What is the most suitable interpolation method (IDW, spline, kriging, natural neigbours) for evenly and dense distributed points (attached picture .jpg)? and if you recommend any can you explain shortly why this method. Thank you.
Which method to use for interpolation usually depends on the distribution of the sample data and the type of surface to be created. The spatial data interpolation methods are all based on the spatial correlation. No matter which interpolation method is selected, for known points (the more data, the wider the distribution, and the more uniform the distribution), the closer the interpolation result will be to the actual situation.
For this problem, although the sample points are evenly distributed, there are differences in distance, and the spatial relationship should be analyzed by combining specific sample values. When there are spatial autocorrelation or directional trends in the sample points, Kriging (geostatistics) is the most suitable interpolation method. Other methods (i.e. IDW, local polynomial, moving average, nearest neighbor) need to refer to the definition of the relevant method to determine whether it is applicable.
https://docs.qgis.org/3.4/en/docs/gentle_gis_introduction/spatial_analysis_interpolation.html
https://pro.arcgis.com/en/pro-app/tool-reference/spatial-analyst/an-overview-of-the-interpolation-tools.htm
Daikun Wang there is no differences between sample points. I‘m using ERA5 and UERRA reanalysis of gridded points 5.5x5.5 km resolution for UERRA and 10x10 km for ERA5. I’m analysing precipitation data from reanalysis with the official meteorological data from stations.
Dainius Frišmantas as I could see the data you are analyzing is the data in the precipitation grid, which I also use, but the GPCC satellite database as well as rainfall stations.
Answering your question, for precipitation data interpolations, there are some studies here in Brazil that claim that the best method to interpolate is Kriging, since they better simulate its spatiotemporal variability.
I hope I helped, if you want some references, see two brief articles I made about spatialization and rainfall trends for a Brazilian state and another for a hydrographic region. You can also look for other papers on the same subject, so consult in the academic setting.
1- https://periodicos.ufpe.br/revistas/jhrs/article/view/240869.
2- https://revistas.ufpr.br/revistaabclima/article/view/61516.
Att,
David Ferreira
Dainius Frišmantas Well received. I'm not familiar with the products of weather reanalysis data, but I think Kriging is the most commonly used. As for the most suitable interpolation method of meteorological station data, it is suggested that you search the published results by keywords.
DEAR SIR
In the Kriging method, searching for semivariograms and determining the best fitted semivariogram model and preparing the required statistical assumptions are time-consuming and trial and error procedure. While in IDW model they were not required to perform these assumptions.
The results also revealed that although the IDW is relatively simple and easy in use, but is less accurate than ordinary Kriging
Testing of various methods and parameters, by comparing and ranking their cross-correlation errors, show that the ordinary Kriging method (with its difficulty) using the different semivariogram models, provides the overall best results compared to IDW method. The Kriging methods were compared and selected following a procedure based on the analysis of cross-correlation prediction errors.
this is from my PhD Thesis CONCLUSION
https://www.researchgate.net/publication/338528091_PhD_Thesis_Computer_Geostatistical_GIS_Remote_Sensing_and_Isotopes_applications_for_Modelling_the_Spatial_and_Temporal_variations_of_the_Environmental_Pollutions_in_the_Amman_Zarqa_Basin_Basalt_Area_J
- Badges
- Science topic