Soil roughness sampling results

Soil roughness sampling results

Raw Data Images Soil Properties Page

The Data
Characteristics

The Files
Format
Name and Directory Information

The Science
Plan
Procedures
Data Processing

Data Access and Contacts
FTP Site
Points of Contact

References

The Data

The surface roughness sampling results include a summary table containing rms surface (root-mean-square) height values, correlation lengths, and the powers of the autocorrelation functions; photographs of the sampled sites; and the raw, digitized data. The original photographs are available for use at the USDA ARS Hydrology Lab. They will not, however, be loaned out. Also included are brief descriptions of the sampling plan and procedures, the data processing methods, and the output parameters.
Characteristics

Soil Roughness Data Parameters
Type Parameter Units Description

Raw data raw rms cm root-mean-square surface height

adjusted rms cm adjusted root-mean-square surface height

length cm correlation length

factor power of autocorrelation function

Image Roughness NA scanned photographs of the roughness grid at the sampling sites

The Files

Format

Type Format Size/file Storage required

raw data ASCII text 1-2KB 277KB

image GIF < 1.4MB ~100MB

summary ASCII text 7839 Bytes 7839 Bytes

Name and Directory Information

Naming Convention for Data and Image Files

With the exception of the summary files, all of the data files are named according to area, field number, and sampling number. The area is represented by the first two characters (cf = Central Facility, er = El Reno, lw = Little Washita) of the filename. The third and fourth characters denote the field number, and the sixth and seventh characters denote the sampling number. The fifth character is a dash, used to separate the field number and the sampling number. The extension is assigned according to file type. For example, the digitized data for sample #2 in field 17 of Little Washita may be found in the file lw17-02.txt. The image file for the same site may be found in the file lw17-02.gif.
The raw data files contain the digitized data from the roughness grid at each sample location. The files are essentially sets of coordinates that record the change in elevation of the soil surface. All of the files have two columns. The first column represents the horizontal distance along the surface, and the second column represents the vertical surface elevation with reference to some arbitrary zero. The raw data files are available in ASCII text format.
Naming Convention for Summary File

The summary file is available in ASCII text format. It is a table that lists the sampling date, number of observations, surface-roughness value, correlation length, the power of the autocorrelation function, and the raw data file and image file associated with each sample. The summary file is named summary.txt.
Directory path
The directory path to the Soil Roughness data is
http://disc.sci.gsfc.nasa.gov/data/sgp97/soil_properties/soil_roughness

The Science
Plan
The sampling plan and procedures are adopted from the Southern Great Plains 1997 (SGP97) Hydrology Experiment Plan and from Lynn McKee (personal communication, Jan 8,1998).
Procedures
Surface roughness was recorded using a photograph of a grid board. The board was 0.5m high and 1.0m long. Each square on the board was 2cm by 2cm. A maximum of four surface-roughness samples was taken per field. The sampling strategy us ed was to obtain well-distributed, random samples representative of the field. Sampling sites also were chosen for minimum disturbance, i.e., no traffic (vehicle or human), no equipment/instrumentation, no holes, and no animal droppings.
The data from the surface-roughness samples were used to describe the height variation of the soil surface. The results for each sample may be summarized by the rms surface roughness (a standard deviation), the correlation length, and the (power of the) autocorrelation function.
Because surface roughness-samples were collected at the same time as bulk-density samples, the task was best accomplished by teams working in pairs. For surface roughness, one person took the picture while the other supported the board and held the label.
The method used to collect surface-roughness samples is given below:

Surface preparation. Once a suitable sample site was chosen, a rectangular area of about 1ft by 1m of grass was clipped to a few centimeters high. If no grass was present, clipping was not necessary.
Grid board placement. The grid board was placed lengthwise at one of the long sides of the clipped-grass rectangle. (The 1m edge of the board coincided with the 1m edge of the clipped-grass rectangle.) The board's gridded side faced the clipped grass.
Grid board anchorage. A hammer was used to pound the board into the ground so that the board stood evenly and perpendicular to the ground. The bottom of the board was "anchored" securely such that no light nor gaps were visible between the board and the ground.
Sample Labeling. A label was made for each sample according to area, field number, and sampling number. Like the naming scheme for the data files, the field number and the sampling number were separated by a dash.
Sample recording. The label was held against the upper-right corner of the board. If necessary, the board was supported by the person holding the label. Two (color) pictures were taken of each sample.

Data Processing
The data processing consisted of several steps. First, the photographs were scanned and digitized to create the roughness-profile files. Next, rms values and correlation lengths were computed from these roughness-profile files. Then, the autocorrelation function of each surface profile was fitted with a general function rho(displacement) = exp(-(displacement/correlation length)^n) It is an exponential function when n = 1, and becomes a Gaussian function when n = 2.
Roughness Profile
The pictures were developed into 4" by 6" color photographs. The photographs were digitized using a software package called SigmaScanPro. This software allows digitization of a scanned image. The following procedures were used to create the roughness-profile data:

Picture scan. The photographs were all scanned using a MicroTek ScanMaker II XE Scanner.
Measurement selection. Due to the "coordinate" nature of the data, the measurement types chosen were x and y data (x represents the distance along the ground and y the vertical height of the ground at that distance). The axes were defined and calibrated using the gridded surface roughness board in the photograph. (Each square represented a 2cm by 2cm area.)
Data collection. Points were taken at about 1cm intervals. For particularly difficult samples (surface obscured by shadow or vegetation), the surface elevation was approximated.
Data conversion. The x and y data were copied into a spreadsheet and saved as ASCII text files. Filenames were assigned using the previously mentioned naming scheme.

Output Parameters
The output parameters are rms surface height, adjusted rms surface height, correlation length, and the power of autocorrelation function. They are presented in table format in the file summary.txt.
RMS Surface Height
The root-mean-square (rms) surface roughness describes the variation in surface elevation. It is also known as the standard deviation of the surface height. The FORTRAN program used to compute the rms surface height was called surface.f. This program was originally written by D.S. Lin of Princeton University, and later modified by A. Hsu for the SGP97 experiment. The equation for this parameter is given as
rms_surface_roughness =((sigma[(z_k)^2] - n(z_ave)^2)/(n-1))^0.5
where
z_k is the surface elevation at distance x_k, or the kth elevation in the text data file;
sigma[]is a summation function, in this case, the function is the sum of (z_k)^2;
z_ave is the average elevation;
n is the number of observations taken per sample.
The equation is valid for random surface components with respect to a flat or horizontal mean (reference) surface.

Adjusted Surface-RMS Height
The adjusted value of rms surface height takes into account a sloped terrain. The above-mentioned Fortran program applied a least-square fit (i.e., a linear regression) to the digitized profile data to remove slope bias, and recorded the corrected (height) data. (The corrected height is the difference between the original observation at some distance x and the estimate of the linear regression at that same distance.) The rms (adjusted) surface height was calculated from the corrected profile data in the same way as the rms surface height was calculated from the original profile data.
Correlation Length (Adapted from Ulaby, et al, 1982)
In addition to the rms surface height, the profile of a random surface may be characterized by its autocorrelation function, which describes the similarity between the height (z) of the surface at some distance x along the surface. For the discrete case, the autocorrelation function normalized by height is given by Ulaby et al., 1982 (see Eqn. 11.11).
As the displacement increases, the autocorrelation function decreases toward zero. This behavior indicates that as the horizontal distance between two surface points increases, the correlation between the heights of those two surface points decreases. The maximum distance at which a significant correlation occurs is known as the correlation length. It is defined as the displacement for which the autocorrelation function is equal to 1/e (~0.36788).
The following is the segment in surface.f that computes correlation length:
     sum4 = 0.0
     do k = 1, np
     	sum4 = sum4+(z(k) - zbar)*(z(k) - zbar)
     enddo
     threshold = 0.367879441

     do j = 1, np
	xprime(j) = j*DX
     	sum3 = 0.0
     	do i = 1, np - j
          	sum3 = sum3+(z(i) - zbar)*(z(i+j) - zbar)
     	enddo
     	rho(j) = sum3/sum4
     	if (rho(j) .le.threshold) go to 9999
     enddo

9999 L = xprime(j) - DX/(rho(j - 1) - rho(j))*threshold - rho(j))
     write (*, 100) f_name, no, sigma, L
100  format (1X, a40, 5x, i3, 9x, f5.2, 10x, f5.2)
Power of the Autocorrelation Function (Adapted from Ulaby et al, 1982 and Oh et al, 1992)
The shape of an autocorrelation function ranges between an exponential function and a Gaussian function. The exponential and Gaussian functions are given by Oh et al., 1992 (see Eqns. 2 and 3). In general, a rough surface tends to have an autocorrelation function close to the Gaussian function whereas a smooth surface tends close to the exponential function. A procedure written in IDL was used to plot the experimental data and the exponential and Gaussian function curves. Curves with power between 1.0 and 2.0 were also plotted to visually determine which power of the general function most closely matched the autocorrelation function of the experimental data.
Data Access and Contacts

FTP Site
The Soil Roughness data set from SGP97 is in the following GES DISC ftp site:
Raw Data ( ASCII) Images (GIF) Summary(ASCII)

Points of Contact
The Principal Investigator for the SGP97 Soil Roughness data is

Thomas J. Jackson
USDA ARS Hydrology Lab
Bldg. 007, Rm. 104, BARC-West
Beltsville, MD 20705

E-mail: tjackson@hydrolab.arsusda .gov
Voice: 301-504-8511

For more information regarding the SGP97 data from GES DISC, contact:
Hydrology Data Support Team
Goddard Earth Sciences

Data and Information Services Center (GES DISC)
Code 610.2
NASA Goddard Space Flight Center
Greenbelt, Maryland 20771

E-mail: hydrology-disc@listserv.gsfc.nasa.gov
Voice: 301-614-5165
Fax: 301-614-5268

References

Lin, D.S., modified by A. Hsu. Nov 13, 1997. surface_all.
Oh, et. al. 1992. Empirical models and inversion techniques. IEEE Transactions on Geoscience and Remote Sensing. March 1992. Vol 30. No 2. pp 372-373.
Ulaby, F.T., Moore, R.K., and A.K. Fung. 1982. Physical mechanisms and empirical methods for scattering and emission. Microwave Remote Sensing: Active and Passive. Vol. 2. Chapter 11.

Last update:Thu Jan 28 09:34:12 EST 1999
Page Info: Hydrology Data Support Team -- hydrology-disc@listserv.gsfc.nasa.gov
Web Curator: Stephen W Berrick
NASA official: Steve Kempler, DAAC Manager -- Steven.J.Kempler@nasa.gov