Checking for non-preferred file/folder path names (may take a long time depending on the number of files/folders) ...

Data and Code for Explainable AI Analysis of Arctic Coastal Groundwater Dynamics
Authors:
Owners: This resource does not have an owner who is an active HydroShare user. Contact CUAHSI (help@cuahsi.org) for information on this resource.
Type: Resource
Storage: The size of this resource is 27.6 MB
Created: May 31, 2026 at 2:47 a.m. (UTC)
Last updated: May 31, 2026 at 5:34 a.m. (UTC)
Citation: See how to cite this resource
Content types: CSV Content 
Sharing Status: Public
Publication Status:
Pending Metadata Review
This resource is under review for minimum metadata and to ensure that it adheres to community guidelines.
Views: 12
Downloads: 0
+1 Votes: Be the first one to 
 this.
Comments: No comments (yet)

Abstract

This repository contains the machine-learning-ready datasets and Jupyter Notebook used in the study of groundwater storage and submarine groundwater discharge dynamics along the Arctic coast of Simpson Lagoon, Alaska. The related manuscript investigated the time-dependent influence of atmospheric, thermal, and oceanic controls on groundwater storage and discharge using three deep learning architectures (1D-CNN, LSTM, GRU).
The original groundwater and lagoon field observations collected during the 2022 thaw and summer seasons (June-September) are archived separately in Demir et al. (2026a). The datasets provided here consist of processed and quality-controlled groundwater lagoon observations merged with atmospheric and oceanic variables obtained from reanalysis products (MERRA-2 and GLDAS) and weather stations. Reanalysis atmospheric variables were obtained from NASA GES DISC (Acker & Leptoukh, 2007) a ~15 km × 15 km region surrounding the site. This repository also includes code for data pre-processing, deep learning model training, model evaluation, and eXplainable Artificial Intelligence (XAI) analyses using SHapley Additive exPlanations (SHAP). These resources enable full reproduction of the analyses used in the related manuscript, and provide a framework for applying XAI to Arctic coastal groundwater systems and other data-limited hydrologic systems.

References:
Acker, J. G., & Leptoukh, G. (2007). Online analysis enhances use of NASA earth science data. Eos, Transactions American Geophysical Union, 88(2), 14–17. https://doi.org/10.1029/2007EO020003
Demir, C., J. A. Guimond, E. Bristol, E. Bullock, J. W. McClelland, M. A. Charette, M. B. Cardenas (2026a). Hydrological and thermal field measurements at an Arctic coastal site, Simpson Lagoon, Alaska (2021-2022), HydroShare, https://doi.org/10.4211/hs.b2f0fcd9f2834f79a4b9216ad717eb69
Related manuscript:
Demir, C., Gomez-Velez, J. D., Guimond, J. A., McClelland, J. W., Charette, M. A., and Cardenas, M. B. Time-varying hydroclimatic and oceanic controls on Arctic submarine groundwater discharge inferred using explainable AI.

Subject Keywords

Coverage

Spatial

Coordinate System/Geographic Projection:
WGS 84 EPSG:4326
Coordinate Units:
Decimal degrees
Place/Area Name:
Simpson Lagoon coast
North Latitude
70.5061°
East Longitude
-149.6360°
South Latitude
70.5014°
West Longitude
-149.6504°

Temporal

Start Date:
End Date:

Content

README.md

Three deep learning models were used:

  • 1D Convolutional Neural Network (CNN)
  • Long Short-Term Memory (LSTM)
  • Gated Recurrent Unit (GRU)

Each model was trained for two prediction targets:

  • Target-Head: groundwater head (proxy for groundwater storage)
  • Target-H-LL: groundwater head minus lagoon level (proxy for groundwater discharge)

Model Files

.keras and .h5 files contain trained model weights and architectures.

SHAP files

*_SHAP_*.npy

Contain raw SHAP values produced using the trained model.

SHAP Plotting Files

*_SHAP_*_plotting.npy

Contain post-processed SHAP arrays used to generate manuscript figures and analyses.

Input Data

This directory contains the machine-learning-ready predictor datasets before and after feature selection.

all_variables_beforeVIF.csv

Complete predictor dataset assembled from field observations, reanalysis products, and weather station observations prior to feature selection via Variance Inflation Factor (VIF) filtering.

Target-*_beforeVIF.csv

Predictor dataset before VIF for each target.

Target-*_afterVIF.csv

Predictor dataset after VIF for each target. This is the data array used in ML train-val-test.

Inter-Model SHAP Outputs

inter-model_average_SHAP_outputs/SHAP_model-average_Target-*.npy

These files contain normalized SHAP values averaged across the CNN, LSTM, and GRU models. These ensemble SHAP outputs were used in the manuscript to reduce model-specific bias and provide a model-uncertainty-aware estimate of feature importance.

Jupyter Notebook

Demir_et_al_DL_XAI_Arctic_SGD_drivers.ipnyb

This notebook contains the complete worflow used in the study, including:

  • Data preprocessing
  • Feature selection via VIF
  • Deep learning model training
  • Model evaluation
  • SHAP analyses
  • Figure generation

Related Resources

This resource is referenced by Demir, C., Gomez-Velez, J. D., Guimond, J. A., McClelland, J. W., Charette, M. A., and Cardenas, M. B. Time-varying hydroclimatic and oceanic controls on Arctic submarine groundwater discharge inferred using explainable AI.

Credits

Funding Agencies

This resource was created using funding from the following sources:
Agency Name Award Title Award Number
U.S. National Science Foundation LTER: Beaufort Sea Lagoons: An Arctic Coastal Ecosystem in Transition OPP-1656026
U.S. National Science Foundation The physical and chemical dynamics of groundwater flow across the land-sea interface in Arctic lagoon ecosystems OPP-1938820
U.S. National Science Foundation The physical and chemical dynamics of groundwater flow across the land-sea interface in Arctic lagoon ecosystems OPP-1938873

Contributors

People or Organizations that contributed technically, materially, financially, or provided general support for the creation of the resource's content but are not considered authors.

Name Organization Address Phone Author Identifiers
James W. McClelland Marine Biological Laboratory, The Ecosystems Center
M. Bayani Cardenas The University of Texas at Austin, Earth and Planetary Sciences
Jesus D. Gomez-Velez University of Iowa, Civil and Environmental Engineering
Julia Guimond Woods Hole Oceanographic Institution, Applied Ocean Physics & Engineering
Matthew A. Charette Woods Hole Oceanographic Institution, Marine Chemistry and Geochemistry

How to Cite

Demir, C. (2026). Data and Code for Explainable AI Analysis of Arctic Coastal Groundwater Dynamics, HydroShare, http://www.hydroshare.org/resource/c74c675f749247a4911679e7a150e383

This resource is shared under the Creative Commons Attribution CC BY.

 http://creativecommons.org/licenses/by/4.0/
CC-BY

Comments

There are currently no comments

New Comment

required