To find the best gridding method for your data, the best thing to do is just to experiment with different gridding methods and grid settings until you get the result you desire. To help choose the best gridding method, here are some suggestions:
- When creating a grid file you can usually accept all of the default gridding parameters and generate a grid file that represents your data well. Under most circumstances, the recommended gridding method is Kriging with the default linear variogram. This is the selected default gridding method because it gives good results for most XYZ data sets. Kriging is flexible and useful with many data types.
- There is a script in the Surfer installation directory, under the \Samples\Scripts folder named GridData_Comparison.bas that runs a data file through several gridding methods. You can run the script and visually choose the “best” method for your data. This is a good way to narrow down the gridding methods to the top 2 or 3, and then you can fine tune the parameters to see which works best for you. To run the script:
- Open Scripter that comes with Surfer, click File | Open, navigate to the Surfer installation directory and click \Samples\Scripts\.
- Select the GridComparison.bas script and click Open.
- Click Script | Run to run it.
- Follow the prompts. A contour map will be generated from 8 of the most popular gridding methods. You may find one or two that you think look the best. In that case, you can focus on those gridding methods and then fine tune the results with the options (2 and 3) below.
- One method is determining the best gridding method based on number of raw data points:
- For very small data sets (less than 10 points), use Kriging
- or Radial Basis Function. If you want only to define the trend of the data, you can use Polynomial Regression. With 10 or fewer points, gridding is extremely fast, so you might want to try the different methods to determine the most effective method for your data.
- With small data sets (<250 observations), Kriging with the default linear variogram, or Radial Basis Function with the multiquadric function produce good representations of most data sets.
- With moderate-sized data sets (from 250 to 1000 observations), Triangulation with Linear Interpolation is fast and creates a good representation of your data. Although Kriging or Radial Basis Function generate the grids more slowly, they also produce good data representations.
- For large data sets (>1000 observations), both Minimum Curvature and Triangulation with Linear Interpolation are quite fast, and both produce good representations. As with most other data sets, Kriging or Radial Basis Function probably produce the best maps but are quite a bit slower.
- If you want your data points honored exactly when the point coincides with the grid node being interpolated, use Inverse Distance to a Power, Kriging (with no nugget effect), Nearest Neighbor, Radial Basis (with no R^2 value), Modified Shepard’s Method (with no smoothing factor), Triangulation with Linear Interpolation, or Natural Neighbor.
- If you want to use the a smoothing interpolator to massage your data and reduce the effects of small-scale variability between neighboring points, you can try Inverse Distance to a Power, Kriging (with a nugget effect),Polynomial Regression, Radial Basis (with an R^2 value), Modified Shepard’s Method (with a smoothing factor), Local Polynomial, or Moving Average.
- If you like the look of a map produced with the Kriging gridding method, but wish to include a fault file containing fault traces, then Minimum Curvature might be the best option for you.
- If your data is already on a regularly spaced grid, you might want to try the Nearest Neighbor gridding method, setting the grid spacing to be the same as the spacing of your data points.
- If you want to create a grid of data statistics, such as density or count within an area, then use the Data Metrics gridding method and set the metric to be used in the Advanced Options.
Once you find the best gridding method, you can fine tune it using the advanced options for that method. Each gridding method has its own set of options. Some of the options are the same or similar to those of other gridding methods, and other options are specific only to a specific gridding method. For example:
- If your data is taken at an angle (not perfectly n-s and e-w), then you may be able to set the Anisotropy to account for this (if the gridding method you choose supports anisotropy).
- You could also try setting the search ellipse ratio and angle. If no data points are found inside the search ellipse, then the grid node is blanked.
- You could use a fault or breakline file to help constrain the gridding, if the gridding method you choose supports faults or breaklines.
For more information about the gridding methods, please see
- Knowledge base articles:
- What types of gridding methods are available?
- A Basic Understanding of Surfer Gridding Methods – Part 1
- A Basic Understanding of Surfer Gridding Methods – Part 2
- What gridding method should I use? Gridding for non-geostatisticians
Updated February 3, 2017