Hybrid Streamflow Modelling using Machine Learning and Multi-Model Combination

Master thesis Applied Data Science, Utrecht University

June 2023

This repository contains the code and data for the research study titled "Hybrid Streamflow Modelling using Machine Learning and Multi-Model Combination."

Data

The raw and preprocessed input data for the study can be accessed here. This dataset includes the necessary input variables for the streamflow modelling process.

The training and validation datasets used in the study are available here.

Code

The code in this repository implements Random Forest and Multiple Linear Regression analysis for streamflow modelling. It takes the provided input data and generates the corresponding results. In this study, Python was used for data preprocessing and visualization of results, while R was used for data analysis and machine learning modelling.

The folder with the Python code is structured as follows:

└── py
    ├── 00_combine_resample_meteo_data.ipynb
    ├── 01_extracting_obs.ipynb
    ├── 02_Upstream_average.ipynb
    ├── 03_Extract_station_data.ipynb
    ├── 04_Combine_files.ipynb
    ├── 05_Data_cleaning_Normalization.ipynb
    ├── 06_Sampling.ipynb
    ├── 07_Combine_CatchmentAttributes.ipynb
    ├── V0_discharge_plots.ipynb
    ├── V1_cor_plot.ipynb
    ├── V2_coefficients_plot.ipynb
    └── V3_KGE_plots.ipynb

The folder with the R code is structured as follows:

└── R
    ├── 1_randomForest
    │   ├── 00_RF_tune.Rmd
    │   ├── 01_Train_test_RF_MMC.Rmd
    │   ├── 03_Train_test_RF__MMC_elbe_maas.Rmd
    │   ├── 04_Train_test_rhine_benchmark.Rmd
    │   ├── 05_Train_test_elbe_maas_benchmark.Rmd
    │   ├── fun_2_1_hyperTuning.R
    │   ├── fun_2_2_trainRF.R
    │   └── fun_2_3_apply_optimalRF_avg.R
    ├── 2_multilinearRegression
    │   ├── 01_MLR_train_test.Rmd
    │   ├── 02_MLR_train_test_one_sample.Rmd
    │   ├── MLR_train_test.Rmd
    │   ├── fun_apply_MLR.R
    │   └── fun_trainMLR.R
    └── 3_visualization
        ├── 0_Vis_HyperTuning.Rmd
        ├── 1_VarImportance.Rmd
        ├── 2_Vis_KGE.Rmd
        ├── 3_Vis_KGE_one_sample.Rmd
        ├── 4_Hydrograph_all.Rmd
        ├── 5_Hydrograph_benchmark.Rmd
        ├── fun_3_4_hydrograph_benchmark.R
        └── fun_3_4_hydrograph_fdc_residuals_all.R

Results

The output and results of the Random Forest and Multiple Linear Regression analysis can be found here. These results showcase the performance and accuracy of the implemented models.

Please refer to the provided DOI links to access the respective data and results.