Help with ArcGis/Spatial Analyst. Anyone?

[klar]

I need to work with ArcGis and have no idea what I am doing. I have gotten advice on other unlikey subjects here before, so.... If anyone feels like giving me a few pointers that would be really great.

I have DEMs of a glacier for different years and want to see how the volume changes for different elevation zones. what I would like to eventually have is tables from which i can see how much volume is lost between year a and year b in the area of the glacier from 2000m-2100m, 2100-2200m and so on.

I would like to cut the raster DEM files along the glacier boundaries. i have not been able to figure out a way to use the clip tool to cut the raster along the glacier boundary polygon...? it will only clip two polygons or rectangles out of a raster.

Using the minus tool on 2 DEMs produces a new raster of differences in elevation/volume. I am trying to use Zonal Statistics as table to look at the actual values. How do I create an input file defining the zones (say increments of 100m of elevation)?

Many thanks if someone can help!

[bagtagley]

In Spatial Analyst Options, set the Analysis Mask to the Polygon shapefile. Then, open the Raster Calculator and bring in the DEM to the calculation and hit Calculate. This will "clip" the DEM.

I'm not fully understanding what you want to do, but with your analysis mask set, the tool will only analyze on the cells of the raster that fall within that mask, so in this regard, the initial "clip" isn't really necessary.

[Tippster]

I heart this place. I bet someone here knows how to fix a Flux Capacitor or build a Light Saber as well....

[klar]

cool, thank you bagtagley! that was easy and i am stupid. i just never find the settings to set, i guess.

i now need to create a file where i define seperate zones of 100m intervals in the glacier, either as raster or shapefile dataset. not sure how to "define a zone" do i just draw polygons?

[Scrubadub]

I did a project very much like this for my M.S. at UWYO.....hypsometric curves anyone?

[hop]

Nuke it from orbit, it's the only way to be sure.

(I ended up doing a shitty job of my GIS stuff in my thesis so for the manuscript, I found a friend and offered co-authorship in exchange for her far more expert than my own GIS skillz.)

[bagtagley]

cool, thank you bagtagley! that was easy and i am stupid. i just never find the settings to set, i guess.

i now need to create a file where i define seperate zones of 100m intervals in the glacier, either as raster or shapefile dataset. not sure how to "define a zone" do i just draw polygons?

I'm assuming you mean elevation contours. This can be done with the DEM directly, as opposed to the clipped data, if your analysis mask is set. In the Spatial Analyst pulldown, select Surface Analysis and Contour. Assuming your DEM is in meters, set the interval to 100.

[donutlynx]

cool, thank you bagtagley! that was easy and i am stupid. i just never find the settings to set, i guess.

Wouldn't say you're stupid. A large % of being a good GIS analyst (aside from writing code) is knowing which buttons to push, and in what order to push them in.

[khakis]

you can do what bagtagley said but if the surface elevation of the glacier changes I think that's going to introduce a confounding variable. I would clip the glacier DEMs based on 100m contours that are not dependent on the glacier, (i.e. shapefiles you make from one glacier snapshot, or that you interpolate based on non-ice surface)

[bagtagley]

I thought he was looking for contours for each individual snapshot of the glacier to compare elevation changes?

[khakis]

I fixated on the "volume" part of the question. If volume is the real variable of interest I would solve it by evaluating changes in glacier volume within a fixed 2d area over time. Showing movement of the 500m (for example) elevation contour from year to year would be an interesting visual but I don't think it would result in volume values that could be compared year to year

[bagtagley]

Yeah, that makes more sense. I missed the volume aspect.

[klar]

thanks a lot guys. trying to do something more along the lines of what khakis is saying. i can see how individual shapefiles would work rather than using contours, which i have been playing around with. very helpful tips from all! maggots rule!

I thought he was looking for contours for each individual snapshot of the glacier to compare elevation changes?

she, btw.

[bagtagley]

she, btw.

Ooops, sorry 'bout that.

[flyandski365]

Bookmarking this thread for future reference, you guys are awesome.

[MMP]

she, btw.

in that case,













pics please.

[geomorph]

i now need to create a file where i define seperate zones of 100m intervals in the glacier, either as raster or shapefile dataset. not sure how to "define a zone" do i just draw polygons?

Another option besides doing contours is to use the raster calculator, and build expressions such as
DEM>=2000m & DEM<2100m.
This will return a raster with 1's and 0's where 1 means the condition is true and 0 means the condition is false. A second raster calculation of this first result multiplied by the original DEM would produce a mini DEM of just the elevation band you're looking for.

Fuck Arc. I hate it but need it too.

[jeffmoore]

yah i can help you out -
ok, click icon; wait, tie your shoes, wait, blow your nose, wait, pick your nose, wait, have a coffee, wait, click around the desktop in frustration, wait, lunch time!, wait, not open shit, wait, get mad, wait, but then cool off, wait, mad again, wait, contemplate murder-suicide, wait, give in and give up, still waiting... die old and happy that ArcGIS couldn't open in time to piss you off any further - in 2D.

[geomorph]

yah i can help you out -
ok, click icon; wait, tie your shoes, wait, blow your nose, wait, pick your nose, wait, have a coffee, wait, click around the desktop in frustration, wait, lunch time!, wait, not open shit, wait, get mad, wait, but then cool off, wait, mad again, wait, contemplate murder-suicide, wait, give in and give up, still waiting... die old and happy that ArcGIS couldn't open in time to piss you off any further - in 2D.

This guy knows what's up.

[Telenater]

Putting aside the ESRI sniping (I not only drink the koolaid, I deal it too)...

There are several ways to "clip" a raster. Setting the analysis mask is one, you can also use the extract by mask took in the toolbox for spatial analyst too. Both will give you equivalent data if done right.

What method are you using for calculating your volume changes? It sounds as though you have a time series of elevation data. If so, the easiest is probably Surface Volume in the 3D analyst toolbox. If you isolate your glacier (and probably a small radius around it) in each time series, you can then calculate the volume of each and do appropriate subtractions to get your change.

I would consider switching up the order so that you do the elevation subtraction prior to the volume calculation. This will let you easily compare the underlying dems to see if there is a systematic error in one or more of the elevation models.

So, define your glacial study area and extract that area from each of your elevation models. Let's call them T1, T2, and T3 for time steps 1-3.

Then subtract the elevation of T2 from T1. This will give you a positive value for how much decrease in elevation there has been. If the edges that should be essentially bare earth have a value of 0 or close to it, you're probably ok on your elevation model accuracy. You can then use the Surface Volume tool to calculate the volume using ABOVE and 0 as your base plane, make sure that you're using an appropriate Z unit conversion for the volume if your Z and XY units are different (ie elevation in feet and XY coordinates in meters, if you're in decimal degrees, go back to step one and reproject to a projected coordinate system, or figure out what the conversion is right where your data is).

You and then repeat this for T1-T3 to look at total change over the entire time series.

My experience has been that it's worth checking whether you get a zero difference between your time series members in areas that should be the same. Overall, it probably won't make a huge difference, but it could be noticeable.

Edit: and if you don't have access to the 3d analyst extension, one thing to try rather than using zonal stats, take the elevation difference raster, and integerize it using the INT function (there isn't a built in round function). Then grab the attribute table from the resulting raster. The value column will be the depth of the difference, and the count will be the number of cells with that depth. Add a new column and use the field calculator (or export it for use in another database) to calculate the new column = Value * Count * Area of Cell *z factor. Where the z factor is a conversion constant between the vertical and horizontal units. Then you can get the sum of the new column from the statistics dropdown (right click on the column name and click statistics). That'll be the total volume of the change.

You could do it with zonal stats, but I think it'd be more difficult.

[Franz Klammer]

I heart this place. I bet someone here knows how to fix a Flux Capacitor or build a Light Saber as well....

It really is stunning what answers this place is able to deliver ...

[klar]

Putting aside the ESRI sniping (I not only drink the koolaid, I deal it too)...

There are several ways to "clip" a raster. Setting the analysis mask is one, you can also use the extract by mask took in the toolbox for spatial analyst too. Both will give you equivalent data if done right.

What method are you using for calculating your volume changes? It sounds as though you have a time series of elevation data. If so, the easiest is probably Surface Volume in the 3D analyst toolbox. If you isolate your glacier (and probably a small radius around it) in each time series, you can then calculate the volume of each and do appropriate subtractions to get your change.

I would consider switching up the order so that you do the elevation subtraction prior to the volume calculation. This will let you easily compare the underlying dems to see if there is a systematic error in one or more of the elevation models.

So, define your glacial study area and extract that area from each of your elevation models. Let's call them T1, T2, and T3 for time steps 1-3.

Then subtract the elevation of T2 from T1. This will give you a positive value for how much decrease in elevation there has been. If the edges that should be essentially bare earth have a value of 0 or close to it, you're probably ok on your elevation model accuracy. You can then use the Surface Volume tool to calculate the volume using ABOVE and 0 as your base plane, make sure that you're using an appropriate Z unit conversion for the volume if your Z and XY units are different (ie elevation in feet and XY coordinates in meters, if you're in decimal degrees, go back to step one and reproject to a projected coordinate system, or figure out what the conversion is right where your data is).

You and then repeat this for T1-T3 to look at total change over the entire time series.

My experience has been that it's worth checking whether you get a zero difference between your time series members in areas that should be the same. Overall, it probably won't make a huge difference, but it could be noticeable.


You could do it with zonal stats, but I think it'd be more difficult.

this is a cool and much more straighforward way but i ended up using zonal stats as table because I need the volume change for around 30 different parts of the glacier, not just the total change.

didn't find a way to do this with surface volume in a way that lets me do all the areas at once, which worked with zonal stats after i made a shapefile with appropriate polygons. i'm calculating volume as mean * cellsize * count and that seems (...) to give reasonable values.

i've also tried using the cut/fill tool for an idea of how much the glacier looses in which areas by applying it to two DEMs from different years but that seems to give really weird results. oh well, thanks again everyone!

[Telenater]

It sounds like for what you needed, zonal stats by table was the right call. Really interesting stuff. Let us know if you publish.

[klar]

bump

not least because of the help from people here i've managed to solve most of my major ArcGis issues. telenater is the man

i now have the following problem: i want to make graphics showing elevation changes in the glacier for different years. i would like to use the same colour coding for all the graphics.

i am displaying classified raster data with a colour bar and can't get the colour bar to stop shifting with each file. i.e. in raster1 value 30 will be pink, in raster2 (which has a higher maximum value) value 30 will be blue. experimenting with .lyr files has not worked so far.

any one want to help?

[Alpenglo]

I guess I'm not exactly sure what you mean. If you're using classified data you should be able to manually select each class(and manually set where the breaks are) and choose the color. If you using stretched data then it can be hard to get different datasets with different mins/maxes to line up. But with it manually grouped into classes you should be able to choose what colors you want. But maybe I'm missing something?

Not that it will help your problem but Colorbrewer is a great resource from a cartographic perspective for choosing color ramps/schemes.