Robert Morris and Jason R. Stowe
Wildfires are dangerous and very unpredictable. "Wildfires [come] from natural and man-made causes and [have little or no] effective human control. During 1990 ...an area larger than the state of Massachusetts [was burned]. ...Effective means of managing fire risk, therefore, are likely to yield significant benifits to society" (Clark, Brass, & Riggan, 1994 p. 1355). Fire has always been a force in the in the sphere of human and other world species development. Lightning strikes a tree and the wood, tinder dry, explodes into a roman candle of rapid chemical change. The landscape bursts into a conflagration that none can withstand. The firestorm sweeps the landscape clean, purging the earth, washing it with waves of self generated fury. Puny man can but watch in despair, and stayout of the way (Jones, & Johnston, 1968, p. 115).
John Varley, the cheif of research for Yellowstone National Park said, "Fire to human beings has been a friend or a foe as long as there have been human beings. The one day it's your friend, and the next day it's your enemy .... But to an ecologist what happened is neither good nor bad. It's just the natural progression of things" (cited in Jeffery, 1989, p. 271).
Today people pay little or no attention to the roll nature plays in their lives. The press of humanity against the boundaries of nature has been growing in speed as well as scope. At the turn of the century the United States was in a massive phase of in-migration and expantion. The new world held the promise of limitless resources and boundless frontiers. The national parks and forest system was inaugurated in the United States as people came to realize that the beauty and resources of our nation were being lost by rapid exploitation and inadequate management. Nature gaves way to the pressure of humanity in our as well.
The growth in townships experienced recently in the Cache Valley has its impact on the natural setting. Townships ar steadily encroaching on federal land and wilderness areas. A contempory revival of Joni Mitchell's song "Big Yellow Taxi" (recently sung by Amy Grant), cautions developers with a reminder of what can be lost. "Don't it always seem to go, that you don't know what you've got 'till it's gone. Pave paradise, (with local appropriateness) to put in a parking lot." Building and expantion of small rural communities has far exceeded past growth and has it's management difficulties.
A natural setting has varied aspects that need to be managed. The encroachment of developement makes management even more difficult. One of these aspects is the developement of policy related to fire prevention for large tracts of government and private lands.
The establishment of fire policy in our national forests from 1920 has not been consistant with nature. In the past, before human intervention, fire was a natural element in the established cycles of forest ecology. The forest became embroiled in economic concerns as the United States expanded. Wood needed to be protected from destruction for the sake of development. The policy therefore, has been to fight all fires to protect lumber assests. In 1972 the park service changed this policy to allow natural fires to burn when no threat is exercised against the human factor (Miller, 1990, p. 401).
The issues of management of range and forest fires are complex. Computers have been used to help managers understand complex issues. In prophetic words, National Geographic staff writers Jones & Johnston said, "One day soon computers will join the battle [against fire]. Fed data on fuel characteristics, winds, humidity, and the like, these machines will tell firefighting commanders when and where to deploy forces" (Jones, & Johnston, 1968, p. 106).
It is very difficult to predict when a wildfire will occur, however it is quite possible to predict where that wildfire has the greatest possibility of occurring. That will be the scope of this project. Our study area will be the Logan Canyon to Blacksmith Fork area (Figure #1). We will attempt to predict possible high, medium and low risk areas within these boundaries.
Figure #1
The first step in beginning our analysis was first to examine different methods used for fire risk predictability. We also had to determine which environmental factors would be the most important in predicting potential fire areas. The factors that we will consider in this report are the following: proximity of roads, proximity of waterways, elevation (slope), and the wetness/dryness of vegetation. Using these factors with the aid of Imagine (a computer-based remote sensing and GIS tool) we will attempt to predict where areas of high, medium, and low fire risk occur. We will then compare our results with the computer generated model that we will create. So, let's begin.
For our first step in the analysis phase we acquired data pertaining to the roadways and waterways of our selected area. This data was obtained from Dr. Douglas Ramsey through a database which he provided to all of the students. At the same time, we obtained our elevation data and the TM images which we would later use to create a wetness/dryness model.
Our first encounter with the Imagine software was a difficult. We attempted to create a 100 meter buffer for the roadways data in Imagine, but we soon learned that we had to build the coverage in ArcInfo (a GIS software package) and then import that coverage into Imagine. Figure #2 shows the roadways with a 100 meter buffer. This vector is overlayed on the TM spectral image of our analysis area.
Figure #2
The next step in our project was to produce a buffered layer of the waterways. This was somewhat easier as we had already completed the necessary steps for the roadway buffer. The waterways were buffered to a distance of 50 meters on either side (Figure #3).
Figure #3
Our reasoning for selecting 100 meters for the roadway buffer hinged on the precept that firecrews would not be able to reach areas farther than 100 meters except on foot. Fire suppression vehicles would not be able to advance further than this point because of slope and possible dense vegetation. For the waterways, we decided that 50 meters would be the area most protected by the waterways. Any further effect from a nearby waterway would be insignificant at a distance further than 50 meters. Figure #4 shows both the buffered roadways and the buffered waterways overlayed upon the TM image. This new image would later be used to help us make our own predictions of potential risk areas. Now, onto the elevation model.
Figure #4
The elevation model was somewhat difficult to create. We ran into problems getting the coverage to maintain its values so that we could make distinctions between levels of elevation. Most of the transformations had to be done in ArcInfo and then imported into Imagine for later use with the two buffered layers. After three or four consultations with Dr. Ramsey we were able to create a usable elevation model. This was then colored for maximum effect and the result became Figure #5. The elevation model was clipped and used as the base map for referencing the layered models for the final conparison with the other maps. The use of the dma elevation model for the entire cache valley was also used to drap our study area for graphic reference in our presentation and for giving the aspect perspective for high south and low north facing slopes.
Figure #5
Figure #6 shows the two buffered layers (roads and waterways) layered with the completed elevation data. We are now one step closer to having all the information that we need to begin making our predictions.
Figure #6
Building the vegetation wetness/dryness model was next on the list. This began by cutting or clipping the appropriate area out of the entire cache county TM image set. After this was accomplished we turned to the Image Interpreter function in Imagine to do the actual classification of the image. Under the Image Interpreter function we went to Spectral Enhancement, Tasseled Cap. These functions would classify the TM image into six different layers.
The first layer (Band 1) would give us brightness values for our area. The second layer (Band 2) would give us greenness values for the area and the third layer (Band 3), the one we are really interested in, would give us wetness/dryness values. The remaining three layers were of no importance to us, so we did nothing further with them.
The classified image now had six bands, but we will only be dealing with the third layer. In this layer pixel values are represented in a gray scale where the brightness of the gray corresponds to the wetness of the vegetation. For example, if the color is almost white, it signifies that at that particular spot, or pixel, the vegetation is very wet. If the color is dark gray or almost black then that would represent a dry spot.
At this point we wanted to recode the pixel values of layer three so that we only had three values. One for wet, one for medium wetness, and one for dry vegetation. Figure #7 is our vegetation model. The wetness/dryness of our study area is depicted by the colors in the legend.
Figure #7
DRY
MEDIUM
WET
Now we are ready to begin to make our own predictions of High, Medium, and Low risk areas. To do this we used Imagine. First we brought up our vegetation wetness/dryness model. On top of that we layered the buffered roads and waterways. We were unable to overlay the elevation data with the others because the elevation data was in raster form. We made several attempts at converting the elevation raster coverage to a vector or grid coverage but in doing so we could never retain the actual data that went with the coverage. All we got were a bunch of lines, contour lines actually, and there were so many of them that they almost completely blacked out the screen. Needless to say we decided to leave the elevation data out of our predictions. However, we could still use it when we asked Imagine to generate it's own prediction model.
To make our prediction layers we used the AOI Tools command and drew onto our vegetation model (which is layered with the buffered roads and waterways) a high risk layer and then a low risk layer. We did not create a medium risk layer because we decided that everything that is outside the high and low risk layers is medium risk area. Figure #8 shows the high and low risk layers that we produced. These two layers are of our own design without any help from the computer except for displaying the proper raster and vector files in the viewer.
Figure #8
HIGH RISK LAYER LOW RISK LAYER
Any area that is not contained within the HIGH or LOW risk layers is considered MEDIUM risk.
Now we have come to the moment of truth. We want to generate a computer model by inputting all of the data (road and waterway buffers, elevation data, and vegetation data) into Imagine and then Imagine would give us it's prediction for high, medium and low risk areas.
It turns out that this was the hardest part of the project to get to work. The first thing that needed to be done was to decide which factors were the most important and then create a heirarchy with the remaining factors. We decided that the proximity of roadways was the most important factor. Next in importance was the waterway buffer, then the elevation data, and finally the vegetation model. We needed this ranking of the factors so that we would obtain the proper results.
Our first attempt involved using the Modeler in Imagine. Using this command we would be able to input all of the data from our four coverages, splice them all together and the resulting image would be one that gave us the areas of intersections of the four coverages. These areas would be the low risk areas, because there was water, roads, low slope and wet vegetation. The areas that would be the high risk areas are those that have nothing in common with the other coverages.
The first problem was that a couple of our coverages did not have matching elements. Sometimes the projection types were mismatched and sometimes the units of the coverage were wrong. We were able to rectify these problems. Nevertheless when we attempted to run the model, it crashed. At this point we turned to Dr. Ramsey for assistance and we were still unable to get the Modeler command to function properly. However Dr. Ramsey did give us another option.
Our new plan of action involved taking all four images, recoding them and then indexing them with the Image Interpreter Index command. The basic process of this operation was that each image was recoded to two classes. One class signified a high risk situation while the other class signified a low risk situation. As the first two images are indexed together there will become a total of three classes: one where neither class is present, one where only one class is evident and a third case where both classes are evident. As each of the remaining images were indexed to the previously created index file, another class is created. By using this process five classes were built by Imagine. Again we ran into mismatches in the image information but we were able to find the problems and fix them. After making four or five attempts we were able to come up with the computer's generated model. Figure #9 shows the fire prediction model that was generated by Imagine.
Figure #9
LOW
MODERATE
MEDIUM
MED-HIGH
HIGH
After having designed our own predicted areas of fire risk for the selected area and having also built the Imagine generated model, we wanted to compare the two results. Figure #10 shows all three of the risk models that we built. As one can see, we were able to come fairly close to the areas of fire risk that Imagine predicted.
Figure #10
HIGH RISK LOW RISK COMPUTER GENERATION
The following images are a result of draping the dma elevation model with the computer's prediction model of fire risk. The perspective has been altered four times to dramatize the effect aspect has on north and south facing slopes. Perspective images
Dramatizes the increased fire risk on south facing slopes.
North facing slope
South facing slope
South facing slope
North facing slope
Throughout the course of this project we (as a team) were able to learn the workings of Imagine and ArcInfo. We ran into many problems that at first we needed help to overcome. As time went on and new problems arose we began to have an idea of where to look to solve our problems. We also became more involved and more interested in the work that we were doing. As we became more comfortable with our computer skills we began to enjoy ourselves and we began to really learn. For our team this project was a great learning experience, and after a while it became a lot of fun.
We believe that we were successful in completing our desired goal of predicting potential fire risk areas in the Logan Canyon and Blacksmith Fork mountainous region. Over the course of this quarter we have worked hard, accomplished more than we believed possible, and above all, we've had FUN learning.