EGU23-9302, updated on 24 Oct 2023
https://doi.org/10.5194/egusphere-egu23-9302
EGU General Assembly 2023
© Author(s) 2023. This work is distributed under
the Creative Commons Attribution 4.0 License.

Assimilating GRACE Total Water Storage Anomalies into hydrological forecasts in order to acquire information for groundwater level changes

Alexandra Gemitzi and Stavros Stathopoulos
Alexandra Gemitzi and Stavros Stathopoulos
  • Department of Environmental Engineering, Democritus University of Thrace, Xanthi, Greece (agkemitz@env.duth.gr)

The present work aims at assimilating GRACE Total Water Storage Anomalies (TWSA) into hydrological forecasts in order to estimate groundwater level changes. The motivation of our research was to provide information for groundwater levels and storage changes along with the rest of hydrological parameters provided usually by hydrological models, i.e., surface runoff, lateral flow contribution to stream flow, groundwater contribution to stream flow, water percolation below the soil profile, soil water and evapotranspiration. Therefore, we investigated a possible approach to acquire information on the groundwater regime of a watershed assimilating downscaled GRACE TWSA along with two variables related to groundwater flow obtained from the application of SWAT model, i.e., groundwater contribution to stream flow and percolation. The methodology was developed and tested in a medium sized river basin (~360 km2) in NE Greece, namely Vosvozis river basin. Initially we checked for possible correlation of the downscaled GRACE TWSA with the groundwater level anomalies for the period 2013 – 2021. Results indicated that downscaled GRACE TWSA can be used as possible predictor for groundwater level changes. Thereafter, two model approaches were evaluated for their predictive ability regarding groundwater level changes. The first approach is a Multiple Linear Regression (MLR) model whereas the second was an Artificial Neural Network Multilayer Perceptron (MLP-ANN) model. Both models indicated a satisfactory performance with R2 values ranging from 0.76 – 0.78 for the MLR model, in the training and testing phases, whereas the MLP-ANN outperformed in both phases the MLR model, with R2 ranging well above 0.8, indicating its predictive ability for groundwater level changes. The methodology can be applied parallel to SWAT model and groundwater level changes can be acquired simultaneously with the rest of hydrological variables.      

How to cite: Gemitzi, A. and Stathopoulos, S.: Assimilating GRACE Total Water Storage Anomalies into hydrological forecasts in order to acquire information for groundwater level changes, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-9302, https://doi.org/10.5194/egusphere-egu23-9302, 2023.