The Lunar Exploration Neutron Detector (LEND) aboard the Lunar Reconnaissance Orbiter (LRO) has been mapping the neutron flux from the Moon since July 2009. LEND has four different types of neutron detectors which allow a comprehensive study of lunar neutron emission: a pair of omnidirectional sensors for thermal and epithermal neutrons, a pair of Doppler filter sensors for thermal neutrons, four collimated sensors of epithermal neutrons, and a sensor of high-energy neutrons. This paper describes the data reduction procedures to convert the raw data into the higher-level PDS products containing neutron-counting rate maps of the Moon.
 Solar System planets and bodies with little or no atmosphere, such as the Moon, Mars, and Mercury emit gamma rays and neutrons from their surfaces. The neutron and gamma radiation occurs because of continuous bombardment by high-energy galactic cosmic rays and episodically by solar energetic particles [see, e.g.,Arnold et al., 1962]. They produce a cascade of secondary particles in the upper tens of centimeters of regolith including secondary neutrons with energies of about 1–20 MeV, a portion of which may then escape and be measured in nearby space. Others are absorbed through capture reactions before escaping or decay because of a finite lifetime. Before escape, neutrons may lose energy in the collisions. The energy spectrum of emitted neutrons has a thermal component and a power law tail from epithermal energies up to the original energy of particles that escape without moderation [Drake et al., 1988]. The energy spectrum of the escaping neutrons depends on the composition of the soil and mostly on the content of hydrogen, because H nuclei are the best moderators of neutrons; even a fraction as small as 100 ppm is known to produce a measurable suppression of about 5% of epithermal neutron flux from the surface [e.g., Mitrofanov et al., 2008]. Areas of enhanced hydrogen abundance (in the form of hydroxyl or water or water ice) may be distinguished by identifying regions with significant suppression of epithermal neutron counting rates. The by-product of neutron leakage from the subsurface is the spectrum of gamma ray nuclear lines produced through the reactions of neutron inelastic scattering and capture in the surface regolith, providing unique information about elemental composition of the subsurface.
 LEND consists of ten neutron sensors (Figure 1) providing registration of neutrons in different energy bands [see Mitrofanov et al., 2008, 2010a]. Eight sensors are identical proportional counters, LND 253123. These cylindrical detectors are filled with 3He gas under pressure of 20 atmospheres and have active zone ∼5 cm at the middle of sensor's tube and dead zones ∼1.25 cm at each end of sensor's tube. These sensors are used for detection of thermal and epithermal neutrons. The other two LEND sensors are combined in one detection system to detect fast neutrons. The inner part of this system is a stilbene scintillator (C6H5CH = CHC6H5) sensitive to gamma rays and for fast neutrons with energies within 0.5–15 MeV range.
 The logic of LEND operations is based on the RTAX series Actel FPGA. It performs separate digitization of counts of all nine sensors, supports an anticoincidence system for the fast neutron detector assembly, makes a data frame with counts from all nine LEND sensors, allows changes of high-voltage and low-amplitude signal discrimination levels for each sensor, accepts and executes commands, and sends telemetry data to LRO telemetry system via a 1553 interface.
2.1. Detector System of Collimated Sensors of Epithermal Neutrons
 The first group of LEND detectors includes four collimated sensors of epithermal neutrons, CSETN1–4, which are installed inside the neutron collimator with walls composed from polyethylene and 10B layers.
 The 3He nucleus has one neutron less than the main isotope of Helium (4He or alpha particle) and it has a large cross section for capture of low-energy neutrons in the reaction n + 3He → 3H + p + 764 keV. The energy released by this reaction is distributed between a triton 3H and a proton with 191 keV and 573 keV energies, respectively, in inverse proportion to their masses. Pulse height spectra are digitized in 16 linear channels corresponding to the energy deposited in the detector volume after a neutron has been captured by the 3He nucleus. Maximum energy is deposited in the detector if both the proton and triton are stopped in the detector's volume. Detection efficiency of these sensors is about 100% in the low-energy range and decreases by 10 times at ∼500 eV. Its numerical simulation using MCNPX code [Pelowitz, 2005] is presented on Figure 2. It is similar to the efficiency estimated for the LPNS neutron detectors [see, e.g., Feldman et al., 2000].
 The collimator absorbs the flux of neutrons impinging on the detectors at large angles (within 20–75 degree off the nadir direction) and provides the capability to measure lunar neutron flux of epithermal neutrons within the narrow FOV (<20 degree around nadir direction). It is illustrated on the Figure 3 with experimental results from the ground calibration and its numerical simulations of the LEND angular dependence. The tops of the collimated counters are nadir (+Z direction) pointing and covered by a Cd filter 0.5 mm thick to reject neutrons with energies <0.4 eV. The x axis in Figure 3 is defined as the angle between nadir and the direction of incoming neutrons. Normalization is done to the maximum in the peak at zero degree. The model is presenting our preliminary attempt to numerically simulate LEND calibration. It has good agreement within 50 degrees and ∼2σ- deviations within range 50–75°. This is good for Moon applications because the angular size of the Moon at an altitude of 50 km is about ∼76°. For larger angles (>75°) the adjustment between numerical simulations and results of calibration is still in progress. We would like to take into account the heavy mechanical fixture supporting LEND at the calibrations (it significantly influences at the backscattering at the large incident angles); see alsoAppendix A.
2.2. Detector System of Doppler Filter
 Sensors STN1 and 2 use the Doppler filter technique for detection of thermal neutrons [Feldman and Drake, 1986]. There are two 3He proportional counters attached to the opposite sides of bottom skirt of the collimator body (Figure 1). STN1 is oriented in +x axis of the spacecraft and STN2 is oriented along −x axis of the spacecraft. Each sensor has efficiency about 100% for thermal energies and lower for epithermal neutrons (Figure 2). The spacecraft velocity (∼1.65 km per second) is comparable with the velocity of thermal neutrons (with energies ∼0.015 eV). The neutron sensor facing in the spacecraft velocity direction detects significantly more neutrons than the detector facing backward. This difference in counting rates strongly depends on neutron energy.
 The spacecraft overtakes neutrons with energy less than 0.015 eV, enhancing detection in the forward detector, and comparably reducing detection in the backward detector. Ideally, in this approach the difference of counts between STN1 and STN2 is produced by thermal neutrons coming directly from the Moon.
 In PDS we have not yet presented the difference of counting rate between Doppler detectors (STN1 and STN2). They are presented as individual counting rates in each detector corrected only for efficiency changes (see below for details). The more complicated data processing including analysis of difference in the counting rates and creating a global map of thermal neutrons (based on this difference) is presented in another paper [Litvak et al., 2012].
2.3. Detector System for Cd Difference
 Two 3He counters, STN3 and SETN, are used to measure omnidirectional fluxes of neutrons above and below the cadmium cutoff energy at 0.4 eV. The first, Sensor of Thermal Neutrons (STN 3), is attached to the upper part of the collimator body (at -Y top edge of collimator body) (Figure 1). This sensor is identical to the Doppler proportional counters having the same sensitivity to neutrons. The second counter is the Sensor of Epithermal Neutrons (SETN) attached to the upper part of collimator body (opposite side to the detector STN3, +Y top edge of collimator body). This sensor is wrapped in thin Cd enclosure (0.5 mm thick) to absorb thermal neutrons with energies below 0.4 eV (see Figure 2). The data from SETN are used to make the global epithermal neutron map of the Moon with low spatial resolution (full width at half maximum is greater than the altitude of the spacecraft). Both sensors measure neutrons above the Cd cutoff energy, so the difference of counting rates between STN3 and SETN (so-called Cadmium difference) also may provide a measure of low-energy thermal component of neutrons <0.4 eV.
2.4. Detector System for High-Energy Neutrons
 Finally, the Sensor for High Energy Neutrons (SHEN) with the stilbene crystal is surrounded by a plastic scintillator for anticoincidence of external energetic particles (Figure 1). The size of stilbene cylinder is 5 cm in height and 2.5 cm in radius. The numerical simulation of its registration efficiency is presented in Figure 4 for the neutrons along the axis of the cylinder (nadir direction).
 The pulse shape discrimination circuit of SHEN separates signals in the stilbene into components of neutrons and gamma rays. Neutrons are detected with long risetimes because of recoil protons; gamma rays have much shorter risetimes produced by electrons. The time versus amplitude signature of scintillation profiles identify neutron and gamma ray components according to risetime selection criteria. The signals from neutron and gamma ray components together with the signal from the anticoincidence sensor are digitized to three 16 channel spectra.
 The SHEN is integrated in the center of the collimator (Figure 1). The whole detector assembly is inherited from fast neutron detection technique of HEND/Odyssey instrument [Mitrofanov et al., 2003; Boynton et al., 2004]. The collimator was not optimized for SHEN as it is done for CSETN1–4, but nevertheless it significantly suppresses the flux of fast neutrons with large incident angles to the nadir direction. The measured angular dependency efficiency curve (during ground calibrations) and its numerical simulations for SHEN inside collimator are shown in Figure 5. The angle on Figure 5 is defined in the same way as for the calibration of the collimated sensors presented on Figure 3 (see section 2.1).
3. LRO Mission Phases
 LRO mission phases are divided up into (1) Launch, (2) Cruise, (3) Lunar Orbit Acquisition, (4) Commissioning, (5) Nominal Mission, (6) Science Mission, and (7) Extended Mission.
 The Launch phase (18 June 2009) began with launch vehicle lift-off and lasted about 90 min until LRO separation from the launch vehicle. The Cruise phase (18 June 2009 to 23 June 2009) began with spacecraft separation and lasted until the lunar orbit insertion (LOI) sequence began. During this phase, the LEND instrument performed early in-flight calibration tasks. It was switched into different configurations (different signal discrimination thresholds and different levels of detector high voltage) to make measurements of spacecraft background induced by galactic cosmic rays (GCR). The cruise time series data for each LEND detector are submitted into the PDS data system.
 Using Cruise in-flight measurements, the optimal configuration was determined for LEND, which has been used for almost all observations. It was found during cruise that all eight3He counters have the low-amplitude component because of contribution of charged particles (seesection 6.2 below). Parameters of this component and the data for spacecraft neutron background in the LEND optimal configuration are used for LEND data analysis (see section 8 below).
 Lunar Orbit Acquisition (23 June 2009) began with the start of the LOI sequence and lasted until the commissioning orbit was attained. The main task for in-flight calibrations in this phase was to measure variations of neutron flux from the Moon and from the local background as they varied with distance from the Moon. The Commissioning phase (23 June 2009 to 14 September 2009) began with the attainment of the 30 × 216 km commissioning orbit. The periapsis was over the south pole. During this phase LEND was nadir pointing. Data from the commissioning orbit is the important part of the instrument in-flight calibration, because it measured at the variable altitude above the Moon. It also provides valuable science output because of low-altitude flybys (∼30 km) above south pole. In this paper we did not discuss these measurements in details and did not use it as part of the data reduction process (seesection 6). Low-altitude data from the commissioning orbit will be combined with measurements gathered during the frozen elliptical orbit (begun in December 2011) and presented later with analysis at a higher level of statistical confidence than with commissioning data alone.
 The nominal 1 yr mapping mission began with the attainment of the mission lunar polar orbit of 50 km with altitude variations from 30 up to 70 km. The Science mission is actually the continuation of the nominal mapping operations in the same orbit and lasts for two years. During mapping, with consecutive orbits of ∼2 h separated by roughly 1.07° longitude, LEND obtains complete mapping of the lunar surface every 2 weeks. Near the pole, >89°, this separation of tracks of the axis of FOV is ∼0.55 km, much less than the instrument resolution. At the equator, this separation is ∼32.4 km.
 The orbital dynamics of a low lunar orbit force LRO to perform periodic sets of station-keeping maneuvers. A passive spacecraft in an initially circular polar orbit about the Moon at an altitude of 50 km would impact the Moon in approximately 41 d. The prime meridian of the Moon is through the mean-sub-Earth point. Constraints on the station-keeping algorithm led to optimal maneuvers occurring at approximately 90° or 270° lunar longitude. Yaw flips of the spacecraft between +X and −X directions have been performed roughly every 6 months.
 An Extended Mission phase follow the Science Mission phase where the orbiter has been placed into a frozen orbit so as to prolong possible mission lifetime.
4. Instrument Modes and Operations
 LEND operates autonomously, collecting data throughout the lunar orbit generating approximately 0.26 Gbits of measurement data per day. The nominal collection interval is 1 s or 86,400 records per day. During this time, the spacecraft covers 1.65 km over the lunar surface, which corresponds to about 6 independent measurements over a spatial resolution footprint (spot with diameter of 10 km).
 LEND has two operational instrument modes: Measurement and Stand-by. While in Measurement mode, high voltage of sensors is “on” and the instrument generates measurement and housekeeping data. In stand-by mode, high voltage of sensors is “off” and housekeeping data packets only are generated, the science data packets contain “zeros.” While the instrument power is “off,” the external heaters from spacecraft only are “on” with provision of external temperature data from the spacecraft system.
 The position of LEND onboard is very close to one of the attitude-control propulsion nozzles. Because of this proximity, the high voltage (∼2 kV) to the detectors is turned off during station keeping maneuvers because of risk of corona discharge during propulsion. In this mode science packets are still generated but all counts in all channels of the detectors are zero. The combination of polar orbit with station-keeping maneuvers near to 90° and 270° longitudes results in a significant inhomogeneity of exposure time and larger statistical errors along this longitudinal belt, but it is the cost for low altitudes 30–70 km of the spacecraft at the mapping stage (seeFigure 6). The LEND map segments most useful for the high surface resolution data analysis are discussed by Mitrofanov et al. .
5. Telemetry Data Stream and Product Types
 Science and housekeeping files are sent during K-band passes from LRO to either White Sands (NASA Space Network), in Las Cruces, NM or Deep Space Network (DSN) and then to Goddard Space Flight Center Mission Operations Center (MOC). This telemetry data stream is received at the ground station, with raw science and engineering data embedded (NASA Packet Data, Committee on Data Management and Computation (CODMAC) Raw, Level 1). Checksums, transmission errors, and duplicates are handled by the MOC and the raw data are passed to the LEND Science Operations Center (SOC) at the University of Arizona's Lunar and Planetary Laboratory (LPL).
 The LEND SOC is responsible for recording and verifying these data. An automated ingest process is running continuously to process science, housekeeping, and real-time Consultative Committee for Space Data Systems (CCSDS) packets and Spacecraft, Planet, Instrument, C-matrix, Events (SPICE) kernels provided by the LRO MOC, while verifying the integrity of the data. The ingest process verifies the consistency of the data, and sorts it into the relevant database tables. This real-time processing uses predictive SPICE kernels received from the MOC. When definitive kernels are received, a separate process is launched to update the spatial information covered by the new kernel. These data packets (e.g., raw voltages, counts) are processed and inserted into an Oracle database at full resolution, time ordered, with duplicates and transmission errors removed and checksums revalidated for every file package (NASA Level 0; CODMAC Edited, Level 2). A detailed description of the LEND Experimental Data Record (EDR) (NASA Level 0) data product is contained in the LEND_EDR_SIS document released with each quarterly PDS release. The EDR product contains uncalibrated neutron counting spectra from LEND, housekeeping, spatial, and ancillary information which will be used in the reduction of higher-level data products.
 The LEND Reduced Data Record (RDR) data products (NASA Level 1) consist of five types of data: (1) LEND converted HK Data (CHK), reduced housekeeping converted to engineering units; (2) LEND rectified science data (RSCI), temporally and spatially rectified science data associated with the collection interval; (3) LEND derived data (DLD), a time series of neutron detector science data that has been processed and corrected for multiple parameters described later in this paper; (4) averaged LEND counts data (ALD), neutron detector science derived data that have been further normalized and averaged to yield a counts/second format for each Lunar map pixel; and (5) averaged LEND flux data (ALF), neutron detector science derived data that have been further normalized and averaged to yield orbital epithermal and fast neutron fluxes. The LEND RDR data products contain individual spatially and temporally rectified or derived LEND spectra and associated data as well as normalized and averaged LEND spectra corresponding to NASA Processing Level 1A (CODMAC Level 3), 1B (CODMAC Level 4) and 1C (CODMAC Level 5), respectively.
6. Physical Effects Accompanying LEND Measurements, Which Should Be Removed From Data Products
 There are a number of systematic effects in LEND measurements, and special data processing procedures are needed to make the necessary corrections to the raw LEND counting rates to generate maps of neutron counts across the lunar surface. In this section we present a summary of the correction, normalization and other data processing procedures applicable to the raw LEND data to create derived DLD and ALD PDS data products.
6.1. Efficiency Saturation Effect
 High-pressure3He sensors detector efficiency changes as a function of time following detector high-voltage turn on. This effect is seen in all eight3He proportional counters used by LEND. They have active volume equivalent to 5 cm diameter, 5 cm length and filled with gas under 20 atmospheres of pressure. Detector length and diameter, high pressure, and significant counting rates lead to the generation of a HV warm up saturation effect. The physics of this effect is thought to be related with the slow increase of the effective volume at ends of the cylinder near the anode thread. The effect of this gradual increase in counting rate has been observed during all ground operations with the instrument. Several ground tests have been accomplished with different proportional counters to investigate this effect in more details. Finally it was concluded that it is related with slow increasing of the effective volume at the ends of the cylinder near the insulator of anode thread because of slow distortion of electric field by accumulating of charged particles on the surface of the insulator. The strength of this effect increases with the increasing of the counter's pressure and incident flux of epithermal and thermal neutrons hitting the neutron detector.
 After the high voltage is turned on the proportional counter counting rate smoothly increases as a function of time until it reaches final saturation (variations may be as large as 20%). The rate of increase depends on the incident flux and the saturation (for LEND type of proportional counters) may take several days (see Figure 7). During instrument development, it was assumed that LEND would operate continuously in lunar orbit, as HEND/Odyssey is currently operating in the Mars orbit, and such “efficiency saturation” effect would not have any significant impact on mapping data. However, during the mapping phase of operations this effect significantly modifies the LEND data because of multiple HV off/on activities. LRO spacecraft needs to be kept in a circular orbit, which requires delta-H momentum unload and station keeping maneuvers approximately every two weeks. During each maneuver LEND HV is turned off and put in Stand-by mode for several hours. As a result, all LEND data are broken into approximately two week time segments, each requiring correction for the efficiency changes.
 To derive the efficiency variations from the data we analyzed each two-week operational interval. We have selected all measurements when the spacecraft is above the 86° latitude both for south and north, and distributed the data from these measurements along a time axis with “zero” time at the beginning of the two week interval beginning right after HV turn on. Our goal was to maximize statistical confidence while minimizing the impact of surface composition differences. We suggest that counting rate variations from one flyby (poleward of 86°N/86°S) to another one are primarily associated with sensor efficiency changes.
 The observed efficiency change (Figure 7, where each data point presents counting rate derived from flybys poleward 86°S or 86°N latitudes) was fitted by a function:
where t is time from the initial HV turn on; A, k and t0 are free parameters for fitting. To get best fit parameters the expression (1)is applied to all two-week data intervals (each such interval is fitted individually) for each3He sensor individually. The fitting curve is used to adjust the number of measured counts at a time t by a correction factor, which is the ratio between the saturation level A and the value of F(t) at the moment of measurement. The similar approach for LEND efficiency correction was also implemented in the work of Eke et al. .
 The example of time profiles of original and corrected counting rates is shown in Figure 8 (omnidirectional epithermal detector, SETN) for several months of LEND observations. We have shown only SETN because it is less noisy than collimated detectors and provide better illustration of trends in counting rates.
 The suggested approach of efficiency correction is not the best one and may be improved based on the additional ground calibrations. But it is good enough to be considered as a first order of approximation. Now it is implemented for all LEND PDS data.
6.2. Distinction Between Counts From Charge Particles and Neutrons in 3He Counters
 Neutrons in LEND proportional counters are detected through absorption by 3He in the 3He(n,p)3 H capture reaction. The standard pulse height spectrum in each proportional counter does not depend on the energy of incoming neutrons and represents the well-known spectrum of deposition of energy from reaction products in the volume of the detector. In the case of LEND, the measured spectra from3He counters are different from the standard shape. The shape of spectrum (measured during ground tests) is presented on the Figure 9 (shown by red color). We tried to optimize it to study impact of charged particles measured by the proportional counter in space conditions.
 Counts from charged particles are primarily associated with Solar Particle Events and Galactic Comic Rays (see section 6.3 for details). These energetic particles, mainly protons, pass through the detector volume, losing a part of their energy through the ionization of 3He gas. The energy deposition from GCR charged particles has a maximum of tens of keV and falls off above this energy. The energy scale (16 channels) for proportional counters is linear and is stretched out to cover energy deposition from all products of nuclear reaction 3He(n, p)3H (with energy release of 0.76 MeV) used to detect neutrons in the volume of counter. So, the energy deposition from GCR charged particles with energies of tens of keV will be registered in low-energy channels (below channel 10). That is why during measurements in space low channels are mainly populated with counts induced by charged particles but not only neutrons. For omnidirectional sensors the fractional contribution from charged particles is quite small because they record much higher number of counts of lunar neutrons. In the case of collimated sensors CSETN1–4 the lunar counting rate is much smaller, and charged particles dominate in the low-energy channels (seeFigure 9).
 A special procedure was developed for deconvolution of 3He counts into the fractions from neutrons and charged particles. The shape of the neutron spectrum in 3He detector has been derived from measurements in the laboratory. The red lines on Figure 9 actually represent a shape of neutron spectra measured in the laboratory. The lab spectra have been normalized to the counting rate measured onboard LRO at channels 14 and 15. The result of this normalization was assumed as a real spectrum of neutrons measured on the LRO orbit. The subtraction of this neutron spectrum (red line) from the total measured spectrum (black line) is shown by blue line of Figure 9and addressed to the registration of GCR charged particles. It has approximately a power law shape in channels 5–15 with an exponential cut off in channels 2–5 because of low-amplitude discriminator in the instrument's electronic.
Figure 9 shows the result of spectral deconvolution for the omnidirectional sensor, SETN, and for the sum of the four collimated sensors, CSETN1–4 at the orbit around Moon (first year of mapping phase).
 After counts of collimated sensors are deconvolved into two separate components of external charged particles and neutrons, one has to determine the best selection of channels for studying neutron signal with highest signal-to-noise ratio (SNR). The highest-amplitude channels are free from the contribution of charged particles and can be summed to get maximum statistic of neutron counts. The lowest ones are populated with charged particles and after their subtraction the difference spectrum has large uncertainties. Especially it is important for the collimated sensors CSETN1–4 where neutron signal is weak because of collimation. In our analysis we have focused on the counting rates accumulated in the neutron spectra (extracted after separation between neutrons and particles; seeFigure 9b) in CSETN1–4 detectors. The result is presented on Figure 10 using the following algorithm:
Number of counts produced by neutrons in the 16th channel (red line in Figure 9b) has been divided by their statistical uncertainty and resulted value has been associated with the last bin (x axis) on Figure 10;
Sum of counts produced by neutrons in 15th and 16th channels have been divided by the statistical uncertainty of this sum and resulted value has been associated with 15th bin (x axis) on Figure 10;
Following this sequence, sum of counts produced by neutrons in channels from ith to 16th have been divided by statistical uncertainty of this sum and resulted value has been associated with ith bin (x axis) on Figure 10.
Finally, the derived curve has been normalized by its maximal value.
 The curve presented on Figure 10 has a distinct maximum because adding counts in higher channels (mostly populated by neutrons) leads to the increasing of SNR just because of accumulation of statistic (right “wing” after the maxim of the curve presented on Figure 10). But adding counts from lower channels (mostly populated by charged particles) decreases signal-to-noise ratio because of large uncertainty of these counting rates (left “wing” before the maximum of the curve presented onFigure 10).
 It has been found that the optimal set of amplitude channels for measurements of neutron counts is 10–16 (see Figure 10). The DLD and ALD (and other) data products for PDS are based on data measured in these channels of 3He counters.
6.3. Data Correction for Variations Due to SPE and GCR
 Another systematic counting rate variation applicable to all LEND detectors is associated with solar activity and variation of Galactic Cosmic Rays. Sporadic Solar Particle Events (SPE) produces variable background from charged particles, which significantly increase the counting rates in all detectors and might fully hide the regional variations of lunar neutrons. The solar activity during the LRO mapping phase is increasing from the extended minimum to a higher level with many powerful SPEs. For some, the counting rate in LEND detectors is increased by several times (Figure 11). For the mapping of lunar neutrons such interval shall be excluded from the data set (see below).
 A more difficult problem is associated with long-term variations (weeks and months) of the average counting rate in LEND detectors because of slow variations of GCRs. The GCRs are the primary reason for the neutron production in the lunar subsurface, and intrinsic neutron emission from the Moon varies with variations of GCRs. Follow this, the all neutron data applicable for the mapping should be corrected for temporal variations because of variations of GCRs.
 It is known that higher solar activity protects the inner part of the interplanetary space from propagation of energetic charged particles of GCRs, resulting in a change in the total lunar albedo of about a factor of 2 between solar minimum and maximum [see, e.g., Castagnoli and Lal, 1980]. LRO was launched during a period of quiet Sun, which lasted unexpectedly long and resulted in a historical maximum of observed GCR flux. This provided the possibility to get significantly higher count statistics because of larger flux of subsurface neutrons. After several months of mapping the flux of lunar neutrons has started to trend down. Figure 12shows time profiles of neutron count rates in sensors STN3, SETN, and at the sum of collimated sensors CSETN1–4. It is evident that the first two profiles are very consistent with each other with similar features of short variations, while the profile from collimated sensors shows the same trend, but is quite different in features of short-time variations. For collimated sensors, the amplitude of these fast variations is much larger than for STN3 and SETN.
 One cannot use neutron data with long-term variations for mapping of lunar emission, because the goal of the mapping is to compare the flux at different surface elements of the Moon when all other conditions are the same. To remove the long-term variations from the LEND data, we have normalized the count rates in all detectors to the values corresponding to the initial stage of the mapping mission at the maximum of GCR flux (the fall of 2009).
 At the beginning of mapping phase, after correction for the detector efficiency saturation effect (see section 6.1), we estimated the counting rates in collimated CSETN1–4 detectors as 1.13, 1.31, 1.25, and 1.38 cps. The total counting rate for all collimated detectors is equal ∼5.1 cps in this case.
 We have used mean values of counting rate averaged for the periods of LEND operations to measure the average time profile for each sensor (before averaging, the correction for the efficiency saturation has been implemented for each detector). Each period of operation was equal to ∼2 weeks, which corresponds to time interval between successive station keeping maneuvers. Two weeks time interval is approximately equal to half a lunar day, and LRO orbit covers practically the whole lunar surface during this period of time. It means that averaged counting rates shall be the same for each time interval. The possible source of difference between them is related with significant variations of GCRs. To remove such variation we adjusted all average values of counting rate to the value measured at the beginning of the mapping phase. In this approach the ratio between counting rate at the beginning of mapping and the average counting rate for given period of operation was used, as the GCR correction factor for all data frames within given operation interval.
 All time profiles of counting rates presented in the LEND PDS are corrected for detector efficiency changes and variations of GCRs.
 The GCRs are also showing short-term variations on a daily time basis. It may influence on the orbital counting rates measured by LEND detectors. In the current edition of PDS such correction is not yet included, butBoynton et al.  investigated this effect and found that GRS induced variations of counting rates may be as large as ∼1%.
6.4. Data Correction for Variation of Altitude
 A correction has to be applied to adjust counting rates to the same orbital altitude. The LRO polar orbit has variable altitude from 30 to 70 km around an average of 50 km. The periodic evolution of the orbit at particular longitude belts during a cycle of two weeks is transformed to the measured counting rate for the standard conditions of 50 km orbit. During each two week period, the altitude varies slightly with latitude but more significantly changes as a function of longitude. For the nearside of the Moon, the average altitude is about 48 km but for far side it is about 5–7 km higher.
 This difference is especially important for omnidirectional sensors STN1–3 and SETN, because they are recording neutrons from the entire visible Moon. On the other hand, the total counting rate measured in collimated sensors CSETN1–4 is slightly varies with altitude (much smaller than omnidirectional detectors). Basically it is explained by the fact that total counting rate in the collimated detectors consist of counting rate measured in the FOV (constant as a function of altitude), GCR induced background (it decreases with decreasing of an altitude) and backscattering of Moon neutrons in the body of spacecraft (it increases with decreasing of altitude) [see, e.g., Eke et al., 2012]. The resulted counting rate is showing very small variations with altitude (<1% during mapping phase). For the 1st edition of the LEND data products we have presented total counting rates in collimated sensors CSETN1–4 corrected for efficiency changes and GCR variations but not divided into the separate components of collimated signal in the FOV, GCR induced background and neutrons backscattered in the body of spacecraft. As a result of such presentation the total counting rate presented in the PDS was not corrected for the for altitude variations. To make corrections for the GCR induced background users of the LEND PDS data should subtract from the total counting rates the values presented in Table 2. Several different approaches and arguments how to distinguish components of collimated signal in the FOV and backscattered neutrons in the body of spacecraft can be found in Lawrence et al. , Mitrofanov et al. , Eke et al. , and Litvak et al. . It is also presented in section 8.4.
 For omnidirectional sensors the neutron flux from the lunar surface is proportional to the solid angle of the visible Moon producing increasing/decreasing counting rate with decreasing/increasing altitude. For each data frame we have estimated an altitude adjustment factor for the counting rates to the standard altitude of 50 km. This simplified procedure does not include several additional effects, such as variations of the real landscape and correction for the limb darkening of neutron emission.
 The correction factor for given altitude A (in km) to altitude of 50 km was
where R is the Moon's radius (1738 km).
 The more accurate procedure for the amplitude correction will be implemented for the products of the 2nd edition of LEND PDS.
6.5. Temperatures Corrections of Data From SHEN
 There are several thermal effects, which should be considered for LEND data processing. First one is associated with neutron counts from the fast neutron detector SHEN. The signal produced in this detector correlates with the temperature of the instrument, which is slowly varying over a wide range from −2 to 22°C depending on the relative orientation of spacecraft, Earth, Moon and Sun. We removed these variations from the data by normalizing counting rate of SHEN to the average value corresponding to the fixed temperature of −2°C. This was the minimum temperature observed during lunar eclipse. We have identified LRO orbits corresponded to the minimal temperature (−2°C) and estimated for them average counting rate. The average counting rates for other orbits were adjusted to this value.
 Another effect is surface temperature. The flux of thermal and low-energy epithermal neutrons from the Moon depends on surface temperature. The numerical simulations using MCNPX code (Monte Carlo N-Particle eXtended [seePelowitz, 2005]) shows that effect is largest for thermal neutrons (<4–5% of signal variations between hot equator latitudes and cold poles) and smaller for epithermal neutrons (∼1% of signal variations between equator and poles) [see, e.g., Little et al., 2003; Lawrence et al., 2006]. However, we do not apply the surface temperature correction for LEND PDS data products (DLD and ALD) but discussed it in other publications [see, e.g., Litvak et al., 2012]. The surface temperature effect could be taken into account at the next stage of LEND data processing, when the 2nd edition of LEND PDS data will be developed.
7. LEND Data Preparation for Analysis and DLD Creation
 The first product to undergo scientific data processing is Derived LEND Data (DLD). As previously mentioned, it represents the series of LEND counts in different detectors. These values are normalized to take into account the different engineering and operational conditions. Before any processing takes place, a number of frames is removed from the time series. We present, below, the reasons these frames are excluded before DLD creation.
7.1. LEND Data Exclusion From the Analysis and DLD Creation
7.1.1. LEND Not in Measurement Mode
 There are two most frequent situations: LEND is in standby or transiting from standby to measurements mode. In both cases, the instrument generates scientific frames with “0” counts which are meaningless. Some other cases, when the instrument is in a nonstandard measurement configuration for calibration purposes or other checkout mode, are also omitted from the DLD data set. Also, a time interval of 6 h after transition to measurement mode is removed because the detectors efficiency curve changes too quickly in this period of time to be accurately normalized. The normalization process is described later in this section.
7.1.2. Data Recorded During a Solar Particle Event
 To exclude data during SPEs, the time profile of LEND data is regularly compared with the proton data from space environment satellites GOES-13 and Advanced Composition Explorer (ACE). All time intervals corresponded to the SPE detected by these satellites were visually inspected in the LEND data and removed, if they have any significant deviation from the usual counting rate measured during the period of quiet Sun. In the DLD data products these time intervals are presented, but assigned as −1. The SPEs are short-time events in comparison with duration of whole LRO mission. Excluding them, one loses a few percent only of the total raw data volume.
7.1.3. Good Fitting Not Found for Efficiency Correction Curve
 The normalization process described later in this chapter implies that a good fitting curve exists for the current power-on period. However, in some cases this period is too short, and an acceptable fitting curve for the count rate of the detectors cannot be found and applied. Without this curve, an appropriate normalization cannot be created for the time series, and thus DLD data is not populated within the time interval in question. On the basis of this issue, for LEND PDS we have selected only operational intervals (between instrument switch on and off) with duration more than 1 week.
7.1.4. Outlier or Off Limit Events
 Infrequent, sporadic, and randomly distributed “outlier” values are seen in the raw detector count rates (3He proportional counters) when summing channels 10–16. These outliers could be produced either because of some microdischarge in HV circuit of a counter, or because of some corruption in the instrument memory. In both cases, individual data frames with outliers are excluded from the mapping data together with time intervals of their detection. Distributions constructed for each detector show an additional bump at higher counting rates which deviates from standard Gaussian shape distribution. On the basis of these results, we set the limits on outlier values for any detector record where the sum of the counts in channels 10 through 16 is greater than the value in the vicinity of anomalous bump on the right wing of the Gaussian distribution. The expected impact to the data set is considered negligible because the amplitude of anomaly counting rates is very high (for example, for collimated detectors it may be 10–15 cps compared with average neutron counting rate of about 1.3 cps) and its probability to occur in Poisson counting statistics is estimated at the level of more than 7σ [see Boynton et al., 2012]. In Figure 13 [see Boynton et al., 2012] we have shown histogram of total detector counts in channels 10 through 16. Records with counts >11 are considered outlier values. They constitute only 0.01% of recorded events.
 Detector records that are defined as outlier events are not used in preprocessing (orbital average) products. They are present in the DLD data products, but are omitted for mapping purposes.
7.1.5. Data for Off-Nadir Measurements
 Occasionally, particular instruments on board LRO have requested off nadir slews for targeted imaging: Lunar Reconnaissance Orbiter Camera (LROC) or exosphere studies and Lyman Alpha Mapping Project (LAMP). The nadir angle (angle between spacecraft to subspacecraft point vector and the spacecraft to boresight vector) is recorded for each RSCI record in the database. The LEND off nadir measurements can be used in the analysis and estimate of spacecraft neutron background using angular dependency of counting rate but it is not appropriate for the primary observations of regional variations of collimated neutron flux. The off nadir activity leads to the deviation of spacecraft's Z axis from the nadir direction. During this maneuver the back side of the LEND instrument (not protected with thick layer of neutron collimator) will view the Moon with its field of view providing detection of additional noncollimated lunar neutrons. For this reason, the off nadir periods are excluded from the primary data set used for global mapping of lunar neutron emission. We used a conservative approach and excluded all time periods when the nadir angle was greater than 1.0 degree. These time frames were omitted from the DLD records. A Boolean field in the DLD record also tracks records that are off or on nadir.
7.2. Statistics of Available and Lost LEND Data
 In Table 1 one can find appropriate percentages of frames that were omitted during different stages of DLD data processing, as defined above.
Table 1. The Summary of LEND Available and Lost Data
Omitted Data Type
LEND not in measurements mode (LRO station keeping maneuvers)
Solar Particle Events
Good fitting not found
Outlier or off limit events
Off nadir measurements
Standard operations in nominal conditions
 One can see that even after removing the data that are not useful for scientific analysis, the instrument has produced about 78% of data with high quality.
8. Background Estimation for Mapping Neutrons From the Moon
 The major goal of the LEND investigation is to measure the spatial variations of neutron emission of the lunar surface, i.e., to map the flux of lunar neutrons. The goal determines the method of observations: the neutron counting rate is measured for elementary surface elements of the lunar map (pixels), and the difference between them characterizes the spatial variations of the lunar emission. There are different methods to identify a count within individual pixel(s) on the surface (see section 9 below), but in all cases each pixel shall accumulate some particular number of counts related with this surface area. To do this, first of all, one needs to separate the counting rate from the spacecraft background.
 There are several similar background components for mapping by omnidirectional detectors of neutrons and for mapping by collimated detector. The first one, BGDGCR, is associated with local neutrons from the body of the instrument and spacecraft, which are produced by energetic particles of GCRs. This component is the only source of neutrons in the cruise stage of flight. It is very easy to estimate in lunar orbit. Indeed, in lunar orbit this component decreases proportionally with increasing solid angle of the Moon, because the Moon shadows a part of isotropic flux of GCR:
where is the value measured in cruise (see Table 2) and ξ is a factor ∼1, which takes into account consumption of the LRO fuel after the lunar orbit insertion. LRO propulsion system uses hydrazine, which is a much stronger moderator of neutrons than the spacecraft structure and payload. We have made preliminary numerical simulation of spacecraft (LRO) with LEND onboard. The numerical simulation of the spacecraft model with variable amount of hydrazine has shown that ξ is close to 0.93.
Table 2. The Estimation of the Spacecraft Background
GCR Background, Counts per Second Measured at Cruise Before Orbit Insertiona
GCR Background at 50 km Altitude, Counts per Second, Calculated Using Expression (2)
 It is necessary to mention here that there are two types of GCR induced background. The first one is related with the producing neutrons in the body of spacecraft. The second one is the detection of GCR charged particles in the volume of proportional counters by ionization losses (see section 6.2). The value measured during cruise includes both neutrons and charged particles. These components also show similar behavior as function of altitude above Moon. The only difference is that value ξ is equal to 1 for charged particles. So, the values estimated for orbital observations according equation (3) and presented in Table 2 contain 100% of GCR neutron induced background and 90% of GCR charged particle induced background. It leads to the small systematic error that we underestimate the total GCR induced background during orbital measurements (for example, for collimated detectors CSETN1–4 it may be as high as 0.04 cps).
 The second background component, BGDbs, is due to backscattering of lunar neutrons by the mass of LRO (including the LEND mass itself) to the sensor. This component is approximately proportional to the solid angle of the visible Moon, ΩMoon/4π, but one has also to take into account the limb-darkening effect of neutrons emission [Lawrence et al., 2006]. In the ideal case of two thick target models for the Moon and LRO interacting with the GCR, one would expect that the decrease of local neutrons because of shadow by the Moon (see expression (2)) would be compensated by backscattering of lunar neutrons by the mass of LRO. In reality (because the mass of LRO is not a thick target), this compensation may not take place.
 The third type of background, BGDOther, may be associated with counts of neutrons (or photons, or charged particles; see section 6), which are not associated with the analyzed energy range. For example, sensor SETN of epithermal neutrons also detects a small number of thermal neutrons because of partial transparency of Cd enclosure around the counter. Also, epithermal neutrons are detected by STN3 sensor of thermal neutrons, and one should consider this as a background component for counts of thermal neutrons.
8.1. Background for Sensors STN1–3 of Thermal Neutrons
 Counting rates from sensors STN1–3 should represent the thermal emission for the Moon. Using the difference from the sensors STN1 and STN2 as a Doppler filter, one gets the difference of counts, Cfront – Cback, as the signal for measuring the thermal emission. One might expect that there is practically no background component for this signal, because background components BGDGCR and BGDbs for individual sensors STN1 and STN2 are about the same. These two sensors are installed symmetrically on two sides of LEND. The mass of LRO is located closer to STN1 than to STN2 but it is a secondary effect [see Litvak et al., 2012]. Direct comparison of the thermal neutron map created from the Doppler system STN1 and STN2 with the map for thermal neutrons from LPNS data shows good consistency between them [see Litvak et al., 2012].
 STN3 is the third sensor of thermal neutrons, which is installed at the top edge of the collimator. Counts of this sensor include the signal for thermal neutrons from the Moon together with the local background components BGDGCR and local neutrons due to backscattering BGDbs. One should take into account that the term backscattering is used here in the broad sense: lunar neutrons of all energies may scatter in the material of LRO, and be moderated. This scattering-and-moderation of neutrons is the diffusion process: background componentBGDbs of STN1–3 may be produced by lunar neutrons over the entire energy range from thermal particles up to MeV.
8.2. Background for the Sensor SHEN for Fast Neutrons
 SHEN background, BGDGCR, is about 0.25 cps at the orbital altitude of 50 km, and the total signal of fast neutrons with energy above 0.5 MeV is about 0.5 cps (see Table 2). The component of backscattering should be quite significant for this sensor because scattering of a neutron in the large mass of LRO may change its velocity vector, and scattered particles could penetrate SHEN from the unprotected side of the instrument; the collimator only protects sensors from direct neutrons from the Moon, but not from the structure of LRO itself. Within its FOV, SHEN is measuring fast neutrons starting from ∼1 MeV up to several MeV. The regional variations of these neutrons are mainly determined by the average variations of atomic mass in the Moon subsurface [see, e.g., Maurice et al., 2000]. The backscattering neutrons detected in the SHEN are originally emitted by the lunar surface with even higher energies, but lose some energy due to collisions with spacecraft body and collimator. They also represent variations of atomic mass in the regolith. As a result all components of the Moon's neutron signal measured by SHEN are showing the same regional variations and we may say that LRO amplify the counting statistic in SHEN but worsening its spatial resolution (because backscattering neutrons came to SHEN outside its FOV).
8.3. Background for the Sensor SETN of Epithermal Neutrons
 The average count rate of SETN is about 11 cps, which includes the GCR induced background component ∼1 cps from GCRs (Table 2). BGDbs component is also need to be taken into account but first analysis shows that it does not significantly influence on the regional variation of counting rates measured by SETN. It is confirmed by the good agreement between global map of epithermal neutrons measured by SETN and epithermal neutron map derived from LPNS (see comparison in the work of Litvak et al. ). No excess of counting rate is visible at equatorial latitudes on nearside of the Moon. It indicates an absence or small fraction of high-energy epithermal neutrons (high-energy neutrons backscattered in the collimator [seeLawrence et al., 2011; Litvak et al., 2012]) in the total counting rate.
 SETN has a Cd enclosure to absorb neutrons below the Cd threshold at 0.4 eV, but the thickness of 0.5 mm is not large enough to stop all of them. Therefore, we consider the third background component at this sensor, BGDCd, which is produced by neutrons with energies <0.4 eV. Again, the spatial variation of thermal neutrons is distinguishable from variations of epithermal neutrons. So, to determine the epithermal neutrons in the data of SETN, one has to separate the signal of epithermal neutrons with their spatial variations from two background components, BGDbs and BGDCd, which also have two different spatial variations. To resolve this deconvolution problem, one may also use the data from collimated sensors CSETN1–4, which also include the counts for epithermal neutrons and fast neutrons (see the discussion in papers by Lawrence et al. , Mitrofanov et al. , and Litvak et al. ).
8.4. Background for Collimated Neutrons CSETN1–4
 The average counting rate (corrected for efficiency and GCRs variations) in the LEND collimated detector system is about 5.1 counts per second (cps). It contains the signal counts of lunar neutrons from the collimated FOV, counts of background counts from lunar neutrons outside the FOV and background counts from neutrons produced by the spacecraft itself.
 Estimation of BGDbs is important for understanding the local background of neutron emission onboard the LRO. Especially it is important for collimated sensors where ratio between signal in the FOV and counting rate produced by neutrons outside of FOV can characterize the instrument's ability to observe local spots on the Moon surface. The Moon partially shields GCRs and as a result the spacecraft background induced by GCRs is decreasing by approaching to the Moon. On the other hand, the surface of the Moon produces neutrons, which scatter and diffuse within the body of LRO and produce the background component BGDbs. In the thick target approximation of LRO, one would assume that scattering and transport of neutrons takes place similarly in the mass of LRO, as it does in the subsurface of the Moon. In this case backscattering of lunar neutrons may compensate the loss of local neutrons produced by LRO because of partial obscuration of GCRs. In reality such an approximation does not work and it is necessary to use complicated numerical modeling together with experimental measurements to distinguish value of backscattering component.
 The LEND capability to collimate neutrons has been the subject of several critical papers. They have suggested that such a collimated neutron telescope is not able to detect any significant neutron counts within such a narrow FOV because it will be polluted with high-energy epithermal neutrons penetrating through the collimator walls [Lawrence et al., 2010, 2011; Eke et al., 2012]. These papers have suggested that the LEND collimated signal is as small as 0.05 cps, which is less than 1% from the total counting rate measured by the collimated detectors. The opposite point of view supported by LEND team claims that counting rate in the FOV is much higher and may be estimated as 1.5–1.9 cps [Mitrofanov et al., 2010b; Litvak et al., 2012; Boynton et al., 2012] and agrees with the original numerical modeling of the instrument [Mitrofanov et al., 2008].
 Below we tried to overview several alternative attempts to estimate weight of the background components in the total counting rate measured by LEND collimated detectors.
 The GCR induced background (both neutrons and charged particles) component, BGDGCR, could be easily estimated for CSETN1–4, as well as for other LEND sensors, from cruise measurements (see beginning of section 8 and Table 2). But the contribution from lunar neutrons outside the FOV requires special consideration. Indeed, the CSETN sensors have a narrow FOV, much smaller than the solid angle from the Moon. The CSETN FOV signal is produced by counts from detection of epithermal neutrons within the FOV, and it varies with change of neutron emission from the surface with a spatial resolution of about 10 km. The energy range of these neutrons is determined by the efficiency of the 3He counters and may be identified as 0.4–500 eV (see Figure 2). However, there are also counts from neutrons, which have been scattered and moderated by the mass of LRO, including LEND itself; this is background component BGDbs of CSETN1–4. These neutrons are emitted by the entire visible Moon and their variations correspond to a much larger resolution scale (as seen by omnidirectional detector), and their energy could be higher than 500 eV, because they are moderated in the mass of the spacecraft before detection. After the value of BGDGCR = 2.4 cps (Table 2) is subtracted from the total count rate of CSETN, 5.1 cps, one gets the count rate of 2.7 cps from lunar neutrons, which is produced by superposition of signal counts, CFOV, of directly propagated epithermal neutrons in the energy range 0.4–500 eV from the instrument FOV, with the background counts BGDbsof scattered-and-moderated neutrons from the entire visible Moon.
 Several approaches could be applied to estimate this component of the background. Eke et al.  have used the altitude dependence of different components to distinguish between collimated signal (does not depend on the spacecraft altitude), GCR induced background (decreases with decreasing of the spacecraft altitude) and backscattering background (Moon's neutrons detected by collimated sensors outside the FOV. This signal increases with decreasing of the spacecraft altitude). By results of the analysis, Eke et al. have concluded that collimated signal is very small (<0.05 cps) and overwhelmed by omnidirectional high-energy epithermal neutrons backscattered in the collimator and spacecraft. It means that total counting rate measured in the collimated detectors is dominated by omnidirectional component and cannot resolve local spots of hydrogen variation on the Moon surface. This conclusion looks to be flawed because it contradicts to LEND observations of PSRs and NSRs presented bySanin et al. , Mitrofanov et al. [2010b, 2012], and Boynton et al. . So, Sanin et al.  have shown that LEND collimated detectors (no assumption about backscattering component has been used, just monitoring of regional variations of total counting rate) observe significant depression of neutron flux inside PSR Shoemaker in comparison with sunlit neighboring areas. Moreover regional variations of total counting rate correlates well with local relief (measured by LOLA/LRO) of Shoemaker crater [see Sanin et al., 2012; Boynton et al., 2012]. For the additional illustration we have added Figure 14, which shows how the significance of Shoemaker PSR detection (by LEND collimated detectors) is accumulated as a function of time (during mapping phase). The significance (number of standard statistical deviations σ) was calculated as a difference between counting rate measured inside Shoemaker and counting rate measured outside Shoemaker (in the narrow latitude belt limited by the latitude size of Shoemaker crater) divided by the statistical uncertainty of such measurement (same technique as has been presented by Sanin et al. ). It is seen that collimated detectors can resolve this PSR and show monotonous increase (gradual accumulation of the counting statistic during mission) of the negative significance (counting rate inside Shoemaker area is smaller than counting rate measured outside Shoemaker area) down to ∼−5σ at the end of mapping phase.
 Another way to get estimates of CFOV and BGDbsfor the collimated sensors, CSETN1–4, could be based on the simple phenomenological requirement: the large-scale variations of epithermal neutrons in the energy range 0.4–500 eV, as measured by omnidirectional sensor SETN, should be the same as the large-scale variations of epithermal neutrons in the same energy range, as measured by collimated sensors, CSETN1–4 [seeMitrofanov et al., 2011; Litvak et al., 2012]. Lawrence et al. have proposed that high-energy epithermal neutrons may pass through the collimator and be detected by3He tubes. It was based on the analysis of global map of counting rate measured by collimated sensors. This map [see, e.g., Lawrence et al., 2011; Eke et al., 2012; Litvak et al., 2012] demonstrates significant depression of neutron flux at the poles together with significant excess of neutron flux at the nearside mare basins. It may be interpreted as a map of high-energy epithermal neutrons which behave both as low-energy epithermal neutrons at the poles enriched with H and as high-energy neutrons at the nearside mares enriched with Fe. Using LPNS data,Lawrence et al. suggested that detection of high-energy neutrons may add up to 2.25 cps into the total counting rate of the LEND collimated detectors. It leaves only ∼0.45 cps for the collimated signal (just compare 2.25 cps with 2.7 cps defined above as a counting rate for all lunar neutrons). LEND team [Mitrofanov et al., 2011] found this approach reasonable as a concept but flawed in details. First of all, this technique provides still small (10% from the total counting rate) collimated signal which contradicts other LEND observations (see discussion above). Second, Mitrofanov et al.  have repeated this approach using LEND fast neutron data and found that backscattering component is significantly smaller ∼1.1 cps which turns into the ∼1.6 cps for collimated signal (again compare 1.1 cps with 2.7 cps). Third, Mitrofanov et al. suggested to use the global latitude variations (Orbital Phase Profiles) of the averaged counts rate over two lunar hemispheres. The Orbital Phase Profiles can be used for comparison of regional variations seen by SETN and CSETN detectors. It may help to distinguish between collimated low-energy epithermal neutrons and uncollimated high-energy epithermal neutrons.Litvak et al. has farther developed this approach and has shown that high-energy epithermal component could not explain the whole observable picture and fast neutrons regional variations need to be involved to interpret global map of collimated counting rate. This analysis resulted in following values of collimated signal (CFOV) and the backscattering component (BGDbs): CFOV ∼ 1.7 cps (∼30% from the total counting rate) and BGDbs ∼ 1.0 cps.
 The measurements goal of the LEND collimated sensors is the experimental testing of spatial variations of lunar epithermal neutrons with an amplitude of about 5% at the distance scale as small as 10 km. The signal-to-noise ratio for these measurementsSNR(col) (5%) with four collimated sensors could be estimated, as a function of total exposure time for some particular testing surface element
For many tested spots at the lunar poles, the total exposure time is comparable to or larger than 1000 s [see Mitrofanov et al., 2010b, 2011], so expression (4) shows that LEND has SNR ∼1 for the measurements of 5% spatial variation of neutron emission at the scale of 10 km on the lunar surface.
9. Map Creation Algorithm and Errors Propagation
 One of the LEND key reduced data products is global maps of neutron counting rates recorded in the different detectors (ALD type data in PDS). The creation of such products requires using special mapping procedure to distribute orbital time series data into a pixel grid on the lunar surface. The size of the mapping grid may be different for various applications. For example in the case of LEND PDS mapping data products (ALD) the size of pixel is varied from 0.5° × 0.5° degrees (spatial resolution ∼15 km along latitude) poleward 80° latitude to 1° × 1° degrees (spatial resolution ∼30 km along latitude) between 80°S and 80°N latitudes. Each LEND counting rate is measured with time resolution of 1 s (standard duration of single data frame). During this time the spacecraft covers about 0.05° of arc or approximately 1.6 km along longitude. We approximate the trajectory projection during a single time bin by a straight line (first degree polynomial function). Going from one data frame to the next one we determine the fractions (k1, k2, …, kn) of the trajectory projection on lunar surface corresponding to a given pixel:
where li is length of a part of trajectory projection which belongs to i-th pixel andL is total length of trajectory projection on the lunar surface (see Figure 15).
 Using estimated fractions ki, the counts and exposure time are accumulated in pixels of the map:
where cj is the number of counts accumulated in j-th data frame,tj is the duration of j-th data frame,kij is the fraction of j-th data frame length ini-th pixel,n is the total number of LEND frames involved in mapping procedure, Ci is the total number of counts accumulated in i-th pixel of map, is the total exposure time for i-th pixel of map, andFi is the average flux in i-th pixel of map (counts per second).
 Taking into account mapping procedure corrections and normalizations (described in the sections above) the common expression of resulting signal, Si, in the i-th pixel of map could be presented as
where Ci is the number of counts accumulated in the pixel i, Ai is the total correction and normalization coefficient between original counts and corrected counts accumulated in this pixel and B is the background estimation. The resulting uncertainty of Si comes from the Poisson statistic of counts Ci and uncertainty of background B:
 This paper presents the first description of LEND data processing, which transforms the data from the initial level of raw measurements to the level of derived counts, representing the neutrons emission from the Moon. We believe that all systematic trends and effects in the data, which are not related with the spatial variations of neutrons over the Moon are explained in the text above.
 The LEND instrument has passed through a set of ground tests and physical calibrations to estimate its sensitivity to neutrons flux. The primary goal of the calibration was to get the instrument response functions (individually for each neutron detector), as a function energy and angle of incident neutrons. This goal has been accomplished by irradiating LEND with a calibrated neutron source located at different distances and angles from the instrument.
 All calibrations were performed in two stages with different units of LEND instrument. The first stage is so-called reference calibrations. It has been done with the flight unit of LEND now flying onboard LRO. For this calibration we have used reduced number of cases with different incident angles and simplified program of background measurements. Such a calibration approach was selected because of the limited time available for a ground test program at the LEND home institution (Space Research Institute) and for system tests at Goddard Space Flight Center (GSFC) as part of spacecraft payload.
 The second stage of calibration activities was carried out with the LEND spare flight unit (it is identical to LEND flight unit). This calibration activity was accomplished after LRO launch and was done in a much more detailed way than the first stage. We compared results from the two stages of calibration using the reference data measurements (done both for flight and spare flight units), which made it possible to propagate calibration results from the spare flight unit to the flight unit onboard LRO.
 Calibrations were accomplished by the LEND science team at the Joint Institute for Nuclear Research (JINR, Dubna, Moscow region), a member institution of the LEND science team. One of the main calibration tasks was to suppress the local neutron background because of backscattering of neutrons from walls, floor and ceiling in the calibration facility. All calibration activities with LEND units were carried out in a special building with huge internal volume (a distance between walls is more than 40 m, distance between floor and ceiling is more than 10 m). The LEND instrument was mounted on a mechanical chassis 3 m above the ground allowing its rotation by 360° with high accuracy for individual positioning (<1°) around the vertical axis and horizontal axis. Measurements at different angles were used to produce the 3-D angular response function of the LEND detectors, as required for collimated sensors CSETN1–4.
 The neutron source was placed at the same height above the floor and at a large distance from the instrument (>3 m). Measurements were performed with a calibrated 232Cf neutron source encapsulated inside a polyethylene sphere with a radius of 7.6 cm. Such a source provides a continuous neutron spectrum quite similar to that of lunar neutrons.
 To estimate the local neutron background from walls, floor and ceiling we made a series of measurements, where the neutron source was located at different distances, R, from the instrument. The count rate dependence on the distance from the source has a component proportional to R−2 and a constant. The first one was identified with direct neutrons from the source, and the second was related to the background. These measurements were repeated for different instrument pointing to take into account variations of background as a function of local geometry. To derive the effect of backscattering of neutrons from the mechanical chassis and on the LEND body, we made measurements, where the FOV of CSETN1–4 was screened by a cone shaped polyethylene shield to prevent detection of direct neutrons from the source.
 The campaign of the second stage of physical calibrations took place during 6 months of 2010. There were ∼107 individual measurements with LEND during this stage with 10 different instrument settings and configurations. These data allowed us to make a full description of LEND, as the physical instrument. Detailed paper reporting on LEND physical calibration is now in progress.
 The LEND team is thankful to the LRO project team, which provides all necessary help and support for LEND investigations onboard the spacecraft. Coauthors of this paper are also thankful to the International Space Science Institute for great opportunity to work on the major subjects of this paper and for kind hospitality in the period of 2009–2010.