Friday, September 26, 2008

Masking the lake for a good stretch

If you make your geologic maps using ArcGIS and work with nicely detailed color imagery, then you already know how useful a stretch is. If not, check this previous posts for dummies:

Now that you are back up to speed, I will share a simple trick I figured out by brute force that eliminates areas that may skew your stretch in an inconvenient way. Namely, large bodies of water. Right now, I am supposed to be finalizing mapping in the Spirit Mtn NW quad which includes parts of Nevada, Arizona, and Lake Mohave. Mapping along the lakeshore in the field is a joy; whereas compiling along the lakeshore is a pain in the neck...particularly when you use the standard deviation stretch restricted to the 'current display extent' which is usually the best option for contrast enhancement. The problem is caused by the black hole of lake pixels that dominate the statistics. The solution? Mask out the lake in a new raster using the 'extract' tool:

Here are the results from my current map area:



Epilogue. Someone with considerably more knowledge in GIS than I once explained to me how I could do this with raster math. I screwed around with that and failed. After numerous scans through Arc Toolbox (haven't you scanned that stuff over and over looking for something?), I finally found some commands that sounded useful. Remember, this is digital geoscience for dummies.

Monday, September 22, 2008

How to eat GeoSpaghetti

Geologic mapping can span many scales of time and space. Some of the most complex linework can result from evaluating a fluvial system in great detail in a small area and over a geologically instantaneous period of time. In my case, this scenario corresponds to the Bill Williams River in Arizona. For several years, I have been compiling detailed geologic maps of channel change on that river since 1953. The result? A heaping plate of GeoSpaghetti.

The image above is an excerpt from a 35 mile stretch of river. Yes. The river has undergone some profound changes in the last 50 years or so. Exactly how and why is beyond the point of this blog. One day I will publish it if it matters to you.

The point of this entry is to describe the various tools and methods that I have employed in ArcGIS to compile the lines in a meaningful way and to turn the resulting spaghetti into a meaningful map or series of maps.

The Project:

Map the bottomland geomorphology of the Bill Williams River at specific points in time using a chronology of orthorectified aerial photographs. At this point, I have mapped six generations of the valley bottom. The resulting plexus of lines is a logistical nightmare to a certain extent, but I believe I came up with a reasonable way to deal with them. If you map similar things and have better ideas or suggestions, please let me know.

  1. Set up a geodatabase...yes you need to know the basics of this fundamental operation. Add your lines as classes in a geology feature dataset.
  2. Determine a boundary to which you will be mapping and stick with it. Note that as you map different generations of lines, you will want to alter the just will. However, unless it is a major issue and you will diligently propagate that alteration through all of your line layers, resist the temptation
  3. Develop a line and polygon attribution scheme that is flexible and systematic. Important: this scheme needs to be logical and transferable to each generation of linework. Certain generations may require specific types of lines and polygons, but try to adhere to a common conceptual base so that it makes sense all the way through. Record the nomenclature in a spreadsheet and update it when you inevitably revise or add to your units. The spreadsheet can be a life-saver if you tend to work on too many projects and put this one down for a few months.
  4. Begin mapping the earliest generation of photograph if possible. It is best to map the images in chronological order for reasons that will soon become clear. Map lines NOT polygons. Starting with polygons is whacked. You can build them from lines in a matter of seconds.
  5. Once the earliest generation is mapped (and you have attributed the ‘proto’ polygons with a point feature coming if this is news to you) and the topology is all correct (you did build and check the topology, right?) copy it and rename it. Use this dataset as a starting point for the next generation of photos. Note: the tediousness is about to set in or get worse.
  6. Yikes. Your map is already a mess. Now you need to mesh the data in a logical way. You have added lines that preclude the existence of some of the previous generation’s lines, right? All of the precluded lines need to be removed (don't worry the originals still exist...remember, you copied them).
  7. Luckily, you have already built and analyzed the topology of your first layer, right? Well now build and analyze the topology of the second layer for laughs. The only rule you really need is the ‘no dangles’ rule. If you have the topology built and analyzed, you can use the ‘Planarize’ tool to break selected lines (even all of them) at each intersection. Then you can sweep through and select and delete all of the (now) superfluous lines.

That sounds easy right? It is easy, but really really tedious. Also, unless you have taken some preliminary precautions, you may lose all of your careful attribution. For better or worse, when you set up the geodatabase, you have many, many, options to ignore or address. Some of these are very useful to know about. One is ‘Default value’. What you choose here is the default attribution given to any piece of data that you enter. In the case of the Bill Williams map, setting the apyear (aerial photograph year) to the appropriate year was essential and useful. In other cases, I bet you can come up with some examples of your own where this would be useful.

You may also find yourself splitting and merging many lines. Unless you establish 'split' and 'merge' policies, you may get some disconcerting total loss of attribution that you didn't find out about until you split 10s to 100s of lines:

It is best practice to attribute your geolines immediately upon drawing them unless it is really ambiguous and you have a firm follow-up plan. Thus, choosing a default value for a line that requires some scientific judgment may not be the best idea. In the (recent) past, I have had a tendency to map many lines without attribution, assuming that I will do it in a ‘second pass’ through the data. Yikes. That is really stupid. For one thing, once you have drafted the line, you have covered it; for another, the ‘second (or third) pass’ idea isn’t very efficient and just effing snowballs up on you.

So what to do?

Option 1: Diligently attribute each line after you draft it.

Option 2: Have the program force you to attribute the line, or point, or poly, once you draft it.

Option 2 is the most efficient way to go. I just discovered this one.

Stay tuned for updates as to the progress of the Bill Williams River map...polygons coming next.

My new digital love, the cheap and rugged Laser Range Finder.

A few years ago, a geopal of mine turned me on to the 'magic box'...a laser rangefinder from LaserTech Inc. It was a little bulky and came in a box so padded that you were reluctant to keep it out of it for long...also, it cost several $1000. Nonetheless, I coveted that device from afar and borrowed it for several months.

On a more recent excursion, a different geopal pulled out a nifty little yellow number from LaserTech that made all of the same measurements (slope distance, vertical distance, horizontal distance, inclination, height) but came in a much smaller, more rugged package.

Turns out they got the thing on my advice several months prior when I scoffed at their use of a Jacob staff to measure several 100 meters of section (flat-lying rocks). More importantly, it turns out that the smaller, more rugged version is also less than $1000!

Damn right I got one. Maybe you should have one too. This little number could change your life if you sketch a lot of strat sections and guesstimate unit thicknesses or otherwise conjure up various spatial dimensions on the fly. With a little extra thinking, you can also construct an accurate cross section in the field if you are so inclined.

I used mine in the field over the weekend and was thoroughly satisfied with the results.

Friday, September 5, 2008

Geo-Rant: My Quads are killing me (and wasting tons of time)!!

No, not because I spend so much time hiking in the desert or because I lugged way too much crap down the South Kaibab Trail last month...not those quads. What is killing me is being a victim of mapping 7.5 minute quads. Mapping 7.5 minute quads is a waste of time. It is efficient only in a clerical sense, not in a scientific sense. Mapping on the basis of 7.5 minute quads amounts to mapping in a rectangular frame with boundaries that are (aside from some amazing coincidence) completely arbitrary with respect to geology.

Obviously, the implied goal of mapping 7.5 minute quads is to allow for a systematic framework for eventually mapping a bunch of officially circumscribed rectangles that cover an entire state or region. The key words here are 'officially', eventually', and 'rectangular'. Morevover, the concept of mapping quads is so deeply mired in the deeply pre-digital history of the USGS and the history of printing that it has become an ultra-anachronism.

I have been foolish enough to map a patchwork series of quadrangles along the lower Colorado River in an attempt to better understand the river's geologic history. Each time I move into a new quad, I learn more about that history (or more variations on it) that inform previous maps. Why in the hell I didn't just try to get funding to map the deposits of interest along the corresponding length of river is beyond me. Eight years later, I am still trying to finish some of those maps (sure, I am a perfectionist, but there are other reasons).

My most ambitious mapping project, the Ivanpah Mega-Map (Ivanpaviathan), is a classic example of how mapping quads can (temporarily) wreck your life. In that case, I stupidly proposed to map the entirety of all of the quads that fell even partly into the boundary of the watershed of interest. WTF? What an idiot. That is how mired I was in the Quad Mapping Model (QMM). I paid and paid dearly for that bit of stupidity.

My job involves mapping a lot of quads in Nevada. My agency has a goal of eventually mapping the entire state. Ha! That is not going to happen at 1:24,000 in my or my kids' (or their kids') lifetimes. In fact, this is simply not going to happen ever! Deal with it. Pick the areas that really matter (for whatever reason you like) and map them. Don't worry, you can still circumscribe the area with a quadrilateral that has easily defineable corner coordinates....