Figure 1.
Figure 2.
Predicting Pedogenic Silica Accumulations in
Geologically Complicated Landscapes
Fall 2007 CEE 6440 Class Project
Colby Brungard
Abstract: Soil classification depends on correctly identifying subsurface soil characteristics. Subsurface accumulations of silica are important subsurface diagnostic horizons in areas dominated by silica rich parent material. Identifying areas of subsurface silica accumulation is important for correct soil classification as numerous land use interpretations are made from such classifications. Here I present two methods for identifying these areas using TauDEM (Terrain Analysis Using Digital Elevation Models) software, a 5 meter DEM (Digital Elevation Model), and soil data collected in the field.
Introduction and Literature Review:
Soil is the base of land ecosystems (Stringham, Krueger, Shaver, 2003). Maintaining rangeland ecosystem health depends on understanding the biotic and abiotic components of that ecosystem. State and Transition models greatly facilitate an understanding of the factors which control ecosystem dynamics (Stringham, Krueger, Shaver, 2003). Land managers rely extensively on state and transition models to maintain healthy ecosystems. As soils are part of the foundation (the state) in state and transition models, correct classification of soil attributes is a necessary input for rangeland management practices (Grazing Lands Technology Institute, 2003).
Pedogenic accumulations of silica affect the potential ecosystem that can develop in a soil. Understanding how silica forms and predicting soil accumulations is useful in delineating these areas. Current theories suggest that silica accumulations form when silica weathers from silica rich parent material and is transported both laterally and vertically by soil water. Given enough time, silica can accumulate to such an extent that a duripan (cemented subsurface silica accumulation) will form. Duripans are typically associated with volcanic parent materials that contain appreciable amounts of volcanic glass (Soil Survey Staff, 1999).
While the association between silica and volcanic parent materials and volcanic glass seems to adequately describe a significant pathway of pedogenic silica accumulations, Boettinger and Southard have described duripans formed in areas without any evidence of volcanic glass (Boettinger and Southard, 1991). They argue that in situ weathering of crystalline silicate minerals provides the necessary silica source, especially in arid and semi arid environments where water for silica dissolution and translocation is limited. Using DEM derived products an attempt is made to identify sources of pedogenic silica and provide insight to the question of soil silica sources.
Numerous researchers have applied DEM derived products to predict soil attributes (More, et al., 1993; Gessler et al., 1995; Gessler et al., 2000). Gessler et al. accounted for 52-88% of the soil attribute variability in their study using easily computed hillslope attributes such as slope and flow accumulation (Gessler et al., 2000). Other attributes used to model soil attribute variability are aspect, specific catchment area, maximum flow path length, plan and profile curvature, wetness indices, stream power indices and sediment transport indices (Moore et al., 1993).
TauDEM (Terrain Analysis using Digital Elevation Models) is a suite of functions designed to calculate various attributes from a DEM and is available as a plug in to ArcGIS® (TauDEM website, 2007). In addition to calculating standard DEM derivations such as slope, flow accumulation and performing basic DEM functions, TauDEM allows the user to calculate several specialized functions useful in terrain analysis. As silica accumulations depend on soil water, these specialized functions can be applicable to predicting silica accumulation.
Methods:
Project Area Description:
Located in South Western Utah’s
Due in part to the north-south trend of the geological uplift, the area receives roughly 8-10 inches of precipitation per year, mostly in the form of snowfall and highly localized convective summer thunderstorms. Elevation ranges from 1500 to 2100 m (4900 – 6900 ft). Range (both cattle and sheep) and, to some limited extent mining, dominate landuse. Typical vegetation/landcover consists of sagebrush steppe and pinyon/juniper forest at higher elevations (SWReGAP, 2005).
Figure 1.
Figure 2.
Data Structure:
Data for this project was originally stored in an ArcGIS® geodatabase. However; as the number of raster layers increased and the need to rapidly move individual rasters and shapefiles to multiple locations on the hard drive (and read them with ERDAS Imagine) became apparent, a file structure provided an easier method of data storage. All data layers are projected to UTM Zone12N NAD 83 and where appropriate NAVD 88.
Data Type, Source and Preprocessing:
10 meter DEM:
Acquired from
5 meter DEM:
Acquired from
Silica Data:
Collected during field work, this point feature class includes attribute data and soil classification. Using data from the field, a column for silica presence was created using the editor toolbar and binary code (1 = silica present, 0 = silica absent). All records with silica present (silica = 1) were selected and exported into a separate point feature class. A 20 meter buffer was applied (Saunders 2007) and converted to raster format using ArcToolbox, resulting in a binary GRID. An important input for the upslope dependence function.
Geological Data:
Initially acquired as a PDF map from the Utah Geological Survey, this was converted to .tiff format, the map collar removed, converted to an .img file and georectified in ArcGIS. Digitizing this geology map resulted in the creation of a polygon feature class with four classes: 1 = limestone/dolomite, 2 = other sedimentary rock, 3 = intrusive igneous and 4 = extrusive igneous. Adding one additional polygon around the outside of the image allowed the undigitized area to be classified as Quaternary fill. This was accomplished by converting the polygon shapefile to an .img and assigning all zero values a value of 5 (5 = quaternary fill) using ERDAS imagines’ modeler tool. Then the image was masked to the study area boundaries. As no topology existed when digitizied, groups of pixels occurred inside contiguous areas. Clumping and elimination of all pixel groups smaller then 50 using ERDAS imagine resulted in a meaningful geology map in raster form. Then using the raster calculator in the spatial analyst tool box of ArcGIS, all areas with a value = 4 (extrusive igneous) were output in binary. 1 = extrusive igneous present, 0 = extrusive igneous absent. Extrusive igneous rock was selected as the only contributor to pedogenic silica based on the idea that the mineral composition facilitates silica dissolution and because of the large spatial extent that it occupies in the study area. This is an important input for the Downslope influence function.
ASTER:
Images obtained from USU’s IRDIAC were initially used in conjunction with ERDAS Imagine’s spectral analysis tool as a method for accurately identifying silica rich geologic members. Upon investigation, however; it was determined that the necessary user knowledge and in depth understanding of the mineral composition of the geological units was required and, at present, beyond the scope of this project
Analysis:
Analytical functions:
Downslope Influence Function:
Initially developed to track the movement of a contaminant the downslope influence function (DI) uses a Dinf flow prediction GRID, and a disturbance GRID (in this case a binary GRID of extrusive igneous geology) to predict downslope accumulations of (in this case) silica. This works by proportioning the amount of flow from known source areas using the Dinf contributing area (calculated from the Dinf flow prediction GRID) and calculating the amount and destination of that flow. The output is a GRID with cell values indicating the portion of flow at that cell originating from the cells in the source area (figure 3). In the following figure cells with darker values indicate a larger portion of the flow from the source area.
Figure 3.
Upslope Dependence Function:
The Upslope Dependence function (DEP) is conceptually the inverse of the DI and is useful for predicting silica sources. Initially developed to determine the source of a contaminant, the upslope dependence function (DEP) uses a Dinf flow prediction GRID, and a disturbance GRID (in this case a binary GRID of known silica accumulations) to predict upslope sources of silica. This works by determining those cells which contribute flow to the known accumulation areas using the Dinf contributing area (calculated from the Dinf flow prediction GRID) and calculating the amount of flow that these pixels contribute. The output is a GRID with cell values indicating the portion of flow at that cell that flows to the cells in the known accumulation area (figure 4). In the following figure, cells with lighter values indicate a larger portion of flow to the accumulation area.
Figure 4
Results and Conclusions:
Downslope Influence:
If silica is indeed derived from extrusive igneous rock and transported by soil water, then the downslope influence function should accurately predict silica accumulations. Active wash channels are frequently disturbed by floods/debris flows. Therefore the distinction between active wash channels and soils with enough stability and time for the formation of pedogenic silica was a necessary consideration. Based on previous explorations of the area and visual interpretations of the landscape from the 5 meter hillshade, the break between active wash channel and stable soil was determined to be at a value of 1000. Additionally it was necessary to indicate a lower threshold of flow and this was determined to be a value of 15. For display purposes the downslope influence GRID was converted to a polyline feature class and buffered 20 meters using ArcGIS.
The
application of this function failed to predict the majority of the silica
accumulations as the following figures (figures 5-7) demonstrate.
Figure five shows a large drainage from extensive areas of igneous extrusive rock but fails to indicate much silica accumulation in the alluvial fan (roughly indicated by red bounding box) as might be expected, although silica is present in the soil.
Figure 5
One of the major conceptual limitations inherent in predicting silica accumulations via the downslope influence function is time. Time is a necessary precursor to the weathering and precipitation of silica and allows for changes in the relative position of water courses. It is reasonable to assume that this landscape has remained unchanged (geologically) for the last 10,000 years and that the flow of water across and underneath the surface has shifted. Figure six indicates much the same type of failure as figure five but demonstrates, more clearly, how time might impact the output of this function. This figure shows a high drainage density across a potentially very old alluvium mantel. As this function only predicts flow down current drainage paths it is anticipated that this entire surface has been influenced by drainage from igneous extrusive rock and that silica accumulation is widespread across this geomorphic surface.
Figure 6
Figure seven shows areas of correct prediction (blue dots) indicating that this function does have some application to silica accumulation prediction. The success at which silica was predicted in this figure indicates that the failure of the downslope influence function to predict the majority of the known silica accumulations is in part due to data constraints such as sampling density.
Figure 7.
Actual
accuracy depends on several confounding factors which were not adequately
addressed in this study. Among those most easily addressed, should this project
continue, are the accuracy of the geological data used and the necessity of
specific sampling in predicted areas.
It is also anticipated that the
downslope influence function might be more useful in the Colorado Plateau as a
predictor of soil accumulations from specific rock types as the soil may have
formed on a more recent time scale.
In conclusion the downslope influence function is a good algorithm for its intended purpose and may be of further use to determine quickly deposited sediment accumulations.
Upslope
Dependence:
If silica is indeed derived from extrusive
igneous rock and transported by soil water, then the upslope dependence
function should accurately predict pedogenic silica sources. Figure eight
demonstrates the applicability of this function. As
indicated by the red dots on the image silica accumulates in both residuum
(soil formed in place) and alluvium (soil formed from water transported
deposits). For the purposes of this analysis the silica accumulations in
alluvium are of interest.
The known
silica accumulations with the seven largest upslope drainage areas (by visual
estimation) are shown in figure 8 as blue dots. It was anticipated that silica
accumulations would have large upslope areas dominated by silica rich extrusive
igneous deposits (the green areas). This however; was not the case. Only one of
the silica accumulations had an upslope drainage area that included extrusive
igneous rock and only a very small fraction of the flow in this area
contributed to the silica accumulation. The majority of the drainage appears to
originate in limestone/dolomite deposits. Limestone/dolomite is typically
thought to contribute insignificantly to weatherable silica. There are several
explanations for this incongruity. Perhaps the most obvious reason is that of
human error but, other possible reasons include; misidentification of pedogenic
silica, incorrect delineation of the geology (due to the scale of mapping),
differing chemical makeup of limestone, and the presence of other siliceous
rock types (granite and quartz monzonite) in the upslope drainage areas.
Further investigation is needed to fully answer these questions.
Figure 8.
Figure nine
is more typical of what was anticipated. Here the selected silica accumulations
(blue dots) have an upslope drainage area that is derived from extrusive igneous
rock.
Figure 9.
While the upslope dependence function appears to perform adequately and provokes some interesting research questions the entire function is dependent on correctly identifying areas of silica accumulation and correctly determining the spatial extent of silica accumulation on the landscape. By applying a 20 meter buffer it is thought that large areas of silica accumulation were omitted, which affected the total upslope drainage area. However; due to time limitations, 20 meters was thought to adequately represent the spatial extent of pedogenic silica for the purposes of this study. Further investigation using this method should result in the quantitative analysis of flow from rock type.
Recommendations:
This has been
an exciting exercise in the application of Geographic Information Systems and
Science to answer questions regarding pedogenic silica. While still in it’s
infancy in pedological studies GIS has the power to enable quantitative soil
investigations.
Further investigation is needed to
determine the applicability of both the downslope influence and the upslope
dependence functions, however; obvious methods of improvement exist: inclusion
of all geological strata, better identification of the boundaries of geological
strata using remote sensing or other methods, application
to soils with differing amounts of silica, and silica driven sampling. Additionally the use of the decaying concentration function in
TauDEM may enable actual estimates of silica accumulation rates.
References:
1. Boettinger,
J.L., and R.J. Southard. 1991. Sources of silica and carbonate for Aridisols on
a granitic pediment, western
2. Best,
B.G., Lemmon, D.M. and Morris, H.T., 1989. Geologic Map of the
3. Gessler,
P.E., O.A. Chadwick, F. Chamran, L. Althouse, and K. Holmes. 2000. Modeling
soil-landscape and ecosystem properties using terrain attributes. Soil Science
Society of
4. Gessler,
P.E.,
5. Grazing
Lands Technology Institute. 2003.
6. Moore,
E.D., Gessler, P.E., Nielsen, G.A., Peterson, G.A., 1993. Soil attribute
prediction using terrain analysis. Soil Science Society of
7.
Saunders, A.M., and J.L. Boettinger.
2007. Incorporating classification
trees into a pedogenic understanding raster classification methodology,
8.
Soil Survey Staff. 2006. Keys to Soil
Taxonomy, tenth edition. USDA
9. Stringham,
T.K., Krueger, W.C., and Shaver,
10. Tarboton,
D. G., (1997), "A New Method for the Determination of Flow Directions and
Contributing Areas in Grid Digital Elevation Models," Water Resources
Research, 33(2): 309-319.
11. USGS National Gap Analysis Program.
2005. Southwest Regional GAP Analysis Project—Land Cover Descriptions. RS/GIS
Laboratory,