It depends on the kind of data. Let me explain, for example: for DEMs derived from lidar, we use a Triangulated Irregular Network (TIN) to fill voids left once you remove buildings and other things from the terrain, it works well. For point data, kriging or a Gaussian process regression may work as a method of interpolation for which the interpolated values are modeled by a Gaussian process governed by statistics or covariances. Use either method with a grain of salt, knowing that the filled gaps are not real data.

Best.

Miguel.

If you have discrete point data (with an attribute value), you can use interpolation methods such as IDW, natural neighbourhood and Kriging to obtain continuous surface grid. Many papers about this topic are available on the web. You can use GIS or spatial modeling softwares such as QGIS (open source), ArcGIS, MapInfo and Surfer.

here are some useful links that may be helpful to you.

http://www.cs.wm.edu/~bnie/publications/2017CNSM.pdf

https://arxiv.org/ftp/arxiv/papers/1605/1605.01038.pdf

Conference Paper Fill-in the gaps: Spatial-temporal models for missing data

Article An effective approach for gap-filling continental scale remo...

Article A Comparison of Methods for Filling Gaps in Hourly Near-Surf...

https://www.nonlin-processes-geophys.net/13/151/2006/npg-13-151-2006.pdf

http://www.scar.ac.cn/hhkxen/ch/reader/create_pdf.aspx?file_no=20160509

Article Geostatistical interpolation of missing data and downscaling...

There was similer type of question asked on research gate earlier and a healthy disscussion was held. hope you find it worth reading.

https://www.researchgate.net/post/What_is_the_recommended_method_to_fill_gaps_of_precipitation_data_set