BayesGWL

1. Study Area & Interpolation Grid

Upload a shapefile (.zip containing .shp/.dbf/.shx, ideally .prj) defining your study area boundary - it appears on the map below immediately. Then choose your settings; SRTM elevations are fetched for the full bounding-rectangle grid first, and if you set a thinning cell size, that SRTM-elevation data is the source that gets thinned. Generate Grid is the last step - it (re)builds the grid fresh from whatever settings are currently set, every time you click it.

Upload shapefile (.zip)

SRTM dataset Grid spacing (degrees)

Thin to cell size (degrees, optional)

Generate target grid over

2. Observation Data

Enter borehole observations as Name, Longitude, Latitude, Elevation (mamsl), Water Level (mbgl - metres below ground level, the standard dip-meter reading). Upload a CSV to fill the table below, or edit it directly.

Upload CSV 0 points

Name	Longitude	Latitude	Elevation (mamsl)	Water Level (mbgl)

Include static water level sites from the South African National Groundwater Archive (2019) within the study area extent

NGA elevation source

3. Semivariogram

Starting values are derived automatically from your data as soon as it loads: omnidirectional (Tolerance 360°), Lag width at about half the mean nearest-neighbour spacing, and enough lags to reach roughly half the maximum distance across the study area. Adjust the sliders to refine - the graph updates live - or use Reset to return to the recommended settings. Lags built from very few point-pairs (usually the farthest ones) get noisy and can be ignored.

Direction (deg) 0

Tolerance (deg) 360

Lag width (m) 500 max

Number of lags 20 max

Log-transform

4. Fit Variogram Model

Model

5. Predict Water Levels (Bayesian Estimation)

Co-kriges water level against surface elevation as a "qualified guess" - water levels usually track topography. A second semivariogram is auto-fit to elevation (same model type and lag settings as step 3) alongside the water-level one.

Nearest neighbors Guess scaling

Interpolate over Interpolation always uses the full grid; this sets how much of the surface is shown.

Clip predicted level to surface elevation (water table can't exceed ground level)

Water level surface opacity 75%

6. Cross-sections

Available once a prediction has been run at the study-area grid (step 5). Surface elevation is shown as a line; predicted water level as a blue-shaded area. The cross-section position is marked on the water level map in step 5.

Show

South-North position

West-East position

7. Report

Generates a full report of the pipeline: dataset, study area and water-level maps, semivariogram model, and LOO error analysis - with graphs and stats.

8. Export

Download the data as CSV. "Grid Data" is the generated target grid with its SRTM elevations; "Water Level Data" is the observation + NGA dataset used for the interpolation; "Interpolated Water Levels" is the predicted result at every grid point.