Abstract
Due to the underlying uncertainty in the groundwater level (GWL) modeling, point prediction of the GWLs does not provide sufficient information for decision making and management purposes. Thus, estimating prediction intervals (PIs) for groundwater modeling can be an important step in sustainable water resources management. In this paper, PIs were estimated for GWL of selected piezometers of the Ardebil plain located in northwest of Iran and the Qorveh–Dehgolan plain located in west of Iran, using bootstrap methods based on artificial neural networks (ANNs). For this purpose, the classic feedforward neural network (FFNN) and deep learning (DL)-based long short-term memory (LSTM) were used as ANN bases and the classic bootstrap and moving blocks bootstrap (MBB) as the bootstrap variations. Monthly GWL data of some piezometers as well as hydrologic data of the related stations from both plains were used for the training and validation of the models. The results showed that the LSTM outperforms the seasonal auto regressive integrated moving average model with exogeneous data (SARIMAX), which is a linear model, and classic FFNN in point prediction task. Moreover, in terms of PIs model performance, the LSTM-based MBB (MBLSTM) achieved an average of 30% lower coverage width criterion (CWC) than the FFNN-based MBB (MBFN) and average of 40% lower CWC than the FFNN-based classic bootstrap (BFN). In addition, PIs estimated for piezometers situated in areas with high transmissivity resulted in 55% lower CWC than PIs estimated for piezometers, which are located in areas with lower transmissivity.
Similar content being viewed by others
Availability of data and material
The data will be available upon request.
Code availability
The code will be available upon request.
References
Bakker M, Schaars F (2019) Solving groundwater flow problems with time series analysis: you may not even need another model. Groundwater 57:826–833. https://doi.org/10.1111/gwat.12927
Bengio Y, Simard P, Frasconi P (1994) Learning long-term dependencies with gradient descent is difficult. IEEE Trans Neural Networks 5(2):157–166. https://doi.org/10.1109/72.279181
Bowes B, Sadler J, Morsy M, Behl M, Goodall J (2019) Forecasting groundwater table in a flood prone coastal city with long short-term memory and recurrent neural networks. Water 11:1098. https://doi.org/10.3390/w11051098
Correa CD, Lindstrom P (2013) The mutual information diagram for uncertainty visualization. Int J Uncert Quantif 3:187–201. https://doi.org/10.1615/Int.J.UncertaintyQuantification.2012003959
De S, Dey AK, Gouda DK (2020) Construction of confidence interval for a univariate stock price signal predicted through long short term memory network. Ann Data Sci. https://doi.org/10.1007/s40745-020-00307-8
Dibike Y, Gachon P, St-Hilaire A, Ouarda T, Nguyen V-T-V (2008) Uncertainty analysis of statistically downscaled temperature and precipitation regimes in Northern Canada. Theoret Appl Climatol 91:149–170. https://doi.org/10.1007/s00704-007-0299-z
Dybowski R, Roberts S (2001) Confidence intervals and prediction intervals for feed-forward neural networks. Clinical applications of artificial neural networks. Cambridge University Press, pp. 298–326. https://doi.org/10.1017/CBO9780511543494.013
Ebtehaj M, Moradkhani H, Gupta HV (2010) Improving robustness of hydrologic parameter estimation by the use of moving block bootstrap resampling. Water Resour Res. https://doi.org/10.1029/2009WR007981
Famiglietti J (2014) The global groundwater crisis. Nature Clim Change 4:945–948. https://doi.org/10.1038/nclimate2425
Farajzadeh J, Alizadeh F (2018) A hybrid linear–nonlinear approach to predict the monthly rainfall over the Urmia Lake watershed using wavelet-SARIMAX-LSSVM conjugated model. J Hydroinf 20(1):246–262. https://doi.org/10.2166/hydro.2017.013
Fathi MM, Awadallah AG, Abdelbaki AM, Haggag M (2019) A new Budyko framework extension using time series sarimax model. J Hydrol 570:827–838. https://doi.org/10.1016/j.jhydrol.2019.01.037
Fischer T, Krauss C (2018) Deep learning with long short-term memory networks for financial market predictions. Eur J Oper Res 270(2):654–669. https://doi.org/10.1016/j.ejor.2017.11.054
Gers F, Schraudolph N, Schmidhuber J (2002) Learning precise timing with LSTM recurrent networks. J Mach Learn Res 3:115–143. https://doi.org/10.1162/153244303768966139
Grant EL, Leavenworth RS (1972) Statistical quality control [by] Eugene L. Grant [and] Richard S. Leavenworth. McGraw-Hill. https://books.google.com/books?id=D4DVswEACAAJ
Hochreiter S, Schmidhuber J (1997) Long short-term memory. Neural Comput 9:1735–1780. https://doi.org/10.1162/neco.1997.9.8.1735
Hu C, Wu Q, Li H, Jian S, Li N, Lou Z (2018) Deep learning with a long short-term memory networks approach for rainfall-runoff simulation. Water 10:1543. https://doi.org/10.3390/w10111543
Jerome Morrissey P, McCormack T, Naughton O, Meredith Johnston P, William Gill L (2020) Modelling groundwater flooding in a lowland karst catchment. J Hydrol 580:124361. https://doi.org/10.1016/j.jhydrol.2019.124361
Khorrami M, Alizadeh B, Ghasemi Tousi E, Shakerian M, Maghsoudi Y, Rahgozar P (2019) How groundwater level fluctuations and geotechnical properties lead to asymmetric subsidence: a PSInSAR analysis of land deformation over a transit corridor in the los angeles metropolitan area. Remote Sensing 11(4):377. https://doi.org/10.3390/rs11040377
Khosravi A, Nahavandi S, Creighton D, Atiya A (2011a) Comprehensive review of neural network-based prediction intervals and new advances. IEEE Trans Neural Netw 22:1341–1356. https://doi.org/10.1109/TNN.2011.2162110
Khosravi A, Nahavandi S, Creighton D, Atiya A (2011b) Lower upper bound estimation method for construction of neural network-based prediction intervals. IEEE Trans Neural Networks 22(3):337–346. https://doi.org/10.1109/TNN.2010.2096824
Kreiss J-P, Lahiri SN (2012) 1-bootstrap methods for time series. In: Subba Rao T, Subba Rao S, Rao CR (eds) Handbook of statistics, vol 30, pp 3–26. Elsevier. https://doi.org/10.1016/B978-0-444-53858-1.00001-6
Kunsch HR (1989) The Jackknife and the bootstrap for general stationary observations. Ann Statist 17(3):1217–1241. https://doi.org/10.1214/aos/1176347265
Liang C, Li H, Lei M, Du Q (2018) Dongting lake water level forecast and its relationship with the three gorges dam based on a long short-term memory network. Water 10:1389. https://doi.org/10.3390/w10101389
Liu RY, Singh K (1992) Moving blocks Jackknife and bootstrap capture weak dependence. In: Lepage R, Billard L (eds) Exploring the limits of bootstrap. Wiley, New York
Martínez-Acosta L, Medrano-Barboza JP, López-Ramos Á, Remolina López JF, López-Lambraño ÁA (2020) SARIMA approach to generating synthetic monthly rainfall in the Sinú River Watershed in Colombia. Atmosphere 11(6):602. https://doi.org/10.3390/atmos11060602
Mohanty S, Jha MK, Kumar A, Panda DK (2013) Comparative evaluation of numerical model and artificial neural network for simulating groundwater flow in Kathajodi-Surua Inter-basin of Odisha India. J Hydrol 495:38–51. https://doi.org/10.1016/j.jhydrol.2013.04.041
Morgan M, Henrion M (1990) Uncertainty: a guide to dealing with uncertainty in quantitative risk and policy analysis
Nash JE, Sutcliffe JV (1970) River flow forecasting through conceptual models part I — a discussion of principles. J Hydrol 10(3):282–290
Nishigaki M (2010) Geotechnical aspects of groundwater control. Soils Found 50(6):893–902. https://doi.org/10.3208/sandf.50.893
Nourani V, Behfar N (2021) Multi-station runoff-sediment modeling using seasonal LSTM model. J Hydrol 601:126672. https://doi.org/10.1016/j.jhydrol.2021.126672
Nourani V, Mogaddam AA, Nadiri AO (2008) An ANN-based model for spatiotemporal groundwater level forecasting. Hydrol Process 22(26):5054–5066. https://doi.org/10.1002/hyp.7129
Nourani V, Rezapour Khangah T, Hosseini Baghanam A (2015) Application of entropy concept for input selection of wavelet-ANN based rainfall-runoff modeling. J Environ Inf. https://doi.org/10.3808/jei.201500309
Nourani V, Jabbarian Paknezhad N, Sharghi E, Khosravi A (2019) Estimation of prediction interval in ANN-based multi-GCMs downscaling of hydro-climatologic parameters. Journal of Hydrology 579:124226. https://doi.org/10.1016/j.jhydrol.2019.124226
Rajaee T, Ebrahimi H, Nourani V (2019) A review of the artificial intelligence methods in groundwater level modeling. J Hydrol 572:336–351. https://doi.org/10.1016/j.jhydrol.2018.12.037
Rußwurm M, Körner M (2017) Temporal vegetation modelling using long short-term memory networks for crop identification from medium-resolution multi-spectral satellite images. Proc IEEE Conf Comput Vis Pattern Recognit Works. https://doi.org/10.1109/CVPRW.2017.193
Samadi S, Wilson CAME, Moradkhani H (2013) Uncertainty analysis of statistical downscaling models using hadley centre coupled model. Theoret Appl Climatol 114(3):673–690. https://doi.org/10.1007/s00704-013-0844-x
Sachindra DA, Kamal A, Shahid S, Perera BJC (2018) Cationary note on the use of genetic programming in statistical downscaling. Int J Climatol. https://doi.org/10.1002/joc.5508
Shrestha DL, Solomatine DP (2008) Data-driven approaches for estimating uncertainty in rainfall-runoff modelling. Int J River Basin Manag 6(2):109–122. https://doi.org/10.1080/15715124.2008.9635341
Singh K (1981) On the asymptotic accuracy of Efron’s bootstrap. Ann Stat 9(6):1187–1195
Srinivas VV, Srinivasan K (2005) Hybrid moving block bootstrap for stochastic simulation of multi-site multi-season streamflows. J Hydrol 302(1):307–330. https://doi.org/10.1016/j.jhydrol.2004.07.011
Srivastav R, Sudheer K, Chaubey I (2007) A Simplified approach to quantify predictive and parametric uncertainty in artificial neural network hydrologic models. Water Resour Res 431:W10407. https://doi.org/10.1029/2006WR005352
Tankersley C, Graham W, Hatfield K (1993) Comparison of univariate and transfer Function Models of Groundwater Fluctuations. Water Resour Res 29(10):3517–3533. https://doi.org/10.1029/93WR01527
Taylor KE (2001) Summarizing multiple aspects of model performance in a single diagram. J Geophys Res 106(D7):7183–7192. https://doi.org/10.1029/2000JD900719
Funding
None.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflicts of interest
There is not any conflict of the interest.
Ethical approval
Not applicable.
Consent to participate
Not applicable.
Consent to publish
Not applicable.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix 1: Details point prediction models
1.1 Feedforward neural network
The target data for FFNN models was the GWL at time step (t) and inputs were selected from the GWL, average monthly temperature and precipitation data of previous time steps (e.g., t-1, t-2, …). The functional form of FFNN model can be presented as:
where \(GWL_{t}\), Pt and Tt are representing GWL, average monthly precipitation and average monthly temperature at time step t, and i, j and k are dominant lag times for each feature which are determined using MI measure between potential inputs and targets which is calculated via Eqs. 18–20.
where X and Y are discrete random variables, p is the probability function and H represents the entropy. However, due to LSTM’s capability to find the optimal dependencies dynamically at each prediction (Gers et al. 2002), optimal lag time selection for input data via MI is not used for LSTM.
1.2 Long short-term memory
The LSTM cell is consisting of four gates, each with an activation function, which are considered as neurons and are trainable. The mentioned gates are input, forget, modulation and output gates (see Fig.
9). The process in a LSTM cell is described via Eqs. 21–25 (Hochreiter and Schmidhuber 1997) as:
where W and b are weights and biases of each gate and h and x are hidden state and input vector, respectively. These gates make it possible for LSTM to decide to whether forget the information (via weights) or pass it to the next cell as:
where ft, it and gt are resultant vectors from forget, input and modulations gates respectively (Eq. 21–23). The cell state ct gets updated with Eq. 25. ⊙ indicates the element-wise multiplication and:
where σt is the non-linear activation function, ot is calculated in output gate (Eq. 24) and the hidden state ht is divided from output gate and passes to next cell (Rußwurm and Körner 2017).
For the point predicting purpose, one sequence input layer is used, which takes columns of input data. It is followed by a LSTM layer which aims to include long-term dependencies, by finding optimal dependency between time steps (dynamically inclusion of lag times). Then a fully connected layer is used to improve the regression and set the number of outputs. Finally, the regression layer is used which calculates the error between the targets and generated values, then back-propagates the effect of calculated error.
1.3 SARIMAX
The general formula of the SARIMAX(p, d, q)(P, D, Q)L model is defined as (Fathi et al. 2019):
where Yt is the value of target time series (GWL) at time t and ɛt refers to the residual at time t; p, q and d are orders of non-seasonal auto regressive, moving averages and differencing, respectively, while P, Q and D are seasonal orders of the same parameters. Also, \(\emptyset\) and Φ are autoregressive coefficients, \(\theta\) and \({\Theta }\) are moving averages coefficients of non-seasonal and seasonal parts, respectively. B refers to the back-shifting operator (BYt = Yt−1). \(\beta_{i}\) is the ith exogenous time series parameter, while Xt,i is the value of the ith exogeneous time series (precipitation and temperature). Finally, s refers to the seasonal period length.
Appendix 2: The concept of prediction intervals
The targets in modeling are expressed by (Khosravi et al. 2011a):
where ti is the ith target, ei is the error which moves the target away from the value that has been estimated by \(y_{i}\). By considering that the error is independent and identically distributed (iid) and \(\hat{y}_{i}\) is the output of model for estimating the mean of regression values we have:
While CIs are only dealing with the variance of \(\left[ {y_{i} - \hat{y}_{i} } \right]\), PIs are handling the difference between observed values ti and the estimated values \(\hat{y}_{i}\). Thus, PIs cover more uncertainty sources and are always wider than CIs and therefore:
Appendix 3: Tables and figures
Rights and permissions
About this article
Cite this article
Nourani, V., Khodkar, K., Paknezhad, N.J. et al. Deep learning-based uncertainty quantification of groundwater level predictions. Stoch Environ Res Risk Assess 36, 3081–3107 (2022). https://doi.org/10.1007/s00477-022-02181-7
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00477-022-02181-7