Data and code from: Counterfactual modelling isolates sand mining impacts, revealing it as a key driver of Mekong delta destabilization
Data files
Jun 09, 2026 version files 858.13 MB
-
Ch3_MiningImpact.zip
858.12 MB
-
README.md
13.58 KB
Abstract
This dataset supports a study on the impacts of river sand mining in the Vietnamese Mekong Delta using Delft3D-FM numerical modelling and counterfactual scenario analysis. It includes selected model inputs, reduced model outputs, spatial datasets, post-processing scripts, and figure-ready results used to quantify mining-induced changes in river morphology, hydrodynamics, sediment transport, and salinity intrusion. The dataset is based on two simulations: a mining scenario representing observed sand extraction and a no-mining scenario with identical boundary conditions but without dredging. The comparison between these scenarios allows the isolated effects of sand mining to be assessed. The archive includes NetCDF model outputs, GeoTIFF difference maps, shapefiles, Python scripts, and post-processing outputs. Some restricted or very large datasets, including full bathymetry, upstream discharge, suspended sediment concentration, and full-resolution model outputs, are not included. Reduced datasets are provided to support reproducibility of the main analysis workflow and figures.
Dataset DOI: 10.5061/dryad.mcvdnckg3
Description of the Data and File Structure
This dataset was generated to investigate the environmental impacts of river sand mining in the Vietnamese Mekong Delta using a process-based Delft3D-FM modelling framework.
Two numerical simulations were performed: (1) a mining scenario representing observed sand extraction, and (2) a counterfactual no-mining scenario with identical boundary conditions. By comparing these simulations, the dataset enables quantification of sand-mining-induced changes in river morphology, sediment transport, hydrodynamics, and salinity intrusion.
The dataset includes model input templates, selected simulation outputs, spatial datasets, post-processing scripts, and figure-ready results. Due to data size and access restrictions, some input datasets including full bathymetry, upstream discharge boundary conditions from the Mekong River Commission (MRC), suspended sediment concentration, and full-resolution model map outputs are not included. Reduced datasets and all post-processing scripts are provided to support reproducibility of the main analysis workflow and figures.
Upstream boundary condition data from the Mekong River Commission (MRC) stored in Model_input/MRC_data are confidential and subject to third-party approval. These data are not publicly available but can be requested directly from the MRC at www.mrcmekong.org.
Files and Variables
File: Ch3_MiningImpact.zip
This archive contains Delft3D-FM model inputs, selected model outputs, Python scripts, shapefiles, and post-processing results used to analyse sand mining impacts in the Vietnamese Mekong Delta.
Folder Structure
1. Code
Contains all Python scripts used for reading NetCDF model outputs, processing variables, computing mining versus no-mining differences, and generating all figures in the study. Each script is named according to the figure or analysis it produces. File paths in all scripts must be updated to match the user's local directory before running.
2. Delft3DFM_modelling
Contains model setup files, simulation outputs, mining rate data, and MRC boundary condition data.
(a) Model_input/dflowfm
Contains all Delft3D-FM model input files required to run the simulation. FlowFM.mdu is the main model configuration file defining all simulation settings. The boundary condition files (.bc) define time-varying forcing for discharge, water level, salinity, temperature, and sediment at the open boundaries. The external forcing files (.ext, FlowFM_bnd.ext) link boundary data to boundary locations. The sediment and morphology files (.sed, .mor) define sediment fraction properties and morphodynamic settings. The boundary polyline files (.pli) define the locations of open boundaries. The spatial initial condition and parameter files (.xyz) define spatially varying initial water level, salinity, sediment concentration, sediment thickness, and friction coefficients across the domain. The model grid file (VMD_coarse1_net.nc) contains the unstructured computational mesh. The sand mining zone polygon file (SM_133_zones.pol) defines 133 individual mining zones. The dredging time series file (dredging_timeseries.ddt) contains time-varying extraction volumes. The DIMR configuration file (dimr_config.xml) controls the model execution workflow. The HPC job submission script (mining_scenario.pbs) runs the simulation on a cluster using 30 parallel processes.
(b) Model_input/MRC_data
Contains upstream boundary condition and validation data from the Mekong River Commission. These data are confidential and not publicly available. Figure1_Input_Boundry_Condition.xlsx contains hourly discharge (m3/s) and suspended sediment concentration (g/m3) at Chau Doc and Tan Chau stations used as upstream model input. Obs_Q_H_Sal_SSC.xlsx is an empty template showing the structure used for model validation data including water level, discharge, salinity, and SSC. Obs_vs_Sim_Bathymetry.csv contains observed and simulated bed elevation data along cross-sectional transects used for bathymetric validation.
(c) model_output
Contains simulation output files organised into three subfolders.
The mining subfolder contains: FlowFM_0000_his.nc with time series at observation points including water level, velocity, salinity, and SSC for the mining scenario; mining_BedLEV_SedThick_FirstLast.nc with spatially distributed bed level (m) and sediment thickness (m) as monthly mean raster grids for the first and last months of the simulation; mining_BedLev_SedThick_Monthly.nc with monthly mean bed level and sediment thickness raster grids for all months of the 2017 to 2021 simulation period; and mining_KeyVariables_Summary.nc with selected variables including initial, average, and final bed level, water level, sediment thickness, salinity, and time-averaged total sediment transport magnitude for quick access and plotting.
The nomining subfolder contains the same file structure as the mining subfolder: FlowFM_0000_his.nc, nomining_BedLEV_SedThick_FirstLast.nc, and nomining_BedLev_SedThick_Monthly.nc, all containing equivalent outputs for the no-mining scenario.
The difference subfolder contains GeoTIFF raster maps of the difference between no-mining and mining scenarios (no-mining minus mining) used as input for spatial impact figures. Bed_Difference_Map.tif shows the difference in bed level (m). Dry_Season_Bed_Shear_Stress_Change.tif and Wet_Season_Bed_Shear_Stress_Change.tif show the seasonal difference in bed shear stress (N/m2). Dry_Season_Salinity_Change.tif and Wet_Season_Salinity_Change.tif show the seasonal difference in salinity (PSU). Sand_Fraction_Change_pp.tif shows the difference in sand fraction (percentage points).
(d) MiningRate
Contains spatial and tabular datasets describing current sand mining activity, all derived or abstracted from Kumar et al. (2024). The annual incision raster files (annual_incision_2017.tif through annual_incision_2022.tif) are annual maps of riverbed incision (m) caused directly by sand mining, where negative values indicate bed lowering. avg_incision_2017_2022.tif shows the average annual bed incision over the full 2017 to 2022 period. Figure1_number of boats.csv contains sand mining vessel detection data used to illustrate mining activity trends; columns include SN (observation number), confidence (detection score 0 to 1), class (0 = barge with cranes, 1 = sand transport boat, 2 = other), name (class label), Date1 (MM/DD/YYYY), and Year. The Current_Mining subfolder contains two shapefiles: Current_SandMining_Zones_131 defining 131 individual active mining zone polygons, and Current_SandMining_Reaches_6 defining 6 river reach polygons used for reach-level mining volume aggregation. All files use UTM Zone 48N (EPSG:32648).
3. Post_processing_output
Contains all figures and extracted data files generated by the Python scripts in the Code folder. All subfolders contain figures in PNG and PDF format and Excel files with the underlying data. Final publication figures were produced by combining these outputs in Adobe Illustrator.
Boxplot_Seasonal_SSC_Q contains seasonal boxplot figures and data for SSC and discharge comparing both scenarios. Cal&Val contains calibration and validation figures and data comparing simulated and observed water level, discharge, salinity, and SSC. Salinity_Intrusion contains figures and data quantifying salinity intrusion extent and seasonal changes; key data files are Daily_Salinity_Intrusion_Comparison.csv and Monthly_Salinity_Intrusion_Summary_Long.csv described in detail below. Sediment_Budget_Extraction contains sediment budget analysis figures and data. Sediment_Transport_Impact contains figures and data on mining impacts on sediment transport. Spatial_Impact contains spatial raster maps of mining impacts on bed level, sediment thickness, sand fraction, and bed shear stress. Spatial_Map contains study domain and context maps. Station_CrossSection_TimeSeries_Comparison contains time series comparison figures and data at monitoring stations and cross-sections. Tidal_River_Analysis contains tidal decomposition figures and metrics for stations along the Hau and Tien river branches.
4. VMD_shpe
Contains all spatial vector data related to the study area used for mapping and figure generation. Clean_Water.shp is the water body boundary of the Vietnamese Mekong Delta used as a base map layer. Provinces.shp contains Vietnam administrative province boundaries. VMD.shp contains the outer boundary of the Vietnamese Mekong Delta study domain. Salinity_Line.kmz contains river centrelines used for salinity intrusion distance computation. The ModelBoundry_SHP subfolder contains the model domain boundary shapefile. The Vietnam_TruongSaHoangSa subfolder contains the shapefile of Vietnam including offshore island territories. Each shapefile consists of .shp, .dbf, .shx, .prj, and .cpg component files. All shapefiles use WGS84 geographic coordinates (EPSG:4326) unless otherwise noted.
Variables and Units
Discharge is in m3/s. Water level and depth are in m. Flow velocity is in m/s. Bed shear stress is in N/m2. Sediment thickness is in m. Bed level is in m. Salinity is in PSU. Suspended sediment concentration (SSC) is in g/m3. Sediment transport rate is in m3/s/m. Salinity intrusion distance is in km. Sand fraction change is in percentage points (pp). Bed incision is in m where negative values indicate lowering.
Missing Values
NetCDF files may contain NaN values where grid cells fall outside the model domain or clipped area. No specific missing value flag is used. Blank cells in Excel template files (Obs_Q_H_Sal_SSC.xlsx) indicate that no observational data are included; the file is provided as a structural template only.
Variable Descriptions for Key Tabular Files
Daily_Salinity_Intrusion_Comparison.csv contains daily salinity intrusion distances for four river branches under both scenarios. Date is the timestamp in DD/MM/YYYY HH:MM format. River is the branch name (Hau, CoChein, Ham, or Mytho). No_Mining_km is the intrusion distance in km under the no-mining scenario. Mining_km is the intrusion distance in km under the mining scenario. Diff_km is Mining_km minus No_Mining_km where positive values indicate greater intrusion under mining.
Monthly_Salinity_Intrusion_Summary_Long.csv contains monthly aggregated statistics derived from the daily file. River is the branch name. Month is the integer month (1 to 12). Month_Name is the abbreviated month name. No_Mining_Mean_km and Mining_Mean_km are monthly mean intrusion distances in km. Diff_Mean_km is the monthly mean difference in km. Diff_SD_km is the standard deviation of daily differences in km. Diff_Max_km and Diff_Min_km are the monthly maximum and minimum differences in km. Diff_P10_km and Diff_P90_km are the 10th and 90th percentiles of daily differences in km.
Figure1_number of boats.csv contains vessel detection data abstracted from Kumar et al. (2024). SN is the observation number. confidence is the detection score between 0 and 1. class is the vessel type where 0 is barge with cranes, 1 is sand transport boat, and 2 is other vessel. name is the text label. Date1 is the observation date in MM/DD/YYYY format. Year is the calendar year.
Code and Software
Delft3D-FM (Deltares): version 2022 or later recommended, Windows or Linux HPC supported, required to run the model using files in Model_input/dflowfm.
Python: version 3.8 or higher. Required libraries are numpy, pandas, xarray, matplotlib, geopandas, rasterio, netCDF4, dfm_tools, xugrid, scipy, shapely, and contextily.
GIS software (optional): QGIS recommended for viewing shapefiles and raster outputs.
Spreadsheet software: Microsoft Excel or equivalent for tabular data files.
Workflow: run Delft3D-FM to generate model outputs, then use Python scripts in the Code folder to process variables, compute differences, and generate figures. GIS software can be used optionally to visualise spatial outputs.
Access Information
This dataset is provided as a standalone archive and is not hosted on additional public repositories. Data were derived from Mekong River Commission (MRC) upstream discharge and sediment data (not included due to access restrictions), and published datasets from Lau et al. and Binh et al. used for bathymetry and sediment information. Users requiring full model reproduction should request restricted datasets directly from the MRC or use equivalent datasets formatted for Delft3D-FM.
License
This dataset is released under the CC0 1.0 Universal Public Domain Dedication. Users are free to copy, modify, distribute, and use the data for any purpose without restriction.
Recommended Citation
If you use this dataset, please cite the associated study:
Kumar, S., Park, E., Tran, D.D., Huang, J., Vo, T.Q., Kästner, K., Kantoush, S., Binh, D.V., Wang, J., and Switzer, A.D. Counterfactual Modelling Isolates Sand Mining Impacts, Revealing It as a Key Driver of Mekong Delta Destabilization. Science Advances.
Contact
Sonu Kumar, Earth Observatory of Singapore (NTU).
Email: nie22.sk2447@e.ntu.edu.sg.
Support is limited to clarification of dataset structure and code usage.
This dataset was generated through an integrated modelling and post-processing workflow designed to quantify the environmental impacts of sand mining in the Vietnamese Mekong Delta (VMD).
Model framework and setup
Hydrodynamic, sediment transport, morphodynamic, and salinity processes were simulated using the Delft3D-FM (Flexible Mesh) modelling system. The model domain covers the full fluvial-estuarine continuum of the VMD, including the major distributary channels (Tien and Hau Rivers) and their coastal outlets. An unstructured computational grid was used to accurately represent complex river geometry, channel bifurcations, and estuarine boundaries.
Model boundary conditions include upstream river discharge, suspended sediment concentration (SSC), and downstream tidal forcing. Upstream boundary data were obtained from the Mekong River Commission (MRC) monitoring stations at Chau Doc and Tan Chau, providing hourly discharge (m³/s) and SSC (g/m³). Downstream tidal boundary conditions were applied at the coastal outlets. Bathymetric data were derived from field surveys and published datasets, supplemented where necessary to ensure full domain coverage. Sediment transport was represented using a two-fraction system covering cohesive and non-cohesive sediment, and morphodynamic evolution was simulated through bed-level updating during model runs. The simulation was run for the period 2017 to 2021 using the DIMR (Delta Interface Model Runner) framework on a high-performance computing cluster with 30 parallel MPI processes.
Scenario design and counterfactual approach
To isolate the impacts of sand mining, a counterfactual modelling framework was implemented using two scenarios. The mining scenario includes spatially distributed sand extraction zones representing observed extraction activity, defined using 133 mining zone polygons (SM_133_zones.pol) and time-varying extraction volumes (dredging_timeseries.ddt) configured within the Delft3D-FM dredging module. A target dredge depth of 15 m was applied with a maximum volumetric extraction rate of 2488.97 m3/s, with extracted material disposed to a designated dump location. The no-mining scenario uses an identical model configuration with the dredging component removed. All other boundary conditions and model parameters were kept identical between the two scenarios. Differences between simulations therefore represent the isolated effect of sand mining on the river system.
Model outputs and data structure
The model produces outputs in NetCDF format. History files (FlowFM_0000_his.nc) contain time series data at predefined observation points and cross-sections, including water level, flow velocity, discharge, salinity, suspended sediment concentration, bed shear stress, and sediment thickness. Full spatial map outputs were generated across the model grid but are not included in this dataset due to their large file size. Instead, the following reduced spatial datasets are provided. Monthly mean bed level and sediment thickness raster files (mining_BedLev_SedThick_Monthly.nc and nomining_BedLev_SedThick_Monthly.nc) contain spatially distributed monthly mean outputs for all months of the simulation period clipped to the mining domain. First and last month raster files (mining_BedLEV_SedThick_FirstLast.nc and nomining_BedLEV_SedThick_FirstLast.nc) contain bed level and sediment thickness for the first and last months only. A selected variables summary file (mining_KeyVariables_Summary.nc) contains initial, average, and final bed level, water level, sediment thickness, salinity, and time-averaged total sediment transport magnitude for quick access and plotting.
Post-processing and analysis
Post-processing was performed using Python-based workflows. Scripts were developed to read NetCDF outputs using xarray, dfm_tools, and xugrid; extract key hydrodynamic, sediment, and salinity variables; compute differences between mining and no-mining scenarios; derive impact metrics such as bed-level change, sediment transport variation, and salinity intrusion distance; and generate spatial maps and time-series plots. Salinity intrusion was quantified by rasterizing salinity fields at 100 m resolution, sampling along predefined river centrelines at 100 m spacing, and identifying the farthest upstream point exceeding a 2 PSU threshold for each time step. Spatial analyses were supported using shapefiles defining river boundaries, model domain extents, and administrative regions. Raster outputs in GeoTIFF format were generated for spatial comparison and visualisation of erosion, deposition, and salinity change patterns.
Derived datasets
In addition to direct model outputs, the dataset includes derived products. Annual and average bed incision maps (GeoTIFF) show riverbed lowering caused directly by sand mining for each year from 2017 to 2022 and as a 2017 to 2022 average, derived from Kumar et al. (2024). Daily and monthly salinity intrusion distance time series and comparison tables quantify intrusion along four river branches (Hau, Co Chien, Ham Luong, and My Tho) under both scenarios. Processed difference maps (no-mining minus mining) in GeoTIFF format show the spatial impact of sand mining on bed level, bed shear stress, salinity, and sand fraction for dry and wet seasons. Figure-ready datasets and plots used in the associated study are provided in the Post_processing_output folder.
Data limitations and availability
Some input datasets are not included due to data access restrictions and file size limitations. Upstream discharge, SSC boundary conditions, and high-resolution bathymetry were obtained from the Mekong River Commission (MRC) and published studies and are subject to data sharing restrictions. Full-resolution spatial model outputs are not included due to their large size. To support reproducibility, reduced datasets and all post-processing scripts are provided, allowing users to reproduce the analysis workflow and figures without requiring the full raw model outputs.
Reproducibility
All figures in the associated study were generated directly from the provided scripts and datasets. By following the workflow of model outputs to Python scripts to processed results, users can reproduce the main analytical steps and visualisations presented in the study. File paths in all Python scripts must be updated to match the user's local directory before running.