Gridded Population of the World (GPW) version 3 is based on 2000 round census data. In addition to a year 2000 grid, it includes 1990 and 1995 grids (based on 1990 round census data and interpolations between 1990 and 2000 round censuses), and projected 2005, 2010, and 2015 grids. These represent continuations of the 1990-2000 population trends, with pixel estimates based on a projection of the trend but capped so that the national level totals do not exceed UN projections. GPW is at a 2.5 minute resolution (approximately 5km at the equator), and a derivative product, Global Rural-Urban Mapping Project (GRUMP) v.1, is at a 30 arc-second (1km) resolution (the same as Landscan), and includes an urban-rural reallocation. GRUMP has 1990, 1995, and 2000 population grids.

The following represents a comparison with Landscan 2009

- Landscan represents daytime “ambient” population, whereas GPW represents night-time population. Depending on the application, this distinction may be important. For example, disaster evacuation plans would need to take into account where the population is at a given time, and exposures to environmental pollutants such as day-time smog or night-time heat island effects, will require different data.

- Landscan is a modeled data set. The data set creators apply multi-layered, dasymetric, spatial modeling approach to reallocate populations based on layers representing land use/land cover,  high resolution imagery analysis, transportation networks, elevation, slope, etc. The precise reallocation algorithm is not documented. GRUMP is a “lightly modeled” data set with an urban reallocation based on a night-time lights data layer.  GPW does not model or reallocate the population, but simply grids data based on the original census units, at the highest spatial resolution (smallest spatial size) for which they are available by country. For applications in which the independent variable is anyone of these layers used in Landscan or GRUMP, users are advised to use GPW, since these variables are exogenous to GPW.  

- Landscan has annual updates, but owing to budgetary constraints, CIESIN and its partners only produce  a new version of GPW following each decadal census round. Earlier Landscan releases may not be consistent with current releases owing to changes in data sources and methods. CIESIN’s GPWv3 has consistent time series for each five year interval from 1990-2015 based on census data from the 1990 and 2000 census rounds.

- Globally, Landscan uses 8,285,172  census inputs, whereas GPW v.3 uses 399,747units. The Landscan number is large because of the way the USA is processed (Ambient Global data is processed using LandScan USA Night/Day inputs). Outside the USA Landscan used 79,590 administrative units for ambient modeling. By contrast,  GPWv3 uses  338,863 units outside of the US. Country-by-country details are all in a spreadsheet available at: http://sedac.ciesin.columbia.edu/gpw/spreadsheets/GPW3_GRUMP_SummaryInformation_Oct05prod.xls.

- GPW is free of charge whereas commercial users of Landscan are charged a fee to use the data.

Further documentation on GPW v3 and GRUMP is available from http://sedac.ciesin.columbia.edu/gpw/documentation.jsp, and further documentation on Landscan is available through the Documentation and FAQ pages at http://www.ornl.gov/sci/landscan/