Improving Reservoir Capacity Loss Prediction with Machine Learning and Sensitivity Analysis of Key Hydrologic Drivers

A revised deep neural network (DNN) model was developed for predicting reservoir capacity loss 5 using an expanded dataset with sites from the Reservoir Sedimentation Information (RSI) system 6 which covers the contiguous U.S. The dataset included reservoirs with as few as two surveys and 7 was subjected to anomaly detection using a Kolmogorov–Smirnov–Efron approach. The DNN 8 model was recalibrated with transformed inputs, including log-scaling, min–max normalization, 9 and principal component transformation, reducing twelve variables to five principal components. 10 Model performance improved relative to prior work, achieving an R2 of 0.97 and reducing 11 prediction errors across multiple metrics. A variance-based global sensitivity analysis using 12 Sobol’ indices was conducted to evaluate the contribution of input variables and their 13 interactions to model output variance. Results indicated that individual variables had limited 14 direct influence (low first-order indices), while total sensitivity indices revealed that higher-order 15 interactions dominated model behavior. Key contributing factors included mean precipitation, 16 channel slope, and initial reservoir capacity. The study demonstrates both the capability of 17 machine-learning models to accurately represent reservoir sedimentation and the importance of 18 sensitivity analysis for interpreting model results and guiding practical decision-making.

Authors

Publication Details

Journal
EarthArXiv (California Digital Library)
Published
2026-11-20
DOI
https://doi.org/10.31223/x5wv50
Primary Topic
Hydrology and Sediment Transport Processes
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Improving Reservoir Capacity Loss Prediction with Machine Learning and Sensitivity Analysis of Key Hydrologic Drivers

Marian Muste, Travis A. Dahl, Alejandra Botero‐Acosta, Paul Boyd et al.
EarthArXiv (California Digital Library)
Hydrology and Sediment Transport Processes
preprint

Improving Reservoir Capacity Loss Prediction with Machine Learning and Sensitivity Analysis of Key Hydrologic Drivers

Marian Muste, Travis A. Dahl, Alejandra Botero‐Acosta, Paul Boyd, Ibrahim Demir, Amanda Cox
preprint en

Abstract

A revised deep neural network (DNN) model was developed for predicting reservoir capacity loss 5 using an expanded dataset with sites from the Reservoir Sedimentation Information (RSI) system 6 which covers the contiguous U.S. The dataset included reservoirs with as few as two surveys and 7 was subjected to anomaly detection using a Kolmogorov–Smirnov–Efron approach. The DNN 8 model was recalibrated with transformed inputs, including log-scaling, min–max normalization, 9 and principal component transformation, reducing twelve variables to five principal components. 10 Model performance improved relative to prior work, achieving an R2 of 0.97 and reducing 11 prediction errors across multiple metrics. A variance-based global sensitivity analysis using 12 Sobol’ indices was conducted to evaluate the contribution of input variables and their 13 interactions to model output variance. Results indicated that individual variables had limited 14 direct influence (low first-order indices), while total sensitivity indices revealed that higher-order 15 interactions dominated model behavior. Key contributing factors included mean precipitation, 16 channel slope, and initial reservoir capacity. The study demonstrates both the capability of 17 machine-learning models to accurately represent reservoir sedimentation and the importance of 18 sensitivity analysis for interpreting model results and guiding practical decision-making.

EarthArXiv (California Digital Library)
Hydrology and Sediment Transport Processes
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon