogxK

U2 - 10.1016/j.jhydrol.2017.10.070

DO - 10.1016/j.jhydrol.2017.10.070

M3 - Article

VL - 555

SP - 938

EP - 955

JO - Journal of Hydrology

JF - Journal of Hydrology

SN - 0022-1694

ER -