I was looking for something like my friends Roger or Chris had, a humble job in a technical area where I could combine my academic training with real industrial experience, I was willing to work up from the shop floor. I already knew I wasn’t Nobel prize material, but I wanted to learn how to do something. Now was the time to learn a trade.

It did not go as quickly as I expected, it wasn’t until about three or four months after I had returned to Florida that I found the job I was looking for, a trainee cartographer for an air photo mapping firm in Lakeland, about 30 miles east of Tampa; right off the Interstate, I could get there in half an hour from my apartment in town. I knew back then that the world was running out of resources, and that managing our diminishing supply efficiently would be important, hence profitable, work. I wanted a good job, I wanted prestige and money and interesting work, but I also wanted to do something useful to others. This was the seventies, remember? The work I was learning was highly skilled labor, and it required about a six months apprenticeship before I could be productive, and it would be two years before I could work without supervision, I jumped for it, even though I started off at minimum wage. My title was ‘compiler’ or ‘stereo plotter operator’, that referred to the mapping machine I was going to be working with. The technology was called photogrammetry, and although it was new to me, it had been more or less perfected by the end of the war, bear with me while I explain.

An aircraft flies over a piece of ground steadily photographing the earth with a very precise and enormous camera (the individual frames are 9×9 inches) in such a way that the images on the negatives overlap about 60%. This means that every point on the ground is imaged on at least two, and sometimes three, of the exposures. After the plane flies down a long flight line, it turns around and flies back in such a way that the exposures on this new flight line ‘sidelap’ the previous line. This process is repeated until the entire plot of ground to be mapped is completely photographed, and every spot on the ground appears on at least two subsequent pictures. This is where the stereo part comes in, if you look at the overlap section of two photos, and if each eye focuses on one of them, you can see the ground in stereo, in three dimensions! You can actually see trees and buildings poking up towards you, and every dip and hump on the ground is clearly defined, in depth. It’s the same principle as those silly three dimensional movies, where you have to wear special glasses to get the effect. Think about an everyday activity that demands depth perception, say, threading a needle. In this case, your eyes actually cross to focus on the same spot, the eye of the needle, and your brain determines, from the amount of crossing, exactly how far away the needle and thread are so you can bring them together. You use the same method to fool the eye in stereo photogrammetry, except that you are looking at two separate photographs and your eyes have to cross to fuse the images together. It’s as if the distance between the two points where the images were exposed was the actual separation between your two eyeballs, hundreds of feet, and if everything is aligned up properly the effect is striking, and as it turns out, extremely precise; accurate elevation measurements can be made from thousands of feet up in the air.

In practice, the process is a bit more involved. First, the film rolls are processed, and the individual negatives are printed onto glass plates painted with film emulsion. After they are developed and dried, the plates are clamped down onto holders mounted on a metal frame, each one being capable of being tipped, tilted, and swung independently. The two plate holders, or ‘buckets’, also slide back and forth so the distance between the two can be adjusted. With this arrangement you can exactly duplicate the geometrical relationship of the film planes in the original photography back at the lab, at a reduced scale. Two colored lamps, usually red and blue, shine through each plate and focus on a small stage free to roll around on a table positioned under the frame. Each lamp casts an image of its own plate on a small white surface mounted on the stage, so that you are looking at two images simultaneously, a red view of one plate and a blue view of the other. To see the effect, the operator must wear a pair of glasses with a blue lens over one eye and a red lens over the other. Each eye sees a different photo, the eyes cross to fuse the two together, and the operator gets the illusion of looking through a magic hole floating in the dark in front of him, with a perfect black and white picture of the ground, as seen from a great height in 3-D, visible through it. A tiny hole is drilled in the center of the stage and a speck of light is visible through it, by raising and lowering the stage (with a knob) the optical illusion is created that a little dot of light is being raised and lowered by the operator, and he can drop it precisely on the ground surface and read the elevation off a counter.

The long training period was necessary because not everyone can see in stereo, and even those who can have to train their eyes to do it with sufficient precision. The other problem is that the photographs are taken from a light plane bouncing around in strong crosswinds, and it takes a long and involved procedure of alternatively tipping, tilting and swinging the plate holders until the aircraft attitude error is removed from the photos and the two images fuse into a ‘model’. A model is that imaginary space formed by the overlap of the two plates after the individual photos are manipulated to remove the aircraft’s yaw, pitch, and roll. And it’s not over yet, once the model is established, the entire model itself has to be rocked back and forth and left and right (using big threaded screws that tip and tilt the entire frame the buckets ride on) until the model matches the elevation of the ground. The buckets can be slid towards or away from each other to alter the scale. Scale and elevation are determined by locating known points visible in the model of established position and height above sea level (from a ground survey conducted earlier). You’ve no doubt seen these ‘aerial targets’ while riding around, usually big “X’s” painted at road intersections. Surveyors set up their instruments on these landmarks and establish their elevation and position, and this information allows the photogrammetrist to calibrate his model. When the job is done, every point on the model can be located to a high positional accuracy and its elevation determined precisely. The stage also has a pen mounted directly under it as well, so the operator can trace buildings, roads, and contour lines. It takes a good operator about twenty minutes to go through this set-up procedure, and it takes him about a year to learn how. Depending on the nature of the terrain and the amount of detail to be mapped, an individual model can take from about half an hour to several days to complete. Most jobs consist of many models, and sometimes many flight lines, and all the edges have to match up. Invariably problems arise, targets are covered up or lost in shadow, or surveyors make mistakes, or the pilot can’t quite control his aircraft to specs, so there’s a lot of fudge work to be done as well. When you finally master the technique, then comes the art, seeing the ground level through the trees, learning to ignore false elevation cues caused by wind blowing vegetation around between photos (remember, the theory expects everything on the ground to not move between pictures) and reading featureless ground like snow or tall grass.

I have described only the simplest plotter I worked on, because it is the easiest to explain, but the more sophisticated models I used could be much more complex and accurate. The principle remains the same, and even the simplest plotters are capable of phenomenal accuracy (less than a foot of error in elevation from several thousand feet up). Photogrammetry is accurate enough to be used for engineering maps, stockpile inventories, and support of strip mining operations. One of our biggest clients were the phosphate mines, who were constantly trying to figure out how to move big piles of dirt around as cheaply as possible. I found the work fascinating in principle, even if a bit tedious in practice. I also got a chance to work with real craftsmen, from the old timer who taught me my craft to the wizards that worked in the photo lab, and the master photographer behind the camera. I gained a real measure of respect for the pilots as well, who have to fly over featureless terrain they can’t see, bucking crosswinds and turbulence while transferring the rigid geometry of the flight lines to the ground below. I got a chance to do all of these things, but map compilation was the only thing I was given the opportunity to get good at. Compilers are harder to get than pilots. I also was overwhelmed by the character of the men I worked with, not just their skill, but their integrity. These men would not tolerate sloppy work, and they would not do a half-assed job; they gave their all for the client. The management, on the other hand, was…well, typical management. Some things never change.

Check it over, hit enter, go to next parcel of land. Repeat for hours.

And it was far from the most tedious job I’ve held.

My first GIS job was also in data entry. Me and three other people created the computerized base map of Charlotte County, Florida, my first job after I left California.

All the Mylar plat sheets for the County were scanned, and the imagery then flashed up on screens (HP Unix workstations), where we used the GIS “heads up digitizing software” to enter and identify every single point, vector and polygon which had been originally created by draftsmen with pen and ink. It was highly accurate survey data, benchmarks, property boundaries and rights-of-way, so when it was all put together it made a consistent electronic cadastral (legal for property demarcation) map of the entire county. This base map then became the single unified coordinate system upon which any subsequent cartography or imagery in Charlotte was rubbersheeted over.

Going from hot-shot Silicon Valley remote sensing engineer to data entry clerk can do funny things to your head. I eventually left that job to become Broward County’s first GIS manager.

It took us about 3 years to digitize those Mylars. And I haven’t been the same since.

This was back before they called them workstations.

But the other techniques will vary depending on the characteristics of the sensor, the property being mapped, the equipment and software at your disposal, the use the map is going to be applied to, and the requirements of the user who will be using it.

Air photo was strictly analog, using mechanical-optical devices, stereo plotters, merging stereoscopic air photo pairs. The primary use was developing contours of terrain for civil engineering and mining support applications. You had to physically see the relief in the terrain and then draw contours, taking into account the ability of vegetation, ground texture, shadows, and other factors in obscuring the 3-d signal.

At 3000′ camera above ground you could detect terrain variations as small as 6″ under ideal conditions, and contours were typically a foot apart, much of which was a judgement call. This allowed road builders to figure drainage and strip miners to calculate overburden and inventory stockpiles. Map scale at that altitude was typically about 100′/inch or 1:1200.

Satellite imagery was more multispectral related. We worked with Landsat and Thematic Mapper imagery to try and deduce ground geology. Pixel sizes were on the order of tens of meters. This was highly automated, (70s and 80s minicomputer era, HP-3000 and VAX platforms) but our interest was not so much on the location of ground objects, but on their reflective properties across the spectrum, which gave us clues as to their mineralogy.

I also worked a lot with wellog data, information from sensor probes lowered into well holes which gave us various physical properties of the strata the probes penetrated. The data was x,y,z and some measurement taken at that location by the probe. This allowed us to generate 3-d maps of underground rock formations.

I also worked a lot on maintaining software for intelligence and military applications. I can’t tell you much about that, but one unclassified project I worked on was the imagery databases used to train pilots how to fly glide bombs and other remotely controlled ordnance.

Late in my career, GIS was mostly used to convert statistical ground truth into thematic maps to guide policy makers in making decisions. For example, given legal requirements as to how close a pervert could live to a school, what areas in the County were available for registered sex offenders to live in. I kid you not. I also worked with 911 emergency dispatch data, maintaing the electronic maps used to send police, fire, and ambulances when somebody called with an emergency. The GPS street data on your car navigation devices came from people like me, constantly updating municipal street and address data bases for 911 purposes.

And of course, I’ve worked as a map user, too, in maritime navigation, and one of my hobbies is astronomical mapping.

The interesting thing is that even though these applications were very differnt, using different technologies and yielding different products, everything I learned at one job cold be easily transferred to, and was useful, at the next. Prospective employers were delighted at my breadth of knowlege, specific techniques and technology could always be taught on the job. Lately, in order to get a job in mapping today, you have to demonstrate recent competence in a particular type of mapping software, not general cartographic principles. Past experience in related fields doesn’t mean squat any more. They just want someone who can work with ‘ArcView version 10.0′. Industry doesn’t need cartographers, it needs workstation jockeys.

I retired just in time.

Outfit was called the “Minnesota Land Management Information System.”

There were multiple databases for the same areas, with categories like transportation, vegetation, population, geology, and whatever else they could think of. Thank God I had nothing to do with figuring out the parameters. I just plugged numbers and letters into little squares on a CRT. As I recall, the unit of area we used was the section (1 square mile).

That kind of map is harder to interpret by non-specialist users (usually, managers, bureaucrats or politicians, who can barely read a map at all). In fact, that is exactly what the density-dot map I also included in my suite does, it gives you a visual cue that it directly correlated to the mapped quantity without breaking it up into distinct finite categories. For this type of data, that is the most precise way of characterizing the distribution. But to the untrained eye it looks messy and incomprehensible, sort of like an omelet. Unless you really know what you’re doing, you can’t see trends or hidden features easily.

Another technique that can be used is a contour map, which works great for terrain shaped by familiar natural forces, but fails perceptually for random, discrete data. But contouring requires a great deal of interpretation and interpolation, and an intimate knowledge of the process which creates the landform. Software can contour discrete data points, but it never looks as good or as realistic as that done by hand. I worked for five years drawing contours on air-photo relief maps. Software just can’t do what even the clumsiest human operator can.

Another disadvantage of the continuous gradation map is that the data is sampled, that is, we had a database telling how many people lived in each TAZ, but we didn’t know where each person lived in that TAZ, and there were too many to map them all. Even with density dots, you have to pick a value (I chose 200 people/dot) so that populous areas of the map didn’t merge into an undifferentiated mass of clumped dots, like snail eggs.

Another problem is that since the data is sampled, as you cross TAZ boundaries, the density may change drastically, but not necessarily at the boundary (which is why dots are placed at random within a TAZ). This will cause linear features like TAZ boundaries to appear in the gradated coloration you use. Seeing a clearly artificial straight line in a continuous natural distribution is very distracting and confusing. It looks like you are either making something up, or your data is discontinuous and erratic. (As all data is, at some level of resolution).

When making maps for policy makers and citizens, the cartographer isn’t just concerned with the highest possible accuracy and precision, but also with communicating the data in such a way that it displays what’s really going on on the ground.
The data analysis capability of the human visual system, and the sophistication of the user likely to be using that product must be considered, which means subjective decisions must be made by the geographer in order to create as truthful a representation of reality as possible. What you leave out is as critical as what you leave in. USGS Quad sheets and UK Ordnance Survey maps are excellent examples of first-rate cartography, but those are not thematic maps. They display physical features, not statistical quantities.

Software can give you tools to do this, but map making is an art. Experience, training, instinct and talent play an enormous role. And if you have an agenda, even an unrecognized one (most people are inherently honest) it can show up. For the professional propagandist, on the other hand, cherry picking, taking facts out of context, and all the other tricks of the paid liar are also available to the cartographer. There is no such thing as totally objective data. Interpretation, analysis, professional consensus, biases (both deliberate and hidden) and social and psychological considerations all play a role.

Kind of like those more nuanced “purple maps” you see for elections. At the very least, if the customer has to have numbers to work with (and they probably do), you could go with more gradations. Here’s a map by county with twenty colors. I haven’t looked at their odd numbers enough to tell what method they used for cutoffs.