Maps & Geospatial

The U.S. Department of Agriculture Aerial Photography Field Office (USDA APFO), is the primary physical repository for color-infrared (natural color and black and white), photography  captured from 1955 until recently when digital capture became the standard. Many of these photos were acquired in support of Farm Services Agency (FSA) and one of its predecessors the Agricultural Stabilization and Conservation Service (ASCS) programs in the post Depression and Dust Bowl era. That is to say that they were not originally collected for public access and consumption, although they undoubtedly have high public interest. Many APFO photographs have been digitized to some minimum resolution and public access provided through agreements with other federal agencies like USGS through the EarthExplorer web interface, as well by through APFO directly. Some institutions (like state government and academic libraries), that hold historic aerial photograph prints have also undertaken large digitization campaigns and posted their results online or in some cases kept them behind secure firewalls for members of their institution only.

For photographs that are already available online, digitization and posting largely took place some years ago when computer storage and internet access speeds made working with the large file sizes of digital images challenging. Furthermore, this was also a time-period when scanner technology was still in its infancy and scanner properties like the ability to preserve the geometric accuracy of the original film or image was not very good. Many projects digitized old photograph prints rather than film and in these cases there’s also the issue of printed material not being very stable — it often expands and contracts over time, sometimes warping in ways that can compromise the integrity of the geospatial information captured on the original film. Film isn’t perfect either, but at least it is the original picture without the compounding negative effects of printing, storing printed material, and then imperfect scanning of prints.

Many APFO photographs that are currently available online (e.g. from USGS EarthExplorer or Penn Pilot), were digitized at 600dpi or less and were digitized with scanners of unknown, or common desktop scanner properties. The image resolution and scan quality of these collections is fine for general inquiry, but the original film contains much more geospatial information than can be deciphered from a 600dpi low-quality scan.

Fast-forward to the present day and the 30+ year old film is starting to deteriorate (it can start to liquefy and smell like vinegar!), and APFO is in the process of moving their headquarter offices. For these (and other?) reasons they are in the midst of a massive digitization campaign with preservation in mind; meaning that they are using the highest quality photogrammetric scanners to digitally capture the original film at over 2,000dpi. To my knowledge there are no plans to post these images online, but they are free to the public if you request them from APFO and pay for the storage (i.e. external hard drives), a nominal fee for their services to load the drives, and shipping to get them to you. That said, they are far from free to produce as the scanners being used often cost more than $60,000 not to mention other supporting computer equipment and staff time. Compared to licensed geospatial data, they’re relatively affordable, but then again, in their native form they’re not really geospatial data at all because they contain distortions and lack geographic reference.

Take a look at the difference between a 600dpi image downloaded from USGS Earth Explorer and the image that I recently acquired directly from APFO. This is the neighborhood where I live in State College, PA and I’m not sure that I can even identify my house on the 600dpi image.

Fig 1. 600dpi image downloaded from USGS EarthExplorer

Fig 2. 2,032 dpi image acquired directly from APFO captured from the original film on a photogrammetric scanner

I can clearly identify my house in the second image above and we can zoom in even more to barely identify individual cars in the figure 3 image. The spatial resolution of this imagery given the original camera used, altitude of photo capture, and digitization specifications is about 0.7-meters, or about 2-feet. That means that each pixel in these images depicts an area on the ground of about 2-feet by 2-feet. Cars are then represented by about 2 pixels wide by about 3 or 4 pixels long. Any object smaller than about 2-4-feet will not be decipherable from this imagery. That said, for going back in time 30-years, this is pretty good! Neighborhood scale to car scale, that’s the difference between a 600dpi scan and a 2,032dpi scan of aerial photography film.

Fig 3. 2,032 dpi image acquired directly from APFO captured from the original film on a photogrammetric scanner

We have acquired the 2,032dpi imagery for all of Pennsylvania — in total they number 2,607 images. The capture dates range from March 27, 1980 to May 13, 1987 with an emphasis on capturing “leaf-on” or vegetative growing season conditions because again, these pictures were largely taken to support Farm Services Agency operations having to do with agricultural production. The complete metadata for the imagery that we acquired is available here and the imagery collection can be searched and downloaded directly from the PASDA Imagery Navigator. To search the Imagery Navigator, zoom and pan to your area of interest and then right-click on the map to see what imagery is available for that place and date. A pop-up window with years will appear and the 1980s color-infrared imagery that is the subject of this post is available for download from the “1980-1987” tab. You may get multiple image tile results for your area of interest because these photos were intentionally overlapped when captured (more on this important property of these photos in a latter post).

To see the approximate footprint coverage area for the 1980s photography in the Imagery Navigator use the “Display Tile Index” drop-down menu and select “Infrared 80-87 (approximate)”. These are approximate coverage areas for each photo because they are not georeferenced. There is no way to search photography coverage for lines or polygons. Pay attention to the footprint for the tile that you download to get an idea of where to start looking for your area of interest after you download the image — this is often a challenge because of how the photos are oriented and because they lack any place names or other labels to help you — they’re just a photograph.

Once downloaded, you’ll have to unzip the file and open it in any image viewing and/or editing software. When in doubt, try double-clicking on the file and then wait patiently for a bit — these are very large files that may bring you’re image viewing editing software and/or computer to a crawl. I’ve had good luck using both GIMP2 and IrfanView softwares, both of which are free and can be found easily with a google search.

If you know the photo tiles that you need, you can bulk download files from the PASDA ftp site.

For assistance searching, accessing, and using these photos, please visit the Donald. W. Hamer Center for Maps and Geospatial Information, or contact Nathan.

We are also in the process of developing efficient workflows to georeference these images and remove some distortions so that they can be incorporated directly into a desktop or web-based geographic information system (GIS) — or so that you can view the imagery as a background image in the PASDA Imagery Navigator.

Next post I’ll detail another new acquisition of historic aerial photography from the APFO —  this time from the 1950s…

Web-mapping tools are great! They have revolutionized the world of maps. From Google and Apple maps to ArcGIS Online to countless other government agency, institutional, and homegrown web-mapping applications that put a startling amount of data and information at your fingertips, web-mapping has changed the world.

If you live in the U.S. (and many other places), you can now easily zoom in to your house and see whether your car, (or your neighbor’s car), was in the driveway when the plane or satellite flew stealthily overhead and took the picture that you (and everyone else), can now see on your computer screen. You can probably also see what date and day of the week the image was captured and possibly even the time?! But most people don’t pay attention to this “metadata” so we mostly avoid any conflicts that this kind of information could create.

You can transport yourself halfway around the world and get as good a “feel” for a place (short of employing virtual-reality), with maps, images and street- and birdseye-views as you could ever expect to without actually visiting the place in person. You can even peruse cartographic depictions of the world as it once was through large repositories of digitized historic maps like the David Rumsey collection that have been posted online and georeferenced (i.e. put in web-map space), by you “the crowd”. With a web-mapping tool at your disposal, connected to a large database of geospatial data the possibilities can seem boundless. Or are they?

I can distinctly remember when GoogleEarth (figure 1) launched the historical imagery time-slider in 2009. They used Las Vegas, NV (or at least my first exposure to it was inspecting the Las Vegas, NV area), highlighting the new functionality of their tool with aerial photography dating back to 1950. It was so revealing to be able to slide the bar back and forth and examine the amazing urban growth that has taken place over that time-period in this area — it was like being able to instantly go back in time 60 years with the detail of very tactile things like individual homes and streets that are often blurred over in most satellite imagery captured prior to about 2000.

I thought this kind of functionality would take off, but it hasn’t?

Here is a screenshot from GoogleEarth Pro (taken March 22, 2018), for my hometown of State College, PA with the time-slider enabled that shows for what time-periods there is imagery available from within this web-mapping tool. In case it’s hard to see in the figure below, it shows 7 imagery dates available going back to 1994, and then nothing…crickets.

Figure 1. A screen capture of Google Earth Pro with time-slider enabled

Maybe I am showing my age, but 1994 doesn’t seem like ancient history? The GoogleEarth Engine Timelapse tool pushes the time-period of imagery at the touch of a button back to the mid-1980s and impressively does so globally, but at a moderate spatial resolution from satellite imagery so that zooming in on things with a personal touch like individual homes is out of reach.

How is it that I can overlay contemporary geospatial data on top of a geo-enabled map of the world from 1812 using the ArcGIS Online Living Atlas collection for example, but I can’t look at imagery of my hometown from the 1960s? What if you study spatial patterns of post-WWII suburban development in the U.S., or merely want to see how big a garden the prior owners of your home maintained in their backyard before your family bought the place in the 1980s? Have you ever noticed how web-mapping tools for most places go silent for time-periods prior to the mid-1990s?

The answer lies mostly in a formatting problem. Like most modern geo-technology applications web-mapping tools are data hungry and they want the data in a specific format and structure. In the U.S., we have a pretty lengthy history dating back to at least the 1930s of well-organized and comprehensive aerial photography missions with multiple revisits for most places. The goal of these missions was to capture high-quality and accurate geospatial data and information (it is often called “vertical photography”), so the photogrammetrists in charge did very sciency things like carefully calibrated their cameras in the lab before going out in planes, did their best to fly at consistent altitudes, and captured multiple overlapping images of the same place so that they could produce images (i.e. maps), of places with as little geographic distortion as was possible. The “data” was captured on film and recently more and more of this film is being digitized and made publically available online. So why aren’t they viewable in web-mapping applications?

The formatting problem isn’t one of analog versus digital, but rather is one that requires 2 steps to correct: 1) Remove the geographic distortions inherent in the native film that arise when you take take a camera up into a cold plane and take pictures of bumpy objects from different angles that are not directly overhead; and 2) Tie those images very accurately to a geographic frame of reference that can talk to web-mapping applications so that things line up the way they should. The methods for resolving both of these issues are well-known, but are labor intensive and you really need to do both steps in order to get very good results. There is also the issue that aerial photographs individually cover small areas on the ground and when digitized to a high-quality (i.e. an image with lots of pixels to show detail), produce very large file sizes — but then again it is 2018 and computers are fast and storage cheap so I really don’t consider these to be insurmountable hurdles.

Consider the following great example (from this source, by Patrick Cines), of a change over time gif animation of Penn State’s campus that manually and accurately overlaid each historic image on top of the modern distortion corrected images. Watch the animation a few times carefully — can you see when the images transition from being kind of jumpy to being much more consistent? By jumpy I mean that distances and directions from one landmark to the next don’t seem quite consistent. It’s a little hard to separate the differences caused by actual landscape change versus imperfections in the imagery, but I hope you see what I do, which is that the distortions in the first 3 images are a major distraction from trying to qualitatively assess actual landscape change over this time-period. In spite of these imperfections, there was a lot of public interest in this article evidenced by it being “liked” on FaceBook over 100,000 times in the first 12 hours after it was posted. I’m not very savvy with FaceBook, but that seems like a lot of interest compared to very humble public interest in most of what I do day-to-day?

Like other examples discussed in this post, the imagery transitions from being in the format of “dumb” digital pictures to being “smart” geospatial data in the image labeled 1992, (even though I think this is mislabeled and should be 1994).

The “data” (i.e. high-quality vertical photography that has largely been digitized), is available for much of the U.S. (and a lot of the world I think?), and for many time-periods prior to the mid-1990s it’s just that the work hasn’t been done to reformat it for the greater good and unfortunately, the people that do know how to do this don’t seem that interested? My first opportunity to learn this lesson happened when I was working as a Postdoctoral Research Associate at Montana State University studying the ecological impacts of climate change (past, present, and likely future), in the Northern Rockies and Greater Yellowstone Ecosystem. One of the ways that climate change was thought to have already impacted the Yellowstone area over the last 50-80 years was by drying of subalpine meadows, which allowed conifer trees to encroach on the meadow grasses and forbs and generally shrink meadow area. This was a perfect project for geographic information systems (GIS) (of which web-mapping tools are a subset), all we had to do was find some old aerial photographs on which we could “see” the subalpine meadows and measure their areas through time. To my surprise at the time, I found a wealth of really fantastic old aerial photographs that had been captured by all kinds of different agencies for all kinds of reasons and we were off and running. With the photos in hand we then had to undertake the 2-steps described above to incorporate them into our GIS for analysis and I already knew how to do the first part so things were still looking promising. The problem was that when I asked people who knew how to do the second part they either told me that it couldn’t be done or was a waste of time. Without the ability to accurately measure things like subalpine meadow area we were a bit stuck. Combine this with the experience of a friend and colleague at the time who had just had a manuscript rejected by the journal Ecological Applications for not having addressed the 2nd step in a similar study and we were unsure how to proceed? Research funding cycles are short and we were forced to move on, not having anticipated this methodological hurdle and lacking the resources to address it. In retrospect this was a missed opportunity, but it did pique my interest in these old aerial photographs and I love it when people tell that something can’t be done — to me that rings bells as an opportunity.

So, with the interests of researchers and the interested public in mind, we are taking up the challenge of turning historic aerial photographs into geospatial data for use in web-mapping tools and GIS. We’re doing this for the state of Pennsylvania by doing 2 things 1) Acquiring high-quality scans of historic aerial photography as they become available from the agencies that curate them, and; 2) We’re developing methods to efficiently georeferenced and remove the distortions from the photographs, capitalizing on every new computer shortcut and automation trick available.

I have two primary hopes for this work. First, I hope that others also take up the challenge for surrounding states and elsewhere. Imagine if someone did the same for New York, and someone else for New Jersey, and another for Maryland? Do you want to know how land use has changed in the Chesapeake Bay Watershed over the last ~70 years? Maybe it won’t be “done” after this kind of effort, but we will be a whole lot closer to being a research question that can be answered within a single funding cycle. The State of Delaware’s Environmental Monitoring and Analysis Center has done some interesting work here. Second, I hope that the availability of these data stimulate and inspire new research questions and new students of geography and geographic data. I don’t pretend to understand all the different ways that these data could be used, but I do hope that people who are smarter and more creative than I am will come up with some great ideas.

Up next are a series of posts about some of the imagery that we are acquiring, what we plan to do with it, and early results of our efforts…