Radar backscatter power from the lunar surface was collected at a
  wavelength of 12.6 cm, using the 305 m Radio Telescope at Arecibo to
  transmit and the NRAO's 105 m Robert C. Byrd Green Bank Telescope to
  receive. The first maps were acquired in May, 2008. Data acquisition to
  fill in most remaining radar-visible areas of the lunar near side
  continues as of the initial publication of this archive.
 
  The data products in the DATA directory correspond to NASA Level 2, or PDS
  Level 5 Derived Data. According to the PDS Standards Reference, these are
  'derived results, as maps, reports, graphics, etc.' Each image has an
  accompanying detached PDS label.
 
  The data set consists of 5 image files for each region mapped during a
  typical 29-minute radar integration period. The images are named as
  follows: XXXXXX_SCP.IMG (same-sense circular polarization map),
  XXXXXX_OCP.IMG (opposite-sense circular polarization map), XXXXXX_ANG.IMG
  (map of beam angle in radians), XXXXXX_INC.IMG (map of incidence angle in
  radians), and XXXXXX_CPR.IMG (circular polarization ratio image), where
  XXXXXX is the center latitude and longitude of the radar pointing target
  over the observation period. This is expressed as XX deg N (for north
  latitude) or S (for south latitude) followed by the east longitude in
  degrees (for example, 20N310 corresponds to 20 degrees north latitude, 310
  degrees east longitude). North is at the top of the frame in all of these
  images.
 
  For the polarized ('OCP') and depolarized ('SCP') maps, the floating point
  values represent the estimated dimensionless backscatter coefficient
  (often called sigma-zero) of the lunar surface at a radar wavelength of
  12.6 cm (2380 MHz frequency). 'Polarized' refers to energy scattered from
  the lunar surface in the opposite circular polarization sense to that
  transmitted (the behavior expected of a flat, mirror-like reflecting
  surface). 'Depolarized' refers to reflected energy with the same sense of
  circular polarization (the behavior expected from a surface with abundant
  wavelength-scale objects that randomize the reflected polarization).
 
  The data were collected by transmitting a continuous-wave, circular
  polarized signal from the Arecibo Observatory, and receiving the lunar
  echoes at the Robert C. Byrd Green Bank Telescope (GBT) in West Virginia.
  The transmitted signal was a 65535-sample pseudo-random code with baud
  length of 0.2 microseconds. The resulting 13.17-millisecond period between
  pulses allows for the full possible range of echo time delays over the
  Moon's surface. The reflected signals were recorded at the GBT using 4-bit
  analog-to-digital conversion, 5-MHz sampling rate, and a 4.4-MHz Gaussian
  lowpass filter. Pulse compression was carried out to recover the intrinsic
  2-microsecond delay resolution. A patch focusing method was used to
  correct for time-varying Doppler changes at lunar surface points distant
  from the radar pointing target. Multiple independent integration periods
  (looks) were added together to reduce radar speckle.
 
  Image power values were normalized to the effective scattering area that
  contributes to each pixel (based on a reference spherical shape) and to
  the thermal background noise measured at the Green Bank Telescope.
  Subtraction of this background noise level can lead to small negative
  values of the backscatter coefficient where the signal-to-noise ratio
  (SNR) of the echoes is close to unity. The effect of the two antenna beam
  patterns on the gain across the lunar surface is also compensated to first
  order, and pointing errors at Arecibo are estimated and accounted for in
  this calibration step. Final calibration to values of the dimensionless
  backscatter coefficient is performed by comparing the GBT thermal noise to
  that of a quasar of known flux at 2380 MHz. To date, these calibrated
  values are consistent over numerous radar observing periods, but are
  systematically lower (by a factor of 6-10 in power) than values measured
  by previous investigators. Ratios between the two circular polarization
  channels are well calibrated, since they depend only upon determination of
  the relative noise power in the two received channels. The 'OCP' and 'SCP'
  images are further normalized to the cosine of the radar incidence angle
  for a reference sphere (given in the 'INC' files) at each location. The
  stated location of the sub-radar point at 2380 MHz is an approximate value
  at the start of the multi-look observing period.
 
  In addition to the 'OCP' and 'SCP' images, the archive also includes 'CPR',
  'ANG', and 'INC' images. The 'CPR' image files are circular polarization
  ratio images formed by averaging pixels over a 5x5 moving window, and
  dividing the averaged 'SCP' data by the averaged 'OCP' data. No other
  normalization is performed, so the polarization ratio images have a strong
  decrease toward the center of the Moon (lower incidence angle). Note that
  in areas of low SNR, occasional negative values of the CPR could occur due
  to the subtraction of the background thermal noise. Values of unity and
  greater are typical of rough surfaces, particularly at high incidence
  angles where the diffuse scattering component dominates the echo.
 
  The other image files ('ANG' and 'INC') display the beam angle and
  incidence angle, respectively, in radians for each pixel. These
  co-registered maps, in floating point format, will allow a user to reverse
  the applied calibration steps if desired. The beam angle value for each
  pixel represents the angle (in radians) between a vector from the observer
  to the surface location and the vector from the observer to the radar
  pointing target given for each map. The incidence angle value for each
  pixel represents the angle (in radians) between a vector from the Moon's
  center of mass (COM) to the surface location and the vector from the COM
  that passes through the sub-radar point. Values near the radar-visible
  limb of the Moon thus approach 90 degrees.

Known Problems =
 
  The calibrated SCP and OCP backscatter coefficient values are consistent
  over numerous radar observing periods, but are systematically lower (by a
  factor of 6-10 in power) than values measured by previous investigators.
  The reason for this remains uncertain. Ratios between the two circular
  polarization channels are well calibrated, since they depend only upon
  determination of the relative noise power in the two received channels.
 
  The Arecibo system has an inherent pointing error that can cause offsets
  of the transmitted beam from the desired location on the lunar surface. We
  have attempted to correct for this offset using an iterative fitting
  process that matches the locations of the first beam nulls in the
  delay-Doppler image, but some errors may remain. These manifest as darker
  or brighter arcuate patches along part of the edge of a mapped area.
 
 
  This data set underwent external peer review from November 22, 2010
  through August 1, 2011.