Clementine Lunar Observation Strategy

The observation strategy during systematic mapping was constrained primarily by the quantity of data that could be downloaded during each orbit. About 100 MB could be returned in a typical orbit. During times of RF blockage (full Moon as seen from Earth), the downlink was reduced to as little as 60 MB. Against these downlink constraints, we had to balance observation and compression strategies to achieve the following objectives: (1) global coverage in 5 UVVIS and 6 NIR bandpasses, (2) continuous LWIR imaging under each orbit strip, (3) HIRES polar imaging, and (4) additional HIRES imaging. Furthermore, we desired double imaging with the UVVIS, at long and short exposure times (or different gain/offset states) in order to acquire the best possible S:N without saturation. Typically, the short-exposure image is unsaturated even for the brightest lunar features, but has a poor S:N over dark regions, whereas the long exposure image has much better S:N over dark regions, but is saturated over bright craters or sunlit slopes near the poles.

We wanted to restrict the DCT compression to the most conservative setting allowed by the CNES/Matra chip: uniform quantization matrix with quality factor 4. This strategy produced compression ratios of about 5:1 for the UVVIS long exposures, 12:1 for the UVVIS short exposures, 2.2:1 for the NIR, 1.6:1 for the LWIR, and 3:1 for the HIRES. These compression ratios vary primarily as a function of high-frequency noise and scene contrast. The UVVIS images are almost entirely free of high-frequency noise, so high compression ratios were achieved (especially for the short exposures). The NIR has about 1% "hot" (bad) pixels and significant high-frequency structure in the responsivity matrix (i.e., flat field), resulting in much lower compression ratios. The LWIR has about 5% hot pixels and 10% noisy pixels. Because of the poor LWIR compression ratio and because the array is small (128 x 128), we decided to acquire the LWIR uncompressed. The HIRES has a poor MTF due to the intensifier, such that a resolution element is equivalent to about 3-4 pixels. Because of this MTF, we expected to achieve high DCT compression ratios (>10:1). This expectation proved false with the conservative compression strategy due to two types of high-frequency noise: (1) a "honeycomb" pattern due to the intensifier, and (2) shot noise due to the low gain states used in flight. Due to lifetime concerns about the photocathode and microchannel plates in the intensifier, we chose to use low gain settings and limit the number of HIRES images acquired at the Moon.

The HIRES was of primary importance to the planned Geographos observations, and of secondary importance at the Moon, so we wanted to insure that high-quality HIRES image of Geographos could be acquired. However, the LIDAR/HIRES was designed to require acquisition of an image with each lidar altimetry fireing, so in order to reduce the data volume, most of the equatorial/LIDAR HIRES images (lat -50 to +10 or -10 to +50) were acquired with the opaque filter in place--resulting in dark frames which compressed to very high ratios (about 50:1). Because the high-frequency (< 3 pixel) image content of the lunar HIRES images is all noise, and because of our downlink limitations, we chose to use the JPEG nonuniform Q-matrix to compress the images, with quality factor 4. This strategy resulted in about 8:1 average compression ratio over the polar regions. We chose to limit the HIRES to primarily monochromatic imaging of the polar regions, where illumination angles are good for imaging of surface morphology. Color HIRES "bursts" were a second priority.

One science team member, Carle Pieters, was concerned about the effects of lossy compression on the multispectral data, and proposed that a 10-degree latitude strip be acquired uncompressed each orbit. The uncompressed strip had to occur at least 40 degrees in latitude away from periselene latitude (approx. -30 or +30), because there was insufficient time to record uncompressed images to the SSRD and keep up with the mapping rate near periselene.

Because of the polar elliptical orbit of Clementine, redundant polar coverage was possible for the UVVIS and NIR. For the UVVIS, we chose to eliminate the redundant coverage, but we did attempt to cover each polar region twice: once nadir-looking at highest resolution, and once looking ahead of the orbit to cover the polar regions at lower phase angles. We chose to image pole-to-pole with the NIR because of S:N concerns at high latitudes (double imaging was not an option due to the frame rate of the NIR). This turned out to be a fortunate decision for a different reason: to constrain the offset history.

To satisfy the constraints and goals outlined above, we proposed the following nominal mapping plan for each systematic mapping orbit:

Observation                           CR    #frames   MB
-----------------------------------------------------------
UVVIS   5-color long exp             5:1    820       18
UVVIS  5-colorshort  exp            12:1    820        8
NIR 6-color pole-to-pole           2.2:1   1044       32
10-deg. lat uncompressed UVVIS/NIR   1:1           add 7
LWIR pole-to-pole                    1:1    870       14
HIRES  750-nm  lat  +/-  50-90       8:1    600        8
HIRES  4-filter,  10  deg. latitude 12:1    400        4
Dark frames/ star cal frames         1:1     68        4
LIDAR altimetry                          N/A         <<1
-----------------------------------------------------------
                          TOTALS:          4622       95

The 95 MB/orbit was easily returned (in the absense of downlink anomalies), except during times of RF blockage. During RF blockage, our recommended strategy for was to reduce the data volume by dropping the HIRES color, compressing all of the UVVIS/NIR, and compressing the LWIR. The actual data acquisition followed the science team recommendations only very approximately due to the severely understaffed nature of mission operations. "Temporary" sequence changes, for example during peak RF blockage, tended to stay in the sequences for additional orbits before someone found time to change them again. There were also many accidents. For example, following each DHU or HKP upset, the sequence tables that were reloaded were not always the most recent or intended versions. Murphy was proven correct many times: anything that could go wrong did go wrong. However, the spacecraft itself proved amazingly resilient to the many abuses it suffered, and nothing really bad (unrecoverable) happened until 3 days after leaving lunar orbit.

The highest priority throughout the lunar mission was acquisition of global multispectral mapping. There were approx. 10 spacecraft upsets or downlink problems during systematic mapping, resulting in loss of all or part of the data for these orbits. Gaps from month 1 were filled in month 2 (at lower resolution in the southern hemisphere), and gaps from the early part of month 2 (longitude 0-100 W) were recovered during the post-mapping period. For the latter parts of month 2 (longitude 0-230 E), a strategy was implemented to fill gaps in orbits immediately following an upset by pointing the spacecraft to the east, taking several orbits to fully recover. Most of the HIRES and LWIR observations were sacrificed during these late recovery efforts. These recovery efforts were largely successful, but there may remain small gaps (< 1% of the lunar surface) in the UVVIS/NIR mapping. There are larger gaps or bad data in certain bandpasses. For example, a filter slip problem late in the mission resulted in loss of several orbits of 750-nm mapping, and bad gain/offset settings resulted in useless 1250-nm images over the Orientale basin region during month 1. Several frames per orbit were typically lost due to timing errors or unknown causes. For the UVVIS, these individual frame dropouts typically affect just one of the two long/short integration pairs.