Data Set Overview : The following description applies to Version 1.0 of the Derived Neutron Data (DND) data set. The data have been completely reprocessed in Version 2.0. See mons2_pds_final.pdf in the DOCUMENT  directory of the archive. The Mars Odyssey Gamma-Ray Spectrometer (GRS) is a suite of three instruments working together to collect data that will permit the mapping of elemental concentrations on the surface of Mars. The suite of three instruments, the gamma sensor head (GS), the neutron spectrometer (NS) and the high-energy neutron detector (HEND), are a complementary set of instruments in that the neutron instruments have better counting statistics and sample to a greater depth than the GS, but the GS determines the abundance of many more elements. A full description of the Mars Odyssey Gamma-Ray Spectrometer instrument can be found in [BOYNTONETAL2004]. See [FELDMAN2002] and [BOYNTONETAL2004] for details of the neutron spectrometer design and operation.  The ODY MARS GAMMA RAY SPECTROMETER 4 DND data set is a time series collection of neutron counting data from the NS  sub-system of the Mars Odyssey Gamma-Ray Spectrometer. The DND are an intermediate data product. These data are the neutron data that must be derived from the raw Level-0 data before a set of calculations can be run that will result in maps of  thermal, epithermal, and fast neutron counting rates. The GRS collects a new spectrum (pixel) approximately every 20 seconds, 360 times per orbit. Approximately 4200 neutron measurements are expected to be received every day. The data are downloaded from the spacecraft by the Jet Propulsion Laboratory (JPL) into the Telemetry Data System (TDS). The TDS sends data to a process that translates data packets and examines instrument health via messages. Data are output to a spooler that passes it to the University of Arizona (UA) database ingest process. The ingest process inputs raw data into the UA database. Neutron data are extracted from the UA  database by the NS data team at Los Alamos National Laboratory (LANL). Neutron data are then processed by LANL through a  number of programs to yield the derived neutron data. The DND product is intended to be the first intermediate data product available for the NS portion of the GRS data. These  data should be useful to those scientists who are experienced in neutron spectroscopy. Parameters : The DND data set is composed of a single data type (DND). The objective of compiling the DND is to create a series of neutron data that have been processed to yield normalized neutron counting rates for all four of the Neutron Spectrometer prisms, from which thermal and epithermal neutron counting rates are  determined.   Each DND product data file will contain a time series of  neutron data collected over a 24-hour time period. The data files are labeled as Earth days. Each DND data file is a table containing 53 columns, with one row of data for each  approximately 19.75 second collection interval. Processing : The products of the neutron data reduction algorithms are a  time-series data set (DND), corrected for variations in the  instrument response and the cosmic ray background, that can be used to produce thermal, epithermal, and fast neutron  global maps of Mars. A summary of these algorithms adapted from Appendix A of [PRETTYMANETAL2004] is provided here. See  [Neutron Spectrometer Processing V1.1], provided in the  Documents directory accompanying this release for details of the neutron data reduction process.   The neutron spectrometer consists of a block of boron-loaded  plastic that is diagonally segmented into four optically  decoupled prisms that are viewed by separate photomultiplier  tubes. During mapping, two of the prisms are aligned along the axis of motion of the spacecraft: Prism 2 (P2) faces in the  forward direction and Prism 4 (P4) faces opposite to the  direction of motion. The remaining two prisms are oriented  along the nadir axis: Prism 1 (P1) faces downward towards Mars  and Prism 3 (P3) faces upwards towards the spacecraft. P1 is  covered with cadmium foil, which absorbs neutrons below roughly 0.5 eV. Consequently, P1 is sensitive to epithermal neutrons. Raw (EDR) neutron data, including spectra, histogram, and  events arrays, are processed through several automated programs to yield normalized neutron counting rates for all four of the NS prisms, from which thermal, epithermal and fast neutron  counting rates are determined.   The first step in the neutron processing is the determination of the 10B(n,alpha) peak centroid and the after-pulse  threshold of each prism averaged over a course time window 720 collection intervals in length. This information is used  subsequently by algorithms that determine the area of the 10B(n, alpha) peak for each prism for each 19.75s interval The  10B(n,alpha) peak centroid is determined using double pulse events (Category-2 events), which are primarily a result of interactions of fast neutrons with the spectrometer. Fast  neutrons are identified by the detection of a characteristic double pulse. The light output caused by proton recoils  produced by the prompt interaction of a fast neutron and the  hydrogen in the plastic provides a measure of the energy of the incident neutron. A second pulse of light corresponding to the absorption of the neutron at low energy by 10B follows a delay in which the neutron is slowing down in the plastic, but does not produce light. This double-pulse signature occurs for neutrons above approximately 700 keV.   Category-2 event mode data (EDR EVENTS_ARRAY) are accumulated over a coarse time window containing 720 19.75s intervals. The overload counts (EDR GCR_CNT) are monitored and events are discarded for intervals for which the overload counts exceed a user-defined threshold, currently set at 7000. For each time window, the second event spectrum is constructed from the accumulated event-mode data. If the number of accepted  intervals is less than 360, then the time window is not  analyzed and the window is assigned the results for centroid and threshold from the analysis of the previous window.  The second event spectrum acquired for each window is analyzed by simultaneously fitting two normal distributions (one for the 10B peak and one for the after-pulse peak) and a constant background (to represent the continuum) to the data. The  fitting algorithm minimizes the sum of the squares of the difference between the model and the data weighted by the statistical uncertainty in the counting data i.e., a weighted least squares fit to the data. The weighted least squares fit yields the peak centroids, the highest of which is taken as the 10B peak, the full width at half maximum, and the after-pulse threshold, which is taken to be the channel for which the model is minimum between the two centroids. For a more complete  description see Neutron Spectrometer Processing V1.1 in the documents directory that accompanies this release. The peak centroid determined by this process is used subsequently to make gain corrections in the next processing step, which is evaluation of the Category 1 single event spectrum.  Thermal and epithermal neutrons are detected through the  10B(n,alpha)7Li* reaction, which produces a distinct peak in  the pulse-height spectrum caused by the deposition of the  recoil energy of the reaction products in the scintillator. This peak corresponds to an equivalent electron energy of 93 keV. For each prism, the net counting rate for this peak was  determined from the Category-1 pulse-height spectrum, a  spectrum of single-interaction events that is recorded every 19.75 s. During this measurement interval, the spacecraft traverses approximately one degree of arc length (roughly 60  km). Determination of peak areas for each prism from the Category-1 spectra involved several steps summarized here:  1) A correction was applied to each spectrum to remove  artifacts of the differential nonlinearity of the analog to  digital converter from the spectrum.  2) The spectra were corrected in gain so that the 93 keV peak always fell in Channel 11. Implementation of this step  required knowledge of the original peak location determined  from the second interaction spectrum. The gain correction  enabled the same region of interest (range of energy) to be used for the peak and background for all spectra.   3) For each prism, selected channels surrounding the peak above and below were fitted to a power law. The power law was determined for spectra in a sliding window 5 x 19.75 s in  length. The length of the window was selected empirically to minimize the uncertainty in the background parameters because of statistical variations in counting rate while avoiding the introduction of bias resulting from integration over surface  features.   4) The peak areas (counts in the peaks for each prism) were  determined for each 19.75 s spectrum by subtracting the  background predicted by the power law from the total counts in the peak region of interest.   The thermal-neutron counting rate was determined by subtracting the peak area for the backward-looking prism (P4) from that of the forward-looking prism (P2). During the mapping phase, the  spacecraft traveled in a circular polar orbit at an average  altitude of 400 km and speed of 3380 meters per second. A significant portion of the population of thermal neutrons at orbital altitudes had speeds much less than that of the spacecraft and the flux of low-energy neutrons incident on the exposed face of P2 was much higher than that of P4. If the return of neutrons from the spacecraft is neglected, both prisms should receive the same contribution from epithermal neutrons. Consequently, subtraction of P4 from P2 yielded a response that was representative of the thermal-neutron population. The epithermal-neutron counting rate was taken to be the counting rate for P1, which was shielded from thermal neutrons by a Cd foil. Note that based on an analysis of cruise data, the cosmic-ray-induced background for Category-1 counting rates for P1, P2, and P4 was found to be consistent with zero.  For fast neutrons, two data products are provided in the raw data set: Event-mode data for which the magnitude of the first pulse, time to the second pulse, and magnitude of the second pulse of a selected number of events (84 per 19.75 s  acquisition period) were recorded; and histogram data for which first pulses with magnitude above a preset, fixed threshold were binned into histograms for early and late time windows. The late time interval (nominally chosen to be between 20 and 25 microseconds) was selected so that it is well beyond the die-away time of neutrons in the spectrometer and thus contains accidental events only. The early time interval (nominally chosen to be between 0.4 and 5.4 microseconds) was selected so that it is sensitive to neutrons slowing down in the spectrometer. Subtraction of the late from early time histograms yields a pulse-height spectrumthat is sensitive to the energy spectrum of fast neutrons.   The fast-neutron counting rate was taken to be the first two channels of the histogram spectrum for P1, which is primarily sensitive to neutrons coming directly from Mars. The lower channels were selected because they give the greatest count rates and are also least susceptible to contamination by after-pulsing. However, because considerable drift in gain occurred, the threshold selected for after-pulse suppression used to construct the histogram data was not always effective. During times of high gain, the threshold was too low and after-pulsing events contaminated the histogram spectrum. Variations caused by shifts in gain were corrected using the BellyBand algorithm described below. Neutron counting rates are sensitive to a number of factors unrelated to the surface and atmosphere of Mars. These  include, for example, instrument drift and variations in the galactic cosmic-ray flux. To eliminate these variations from our data set, we developed an algorithm to adjust our  time-series counting data so that counting rates at equatorial latitudes did not change with time and were equal to the  average counting rates observed over a normalization interval, which was arbitrarily selected to be from LS equals 329 degrees to LS equals 131 degrees. The region between latitudes plus to minus 30 degrees (the belly band) was used for normalization. Minor variations in counting rate due to seasonal changes in atmospheric mass were initially ignored, but were later  restored to the time series.   The belly-band normalization algorithm removes gain and offset variations caused by a number of effects, including variations in the galactic cosmic-ray background and errors in the  Category 1 background subtraction algorithm, with a linear correction (CAT1_PRISMX_NORM_SLOPE, CAT1_PRISMX_NORM_OFFSET) However, the normalization algorithm also removes variations that would otherwise be present as a result of seasonal changes in atmospheric mass at the equator. To restore the effect of  atmospheric mass, we combined maps of atmospheric mass at  equatorial latitudes (derived from the ARC-GCM) with  expressions for counting rate as a function of atmospheric  thickness [PRETTYMANETAL2004] to determine the relative  variation of belly-band-averaged neutron-counting rates as a function of time. The normalized counting rates were  then multiplied by the relative variation to restore the  atmospheric effect. Use of pressures from the ARC-GCM is  justified because the ARC-GCM fits the Viking 1 and 2  landing-site pressure data that are representative of  equatorial to midlatitude pressure variations and were found to be reproducible year after year. This procedure was applied to the fast and epithermal counting data that show the largest effect. Thermal neutrons vary negligibly with atmospheric mass when the water abundance of the surface is less than about 15 percent, which is representative of the belly band. The fast neutrons, which have the largest variation, show a 5 percent peak-to-peak variation in counting rate over an entire Martian year. Therefore, the correction is relatively minor.  Data : The DND data set is composed of a series of date stamped files that contain 24-hours worth of data. Derived Neutron Data -------------------- Derived Neutron Data are composed of the derived neutron data and the associated timing, spatial, and engineering information for each ~19.7 second collection interval. The derived neutron data consists of 53 columns of data that are output from the derived neutron data processing. The timing and spatial data provided with the DND includes spacecraft clock values and spacecraft geometry data. The sc_ev_time, UTC time and spatial fields are all recorded at the center of the collection interval. The associated engineering data is obtained from a linear interpolation of the engineering data packet received just before and just after the neutron data collection interval of interest. Ancillary Data : None at this time. Coordinate System : The coordinate system used for all GRS data is a Mars aerocentric system following the IAU convention [SEIDELMANNETAL2002], with east longitudes from 0 to 360. Software : A library of source code to parse the DND data product files is included in the software directory. This library allows a programmer to build applications that display or manipulate DND data. This source is written in the Java language, and requires version 1.3 of the Java Runtime Environment (JRE) or Java Software Development Kit (SDK). Documentation for the code is located in the software directory in the file GRS_CODE_DOC.ZIP. The contents of this file are described in the label GRS_CODE_DOC.LBL and the source and the binary classes that make up the library are in the file GRS_CODE.JAR. Media/Format : The DND will be delivered using CDROM media. Formats will be based on standards for such products established by the Planetary Data System (PDS) [PDSSR2001]

Confidence Level Overview : The data presented in the DND is intended to be the first intermediate data set released for the neutron spectrometer sub-system of the GRS. Data presented here are a highly processed representation of the neutron data as received from the spacecraft. It is possible that changes will be made in the neutron processing algorithms if errors are found. If errors are found, the data will have to be regenerated from the raw unprocessed neutron data set. Review : The DND was reviewed internally by the GRS team prior to release to the PDS. PDS will also perform an external peer review of the DND. Data Coverage and Quality : In order to consider data for inclusion in the DND time series, we required that the instrument be aligned along the bore-site vector of spacecraft motion and nadir orientation. This eliminated times such as when the spacecraft was in safe mode for which P1 was not pointed in the nadir direction. We also required that the data in the 5 x 19.75 s background buffer be contiguous, separated by no more than 20s of UTC time. This avoided the introduction of errors in the counting data when orbital data were missing. Finally, we did not process data for time intervals in which the overload monitor of the neutron spectrometer exceeded 7000 counts per second, which  corresponded to solar-energetic particle events. The solar- energetic particle event overload threshold for data collected after April 1, 2004 was raised to 7500 counts per second to account for a rise in the background galactic cosmic-ray flux  due to the solar minimum. The solar-energetic particle event  overload threshold for data collected after April 1, 2005 was raised to 7800 counts per second to account for a rise in the  background galactic cosmic-ray flux due to the solar minimum.  The solar-energetic particle event overload threshold for data collected after June 30, 2005 was raised to 8500 counts per second to account for a rise in the background galactic  cosmic-ray flux due to the solar minimum. The solar-energetic particle event overload threshold for data collected after  October 1, 2005 was raised to 9000 counts per second to account for a rise in the background galactic cosmic-ray flux due to the solar minimum. The solar-energetic particle event overload  threshold for data collected after January 1, 2006 was raised to  9400 counts per second to account for a rise in the background galactic cosmic-ray flux due to the solar minimum.   Neutron spectrometer data collected in between July 13 and October 4, 2005 are suspect due to low voltage on the high voltage power supply. A command was sent on October 4, 2005 to change high voltage power supplies. The result of this command was an increase in the voltage that allows for more accurate  ground data processing of the NS data. Derived data products  were not generated for data collected between October 4, 2005  and October 13, 2005. These data may be filled in at a later date.  Limitations : The major limitation of this data set is that it is derived  from the raw minimally processed data. The data is received from spacecraft telemetry, and ingested into a database. If gaps exist in the telemetry, data is lost. Timing and spatial components of the data set rely on the accuracy of the NAIF SPICE kernels. The validity of the derived neutron data is  based on the 'correctness' of each step in the processing.  Data Compression : No compression is used on the DND data set.