Soil Science Society of America Journal 64:679-680 (2000)
© 2000 Soil Science Society of America
DIVISION S-5-PEDOLOGY
Update and recorrelation of soil surveys using gis and statistical analysis
G.R. Brannona and
B.F. Hajekb
a USDA-NRCS, P.O. Box 311, Auburn, AL 36830 USA
b Agronomy and Soils Dep., Auburn Univ., Auburn, AL 36830 USA
greg.brannon{at}al.usda.gov
 |
ABSTRACT
|
|---|
The introduction of U.S. soil taxonomy and the increased pressure on land use and development has generated the need to update soil surveys that were published before 1965. A portion of a pre-1965 soil survey from Montgomery county in Alabama was selected to evaluate an update approach using geographic information system (GIS) and statistical analysis. The update included map recompilation, correlation, interpretation, and presentation methods. Sampling points were identified with a stratified random sampling and data obtained at each point were analyzed by traditional statistical methods. The taxonomic accuracy was 75 to 83% at a confidence level of 90%. Interpretative reliability was 90 to 95% for dwellings without basements, 95 to 98% for septic tank absorption fields, and 93 to 98% for local roads and streets. Updating old soil surveys by using GIS technology and statistical evaluation can produce a quality soil survey that meets or exceeds National Cooperative Soil Survey (NCSS) standards. Using this method, an experienced soil scientist can update, recorrelate, and recompile 
40180 ha (100000 acres) yr-1. This is an increase in production of 22090 to 24100 ha (55000–60000 acres) or 120 to 150% compared with conventional remapping methods.
Abbreviations: GIS, geographic information system • MLRA, Major Land Resource Area • NCSS, National Cooperative Soil Survey
|
INTRODUCTION
|
|---|
THE PRODUCTION of soil surveys is undergoing a fundamental change as we approach the 21st century. The NCSS traditionally focused on producing soil surveys by counties or political regions (Indorante et al., 1996). The NCSS is moving away from a county basis to the administration, correlation, maintenance, and production of soil surveys on Major Land Resource Areas (MLRAs), which includes use of GIS. This approach improves reliability by applying consistent standards to the entire geographic area while providing easy access to data and products. (McLeese et al., 1991; McLeese, 1992; Soil Survey Staff, 1993a). In addition, many published soil surveys were completed before 1965 and need to be updated using current U.S. soil taxonomy (Soil Survey Staff, 1975).
In October, 1995, 17 MLRA offices were created in a major reorganization of the NCSS to implement and manage the Soil Survey by geographic regions. Reductions in funding and numbers of personnel make this challenge increasingly difficult by forcing the federal government to do more with less. The reinvention of government initiatives set forth in 1993 by the National Performance Review called for a major reduction of the federal work force (Kettl, 1994).
Due to constraints of personnel and resources, alternative methods to remapping are needed. In this study, a combination of GIS and statistical analysis using stratified random sampling and binomial statistics were used to evaluate remapping and recorrelation of a portion of a pre-1965 county soil survey.
|
Materials and Methods
|
|---|
The study area is in eastern Montgomery County, Alabama and extends west from the Bullock and Macon County line to 86°7'30'' N, and north of Pike County to the Tallapoosa River. The area is in MLRAs 133a and 135, the Southern Coastal Plain and Alabama, Mississippi, and Arkansas Blackland Prairies, respectively. The study area encompasses 62306 ha (153840 acres), comprising about 30% of the county (U.S. Department of Commerce, 1990).
The legend was evaluated by an experienced soil scientist and soil map units with similar profile characteristics, landscape positions, slope, and geology (parent material) were combined in an office recorrelation. Lines separating soil map unit delineations were transferred by hand from the published soil survey atlas sheets (Burgess et al., 1960) to the corresponding 7.5' U.S. Geological Survey topographic quad sheets. Soil map units that were combined in the office recorrelation were merged on the topographic maps by combining the map units while transferring the lines. The soil map unit polygons were adjusted to follow map features.
A field reconnaissance of the study area was done to check the accuracy of the office correlation, line placement, validity of the soil map units, and to identify potential problem areas. This reconnaissance was restricted to primary and secondary roads. Very few off-road investigations were conducted since this study was not designed for intensive field investigations and remapping. Soil map units were added as needed.
The revised soil lines contained on the topographic maps were traced onto mylar sheets and scanned on an LDS 4000 Plus Scanner (Summagraphics, 1991). The digitized data then were loaded down into a Unix SUNSPARCstation 1+ work-station system (Sun Microsystems, 1994) and imported into the LT4X Release 3.21 for Sun (SunOS/SOLARIS) program (Infotec Development, 1993) and registered. The individual maps were cleaned, edited, and the lines along the map edges were adjusted to match the joining maps. Individual maps were then imported into ARC/INFO (Environmental Systems Research Institute, 1994) for joining into a composite map and conducting further editing, which primarily consisted of dissolving the map edges and matching the adjoining soil map unit polygons.
A stratified random sampling method was used to select the points to check the accuracy of the recorrelation. Points were stratified by map unit. This method is frequently used by geographers and some of the early applications were for land use studies (Wood, 1955). This method has been found useful when the study area is extremely variable (Cole and King, 1968). This method has two advantages: (i) it is possible to sample in proportion to the size of the units sampled and (ii) it increases the precision of sampling (Shaw and Wheeler, 1985).
The composite soil map was imported from ARC/INFO into IDRISI (Clarke University, 1992), a DOS-based GIS software package. The stratified random sampling program in IDRISI generated the points that were used to evaluate the revised soil map. Ninety-eight sample points were randomly selected throughout the study area. The UTM coordinates for the points were identified, recorded, and plotted on the corresponding topographic maps. Three subsamples, or pedons, were randomly selected within the polygon at each point and were checked for taxonomic accuracy, slope, and series identification. A total of 294 observations were evaluated. Series identification was accomplished by using Keys to Soil Taxonomy (Soil Survey Staff, 1996) and the Soil Survey Manual (Soil Survey Staff, 1993b), and the corresponding official series descriptions. Each subsample was also evaluated for the following interpretation ratings: dwellings without basements, septic tank absorption fields, local roads and streets, and camp areas. The subsamples were analyzed using soil interpretations ratings guides to determine whether the observations had the same interpretation rating as the correlated map unit.
|
Results
|
|---|
The interpretative data were assessed using statistics applicable to binomial data (Steel and Torrie, 1960). A 90% confidence level was selected to evaluate the data. The initial evaluation for taxonomic accuracy revealed that 91 of the 294 subsamples, or 31%, did not meet taxonomic accuracy (Table 1)
. Additional review of the data point observations resulted in renaming four soil map units to better reflect the point observation on map unit composition. The change in soil map unit names resulted in a decrease to 63 subsamples that did not meet taxonomic accuracy, or an increase to 78.6% in taxonomic accuracy of the map units.
View this table:
[in this window]
[in a new window]
|
Table 1 Confidence intervals at a confidence level of 90% of the subsamples before and after the soil map legend was adjusted to reflect the observations in the field
|
|
The interpretations were evaluated and the results are listed in Table 2
. The confidence intervals ranged from 89.7 to 97.6%, reflecting a high degree of confidence in the recorrelated map units.
Using this method of soil survey, in this region, an experienced soil scientist can recorrelate and recompile 40180 ha (100000 acres) in a fiscal year. This is an increase in production of 22090 to 24100 ha (55000–60000 acres) or approximately 120 to 150% more than conventional remapping methods.
The study was designed to address updates of pre–U.S. soil taxonomy soil surveys made and published on a photobase in MLRAs 133a (Southern Coastal Plain) and 135 (Alabama, Mississippi, and Arkansas Blackland Prairies). They will have to be tested and revised as needed to address local needs, existing data, and quality of existing soil surveys.U.S. Department of Commerce 1990
Received for publication April 5, 1999.
|
REFERENCES
|
|---|
- Burgess L.H., Wilson L.S., McBride E.H., Anderson J.L., Dahms K.E. Soil survey of Montgomery County, Alabama. Washington, DC: U.S. Gov. Print. Office, 1960.
- Clarke University. Graduate School of Geography IDRISI Version 4.0 user's guide. Worcester, MA: Clarke Univ, 1992.
- Cole J.P., King C.A.M. Quantitative geography. New York: John Wiley & Sons, 1968.
- Environmental Systems Research Institute. 1994. ARC/INFO Version 7.0 user's guide. Copyright 1990–1995. Environmental Systems Research Institute, Redlands, CA.
- Indorante S.J., McLeese R.L., Hammer R.D., Thompson B.W., Alexander D.L. Positioning soil survey for the 21st century. J. Soil Water Conserv. 1996;51:21-28.
- Infotec Development. 1993. LT4X user's guide. Copyright 1992. Infotec Development, Santa Ana, CA.
- Kettl, D.F. 1994. Reinventing government: Appraising the national performance review. Center for Public Management. CPM rep. 94-2. The Brookings Institute, Washington, DC.
- McLeese R.L. Updating and maintaining the soil survey in Illinois: A major land resource area approach. Illinois GIS and Mapnotes 1992;11:36-38.
- McLeese, R.L., S.J. Indorante, D.R. Grantham, D.E. Calsyn, and D.J. Fehrenbacher. 1991. Soil survey updates by major land resource areas or Illinois soil surveys: The next generation. p. 316. In Agronomy Abstracts. ASA, Madison, WI.
- Shaw G., Wheeler D. Statistical techniques in geographical analysis. New York: John Wiley & Sons, 1985.
- Soil Survey Staff. 1975. Soil taxonomy: A basic system of soil classification for making land interpreting soil surveys. USDA Handb. 436. U.S. Gov. Print. Office, Washington, DC.
- Soil Survey Staff. 1993a. Soil survey by geographic area. National Soil Survey Center. U.S. Gov. Print. Office, Washington, DC.
- Soil Survey Staff. Soil survey manual. Washington, DC: U.S. Gov. Print. Office, 1993 b USDA Handb. 18..
- Soil Survey Staff. Keys to soil taxonomy, 7th ed Washington, DC: U.S. Gov. Print. Office, 1996.
- Steel R.D.G., Torrie J.H. Principles and procedures of statistics with special reference to the biological sciences. New York: McGraw-Hill, 1960.
- Summagraphics. 1991. LDS4000Plus operation manual. Copyright 1990, 1991. Houston Instrument Div. of Summagraphics Corp., Houston, TX.
- Sun Microsystems. 1994. UNIX SUN SPARCstation 20 Workstation System. Sun system user's guide. Sun Microsystems. Mountain View, CA.
- U.S. Department of Commerce, Bureau of the Census. 1990. U.S. Gov. Print. Office, Washington, DC.
- Wood W.F. Use of stratified random samples in land use study. Ann. Assoc. Am. Geogr. 1955;48:350-367.
TOP
ABSTRACT
INTRODUCTION
