The Boise Hydrogeophysical Research Site (BHRS) is an in situ field laboratory (research wellfield) that is being developed by CGISS in a shallow, coarse (cobble-and-sand), alluvial aquifer. Our goal is to develop cost-effective, non-invasive methods for quantitatively characterizing the distribution of permeability in heterogeneous aquifers using hydrologic and geophysical techniques. Permeability is the most significant aquifer parameter for quantitatively describing or modeling groundwater flow and contaminant transport, and for designing remediation systems.
The BHRS is located on a gravel bar adjacent to the Boise River about 15 km southeast of downtown Boise. Eighteen wells were emplaced at the site in 1997 and 1998. The stratigraphy consists of coarse, cobble-and-sand, braided-stream fluvial deposits that overlie a clay layer at about 20 m depth. The design of the wells and the wellfield provides for a wide range of surface, single-well, cross-hole, multiwell and multilevel hydrologic,geophysical,well log and combined hydrologic-geophysical tests.
Locations of the 18 wells are superimposed on this
aerial photograph of the BHRS. Flow in the river at this location is
to the northwest.
Quarry exposure of analogous
coarse, braided-stream deposits showing disconnected sand lenses (S)
and a variety of cobble-dominated facies ranging from poorly-sorted
massive units (Gm), to moderately-sorted
horizontally-bedded units (Gh) and trough crossbedded
units (Gt). Heavy lines identify bounding surfaces
between depositional sequences. Prime Earth quarry northwest of Boise,
Idaho. For scale, quarry face is approximately 12 m high.
Example of core recovery from 34 to 42
ft BLS in well X1 at the BHRS. Changes in cobble size are evident. Sand
lenses (e. g., at ~37.6 ft - 38.4 ft) and sand matrix with cobbles are
captured in place. Unrecovered intervals (places held by foam with X
marks) are arbitrarily assigned to the top of a given cored section
during collection in the field.
Return to the table of contents
Return to the top of the page
At
the BHRS our intention
is to thoroughly characterize the wellfield as a control volume with
"known" 3-D distributions of sedimentary facies, geophysical
parameters, and hydrologic parameters. Then responses from geophysical
methods, alone and together, can be correlated against known
parameter distributions to develop techniques for mapping the 3-D
distribution of permeability with non-invasive geophysical methods.
Initial efforts will concentrate on three generally accessible
geophysical methods: seismic, ground penetrating radar (GPR) and
transient or time-domain electromagnetics (TEM). Also a variety of
borehole geophysical logs are being run in all wells.
Issues such as resolution, depth of penetration, and the ability to measure or image desired parameters will determine the usefulness of different geophysical methods. For example, GPR provides high resolution subsurface reflection images, but does not penetrate to the deeper parts of the aquifer at the BHRS. TEM, on the other hand, measures the 1-D electrical resistivity distribution with depth throughout the full aquifer thickness, but with less vertical resolution. Additional important data sets from the BHRS are the core analyses and geophysical well logs from the 18 wells at the site. We are using these data to verify and calibrate our geophysical interpretations. The many geophysical methods and acquisition geometries, combined with the well control, will provide an outstanding data set to characterize the heterogeneity of the alluvial aquifer and to develop techniques for mapping permeability with non-invasive geophysical methods.
Return to the table of contents
Return to the top of the page
Crosshole GPR Tomography Results
|
|
||
Tomography Acquisition |
||
Level run, including the porosity logs
plotted on the left side of the figure. Note the strong correlation
between high porosity and travel time delay. |
A variety of acquisition geometries are
available with borehole experiments to determine the velocity structure
in the subsurface. Level runs (figure to the left), where the source
and receiver are lowered down the borehole at the same depths, provide
quick, one-dimensional information about the horizontal velocity.
Vertical profiles, like VRPs and VSPs, provide information about the
vertical interval velocities. Single hole reflection studies are used
primarily in consolidated sediments or hard rock to detect fractures in
the subsurface. Tomography (figure to the right) provides a detailed,
two- or three-dimensional image of the subsurface, but is very time
consuming to acquire and process. At the BHRS, we have acquired VRPs,
level runs, and tomography data sets, The level runs provide quality
control and help calibrate origin times in the tomography data.
Velocities obtained from VRPs can help constrain tomographic
inversions. Tomographic data, although expensive to acquire, provides
detailed information about both vertical and lateral velocity changes
in the subsurface. |
Tomography receiver gather. The receiver
marked by the star and is located at 11.2 m depth. |
Tomography and Inversion |
||
|
Crosshole tomography provides a method to determine the velocity structure between two wells. CGISS uses a two-dimensional, finite-difference approximation to the eikonal equation to generate ray paths through the model. In the inversion, we use a weighted, damped, least squares approach to solve the ill-posed, ill-conditioned problem. We are investigating different ways to incorporate constraints on the problem to compute a more accurate subsurface velocity distribution. The most common method to constrain the solution invokes Occam's razor. Occam's razor states that in the absence of other information, use the simplest solution. For tomographic inversion, the procedure imposes the simplest structure on the model that fits the data. Thus smoothness constraints are used when inverting the data. Another approach that is gaining acceptance is Bayesian inversion. In the Bayesian case, the inversion routine incorporates a priori information to constrain the model. Bayesian inversion can update the inverted model as more information about the subsurface is discovered. Formal inversion methods enable an error analysis of the results. Unfortunately, most routines use a method that solves for the model directly, without computing the inverse solution. We are modifying inversion algorithms to compute or approximate the matrix inverse, so that we can assess the resolution and uncertainty in our models. Error analysis is a crucial aspect of inversion that warrants further research into both efficient error computation and interpretaion of error distribution in the modeled space. |
Site map showing the geometry of the well locations. |
|
Results and Analysis |
||
Velocity tomogram. The solid lines to the left and right of the wells are the neutron-derived porosity logs from the wells. The wells are deviated. |
The figure on the left (click to enlarge) is the tomogram for well pair C6 and C5 at the BHRS. The receiver is in well C6 (stars) and the transmitter is in C5 (circles). The wells are deviated from the vertical. I account for the deviation in the tomographic inversion. The color scale corresponds to the subsurface velocities. Above the water table, the velocities are off scale so that the variation in the saturated zone is easier to see. The solid curves to the left and right of the wells are the neutron-derived porosity logs. The plot to the right is the distribution of the travel time residuals for this inversion. The source depths are plotted vertically and the receiver depths are plotted horizontally. The largest residuals are near the water table, where the forward routine has difficulty accurately modeling the wave propagation. |
Travel time residuals plotted in relation to the appropriate transmitter-receiver pair. |
Rays traced to every 10th receiver showing the curved paths of the waves. |
To show how the rays travel through the model, I have plotted the ray paths for every tenth receiver location. Many inversion algorithms assume straight ray paths between sources and receivers. The left plot clearly shows that the rays are quite curved. The right figure plots the ray density at each grid cell in the model. Ray density is a measure of how many rays traverse each cell normalized by the cell's area. Larger ray density values indicate that more rays cross that cell. Often, we assume that larger ray densities imply more accurate model parameters at that cell. |
Ray densities are measure of the number of rays per grid cell. |
Diagonal elements of the resolution matrix. |
I have plotted the diagonal elements of the resolution (left) and the covariance (right) for the tomogram. Resolution measures how well the model locates the parameter. The resolution plot is very similar to the ray density plot. Cells with high resolution values correlate with cells with high ray density. This correspondence justifies using the ray density as a qualitative approximation for the model resolution. Covariance measures the uncertainty in the inverted parameter value. As expected, the covariance values are correlated with the resolution values. Large covariance values indicate highly resolved cells. As more rays traverse the cell, we are more assured of its position. However, more variance is associated with the increased number of rays, so the covariance increases. For those cells with only a few ray crossings, the variance will be small, but the resolution will be large and the value poorly constrained. |
Diagonal elements of the covariance matrix. |
We are also collaborating with researchers from Lawrence Berkeley Laboratories (Ernie Majer, John Peterson, Ken Williams, and Tom Daley) to develop crosshole radar and seismic tomography methods for imaging the zones between wells and estimating dielectric and seismic properties at the site. These material property distributions will then be compared to the porosity and permeability distributions derived from hydrologic well tests. Here are some examples of preliminary radar tomographic inversions performed by LBL for three wells in a triangular pattern.
The first panel is the 200 MHz radar tomogram for the region between wells B1 and B2. |
This panel is the 200 MHz tomogram for the region between wells B2 and C1. |
The third panel is the 200 MHz tomogram for the region between wells B1 and C1. |
Here is another tomogram for wells B1 and C1, but using 100 MHz antennas. Note the similarities and differences between this tomogram and the 200 MHz tomogram for the same well pair. |
For more information about this research, and preliminary results using the LBL high-frequency piezoelectric borehole seismic source at the BHRS, see the SAGEEP'99 paper by Peterson et al. below.
Return to the table of contents
Return to the top of the page
Return to the table
of contents
Return to the top
of
the page
Return to the table of contents
Return to the top of the page
Photo Gallery
Check out photos
of the BHRS and the people involved with the project.
Return to the table of contents
Return to the top of the page
This site has been accessed
times.
Last updated on February 24, 2005
Return to the table of contents
Return to the top of the page
Return to the CGISS
page
