help me! (light pollution help)

[WilliamC]

I am doing a science fair project related to light pollution. I plan to zoom in on 'earth at night' maps located on the Internet and count the number of pixels in a certain 'area' (e.g. mid sized city) and see if light pollution naturally occurs in areas where there is population. Unfortunately, few (or no) maps on the Internet are reliable enough for me to look at (because of low res, no state or country boundaries, chance of being Photoshopped to make the image look 'better', etc...) and the process seems too tedious (city boundaries hard to 'distinguish', takes time to counting pixel, etc...). Can anyone help me, like introduced me to some special software or a good image (ahem...I mean really good, and better than the IDA stuff or the World Atlas of Night Brightness)? And if you can find the latter (an image) please give me a suggestion on how I can make the project less tedious. If you think I should have a new project, please keep it so I don't have to go outside during the nighttime or have to buy too many things. Thank you!

[deleted-2574]

Hi WilliamC,

The following web site may help:

http://saveourstarryskies.homestead.com ... ionQA.html

It has pointers to other resources and a provision to get your question answered.

Please excuse if you already found this site.

[ghariman]

Hi there,

I found a website from NASA:
http://visibleearth.nasa.gov/view_detail.php?id=1438
There you can find an image (TIF file) that is 16384 x 8192 in resolution (the file size is almost of 40MB - quite big).
I believe this image should be sufficient for your undertakings.

I am not an expert in Image Processing. But if you are familar with the tool MATLAB and have access to a full version then there is a very nice Image Processing Toolbox that comes along with it.
What I would do (conceptually speaking as I have never done this before) is feed the image into MATLAB (how ? - not sure) and do some Image Processing and then produce a 3D plot where the Z axis shows the intensity of color that is closer to WHITE.
MATLAB works in terms of matricies. So you would be able to build a database of the intensity of WHITE color on the map.

Not much help on the sotware side. But hey now you have a huge night-light map to work it. Thanks to NASA !

[deleted-2574]

Hi George,

One of William's problems was that the data available on the internet did not map to city/state. I'm not sure if he was aware of the NASA data until you pointed it out to him. But in any case, the NASA data does not apparently map to city/state.

The exception being if there are only a few hits on the NASA data, and the mapping of a set of data from NASA to city/state is easy. Any thoughts?

[WilliamC]

Somehow I feel my project is too simple and too difficult at the same time. I can't get or access MATLAB (Do you think my parents are going to buy me something that expensive? ). I could layer images in a GIS but they're hard to use and ArcGIS costs $1,900 and $4000 with their Spatial Analyst. I need to learn some really weird code for GRASS or with other free GISs. I don't know enough programming to code a pixel counter/analyzer and the list goes on and on and on.

Can anybody suggest to me a similar but less tedious experiment? Once again, I would be great if the time spent outside is minimized. And the amount of materials to buy minimized.

Note: Don't answer unless you read the whole thing. Doesn't take 30 seconds.

[ghariman]

Hi David,

Ah I see his point now. Well I would go for the manual route and use an atlas of the world to approximately map the location of the lights on the night-lights map from NASA.

He could print out the NASA night light image on a big sheet of paper.
And then print a scaled version an the world atlas (so that it overlaps the night-light printout) and then do a manual "mapping" of where certain big metropolitan sit.

Crude but do-able. What do you think ?

Hi George,

One of William's problems was that the data available on the internet did not map to city/state. I'm not sure if he was aware of the NASA data until you pointed it out to him. But in any case, the NASA data does not apparently map to city/state.

The exception being if there are only a few hits on the NASA data, and the mapping of a set of data from NASA to city/state is easy. Any thoughts?

[ghariman]

Hey again,

Well I think there is a trade-off between money and time here (as always in this world of ours).
This problem is a complex one. And so to do a "less tedious" amount of work would translate to "more cost".

By the way if you are a student you can get a STUDENT version of MATLAB that is only $100-200 (the last time I bought mine).

But if you read my reply in the last posting above there is a very simple and cheap of doing it, but yes you need to spend more time doing it. Print both the night-light map and a world atlas to a big (and same) scale. Then lay them on top of each other.
Put a plastic sheet on top of the two (like saran-wrap) and then 'draw' out the boundary of the city you see on the world atlas map, and the pressure you put on the plastic on the top most layer should imprint on the night-light map on the very bottom.

Does this sound reasonable to you ?

Somehow I feel my project is too simple and too difficult at the same time. I can't get or access MATLAB (Do you think my parents are going to buy me something that expensive? ). I could layer images in a GIS but they're hard to use and ArcGIS costs $1,900 and $4000 with their Spatial Analyst. I need to learn some really weird code for GRASS or with other free GISs. I don't know enough programming to code a pixel counter/analyzer and the list goes on and on and on.

Can anybody suggest to me a similar but less tedious experiment? Once again, I would be great if the time spent outside is minimized. And the amount of materials to buy minimized.

Note: Don't answer unless you read the whole thing. Doesn't take 30 seconds.

[WilliamC]

I don't think I have a big enough printer though and I'm uneasy about tiling it.

Maybe I should do a new but (maybe) similar experiment with data available on the Internet I could find without too much digging (I hated spending 10 hrs digging only to find myself scratching my head)?

I could compare two, three, or even more sets of data...

[WilliamC]

If indeed I should print, if ignoring my thoughts about a new experiment, I might have to:
1) Get a new printer (Parents would never let me)
2) Use some other guy's printer.

Anyways what is 'big' in this case? And printers these days don't make that great imprints, even on saran wrap.

Also the images I have/know of generally don't tell their scales or projections.

I need a new experiment...(???)

*thinking*

[ghariman]

Hi,

You don't make an imprint with a printer.
This is what you do (if you want to do it)

1) print a night-light map
2) print a non-night-light map of the same scale and size
3) put the night-light map on the bottom
4) put the non-night-light map on top of the night-light map
5) take a transparent plastic (like saran-wrap) and put that on top of the non-night-light so that now you have a 3 layer sandwich.

Take a blunt instrument (such as a pen) and "draw" out the outlines of the city of interest on the plastic (saran wrap). If you press hard enough the imprint will show as an indentation (not as an actual inked print) on the most bottom layer which is your night-light map. So now you have an outline (indentation in paper) of the city of interest on your night-light map.

Now in terms of size, if you are only interested in a particular city (say San Francisco), then only print the night-light map (and non-night-light) map of that city on a regular letter sized paper.

I think this crude method should work. If not then perhaps someone else in this forum can give more suggestions ?

If indeed I should print, if ignoring my thoughts about a new experiment, I might have to:
1) Get a new printer (Parents would never let me)
2) Use some other guy's printer.

Anyways what is 'big' in this case? And printers these days don't make that great imprints, even on saran wrap.

Also the images I have/know of generally don't tell their scales or projections.

I need a new experiment...(???)

*thinking*

[WilliamC]

[thinking about it]

[WilliamC]

it would certainly be helpful if somebody could respond

i could of course just count a number of pixels around the city centers on the computer screen

[ghariman]

Hi again,

Hmmm...I'm sorry that no other experts are trying to help you out. But my stance is that counting pixels on a computer screen is 100x more difficult than the method I pointed out before and approximating area size on the papers.

[WilliamC]

if i do count pixels, then in what way should i do it?

it is difficult, but probably most convenient for my own time (let's just says i don't have to time to go to a print shop and print two giant maps; btw, it's not that i'm lazy, just some-stringent parents that are never home)

the only big problem i have with counting pixels off a screen is that hard to standardize the amount of pixels counted around the city, more so than where it should be centered on the city

is there anyways i can standardize the amount of pixels around a city in not-so-super-tedious way? (that's why it's difficult i guess)

[Louise]

if i do count pixels, then in what way should i do it?

it is difficult, but probably most convenient for my own time (let's just says i don't have to time to go to a print shop and print two giant maps; btw, it's not that i'm lazy, just some-stringent parents that are never home)

the only big problem i have with counting pixels off a screen is that hard to standardize the amount of pixels counted around the city, more so than where it should be centered on the city

is there anyways i can standardize the amount of pixels around a city in not-so-super-tedious way? (that's why it's difficult i guess)

I think the map idea is best. You suggested tiling it- I think that is probably a good solution. I don't think there are too many ways to do this, and most of them have been ruled out as not feasible (using image processing softwared, etc.)

Louise

[WilliamC]

SO how do I do it without taking centuries (tiling) and then having to start over if I mess up (forget)?

Well, it all depends on what 'tiling' means.

[WilliamC]

SO how do I do it without taking centuries (well actually i could do it all in a few weeks) and then having to start over if I mess up (forget)?

Well, it all depends on what 'tiling' means.

[ghariman]

It has already been 2 weeks since you first posted the question. I am pretty sure if you had started with the "tiling" then that by now you are done with your experiment.

At some point in time you will need to just start the experiment.

SO how do I do it without taking centuries (well actually i could do it all in a few weeks) and then having to start over if I mess up (forget)?

Well, it all depends on what 'tiling' means.

[deleted-71447]

Hi William,
I would have chimed in earlier but I had forgotten about this free graphics program that could be useful for your type of project.
http://www.gimp.org/
With the "magic wand" tool, you can select contiguous regions of the same brightness (within a certain tolerance). For a selected area, using Dialogs -> Histogram will provide a count of the number of pixels selected as well as the mean, median, and standard deviation of the brightness. You can also overlay layers (such as a night and day map) and you can make the overlying layer semi-transparent so that you can see the underlying image. There are many other functions that could also be helpful. If you have some graphics experience, the program will probably be somewhat familiar. If not, it might take a lot of time and energy to figure it all out, and you might be better off with another approach.

I tried GIMP with the smaller JPG images from the NASA site and it worked fine. With the bigger GIF images from that site, GIMP ran out of memory (with 1Gb of RAM) and bogged-down. If you decide to use GIMP, you might need to find a way to cut some of those bigger images down into smaller blocks.

Regarding your topic, I suggest including per capita resource consumption or per capita income in addition to population density.

I agree with George that it is time to get started. Whatever method you choose is going to involve a fair amount of work. Also, the more the we see that you are committed to the project and are making progress, the more enthusiasm we will have to provide assistance.

[WilliamC]

Now comes the problem of figuring out the population of my 'small' image!

[Louise]

Now comes the problem of figuring out the population of my 'small' image!

I don't understand what the problem is... you're looking at a whole city, correct? Can't you just look up the population? There are also detailed poppopulation density maps that you can find if you google.


Louise

[WilliamC]

how do I find out the city boundary?

[Louise]

how do I find out the city boundary?

Isn't this exactly the same question you asked before? Use the same map you use to define city boundaries for the lights.


Louise

[WilliamC]

Sorry if I'm becoming a pain, but where exactly do I find a map with projected boundaries (with the same scale as the light pollution maps) so I can effectively measure the correlation of light pollution/population.

I NEED to know the population and light pollution of whatever city/county/state that I'm measuring.

Please ignore any contradictions I have made concerning this post and previous ones.

One map I am using: http://cleardarksky.com/lp/SaBarbCAlp.gif

[deleted-71447]

From here:
http://cleardarksky.com/lp/SaBarbCAlp.html
select "link roadmap" then click on the map
It opens a google map of the same region with a scale bar.
Adjust the slider on the left side of the screen until you have approximately the same area.
Use your "prtscrn" button to copy the screen image of that map.
Paste it into a graphics editing program like GIMP.
Resize one of your two maps so that both maps have the same scale.

If the google map is not sufficient for what you want to do, then do the same procedure with the population (or whatever) map that you need.

If it seems too difficult to do this at the resolution of cities, consider doing it at the resolution of states, or countries, or continents.