6.2 Formats for spatio-temporal environmental data

In order to process spatio-temporal data effectively, a GIS needs an efficient system for storing spatial data from different time periods, such as the air temperature for different months of the year. This object describes the different methods available for storing such data. Most commercial GIS systems were designed to handle spatial data and are weaker at handling spatio-temporal data. However, there is active research to find better ways of handling spatio-temporal data. Several software companies – notably TerraSeer and Discovery Software’s STEMgis – have produced GIS software that is specifically designed for temporal data.

Several methods have been proposed for handling temporal data. One idea is to ‘time-stamp’ items of data – providing each data item with a start date and an end date. From a database point of view, approaches to handling data may either involve time-stamping whole tables, time-stamping rows within tables, or time-stamping individual columns of attributes. These concepts are illustrated below.

From a spatial database perspective, Langran (1992) describes three approaches to handling temporal data:

Sequential snapshots : this involves storing separate map layers for all time periods. For example, to represent monthly precipitation, each month’s data might be stored as a separate raster map layer. There are several disadvantages to this approach. Where no change occurs, the same data are stored separately for different months. The method is therefore somewhat inefficient and introduces some redundancy in storage. Furthermore, there may be discrepancies between the rate of real world change and the period between consecutive map layers. Precipitation varies from hour to hour, yet is only stored every month. Despite this, sequential snapshots remain the most widely used means of storing temporal data.

Data formats are emerging that provide a means of handling sequential snapshots within a single, integrated file format. Most notably, in the environmental sector, one such format is known as Network Common Data Form (NetCDF). This format originated in the atmospheric science community under the Unidata program of the University Corporation for Atmospheric Research. NetCDF provides a way of handling both raster and vector spatio-temporal data sets within a single, platform-independent file format. This open standard (for which full documentation is provided to all) has three key characteristics:

  • dimensions: With spatio-temporal data, there are generally 3 dimensions – time, x co-ordinate and y-coordinate, but the format is capable of handling more than 3 dimensions (e.g. a 4 dimensional NetCDF file could include time, X, Y, as well as different wavebands for remotely sensed data).
  • variables: There may be just one attribute, but there could be multiple attributes, all held within the same file (e.g. 10 attributes representing the presence or absence of 10 different species, with presence coded as one and absence coded as zero).
  • header: there is also provision to record key characteristics of the data file within a header.

Packages are available that enable the user to ‘slice’ a NetCDF file in different directions, or to animate attribute changes over time. For example, there are packages that enable the user to view attribute values for a particular time period, or take a specific point in space and view the changes that occur over time at that one point in space.

Space-time composites : This involves combining data from all time periods into a single map layer, which is usually held in vector format. A space-time composite can be created by overlaying the data from all time periods. In a space-time composite map layer, the area inside each polygon shares the same characteristics in every time period. Again, this method of storage has some disadvantages. When many time periods are involved, the resultant composite map may contain numerous small polygons and ‘slivers’. Furthermore, it can be difficult to update such composite map layers to incorporate data from new time periods.

Space-time objects : In a space-time object model, data are decomposed into a series of objects, which may represent any basic geographic feature (point, line or polygon). The object is time-stamped with a start and end date and its spatial extent at the start period is described. The identifier of any preceding object in the previous time period is stored, together with a link to the identifier of any successor object in the next time period. For example, the location of a whale may be represented as a point in a space-time object model. The whale’s location is time-stamped, so an attribute field records when the whale’s location was measured (e.g. Christmas day 2004). The whale’s location is also given a unique ID number. The whale’s location on Christmas Eve and Boxing Day is also stored as separate points, also with unique ID numbers. The unique IDs for this earlier and later point are stored as attributes of the whale’s position on Christmas day, so that the different points representing its pattern of movement can be pieced back together.

Where gradual changes are being described, the object may move between start and end date (for example, as a patch of bracken slowly expands into neighbouring vegetation patches). In some object models, this gradual movement during the time period can also be described and stored with the object as a distance and direction. A more detailed description of a spatial object model may be found at the TerraSeer web site (see below). In practice, this approach is most widely used with point data, particularly GPS data representing (for example) animal movement.

In practice, space-time objects and space-time composites are seldom used with environmental data. The data structures used generally reflect the sources of data, so for example meteorological and land cover time series are typically stored as sequential snapshots of raster grids. This situation may change in the future as demand for efficient handling of spatio-temporal data increases.


Examples of spatio-temporal data storage in GIS (sequential snapshots and NetCDF)

Download the zip file, which contains land cover data from the Mafungabusi forest reserve in Zimbabwe . View the files it contains within your GIS software. Of the three approaches described above (sequential snapshots, space-time composites, and space-time objects), which one has been used to store the data here?


Sequential snapshots – the land cover data for each time period have been stored as separate map layers.



If you are using ArcGIS Desktop, download the pdf file and follow the instructions for the practical exercise, which involves downloading and manipulating some global climate data in NetCDF format.  If you are using ArcGIS Pro, download this pdf file instead to follow the exercise.

References (Essential reading for this learning object indicated by *)

This article describes concepts of spatio-temporal data storage such as sequential snapshots:

Godchild, M. (2013) Prospects for a space-time GIS. Annals of the Association of American Geographers 103 (5), 1072-1077. http://www.tandfonline.com/doi/full/10.1080/00045608.2013.792175

The STEMgis software, which handles spatio-temporal data, is available from here: http://www.discoverysoftware.co.uk/STEMgis.htm

Langran, G. (1992) Time in Geographic Information Systems. Taylor and Francis, London.

The exercise above uses NetCDF format data available from the Climate Research Unit at University of East Anglia: http://www.cru.uea.ac.uk/cru/data/temperature/

Comments are closed.