Assessment of Erosion Potential from the Ponil Fire, Philmont Ranch, New Mexico

Nick Sommer

Dec. 6, 2002

GPS/GIS, GEO 327G, Dr. Mark Helper

University of Texas at Austin

Problem Introduction

In June of 2002, a lightning storm sparked four separate fires in the northeastern New Mexico backcountry. These fires were located on or near the Philmont boy scout ranch, northwest of the city of Cimmaron. The fires spread until the original four blazes united to form one large fire. Erosion control is necessary not only to retain soil for re-growth of vegetation but also to keep soot and ash out of the nearby Ponil River, which serves as the water source for the small city of Springer, NM. Is it possible to predict which areas in the fire zone would be most prone to become severely altered by erosion due to destruction of natural vegetation?

Infrared image one day after the fire began

In this project I hope to identify the burn area on a map and asses the erosion potential on and near the Philmont ranch for this newly burned landscape by creating an erosion potential map as my end-product. I believe this goal is achievable and worthwhile results may be created by following this outline.

Project Outline

I) Collect representative datasets necessary

a) Elevation, geology, soils, roads, cities, fire map, ranch map

II) Preprocess data

a) Unzip compressed files and organize data folders

III) ArcGIS processing of data

a) Add data to workspace

b) Use ArcCatalog to create three new polygon shapefiles for the project extent, fire extent, and ranch property line

c) Create project extent and clip data to project extent

d) Merge separate blocks of the same data together

e) Edit fire and ranch polygons using the editor toolbar

f) Use ArcToolbox to project data to UTM coordinate system, zone 13N

g) Using Spatial Analyst

1) Create a slope profile of elevation data

2) Convert vector data (geology, fire extent) to raster data

3) Use raster calculator to evaluate slope, soils, and geology with an appropriately weighted formula to produce the final erosion potential map.

IV) Conclusions

a) Areas of high potential – are they predictable?

b) Possibilities of improving accuracy of results

Data Collection, Project Assumptions, and Data Preprocessing

Data Collection

To perform this analysis, I must create an erosion potential map as an end product. So, erosional factors will control the datasets necessary for this project. Upon contemplating the factors involved with erosion, I came up with the following list:

1) Steepness of terrain

2) Geology of the area

3) Soil cover of the area

4) Burn intensity

5) Precipitation

6) Vegetation

7) City/Town locations - Census 2000 data

8) Transportation - infrastructure

Each of these factors must have a representative dataset to become integrated into a GIS project. It was necessary for me to utilize a few key resources to accumulate my project data. All of this data I acquired free of cost:

1) Steepness of terrain

I choose to use NED (National Elevation Data, 28.5 meter resolution) data which I collected from the USGS data extraction page in ArcGrid format with the generous help of my T.A. Tim Pierce. This is a seamless dataset which covers the entire United States; simply draw a box around the area of interest on the data delivery page and follow simple download instructions. My working area was covered in two rectangles of data, so I later clipped and merge these two blocks together. The data is in the NAD83 geographic coordinate system, with x and y coordinates in decimal degrees. Source: (http://edcnts12.cr.usgs.gov/ned/default.asp for main page, http://seamless.usgs.gov/viewer.htm for direct download of chosen area).

2) Geology of the area

Geology data was available from the New Mexico Resource GIS Program page (http://rgis.unm.edu/data_entry.cfm main page, http://rgis.unm.edu/loader_div.cfm?theme=Geology for geology section). This page also was useful for gathering data later used for creating polygons of the project extent, the fire extent and the ranch property boundary. The Geology shapefile is in NAD27, differing from the rest of my data except for the soils shapefile. It was also necessary to download a zip file of documentation (.E00 format once unzipped) to find rock unit names and descriptions, as well as information about the creation of the data.

3) Soil cover of the area

Soil cover was also available from the New Mexico Resource GIS Program page (http://rgis.unm.edu/loader_div.cfm?theme=Soils ). GCS is NAD27 also.

4) Burn intensity

Burn maps are available on a variety of sites. Maximum extent of burn is what I will consider. Intensity levels will not be differentiated (see assumptions).

5) Precipitation data is not necessary for this analysis (see assumptions).

6) Vegetation

Vegetation has all been burned (see assumptions). Land cover data are available with NED data, but not used in analysis.

7) City/Town locations – Census 2000 data

This data was available on the New Mexico Resource GIS Program page (main page http://rgis.unm.edu/data_entry.cfm, census 2000 page http://rgis.unm.edu/loader_div.cfm?theme=Socioeconomic%20Data ). This data was used for the city locations as reference points; they help in editing the ranch and fire boundary polygons. I mapped only the cities pertinent to my project area.

8) Transportation

Transportation structure was used for the same purpose as cities. Roads provide useful reference lines for editing boundary polygons. Available on the New Mexico Resource GIS Program page:

http://rgis.unm.edu/loader_div.cfm?theme=Transportation

I symbolized and mapped only the roads which were pertinent to the project area.

Project Assumptions – Practical Analysis and Limitations

I made a few assumptions to suite this analysis to my working knowledge of ArcMap, data availability, and this backcountry burn site:

A) Burn intensity was equally severe throughout the burn area.

No data was available to me which designates differential burning of certain locations. All vegetation is assumed to have perished.

B) Precipitation affected this locality evenly throughout each passing weather front. This is a fairly arid region and I shall assume that the subsequent rainstorms pass evenly through the area and affect all terrain on the same level of intensity.

C) Fire and property boundaries are closely estimated – not exact.

No exact vector data of the ranch property line or the fire’s maximum extent were available for my use. It was necessary for me to create and edit my own data for these boundaries (to be discussed later) from map drafts accessed on the internet.

So, the factors to be used in the final analysis after making these assumptions are:

1) Steepness of terrain

2) Geology of the area

3) Soil cover of the area

Data Preprocessing

The data available to me were in formats such that preprocessing was not a major task. All data was downloaded in .ZIP format, so it was necessary to unzip these compressed files. As I did this, the files were uncompressed into their own new folders, which were automatically created. I organized these with logical names in a hierarchical structure to provide for simple file storage.

Unzipping of NED data block

Data Processing

Elevation, census, and transportation were in NAD83 GCS and geology and soils were in NAD27. However, they all may be incorporated into the same workspace under NAD83 by having ArcMap project soil and geology shapefiles on-the-fly to this (NAD83 of the elevation data) datum. The NED data, however, required much manipulation before it was ready to be integrated into the final workspace.

Initial Step – Set Project Extent

The first step I took was deciding what the extent of my project analysis area would be. To do this I visually inspected maps with fire and ranch boundaries as well as cities and roads. From here, I created a new polygon shapefile in ArcCatalog (working_extent.shp) and created a new polygon feature using ArcMap’s editor toolbar. The result of this process is a shapefile with one large rectangle which encompasses more than enough area to analyze the burn site.

Now that the project extent had been set, I used this rectangle to clip my remaining data to the project site rather than all of New Mexico. This step involved clipping both vector and raster data, with the later being a much more exhausting task.

Clipping Geology and Soils shapefiles – Vector Data

To clip the vector data (geology and soils), I used ArcMap’s GeoProcessing Wizard. I used the extent layer which I created to first clip the geology and then in a separate operation the soils. This was a simple step and the results were just what I desired.

Clipping interface – geology shapefile to be clipped

Geology shapefile clipped to project extent, with units differentiated

Clipping NED rasters

Clipping the raster elevation data was a more difficult step. The two NED data blocks were huge datasets (about 12 million cells each!). For this reason, I had trouble due to long calculation times, program errors in ArcMap, and even available drive space. I used spatial analyst and the raster calculator to perform the clipping of each of the NED data blocks.

The same extent shapefile was used to specify the Analysis Mask (under “general” tab) and the Analysis extent (under “extent” tab) in the options menu of the Spatial Analyst toolbar:

The final step is to have the raster calculator evaluate the raster of interest and this performs the clip.

This must be performed on both blocks of the NED data.

Mosaic of the two NED data blocks

Before I did this, I first changed each raster to a classified rather than stretched appearance under to symbology tab of the layer properties window. I used the natural breaks (Jenks) default option for the classification type and divided the values into five classes. This changes the raster representation value to fall into one of the five classes as opposed to a stretched color ramp representation.

To mosaic the separate, clipped NED rasters into one file it is again necessary to use the raster calculator of the spatial analyst. I used a mosaic formula form the ESRI online support technical article on mosaicing images and grids (mosaic_images_and_grids.txt). Simply enter the formula in the space provided and press evaluate.

The desired result was achieved, a mosaic of the two clipped NED raster data blocks into one classified raster block, called Band1 and shown in the following graphic:

Mosaic of the two clipped NED data rectangles

From here, the next desire is a slope profile. But before this is possible, I needed to reproject the mosaic so that x, y, and z data are all in the same units.

Projecting NED data

Projecting was achieved using ArcToolbox. This step is important because x and y data are currently measured in decimal degrees and elevations are in meters. In order to perform the surface slope analysis, all data must be measured in the same units. I used the projection wizard for grids and projected the data to UTM zone 13 north. The datum for the data was still set to NAD83. I choose a Bilinear resampling method, as advised by my professor Dr. Mark Helper. For elevation data, the influence of the surrounding cells does not need to be heavily accounted for, so a nearest neighbor or cubic resampling method would unnecessarily use up too much memory.

Now all data (x, y, and z) are measured in meters and the elevation data is ready to be analyzed for a slope profile.

Creation of Slope Profile using Spatial Analyst

This also may be done using the 3D analyst, but I choose to again use the spatial analyst. Under the surface analysis heading, choose the slope option. I let the options remain at the default values, including the z factor, which is the desired vertical exaggeration.

Slope profile of project extent elevation data

Creating Philmont Ranch Property Boundary

This requires using the ArcEditor toolbar and creating a new polygon feature in a new shapefile created by me (similar to creating the project extent, but a much more complicated polygon than a simple rectangle). Recalling my assumption C) from before, the property boundary I created is estimated, not measured data. My source is a crude map which I pulled off of a boy scout troop website (see below).

Using the cities and road intersections as reference points, I created a new polygon mimicking this source map. This is the same process which will be followed for editing the maximum fire extent polygon.

Creating Maximum Fire Extent Boundary

Using a source map, create a new polygon feature in the fire_max_extent shapefile with the editor toolbar. Again, boundary is estimated.

Source map for fire extent – ranch boundary shown also

Edited Fire Extent Polygon

Spatial Analyst Analysis

Factors considered:

1) Steepness of terrain

2) Geology of the area

3) Soil cover of the area

Now, I must assign weights to each of these for the final analysis. I believe that slope will be the most deciding factor concerning the severity of erosion potential. Here are the values decided upon for weighting the nine classes of slope (slope measured in degrees):

Slope Weighting

Scale 0 – no potential to 10 – great potential

Class Range Weight

1 0 – 3.25 .2

2 3.25 – 7.58 .5

3 7.58 – 11.9 1.2

4 11.9 – 16.24 2

5 16.24 – 20.57 3.5

6 20.57 – 24.9 4.5

7 24.9 – 29.5 5.5

8 29.5 – 34.9 7

9 34.9 – 69.3 9

Geology Weighting

Before I assign weights, I desire to know which geologic units underlie the burn area. This is easy to figure out; make a “select by location” query for features in the geology shapefile which intersect the fire extent boundary. It turns out only seven different units intersect the burn area, and all are either Quaterenary or Tertiary in age.

The geology shapefile has only abbreviations, not full unit names, in it’s attribute table. So, for me to be able to assign values, I must find out what the abbreviations stand for and assign values accordingly. More crystalline, cemented rocks will have a low erosion potential value and more unconsolidated sediment will have the highest values. It was necessary for me to consult the documentation for this shapefile which I downloaded.

Select by location query interface and Geologic units intersecting the burn area

Scale 0 – no potential to 10 – great potential

Abbreviation Name Rock type Rank

Qa Quaterenary Alluvium sediment 9

Qp Piedmont Alluvium sediment 9

Tui ? Intrusive (crysytalline) 1

Tpc ? Continental (Sedimentary) 3

Tkr Ringbone Fmn. Continental (Sedimentary) 3

TKpr Picuris Fnm. Continental (Sedimentary) 3

Kut Trinidad Sandstone Sedimentary 3

Soil Weighting

Again, I select by location the soil types which underlie the burn site. It turns out only one soil type, eutroboralfs, underlies the burn area. For this reason, soil may be omitted from the final analysis (since it will have only one weight, and that would affect all cells the same). Eutroboralfs is a glacial till derived soil.

Converting Vector to Raster

Now, in order to perform the final analysis, it is necessary to convert the vector files involved in the calculation to a raster format. This applies to the fire extent file and the geology file. This is done by using the spatial analyst’s convert features to raster option. All data must be in raster format in order to perform the weighted analysis using the raster calculator.

Reclassifying Data

This step is necessary so that the data values may have map algebra performed on them for the final analysis. Reclassifications may be performed using the reclassify option on the spatial analyst toolbar dropdown menu. It is not necessary to reclassify the fire data: it is a Boolean operator essentially, and will have a value of either 0 or 1 corresponding to not burned or burned land, respectively. The slope data will be reclassified into 10 classes to match the weights I have decided on for slope ranges (higher class numbers representing the steeper terrain).

Spatial analyst reclassify window interface – slope reclassification with old and new values shown

The geology raster also was reclassified on a 1 to 10 scale. However, due to repeated rock types, only values of 1, 3, and 9 occur in the final weighting.

Final Analysis using the Raster Calculator

To perform the final calculation, I must decide which factor I wish to weight more heavily in the actual map algebra. I believe that slope is a more vital factor than geology; for that reason, I choose to weight the slope value with a factor of .65 and the geology factor with a value of .35. The fire extent value (0 or 1) is multiplied by the sum of the weighted geology and slope factors so that a fire value of 0 (no burn) has no erosive potential in this analysis (recall I choose to only asses the erosion potential in the burn zone).

The output of this formula is the final erosion potential map.

Conclusions

First of all, I would like to thank the people who helped me along with this project. My professor Dr. Mark Helper and my teaching assistant Tim Pierce undoubtedly saved me multiple hours of time wrestling with ArcMap by offering me instructions and advice. It would be impossible for me to perform this analysis without the knowledge I have accumulated over the past months in this class. I would also like to thank a friend of mine who has worked on the Philmont ranch for two years now, Nick Soorholtz. After being a backcountry “chef” last year, he has been relocated to the erosion control crew this summer and fall. Without personal communications with you, I again would have wasted hours searching for general information. So thank you Nick.

I believe that the initial question, is it possible to predict which areas in the fire zone would be most prone to become severely altered by erosion due to destruction of natural vegetation, has been answered by this erosion potential map. Upon initial visual inspection of the erosion potential map, it is clear that the majority of the land in the fire zone which has a high erosion potential lies in the Philmont ranch property. The southern tip of the burn extent area has the greatest potential to become severely eroded.

It is also clear that the severe potential areas are located closest to the local tributaries of the Ponil river. I suspect this is true due to increased slope and presence of more alluvium in these areas. From personal communication with erosion control staff, he mentioned two areas on Philmont land which had severe erosion occurring after the fire: Dean Canyon and Chandler Canyon. I will symbolize these features (from the New Mexico Resource GIS Program page data) on my map and see if my prediction included these areas in a high potential area.

Chandler Canyon does lie in an area which has high potential, but Dean Canyon appears to be downstream, out of the burn zone according to my map.

In my analysis the areas of greatest erosion potential lie in canyons near the Ponil river. This is a logical result considering the geology of these areas will be more sediment derived and also that slopes near the river are probably increased as compared to the surrounding highlands. The most severe potential lies in the topographic low drainages where rainwater washes away the alluvium soil.

I believe this analysis has been successful. I would be curious to witness this area and see for myself where the erosion has been severe so I could verify my results. However, the overall results of this analysis are quite believable even if more factors (i.e. precipitation, varying burn intensities) should be included in order to obtain a more accurate erosion potential map. This map does indicate which areas have the greatest potential to become severely eroded, so I feel this exercise has produced useful information.

Back to the Semester Projects Page