Skip

Recompilation Methods

Recompilation Methods

Why are those lines placed where they are?: An investigation of soil map recompilation methods.

 T. P. D'Avello and R. L. McLeese

 Abstract

 Recompilation of soil maps to acceptable base maps is currently a major activity for soil scientists. The wide-spread use of GIS has accelerated this activity in the past five to ten years with the emphasis on soil survey updates and SSURGO development. This study was initiated to evaluate methods of recompilation common to the region and determine the effect of base map, and method on the resulting placement of recompiled soil lines by soil scientists. Three common recompilation procedures were utilized. Two used enlarged USGS topographic quadrangles as a base map and one used a USGS orthophoto as a base. Enlarged soil film transparencies were used for two of the procedures and a projected image of a soil map for the other. Five different soil scientists performed the recompilation. Data was digitized and GIS used to analyze, summarize and visualize the data. Results from Chi-Square analysis and Tukey's Multiple Comparison test indicated no distinct relationship between methods. As was expected, repeatability among soil scientists increased as map complexity decreased.

Background

 The standard base map for soil surveys in much of the country has been a rectified, high altitude aerial photograph. These aerial photographs do not possess the properties of a map, and are not acceptable for use as a base for digital conversion of soil maps (Natural Resources Conservation Service, 1996). Current standards of the Natural Resources Conservation Service (NRCS) require soil line work to be compiled on an acceptable base map prior to digitization. Acceptable base maps include orthophotographs or USGS 7.5’ topographic quadrangles. The intermediate step of recompilation to a suitable base map must precede digitization. There are a number of factors influencing the resulting placement of soil lines on a new base map, including the type and scale of the base map, the scale of the original soil map, the detail of the original soil map and the skill and ability of the individual performing the recompilation. Ideally, placement of soil lines for an area would be coincident and independent of the base map and the individual performing the recompilation. A lack of documentation related to this problem prompted this investigation.

Recompilation Methods

The study area was approximately 2,000 acres in size and located in JoDaviess County, Illinois (Figure 1). The area is unglaciated, with local relief of about 400 feet.

The study design included three methods, five soil scientists and two soil maps of differing complexity, as shown in Figure 3.

Method 1 is familiar to all that have recompiled without the aid of orthophotography. In this study, a mylar of the USGS 7.5’ topographic quadrangle, enlarged to a scale of 1:12,000, was used as the base map. Ratioed film transparencies of the soil maps were used as the source document. Registered mylar was placed onto the topographic quadrangle to serve as the new source document. Soil scientists placed this "stack" over the enlarged soil maps, adjusted the stack in relation to the soil map to get an acceptable fit for an area, then traced and labeled soil polygons. Adjustments were made periodically and recompilation continued in this fashion until the study area was completed. Enlarged topographic quadrangles were used for three reasons:

    1. Mapping density is too great to adequately draft at a scale of 1:24,000
    2. Scale should accommodate soil scientists, not the inverse
    3. Assumed errors inherent in enlargement procedure are negligible.

Method 2 was identical to method 1, except the original soils were projected from a slide image of the hard copy soil map using an overhead projection system, commonly called a "goal-post" system, and a right angle lens. An advantage of this system is that altering the projection distance of the overhead projection platform can modify the scale of the projected image. (Personal communication, Wayne Gabriel 1993).

Method 3 was identical to method 1, except the base map is a 1:12,000 scale, 3.75’ USGS half-tone, orthophoto quarter quad.

Five different soil scientists participated in the study over the course of 4 months. Their average length of soil survey experience was 19 years. Some soil scientists compiled the same soil map more than once, as indicated by Figure 3. In those cases, 2 to 3 months separated the compilation episode in an effort to minimize bias. Two soil maps of JoDaviess County were chosen, one with detailed mapping with many complex polygons, and one less detailed with simpler polygons.

GIS Methods

Soils were scanned, vectorized and attributed. Soil polygons were given common numeric attributes between maps. Complexity of soil maps was described via attributes of soil polygons. Complexity of mapping can be characterized in terms of the total number of delineations per unit area, number of map units on the identification legend for the area or the shape of polygons mapped. For this study, the primary difference between the two soil maps was polygon shape. Shape complexity was characterized using the following method (Hole, 1978):

I = P  /  (A /pi )1/2 * 2pi)

 

I = Shape Index

P = Perimeter of polygon

A = Area of Polygon

pi= ~3.14159

Therefore, a circle will have the least complex shape with an index of 1. Complexity increases as the ratio of perimeter to area increases. Classification of shape complexity, as suggested by Hole (1953), was used to develop the histogram describing complexity of mapping for the two soil maps used in this study (Figure 2). Difference maps were created for all combinations of respective soil maps using simple GIS subtraction techniques. Matrices were developed showing the proportion of pixels in agreement for all combinations of respective soil maps. Arcsine transformations were performed on the data and evaluated using Tukey's Multiple Comparison test.

Results and Discussion

 The most distinct relationship observed was that precision, hereafter termed agreement, was inversely related to the complexity of the soil map being recompiled. Pooling all values for each method of respective soil maps indicated 86 percent agreement for soil map 1, and 64 percent agreement for soil map 2 (Tables 1 and 2). The range was much narrower for soil map 1, emphasizing the inverse relationship between agreement and complexity. The number of polygons for both soil maps was similar. The primary difference between the two soil maps was the pattern of polygon distribution. As indicated by Figure 2, soil map 1 is dominated by relatively simple polygons, while soil map 2 is dominated by more complex polygons. Thought of in terms of rate of change, with soil map unit being the unit of measure, soil map 2 would have a greater rate of change than soil map 1. From a practical standpoint, this means discrepancies are more likely, and accentuated as shape complexity increases. The same relationship would be expected as polygon density increases.

High polygon density is the primary problem encountered when recompiling to a base map that is at a reduced scale from the original soil map. Soil maps in much of the Midwest were mapped at a scale of 1:15,840. Recompiling to a base map at 1:24,000 scale requires an approximate reduction of 1.5 times for the soil map and increases the density of line-work 2.25 times per unit area. While this study did not specifically compare the effect of base map scale on recompilation, the same relationship observed between soil map 1 and 2 would be expected. Soil scientists in Illinois have expressed a lack of confidence in their work when compiling to a base map of reduced scale from the soil map. Because of this "confidence factor", enlarged topographic quadrangles have been used as a base map for those counties that did not have orthophotography.

Some of the highest levels of agreement in the study were observed among soil scientists (Tables 1 and 2) that compiled soil lines to orthophotos. However, Tukey's Multiple Comparison test indicates considerable overlap among all methods tested (Tables 3 and 4). For soil map 1, numerous examples yielded similar results when comparing the two methods using topographic quadrangle base maps with the orthophoto method. In fact, the example comparing two topographic quadrangle "stacks" was similar to two of the three results of the orthophoto method. No method stands out as being distinctly superior, and when values are compared by soil scientist, all soil scientists have similar levels of agreement.

The results for soil map 2 were highly variable (Table 2). In general, methods using topographic quadrangles as base maps out performed those using orthophotos. However, as observed with soil map 1, no method distinguishes itself from the others. What is more apparent with this map is the performance of the soil scientists. Observing the values of each soil scientist indicates a wider range of values compared to soil map 1. It is apparent that soil scientists 2 and 4 had far lower levels of agreement than the other soil scientists (Table 2). There are a number of possible reasons for this, but it is primarily due to the fact that recompilation is a subjective, time consuming, tedious procedure that is fraught with error and subject to the vagaries of the personnel performing the task. Where a line is drawn is dependent primarily on where a person lines up the soil map, base map, and co-registered mylar overlay. Discrepancies can occur when the mylar overlay is shifted in relation to the base map or when the base map and soil map is not lined up properly. What designates a good fit between the soil map and base map, and when readjustment of that fit is necessary, is dependent on the knowledge and skill of the personnel performing the recompilation. This study demonstrates the unpredictable performance of people, and indicates that it is the most important variable in the quality of a recompiled soil map.

Soil scientists preferred the following methods, listed in order: orthophotos, topographic quadrangles, and the projection method. Compilation rates were similar for all methods, but familiarity with using a photographic image as a base map added to the confidence of the compilers.

An issue to consider when evaluating recompilation is the lack of a control. Verification of proper boundary placement is subjective. Few polygon boundaries are indisputably placed, with the exception of water bodies and areas altered by humans. Since the original soil lines are based on a map of unknown geometry, the original map can only be used as a reference to verify coding and inclusion of all polygons in the final product. Determining the geometry and camera parameters of the photographic base of soil maps is usually not possible, as they are cut from full exposures, eliminating all fiducial marks. These limitations have prompted some to utilize other means to convert soil maps to a digital format. Ventura (1989) achieved successful results using mathematical transformations, commonly termed "rubber-sheeting", to directly convert and "fit" the soils of Dane County, Wisconsin. Barnes and Vonderohe (1985) have suggested techniques used to create digital orthophotography as a means of converting and correcting soil maps. Ventura and Savory (1993) described a decision support tool to evaluate methods of digitizing soil maps by evaluating relief and ground control points.

We have had successful experiences using digital soils that have been directly converted and transformed in parts of the state with nearly level topography. For areas with nearly level relief and adequate numbers and distribution of ground control points, we have found mathematical transformations to be more precise and efficient than recompilation. The transformed soils are treated as the first step in developing an acceptable digital soil map. Some soil scientists have found it preferable to edit transformed, digital soil maps, rather than recompile. They have been satisfied with the fit of line work, and in cases where adjustments are needed, perform the adjustments on-screen with a digital orthophoto or digital raster graphic as a backdrop. This allows the soil scientist to spend less time on the error prone, imprecise method of recompilation, and more time on pedology

Acknowledgments

We thank the following for their assistance in support of this project: Dan Withers, Dave Preloger, Dale Calsyn, Steve Elmer and Randy Leeper, and thank Wayne Gabriel for sharing his experience and insights in the use of the "projection" method of recompilation.

Citations

Barnes, G., and A.P. Vonderohe. 1985. An analytical technique to account for photographic distortions

in natural resource records. Proceedings, 23rd Annual Meeting of the Urban and Regional Information Systems Association.

Hole, F.D. 1978. An approach to landscape analysis with emphasis on soils. Geoderma. 21:1-23.

Hole, F.D. 1953. Suggested terminology for describing soil as three-dimensional bodies. Soil Sci. Soc.

Am. Proc. 17:131-135.

Natural Resources Conservation Service. 1996. Digitizing specifications. Soil Survey Handbook

Part 647. U.S.Govt. Printing Office, Washington DC.

 

Ventura,S.J.1989. Conversion of geographic data to decision-making information: The Dane County,

Wisconsin Land Conservation Department example. Unpub. PhD. Thesis. University of Wisconsin.

Ventura, S.J. and D. Savory. 1993. A decision aid for soil survey map digitizing. Journal of Soil and water

Conservation. 48:484-488.

 

Tom D'Avello and Bob McLeese are former Illinois Soil Scientist and retired State Soil Scientist respectively. USDA-Natural Resources Conservation Service. 2118 W. Park Court, Champaign, Illinois 61821.