Wilkinson Microwave Anisotropy Probe
The data made available through this page has been updated. The most recent version of this data may be accessed through /product/map/current/
0.23°- 0.93° (frequency dependent)
WMAP maps of the sky are supplied in FITS files, with the maps contained within binary table extensions. The maps are stored in a nested HEALPix projection and are supplied at three resolutions: WMAP's standard Res 9 (NSide=512), Res 4 (NSide=16), and Res 10 (NSide=1024) for high-l TT analysis. Stokes I (temperature), Stokes Q and U (Polarization), and bandpass mismatch (a so-called S map) maps may be supplied in a file. Also included in the maps is an Nobs map describing the number of observations used to construct each pixel. A detailed description of how these files are formatted is available here.
Iterative algorithms are used to create skymaps from the calibrated differential time-ordered WMAP data for each of the ten differencing assemblies. Each pixel in a map represents a sky temperature for the bandpass appropriate to the differencing assembly. The CMB dipole has been removed from the Stokes I maps.
Maps for seven years of WMAP data have been supplied to LAMBDA. Full seven year maps were produced by performing a weighted, pixel-by-pixel, mean of the seven single year maps; the Nobs measurement was used to weight this mean. For the Stokes I maps, the mean temperature outside of a preliminary version of the temperature analysis/KQ85 mask was subtracted from each single year map before the seven year weighted mean was calculated. The zero point of each seven year Stokes I map has been set using a Galactic csc |b| model. The zero level of each single year map has not been set, and is not a meaningful quantity. The Nobs measurements were added together to produce the final Nobs measurement. These maps have also been smoothed to 1° resolution; these smoothed maps have been supplied as a separate data product.
The maps of the individual differencing assemblies of a single frequency band (Q1 and Q2; V1 and V2; W1, W2, W3, and W4) have been combined to produce coadded maps for each frequency band.
Many of the files that contain polarization maps also contain an extension containing the polarization noise covariance matrices for the maps. See the detailed description mentioned above for formatting details. While available for the Res 4 maps, these matrices do not fully characterize the pixel-to-pixel noise in these maps. Therefore a more accurate and much larger inverse covariance (or N-1) matrix is supplied as a seperate product.
Pixel noise in units of mK may be evaluated from Nobs with the expression σ = σ0 / sqrt(Nobs) where
Foreground reduced maps have been produced by removing a foreground model from the 'unreduced' maps using the Foreground Template Model discussed in Hinshaw, et.al., 2006 and Page, et.al, 2006; see those papers for details. Briefly, synchrotron, free-free, and dust emission templates were modeled and then subtracted from the single year 'unreduced' maps. The goal was to produce a set of maps with the foreground removed while retaining WMAP's noise characteristics.
Because the K maps were used to produce the foreground models, only Ka, Q, V, and W band foreground reduced maps were created and delivered to LAMBDA. In some cases, Ka was also used in the foreground model; in those cases, the Ka Stokes I maps have been zeroed out.
As with the 'unreduced' maps, full seven year maps were produced by performing a weighted, pixel-by-pixel, mean of the seven single year maps; the Nobs measurement was used to weight this mean. The Nobs measurements were added together to produce the final, seven year, Nobs measurement.
The same foreground template model was used to produce foreground reduced high resolution (res 10) V and W maps.
The low resolution (res 4, NSide=16) maps were created from the res 9 maps. The individual year maps were computed through a least-squares weighted boxaverage, treating the Stokes I maps seperately from the Stokes Q, U, and S maps: QU, QS, ans US covariances were taken into account but IQ, IU and IS covarainces were omitted. The Nobs vectors were replaced with the appropriate N-1 diagonal elements scaled by σ02. The seven year low resolution maps were created through weighted co-addition of the individual year maps.
When the multi-year, multi D/A, reduced resolution maps are formed, full inverse covariance matrix weights are employed in the averaging. Thus the non-cleaned co-added maps have been recomputed with the new matrices. The previous, V4P0, maps were correctly characterized by the V4P0 covariance matrices, but the new maps and matrices more fully capture our knowledge of the instrument and data. Please note that only the Stokes Q and U maps without foreground cleaning were affected.