The Nature of Urban Growth - Cache Valley

In the Beginning...

As we began this project our first question was, "What do we want to accomplish?" Since Travis lives in an area being rapidly encroached upon by huge new homes, the answer came easily. We decided to study trends in Cache Valley's urban expansion. We were already aware of TM data for the area from 1988, as well as SPOT imagery from 1994. In our nieve beginings we assumed this to be quite sufficeint and exhaustive. Our original plans turned out to be far too encompassing and comprehensive. Fortunately, we understood what we wanted to get in the end, which allowed us to side-step monstrous delays and problems that could have, and did, encompass entire projects. Several of these junctures will be noted throughout the report. Of all that we learned, we learned best just how much there is to learn.

The first step in our learning process was an introduction to our data sets. We found SPOT imagery titled 'south-cache' to represent the area of our interest. We had decided to focus on Logan and its' satellite communities because obvious growth had taken place there. Our Area of Interest (AOI) included Smithfield and Hyrum and was wide enough to cover all of Logan proper. The imagery from SPOT at 10 meter resolution was nearly perfect for our uses, being noticeably more detailed than the TM imagery. It's greatest drawback is that it is panchromatic, making classification almost meaningless. Our efforts only accomplished a rough density slice that defied formal classification, let alone meaningful interpretation. The difference in resolution and projection from the TM imagery also caused headaches. Had our intent been to provide a detailed, accurate, fully labeled classification of the area, our work would have been cut out and further results there would have been achieved. But, our goals lay more with intuitive deduction than hard, fast, numbers, so we moved on. The level of classification accuracy we had hoped for was not available in only one band.

We next focused our efforts to the Landsat imagery from 1988. Since this data was so easy to manipulate and derive information from, we proceeded with a series of unsupervised classifications. First we created 21 classes. It was informative but we wanted more which lead to our two further classifications, adding about ten classes each time. In the end we choose the image of 30 classes for evaluation. From this classified version we determined five major classes; water and shadow, healthy vegitation, dry vegitation, bare ground, and urban areas - including roads and buildings. Unfortunately, our level of accuracy was less than excellent, but then again, sufficient for our needs.

Blue indicates water and shadow, while black is supposably urban, green is healthy vegetaion, yellow is dry vegetation and white is bare ground.

Looking at our classified image closer, you can see that Smithfield and Hyrum have been eliminated as well as part of Hyde Park. After some road surveys, we determined that the most noticable growth had occured closer to Logan. Key landmarks are quickly identified, such as Logan's Sewage Ponds, the airport and highway 89/91 bending south-west towards Wellesville just outside of Logan. The shadow and dark vegetation of the east facing slope of the mountains was mistaken as water, shadow and urban areas by the unsupervised classification. Without implementing several vectors, including elevation data, this could not be eliminated, and without a comparison (TM) from another year, we decided it was accurate enough.

An Era of Side-Stepping...

We had hoped to create a psuedo color image from combining SPOT imagery with the corser Landsat data to yeild a 20 meter resolution, multi-band image. Rather than changing our project to accomplish the rectification of the SPOT image, which was already underway, we decided to again move on. Instead we took both images and cut out AOI's along the bench to highlight clusters of devlopment, again pre-determined by road surveys. Next we found the archaic roads data, ten or so years older than our oldest data set. We overlayed the roads vector with our TM imagery; (both our classified TM and our original TM). Again, not having a compatible SPOT image prevented overlaying roads to it. The process was not simple as the roads data, even after projection correction, was obviously askew. The error proved to be linear, which thankfully made its rectification easier. We solved for the error by adjusting the values of the 'x' and 'y' coordinates in the image. In essence, it was like adjusting a picture in the frame. By adding 30 to the 'y' coordinate (30 = 30m for the resolution of one pixel) the entire image raised one pixel under the roads vector. The process was quite simple but required several tries to get it right. (At least we think it's right) The roads were off by less than 5 pixels in both the 'x' and 'y' coordinates, meaning that the roads vector was no more than 150 meters wrong at any point,...That is except, of course, where the roads data was not even present. (See 10th West, the road leading straight to the intersection of the airport runways).

Ground Truth

With our images as compatable as was reasonably possible, it was time to 'hit the road' again. Ground truth proved informative as well as manotnous. Overall impressions were of rapid, new growth areas all along the bench in the form of large, high priced homes previously foriegn to the area. The densest developments were centered closest to Logan's east bench. The most expanding growth areas were located south of Logan in the Providence area and north in Hyde Park. Here the rapid growth is evident, with vast expanses of open land topagraphically suitable for future growth. At this time there are road networks being cultivated in all areas in anticipation of more growth. In Providence community's east bench, foothill development was still the standard. A large wealthy neighborhood just above the older growth area was evident. It spanned out, following Spring Creek into the boundry of Providence Canyon. An intersesting note in this area was the subdividing of older lots. New homes were located within the original communities between long term residences. Milleville showed less significant growth, reaffirming our ealier decision to bring in our AOI borders.

With the help of local maps and road surveyed ground truth, topographic features that may hamper future development became evident. Judging from the placement of some homes, such as in the east Logan area or Greystone, elevation, drainage and common sense aren't always a prerequisite or a limiting factor.

Boundries of ownership, city limits, and zoning. were not largely evaluated in our study as such boundries are subject to change with urban development and politics. Earthquake zones are a factor in the area but seem only to enhance the value of property and the desires of wealthy people to build there. These factors, being covered thoroughly by our collieges were also omitted from further research. Availability of water is one remaining controlling factor that may yet have a limiting effect on development, being a project in and of itself we bypass this issue also. Encroachment upon the ecosystem of the bench area is worth mention, not that development would be altered, rather effects on wildlife and associated problems to home owners. Animals previously using these lands for winter grazing will be shut out and viewed as nuceinces. Further, the threat of interface fire (wildland/urban mix) will be intensified as residence climb steeper upslope. Add to this the single entry neighborhoods cuasing additional dilemas in disaster management.

Harvesting Information from Data Sets...

Using information obtained by our road surveys, as mentioned earlier, we decided to select four particular samples or areas to concentrate our efforts. These were areas of obvious new growth. Some were nestled next to established residential areas while others stretched lengthwise into the hills of the bench. Their unique spectral signatures made them easy to locate on both SPOT and TM imagery, revealing the nature of their incoming existance. We will discuss each area in detail.

Hyde Park is our northern most site, and eventually chosen, as our northern border because of its proximity to Logan. Growth in the area may eventually envelope Hyde Park, melding it into Logan City proper, then only a street sign or a map would delineate the difference. We included Providence on the south side for the same reasons. The first picture is of our selected AOI in Hyde Park, which is on the south-east side of town. The new growth is stretching away from town into the bench. This area of development is known as Greystone. The inflated pricing of lots as well as zoning restrictions have made this area rather exclusionary.

Our first picture is taken from TM data, obtained in 1988. Though small, it is easy to see that growth is VERY limited beyond the developed subdivisions in the south-west corner. There is no visible growth on the entire eastern half of the image. The bright blue color may designate areas under current development or very recent growth, shown at the "x" and to the west and north-west. The bright square to the south-east is likely a recently plowed, fallow field. The sparce vegetation of the Cache bench gives the entire area a mild, light blue cast, as opposed to the irrigated lawns and fields to the west.

The next picture shown is of approximately the same area, taken by SPOT in 1994. The most noticeable differences are the clarity of roads, which are defined much better by the 10 meter resolution. Residential areas are seen usually as white, including roads, due to their high reflectance values. Agian, the south-west corner appears to be well established. What is made obvious to us now is the artery of growth spreading eastward onto the bench. Land plots, as surveyed from the road, are quite large and are not established into uniform blocks. This pattern helped us to determine expansive growth in all subsequent areas as it is different from the older, established areas. This change appears to be a local trend, leading away from the structured block-grid coordinate system established over a century ago. Another reason for the seemingly chaotic pattern of roads is due to elevation factors. The hilly benches do not easily allow for the conventional block coordinate system. Neither does the developers greedy checkbook. This aspect has been tolerated because people are placing more importance on owning property and wanting to keep their neighborhoods quiet and safe. Consequently, prices for lots in this area were found to be obscene!

Below is our 1988 TM image with the ancient roads vector overlayed. The roads are yellow because it appears to be a color not occuring naturally in the false-color satellite image. When comparing the TM w/roads against the 1994 SPOT image, (Thankfully the SPOT image has a high enough resolution to show roads quite clearly, eliminating our need to rectify it first :*) you can easily see where the most activity has taken place. The road leading into the Greystone estates appears to have been a winding access road to either dry farms or range land in 1988. In 1994 it has been straightened, brached and extended for miles. The relatively mild slope of the Hyde Park bench will allow continued growth as long as the economy permits.

Again, the color image is from 1988 TM data. The well established residential area is shown as mostly purple, with bright red designating Lundstrom Park in the north-west corner. The road dividing the park going north-south is difficult to make out, but is there. This single feature demonstrates the weakness of using 30 meter data for studies in urban development. Too much important data can be lost, especially when trying to classify with high degrees of accuracy. The areas of light blue near the "x" in the center and also to the north show beginning signs of develeopment. What is interesting is that the largest area of apparant development in the TM image is absent in the SPOT image. Our speculation is that devlopment here was not for residential purposes, but rather possibly having to do with the canal that skirts the bench or maybe even preliminary development that has yet to see fruition. Construction, in such cases, would allow the rebound of vegetation in a few short years, if left unmolested afterwards.

Logan TM 1988...

Logan SPOT 1994...

Logan TM w/ Roads Overlay...

The greater resolution of the SPOT image makes it difficult to reconcile it with the TM image at first. The park is again clearly defined, seen here as black. The road dividing Lundstrom Park into East and West sections is plainly visible. To the south-east of the park three succeeding areas of growth stand out. There is no sign of them on the TM image. Due to elevation constraints further growth will likely extend either north-east or south, not due east. Development there has already exceed common sense several times. Again, notice the limited access to this area, making its property value rise.

We seemed to find a pattern about this time in our obsevations. The TM image was relatively easy to manipulate and classify, but proved difficult to glean information from in non-statistical ways. The road vector was invaluable in helping us reconcile the TM image to the SPOT. We ended up studying the SPOT image first, searching for key landmarks and disturbed ground, then compared it against the TM image with roads and finally against the normal TM image. At this stage our classified image was almost completely abandoned.

This area has probably seen the most eye-catching growth of any that we've studied. An obvious choice due to the visibility of the growth that has occured. Again, the TM imagery shows the first signs of construction, but fails to indicate accurately where the growth will occur over the next six years. It is interesting to note that some of the areas of light blue indicating bare ground, meaning likely growth, are also still evident in the '94 SPOT image. The larger area is the main road providing access to cliffside from the island area. It is possible that efforts to revegitate the area were not implemented or else construction there is continuing. We suspect that problems with erosion may develop if vegetation does not recover the hillsides there, making life on the island a dirty business. The smaller area in the north-east corner is the beginnings of an elbow to elbow neighborhood of mansions and even a castle. Comparing this to the SPOT image shows its progression, molding to the hillside along the bench.

Looking at the SPOT image now and focusing in the north-east corner, we see a comma shapped white area. This new residential area is wrapping around the mountain and will not likely experience much further growth due to elevation constraints. Directly south of there is a different story. A network of roads has begun to appear, hinting at plans for continued future development. When comparing this to the TM image with the roads overlay, the expansion takes on its full dimenion. Even notwithstanding the raod data being old, the number of unaccounted sudivisions and roadways is astounding. Because of the limited access to this area, I would suspect that traffic congestion will become a problem on the road leading down into the island in the next few years. Hopefully city planners are already aware of the impending problem.

Of the areas we chose to focus on, Providence holds the most growth. Unlike the growth in Hyde Park that was thin but expansive, Providence has seen the development of entire neighborhoods and subdivisions. Also unlike Hyde Park, the growth was already underway in 1988, as is evedenced by large areas of light blue in the TM data.

By making a close comparison to the newer SPOT data, the progression of growth becomes obvious. The areas of concentrated growth in the TM image have settled down somewhat in the SPOT image while new areas almost fill the north-east corner. In the south-center portion of the Spot image there appears to be continued building. Lots that were empty and untouched in 1988 are now getting homes while others, built in the last 5 years, are becoming established; ie lawns and trees. Vegetation is the greatest key in not only determining present growth but also in deciphering the relative age of recent growth. Expensive areas may defy this to some extend with instant turf, but large trees are still missing. In another six years we would expect this area to be relatively quiet as urban expansion and growth leave it behind. By then it may be difficult to tell it from areas established decades ago.

Summaray and Notes - Future Reference

Our study proved to be anything but comprehensive or exhausting of data, but we feel like much was gained. Given time to extend our study, we would focus into several new areas. They are...

Additional Spectral Research - including manipulation of various layers. Different band combinations may yield better results, making classification easier. Also it may highlight key features so as to be more noticable for interpretaion puposes.

Additional Vector Overlays - including vector sets of elevation and soil type. These would be partiularly helpfull in determining geographic limits to expansion as well as the agricultural value of the land being lost. Various water vectors would also be quite informitive.

Higher Resolution Data Sets - We found the 30 meter TM data to be inadequate for detailed research applications of urban growth. Most two lane roads aren't 30 meters wide, and if split between pixels they can be lost completely. That says nothing of smaller access roads etc.
We also felt the need for multiple bands in our 10 meter Spot data, which is of course panchromatic. Classification at 10 meters would yield tremendous results, but is as yet beyond our reach.

Hyper-Supervised Classification - Classification of urban growth and development defies both supervised and unsevervised classification methods. Urban areas can incorporate healthy vegetation, dry vegetaion, shadow as water as well as roads and building signatures. Without labeling entire polygons, the statistics only reveal density, not true growth. This is particularly important when considering ecological factors.

Increased Urban Density - While expansion is happening throughout the Logan bench area, housing density has also incresed inside what is considered developed areas. Again, with higher resolution imagery this aspect of growth would be very revealing to trends within the city. This growth is also going to be a function of money...people in the the bench suburbs have it and the city residents do not.

Conclusion -

Knowing that growth has been occuring at a fevered pace, we focused on the aspects of that growth. Our findings are primarily a result of intuitive deduction coupled with field work to verify our accuracy. We also speculated as to future concerns, even makin predictions at times, but only briefly.

Growth has occured in three key ways, two of which we addressed in this study. The third was increased urban density, a factor that we feel requires higher resolution imagery than we currently have. The first aspect of growth was displayed in the Greystone area, where development is taking the fast road out of town. New subdivisions are spacious and distanced from the older, denser areas. The second type was seen best in either the Logan or Cliffside images, where development has used land efficiently and is in close proximity to the existing development. The Providence area appears to be a compromise between the two.

The growth also took on a new look, as developers have made an effort to move away from the locally traditional block system. Instead we see more Caul de Sac's and looping roads. Having a sigle road access into an entire neighborhood seems to be a trademark of modern growth. Reasons for this are numerous, but headlined by the elimination of through traffic.

Continued growth is also projected in all four areas, with the Logan area having the most natural (geographic) restrictions. We should mention other restricitng factors, such as availability of water, eathquake risk and likewise, but no one else seems to be paying any attention to them. Greystone in particular has plenty of room to accomodate wholesale expansion, but is nestle between a spliting fault and has no natural water supply in close vicinity. With people living further from Logan City, commerce is also sure to follow. Larger stores and improved city office buildings will likely be a trend in the near future. These towns will also start creating a sizable tax base, helping them to absolve any dependancies and financial blockades they are facing now.

We fully expect to see Cache Valley begin to mirror the areas surrounding Salt Lake City.