The Sampling Design Tool for ArcGIS randomly selects people, animals, objects, or places in a GIS environment to infer status or change. The tool was developed to create and implement a sampling design based on randomized sampling and test different sampling designs to see which is most precise and least costly. For example, the tool has been used to identify stratified random field sites for measuring fish and coral in support of the National Coral Reef Monitoring Plan.
What We Did
The Sampling Design Tool was developed in response to a need by scientists developing sampling strategies in marine environments with limited data. The tool was produced as part of an iterative process of sampling design development, whereby existing data inform new design decisions. The objective of this process, and hence a product of this tool, is an optimal sampling design that can be used to achieve accurate, high-precision estimates of population metrics at a minimum of cost. Although NCCOS has used the program for marine habitats and some examples in the tool’s associated documents reflect this, the tool can be used to sample any type of population defined in space, be it coral reefs or corn fields.
The Sampling Design Tool has two main functions: 1) to help select a sample from a population, and 2) to perform sample design analysis. When both of these functions are combined in an iterative manner, the tool effectively and simply achieves the goal of sample surveys—to obtain accurate, high-precision estimates of population metrics at a minimum of cost.
Key Features of the Tool
- Spatial sampling: Sampling and incorporation of inherently spatial layers (e.g., benthic habitat maps, administrative boundaries), and evaluation of spatial issues (e.g., protected area effectiveness).
- Scalable data requirements: Data requirements for sample selection can be as simple as a single polygon outlining the survey area, or as complex as a stratified sample frame with existing sample data and a stratified sample frame for optimally allocating samples.
- Random selection: Eliminates sampling biases and corresponding criticisms encountered when samples are selected non-randomly. A random seed number can be entered in the ArcGIS 10.X version.
- Multiple sampling designs: Simple, stratified, and two-stage sampling designs.
- Sample unit-based sampling: Points or polygons are selected from a sample frame.
- Area-based sampling: Random points are generated within a polygon.
- Analysis: Previously collected data can be used to compute sample size requirements or efficiently allocate samples among strata.
- Computations: Mean, standard error, and confidence intervals for sample data and inferences of population parameters with known certainty.
- Output: Geographic positions in output simplifies migration to global positioning systems, and sample size estimates and sample statistics can be exported to text files for record keeping.
Software Requirements: There are two versions of this tool: one for ESRI’s ArcGIS 9.2 Service Pack 6 or higher, and another for ESRI’s ArcGIS 10.0 or higher.
Projects That Have Used the Tool
- Research to Support Management of the Flower Garden Banks National Marine Sanctuary
- Caribbean Coral Reef Ecosystem Monitoring Project
- The State of Coral Reef Ecosystems of the United States and Pacific Freely Associated States
- National Coral Reef Monitoring Plan Implementation: Biological and Socioeconomic Monitoring
Reviews of the Tool
- “I am using your tool to randomly select vegetation training sites for input to dataminers such as RandomForests and SVM via Weka to model vegetation maps. I just finished a project where I mapped a 60,000 acre refuge in Oklahoma using 2000 training sites and achieved accuracies that ranged from 83% to 100%! And random sampling is key to getting these kinds of accuracies. So your tool is key to the work that I am doing. “ – National Wildlife Refuge System
- “We have used the NOAA tool to create stratified, random ground-truthing points for plant community mapping in the St. Johns River Basin. Many packages have the ability to do simple random points, be we need the ability to stratify our points by vegetation categories.” – St. Johns River Water Management District
- “I have used it to generate the sampling design for the Florida Reef Resilience Program Disturbance Response Monitoring that takes place every summer. I realize that there is a new ‘Create Random Points’ tool in ArcGIS 10, but the Sampling Design Tool is much more flexible. “ – The Nature Conservancy