The survey design is a plan for selecting the sample appropriately so that it provides valid data for developing accurate estimates for the entire population or area of interest.  At this point in the design process it is assumed that the folloing have been completed:

1. Established the monitoring objectives
2. Identified Resource Characteristics
3. Identified Institutional constraints requiring consideration
4. Defined the target population
5. Specified Subpopulations, Domains and Stratification
6. Created a sampling frame for all units of the target population from which to select the sample
7. Selected a GRTS survey design for the selection of sample of units from this frame
8. Selected a strategy for allocation of sampling efforts in space and time
9. Established sample size(s)

Additional information on each of these items is available on Design Specifics

Luckiamute Watershed Stream Example

Purpose: Example linear GRTS survey design for Luckiamute Watershed streams

This information is presented as an example and recommended format for documenting a GRTS design. It is recommended that this example be followed, prior to initiating efforts to produce a design to meet a particular objective. Due to the randomization scheme within psurvey.design, each execution of the R code will produce a different set of sampling points and output files and will not exactly reproduce the example files.

Materials needed for these examples:

• R version 2.2.1
• psurvey.design package (Version 2.2.1)
• R script or text file: luckiamutedesign.R
• Frame materials as ArcGis files:
• luck-ash.dbf
• luck-ash.prj
• luck-ash.sbn
• luck-ash.sbx
• luck-ash.shp
• luck-ash.shx
• Download Example Files

Examples includes:

An ArcGis map is included, Luckiamute Design.pmf, for the resulting designs for this example (requires ArcReader software to view, explore and print).

Design File Documentation would routinely begin here:

Contact:

Description of Sample Design

Survey Design: A Generalized Random Tessellation Stratified (GRTS) survey design for a linear stream resource is used for this example. The GRTS design includes reverse hierarchical ordering of the selected sites.

Target population: All perennial and intermittent streams/rivers in the Luckiamute Watershed Council basin. Watershed boundaries defined by Luckiamute Watershed Council.

Sample Frame: GIS stream network coverage from PNW portion of RF3. Watershed Map

Sample Frame Summary

The total stream length (km) is 822.4882 km (including intermittent).

Multi-density categories: Strahler order categories: 1 st, 2 nd, and 3 rd+.

  Go To Top

I. Equal Probability GRTS Survey Design

Stratification: None

Expected sample size: 50 sites

Oversample: None

Site Use: The base design has 50 sites. Sites are listed in SiteID order and must be used in that order. All sites that occur prior to the last site used must have been evaluated for use and then either sampled or reason documented why that site was not used.

Site Selection Summary

Map of Equal Probability Sites

  Go To Top

II. Stratified GRTS Survey Design with Over Sample

Stratification: Stratify by Perennial and Intermittent streams .

Expected sample size: 50 sites for Perennial stratum and 50 sites for Intermittent stratum, plus in each stratum, sites selected with equal probability within stratum.

Oversample: 50 sites for Perennial stratum and 50 sites for Intermittent stratum.

Site Use: The stratified design has 50 sites, plus an over sample of 50 sites, in each stratum. Sites are listed in SiteID order and must be used in that order. All sites that occur prior to the last site used must have been evaluated for use and then either sampled or reason documented why that site was not used.

Site Selection Summary

Map of Stratified Sites

Go To Top

III. Unequal Probability GRTS Survey Design with Over Sample and Stratification

Stratification: Stratify by Perennial and Intermittent streams .

Expected sample size: 75 sites with approximately an equal number within each Strahler order category for Perennial stratum. For Intermittent stratum, 25 sites with an expected sample size of 17, 5, and 3 for Strahler order categories 1 st, 2 nd, and 3 rd+, respectively.

Oversample: 36 sites for Perennial stratum and none for Intermittent stratum.

Site Use: The unequal probability design has 75 sites, plus an over sample of 36 sites, for the Perennial stratum and 25 sites for the Intermittent stratum. Sites are listed in SiteID order and must be used in that order. All sites that occur prior to the last site used must have been evaluated for use and then either sampled or reason documented why that site was not used.

Site Selection Summary

Map of Unequal Sites

  Go To Top

IV. Panels for Surveys Over Time with Unequal Probability GRTS Design with Over Sample and Stratification.

Stratification: Stratify by Perennial and Intermittent streams .

Panels: For Perennial stratum, three panels. Year1 sites will be visited in year 1 and then year 3, year 5, etc. Year2 sites will be visited in year 2, year 4, year 6, etc. YearAll sites will be visited every year. Intermittent stratum has a single panel of sites that will be visited only once.

Expected sample size: 50 sites with approximately an equal number within each Strahler order category for Perennial stratum. For Intermittent stratum, 25 sites with an expected sample size of 17, 5, and 3 for Strahler order categories 1 st, 2 nd, and 3 rd+, respectively.

Oversample: 50 sites for Perennial stratum and none for Intermittent stratum.

Site Use: The base design has 50 sites in Perennial stratum. Sites are listed in SiteID order and must be used in that order. All sites that occur prior to the last site used must have been evaluated for use and then either sampled or reason documented why that site was not used. As an example, if 30 sites are to be sampled in the watershed, then the first 30 sites in SiteID order would be used.

Site Selection Summary

Perennial Sites

Intermittent Sites

Map of Panel Sites

  Go To Top

Description of Sample Design Output:

The sites are provided as a shapefile that can be read directly by ArcMap. The dbf file associated with the shapefile may be read by Excel.

The dbf file has the following variable definitions:

Luckiamute_Streams is an ArcMap that displays the stream network and sampling sites for the 4 example designs.

Projection Information

PROJCS["Clarke_1866_Albers"
GEOGCS["GCS_Clarke_1866",
DATUM["D_Clarke_1866",
SPHEROID["Clarke_1866",6378206.4,294.9786982]],
PRIMEM[" Greenwich",0.0],
UNIT["Degree",0.0174532925199433]],
PROJECTION["Albers"],
PARAMETER["False_Easting",0.0],
PARAMETER["False_Northing",0.0],
PARAMETER["Central_Meridian",-96.0],
PARAMETER["Standard_Parallel_1",29.5],
PARAMETER["Standard_Parallel_2",45.5],
PARAMETER["Latitude_Of_Origin",23.0],
UNIT["Meter",1.0]]

  Go To Top

Evaluation Process

The survey design weights that are given in the design file assume that the survey design is implemented as designed. That is, only the sites that are in the base sample (not in the over sample) are used, and all of the base sites are used. This may not occur due to (1) sites not being a member of the target population, (2) landowners deny access to a site, (3) a site is physically inaccessible (safety reasons), or (4) site not sampled for other reasons. Typically, users prefer to replace sites that can not be sampled with other sites to achieve the sample size planned. The site replacement process is described above. When sites are replaced, the survey design weights are no longer correct and must be adjusted. The weight adjustment requires knowing what happened to each site in the base design and the over sample sites. EvalStatus is initially set to "NotEval" to indicate that the site has yet to be evaluated for sampling. When a site is evaluated for sampling, then the EvalStatus for the site must be changed. Recommended codes are:

Statistical Analysis

Any statistical analysis of data must incorporate information about the monitoring survey design. In particular, when estimates of characteristics for the entire target population are computed, the statistical analysis must account for any stratification or unequal probability selection in the design. Procedures for doing this are available from the Aquatic Resource Monitoring web page given in the bibliography. A statistical analysis library of functions is available from the Aquatic Resource Monitoring web page to do common population estimates in the statistical software environments R and S-PLUS (psurvey.analysis)

  Go To Top

For further information, contact

Anthony (Tony) R. Olsen
USEPA NHEERL
Western Ecology Division
200 S.W. 35th Street
Corvallis , OR 97333
Voice: (541) 754-4790
Fax: (541) 754-4716
email: Olsen.Tony@epa.gov

Bibliography:

Diaz-Ramos, S., D. L. Stevens, Jr, and A. R. Olsen. 1996. EMAP Statistical Methods Manual. EPA/620/R-96/002, U.S. Environmental Protection Agency, Office of Research and Development, NHEERL-Western Ecology Division, Corvallis, Oregon.

Stevens, D.L., Jr. 1997. Variable density grid-based sampling designs for continuous spatial populations. Environmetrics, 8:167-95.

Stevens, D.L., Jr. and Olsen, A.R. 1999. Spatially restricted surveys over time for aquatic resources. Journal of Agricultural, Biological, and Environmental Statistics, 4:415-428

Stevens, D. L., Jr., and A. R. Olsen. 2003. Variance estimation for spatially balanced samples of environmental resources. Environmetrics 14:593-610.

Stevens, D. L., Jr., and A. R. Olsen. 2004. Spatially-balanced sampling of natural resources in the presence of frame imperfections. Journal of American Statistical Association:99:262-278.

  Go To Top