FillValue

FillValue is a special marker used in data arrays and file formats to indicate that a data element is missing, unknown, or invalid. It functions as a sentinel value that allows software to distinguish gaps from valid measurements and to preserve array shapes in storage and processing.

In NetCDF and related conventions, a variable attribute called _FillValue specifies which value represents missing data for that variable. Older practice sometimes references a missing_value attribute as well; modern CF conventions standardize the use of _FillValue. When files are created or read, software may initialize or replace data with the fill value, and readers often expose missing data as masked values or as NaN.

Fill values must be representable within the variable’s data type and are chosen to be unlikely to occur in valid data. For integer types, a distinctive sentinel such as -9999 is common. Floating-point variables frequently use NaN as the fill value, with many libraries automatically mapping _FillValue to NaN on read.

In data processing, fill values are ignored or masked during statistical calculations and analyses to avoid bias. When writing data, authors specify the fill value to maintain compatibility across software that consumes the file. Some formats, such as HDF5, provide a dataset creation property to set an explicit fill_value that pre-fills memory with the sentinel value. Related topics include missing data, NaN, NetCDF, CF conventions, and HDF5.