Spatially Balanced Survey Designs

Spatially balanced sampling designs aim to spread sampling points uniformly across the spatial domain of interest.

The Generalized Random Tessellation Stratified (GRTS) sample selection algorithm was developed by Stevens and Olsen (2004) and has been used for regional monitoring programs, including several water quality programs funded by the EPA, Bonneville Power Administration, and Washington Department of Ecology, to name a few.

In 2012 the Integrated Status and Effectiveness Monitoring Program provided a free workshop on GRTS-based sampling designs and analysis, which was led by Don Stevens, one of the developers of the GRTS algorithm.

GRTS Workshop Training Materials (August 22-23 2012, Portland, OR)

Instructor: Dr. Don Stevens

Red Lion Hotel-Portland Convention Center

1021 NE Grande Ave.

Portland, OR 97232

Background Literature

Several key manuscripts are located in GRTS Workshop 2012/BackgroundArticles

DAY 1: Introduction to GRTS and Study Design Development (8:30 am-5pm)

Sections of presentations in GRTS Workshop 2012/BackgroundPresentations supported Day 1 topics. We recommend viewing these background presentations in addition to any ISEMP-specific presentations that are listed in the agenda. All presentations in the Background Presentations folder were generated by Don Stevens.

Section 1. Overview of GRTS (8:30-10am)

History and origin of GRTS and overview: *GRTS Workshop 2012/Day 1/GRTS_grts_ISEMP.pptx

Stratified and variable probability sampling

Inclusion probabilities

Site evaluation

Weight adjustments

Section 2. Designing a GRTS sample for stream/river networks (10-12am)

Introduction to stream network sampling as continuous, discrete or area based samples and the implications of how a response design is set up in these various scenarios

Limitations to application of GRTS along linear networks: GRTS Workshop 2012/Day 1/Sampling a Stream Network.pptx

Specifying objectives: What do you want to estimate? How does design tie to objectives?

Function of the response design and relationship to spatial/temporal design

Spatial design: target population, frames, stratification

Temporal patterns—panel designs

Sampling details: legacy incorporation, densifying the master sample, power analyses GRTS Workshop 2012/Day 1/Merging Legacy Sites with a GRTS Sample.pptx

Extract sample from a master sample

Generating initial weights, probabilities and densities

Documenting the design and generating a design file: what should it contain?

Lunch (on your own, 12-1pm)

Section 3: Analysis (design based inferences) (1-5pm)

Now that I’ve got my files together, site evaluations, and site metrics, how do I make inferences based on my design, and what do I get (theoretical meaning of results)?

See GRTS Workshop 2012/BackgroundPresentations Olsen_Day 3 1. Spatial Survey Analysis.pptx

Distinguish design based inference from model based inference

Develop adjusted weights (how is this done, what is considered)

Categorical vs continuous estimates

Estimates and indicators: totals (e.g, how many fish do I have?), densities, frequency distributions

Sampling uncertainty and precision of estimates

Analyzing data using post-hoc stratification (how does this work?)

---

DAY 2: GRTS functions in R (8:30am-12pm)

Section 5: Description and demonstration of functions available in spsurvey library or recent scripts developed by D. Stevens (e.g. CHaMP scripts) (8:30-12pm)

See GRTS Workshop 2012/Day 2/Nahorniak_GRTS Workshop Presentation.pptx and Starcevish_082312.ppt

GRTS function for linear resources:

Adjwgt function

Cont.analysis function

Cat.analysis function

Don’s specific functions (non-spsurvey version of GRTS code)

Use of GRTS for temporal patterns

Other:

Generating a Master Sample (spsurvey)

Development of web based tool for extraction of samples from master samples: what will it contain and what is its status? GRTS Workshop 2012/Day 2/Wep Tool for.pptx