Welcome to the main repository for the RSI DNN Model for estimating reservoir capacity loss. This model estimates reservoir capacity loss (acre-ft) from twelve input parameters using a progressively increasing deep neural network model. It was developed using a dataset containing 467 records (sets of consecutive surveys) from 174 reservoirs located across the U.S. territory. The model was developed based on a 70/30 training/testing split and once identified as the best-fit model, it was recalibrated using the full dataset. This model includes the calibration from the full dataset which has the following performance metrics: R2 = 0.81 and mean absolute percent error (MAPE) values = 48%. Additional details on the model development and performance are provided by Cox et al., 2023 (https://doi.org/10.31223/X5PH3X).
Two files are provided for application of the DNN model: RSI_Capacity_Loss_Predictor.ipynb and calibrated_best_model.hdf5. The .ipynb file includes the python script for executing the model and the .hdf5 file is used as the calibrated input file for the script.
Twelve input parameters are required for the model which are developed based on the time period between two dates: mean monthly precipitation (in/mo.), maximum monthly precipitation (in), cumulative precipitation (in), drainage basin area (mi2), average basin latitude (°), drainage basin relief (ft), channel slope (ft/ft), average drainage basin curve number (dimensionless), total upstream normal storage (acre-ft), total upstream dam height (ft), drainage basin hydraulic length (ft), initial reservoir capacity (acre-ft). Additional details on how input parameters can be derived are provided by Cox et al. (2023).
The following provides a sample set of input data:
mean monthly precipitation (in/mo.): 1.348 maximum monthly precipitation (in): 5.529 cumulative precipitation (in): 250.4 drainage basin area (mi2): 262,200 average basin latitude (°): 46.09 drainage basin relief (ft): 12,500 channel slope (ft/ft): 0.001035 average drainage basin curve number (dimensionless): 76.33 total upstream normal storage (acre-ft): 108,800,000 total upstream dam height (ft): 147,000 drainage basin hydraulic length (ft): 12090000 initial reservoir capacity (acre-ft): 6,280,000
If you use the model, please cite the following publication:
Cox, A.L., Meyer, D., Botero-Acosta, A., Sagan, V., Demir, I., Muste., M., Boyd., P., and Pathak., C. (2023). Estimating Reservoir Sedimentation Using Deep Learning. EarthArXiv. DOI: https://doi.org/10.31223/X5PH3X
Feel free to send us feedback by filing an issue.
This project is licensed under the MIT License - see the LICENSE file for details.
This model was developed by Saint Louis University and the University of Iowa Hydroinformatics Lab (UIHI Lab, https://hydroinformatics.uiowa.edu) with input from the U.S. Army Corps of Engineers. This research was supported by the U.S. National Science Foundation (Award # 1948940) and the WATER Institute at Saint Louis University.