You are here:
A SOIL SPATIAL DATA FRAMEWORK FOR ENVIRONMENTAL MODELING IN THE CONTIGUOUS US
Citation:
Kern, J. S., W E. Hogsett, AND J A. Laurence. A SOIL SPATIAL DATA FRAMEWORK FOR ENVIRONMENTAL MODELING IN THE CONTIGUOUS US. Presented at 4th International Conference on Integrating GIS and Environmental Modeling, Banff, Alberta, Canada, September 2-8, 2000.
Description:
A suite of soil and related data-layers have been developed for environmental assessments of the effects of tropospheric ozone exposure and nitrogen deposition on forests, and global change (soil C pools and landuse impacts, water balance modeling). These spatial data depict soil particle size analysis [PSA], organic matter, [OM], bulk density [BD], rock fragments [RF], soil depth, soil water retention, organic carbon (OC), N (SN), N availability (as N-mineralization), exchangeable bases, and acidity. These datasets provide direct information such as C and N pool size and they also are input data to hydrologic and vegetation models.
The geographic basis is the 1:250,000 scale USDA Natural Resources Conservation Service (NRCS) STATSGO database which can be linked to STATSGO tabular data by map unit component or linked to soil laboratory measurements by soil classification of the components. The STATSGO tabular data were used to develop datasets of PSA, BD, OM, RF, soil depth, salinity, carbonates, gypsum, depth to impermeable layers, and water holding capacity (WHC). The areal extent of mineral soil, organic soil, miscellaneous area, and water were also estimated. STATSGO-based WHC was not used because it did not vary systematically in relation to PSA and OM. Rather, WHC was estimated with empircle equations (pedotransfer functions) that use PSA and OM stratified by soil classification. These results are adjusted for actual volume using RF and soil depth. We developed new pedotransfer functions for water retention that treat volcanic soil materials separately. The lithology and degree of weathering of soil rock fragments were characterized to better adjust soil water retention estimates.
The STATSGO spatial data and the NRCS National Soil Characterization Database (NSCD) laboratory data were linked by using soil series name or soil classification. In earlier work using the less detailed NATSGO database and the NSCD we found that the great group level of classification was useful because it reflected many soil forming processes related to OC accumulation and data completeness was greater than when using soil series. To spatially distribute soil laboratory measurements made on a weight basis to a real-world volume basis, one needs BD, RF, and soil depth data on a pedon basis. Missing BD were estimated based on regression equations by great group based on the NSCD which greatly increased the available data. Rock fragments and soil depth were used from the STATSGO tabular data because the NSCD is incomplete.
The NSCD was used to make series and great group level estimates of soil of OC, TN, BD, exchangeable bases and acidity, pH, and salinity. The spatial datasets expressed these parameters as mean, median, minimum, and maximum values. Maps of the coefficient of variation (CV) of the mean values were made. In areas of high CVs for critical model parameters one can choose another measure than the mean (minimum, maximum. mode) or run the model for each map unit component and then aggregate the results. Some soil parameters, such as hydraulic conductivity which does not vary linearly with changes in PSA, do not lend themselves to averaging. The availability of N in soil and the forest floor in the eastern US was estimated based on N mineralization. A literature search was conducted to find relationships between N mineralization and other variables useful for broad scale mapping. Vegetation and landuse greatly affect N mineralization rates in addition to TN, soil moisture, temperature, and N deposition. The spatial patterns of N deposition are being characterized to study its effect on N mineralization.
We are exploring alternatives to expressing the soil data as fixed depth intervals including aggregating data by soil genetic horizon. The variability for horizon-grouped data was expected to be less but many horizons in the NSCD lacked horizon designations. The fixed depth intervals used were 0- to 20-, 20- to 50-, 50- to 100-, and 100- to 150-cm. A limitation to these data is soil depth because soil surveys generally focus on the surface meter or two there is a lack of information to evaluate deeper soils.
These soil survey data have enormous possibilities. Not only are there a large number of variables that can be mapped but one can tailor soil spatial datasets to fit specific studies. Soil classification provides much information for identifying areas of interest. Mean soil properties provide useful information but even more detailed information is obtainable in areas where soil classification, slope, and/or aspect can be used to subdivide map units. The location of a component within a map unit delineation is not directly given but it can be approximated if the landscape position it occurs on can be inferred using ancillary data such as digital elevation models.
This collection of soil spatial datasets provides a wide range of soil parameters expressed several ways (minimum, maximum, mode, mean) with a measure of spatial variability (CV). The datasets are useful for many modeling applications and the methodology can be modified to produce datasets specific to project needs. The scale of the spatial data (1- to 4-km cells) provides much useful information while not requiring massive computer resources to manipulate and analyze.