[X] My Assistant

[Rick Sternbach]

Okay, It's June 15. Where's the updated LOLA data? I keep finding the stuff from March in all the usual places. Okay, I'm sure it may take a few days, but does anyone have a clue if the LDEMs are going to get cleaner and crisper this time around?

I've already seen a resin casting of a moon globe made from the LDEM_64 data, but I wanted to wait a bit until more blanks were filled in.

Rick

[Phil Stooke]

It takes time to get it all fixed up... but people, if you want to see how unbelievable LOLA is going to be when it's all done, check out this amazing presentation from the NASA Lunar Science Forum, held at NASA Ames last month. This is by Maria Zuber, and - alas - it didn't survive the PDF-making process properly. I have asked if it can be fixed. But even so, it looks good. Check out the LOLA map of the floor of Shackleton on page 15. As I say I've asked for it to be fixed, so we'll see.


http://lunarscience2010.arc.nasa.gov/sites...files/Zuber.pdf

Other pressies here:

http://lunarscience2010.arc.nasa.gov/agenda

Lots of goodies.

Phil

[James Fincannon]

"It takes time to get it all fixed up... but people, if you want to see how unbelievable LOLA is going to be when it's all done, check out this amazing presentation from the NASA Lunar Science Forum, held at NASA Ames last month. This is by Maria Zuber, and - alas - it didn't survive the PDF-making process properly. I have asked if it can be fixed. But even so, it looks good. Check out the LOLA map of the floor of Shackleton on page 15. "


Yes, it looks good, and is much better than what we have had and they are doing great work.

But there are some points that should be emphasized (and I do not think they are sufficiently)........

For a 25 m/pixel grid for within 25 km of the South Pole,
(1) Only 72% of the grid elements have at least one laser data point. This means 28% are empty, but the DEMs show them filled (interpolated). This is a concern to me because although it creates a nice continuous image/DEM, it needs an accompanying error map to help a user to understand the missing data, interpolation error, etc.
(2) The average number of laser data points in this grid is 1.5 +- 1.4. For a 25m by 25 m pixel, you would like around 25 laser data points to get good coverage with 5 m diameter spots. This means when the DEM is constructed, the height for that pixel is supposed to be an average height of the surface, but really it is the height average of from 0% to 12% of the surface area within the pixel.
(3) How does the laser data point treats the area it "paints"? Is this the average height within the 5 m spot or the highest spot or what?


With coarser grids (240 m by 240 m/pixel), the percentage of surface area with laser data spots is around 8%.

Thus the magic of creating the DEM (i.e. sausage making) has alot of aspects that people need to realize and see if it applies to their usage. I have been stymied from doing illumination analysis because of these concerns. Sure I can do it and have done it with my analysis tools and use either the DEMs or the actual laser points, but I cannot create an error bar, so I have to reassess this laser data DEM.

[zeBeamer]

Thus the magic of creating the DEM (i.e. sausage making) has alot of aspects that people need to realize and see if it applies to their usage. I have been stymied from doing illumination analysis because of these concerns. Sure I can do it and have done it with my analysis tools and use either the DEMs or the actual laser points, but I cannot create an error bar, so I have to reassess this laser data DEM.

James,
we discussed that offline, but I do not agree that we need to paint the whole Moon to have a realistic map at say 25m resolution. Indeed the current filling ratio of the 25x25m near the poles was 72% when we discussed that in June, but it will keep on improving.
The September release in a couple of weeks will have side products for each of the DEMs containing the counts of laser shots in each pixel. People can use that as a mask to see where you can be more or less confident in the measurement averaging (actually a median).
But I am not sure that is what will capture the interest of most people here. And having "gaps" in the DEMs to reflect the actual sampling would not necessarily make it better; I would expect most people want a full map, and do not want to do their own interpolation (they may not be familiar with the tools to do so) when they want to render a given region.
To reassure you, the data is not put through magic black boxes, and the workflow is actually pretty straightforward. It just gets messy to deal with when you have billions of points and those high resolutions.

With coarser grids (240 m by 240 m/pixel), the percentage of surface area with laser data spots is around 8%.

Do you mean globally? In June, polewards of ~85deg, we had ~90% coverage at that resolution.

All,
The September release is coming very soon, and the DEMs will be updated this time, with more than 2 billion (good) points which went into them.
I updated the Celestia products, and they are already available here: http://imbrium.mit.edu/EXTRAS/CELESTIA/
They were made from the to-be-released 128ppd grid. Annoying seams should be gone (note to djellison and John). We are also releasing a 256ppd map (in four tiles), but I do not have the time currently to do it from that source (that would bring us to level6). And currently, it might be overkill. Others are welcome to try it out!

As for the new polar maps you saw in Maria Zuber's Ames presentation, this is not exactly what is going to be released. I'm not going into details here, but basically, the PDS release will still show some (reduced compared to before) orbit streaks near the poles. I will try to provide a better image of the South Pole, as it seems to be of interest here

Erwan

[James Fincannon]

"I do not agree that we need to paint the whole Moon to have a realistic map at say 25m resolution."

What is the meaning of the word "realistic"? It may be realistic for people who want to render topographic surfaces for illustrative purposes or animations or simulations that do not require interacting with the surface. But I think one would not consider it realistic if one is planning a rover path or doing certain types of illumination analysis.

I think you and your team are doing great work, but I think some accuracy/realism aspects go over most users heads and are ignored.

"Indeed the current filling ratio of the 25x25m near the poles was 72% when we discussed that in June, but it will keep on improving."

This is the number I quoted before and is what bothered me because the DEMs show continuous surfaces where they cannot really be definitive as such (other than interpolation). Yes, given enough time you should get "100%" (meaning, to other readers, at least one 5 m diameter laser spot within all of the polar 25m by 25 m areal elements). But I have a little trouble with the rationale of interpolating an entire pixel height based on maybe <4% of the area being painted by laser light (i.e. one laser spot for a surface that needs 25 laser spots to fully define). Maybe the LOLA data is meant just for a certain purpose and I am trying to use it for a purpose that was not intended and I should really use the stereo imagery derived terrain instead.

>The September release in a couple of weeks will have side products for each of the DEMs containing the counts
>of laser shots in each pixel. People can use that as a mask to see where you can be more or less confident in the
>measurement averaging (actually a median).

This will be helpful.

>And having "gaps" in the DEMs to reflect the actual sampling would not necessarily make it better; I would
>expect most people want a full map, and do not want to do their own interpolation (they may not be familiar
>with the tools to do so) when they want to render a given region.

Gaps would be no good in the DEM, but what I think some users would like is the number of shots per pixel (0 shots would tell them it is pure interpolation). It would also be nice to have an error estimate for each pixel (maybe based on the difference between the interpolation pixel height and the average of the heights of the laser spots within the pixel).

>To reassure you, the data is not put through magic black boxes, and the workflow is actually pretty straightforward.

Interpolation is kind of magical in that it is hard to intuitively know how the interpolation will work out all the time for every set of points. You are not simply drawing a straight line between points, it is much more complex.

>>With coarser grids (240 m by 240 m/pixel), the percentage of surface area with laser data spots is around 8%.
>Do you mean globally? In June, polewards of ~85deg, we had ~90% coverage at that resolution.

By this I mean, for a 240m by 240 m DEM, I created a grid of 5 m non-overlapping spots to fill it which gives you 48 by 48/ 5 m spots or 2304 laser spots needed to cover the whole pixel. Then using your average number of laser spots/pixel within 25 km of the south pole (136+-63), I get a maximum of 8.6% and an average of 6% of the area painted by laser light. Sure, the number of 240m by 240m pixels that have at least 1 laser spot is ~100% in the case, but what I am saying is the kind of interpolated height based on 6-8% of the surface area needs some sort of error bar associated with it, since it is very hard for any of us to figure it out just from the raw data.

>The September release is coming very soon, and the DEMs will be updated this time, with more than 2 billion (good) points which went into them.

This is great! Still, globally, this means you are covering ~.1% of the lunar surface with laser light. So will LRO be merging the stereo imaging with LOLA data for significant regions? Are you working with those guys?