This application is for selecting oyster reef sample sites. Sites are chosen using the Generalized Random Tessellation Stratified (GRTS) algorithm, which is a de facto standard used by the EPA. Site selection is done randomly to avoid statistical bias and to reflect the spatial arrangement of available sites. Because we have data at some sites, we can use a model to predict parameters at unsampled sites. The GRTS algorithm uses uncertainties in the prediction model to guide site selection – those sites with greater prediction uncertainty are more likely to appear in the selected sites. Selections use prediction uncertainties for three parameters: shell height, percent live and density.

There are map layers for bathymetry (depth in meters) provided by the National Oceanic and Atmospheric Administration (NOAA) National Centers for Environmental Information (NCEI), existing oyster reefs compiled by the Fish and Wildlife Conservation Commision (FWC), prior sample sites, and prediction uncertainties for shell height, percent live and density. The larger the points on the map, the greater the uncertainty and the more likely they will appear in the GRTS generated sample sites. GRTS sample sites can be displayed after they have been generated by entering the desired number of sites.

How to use this application. Inspect the graphs in the center panels. The one on the left shows improvement as a function of sample size within a reef. The one on the right guides the number of sites/reefs to select. Follow the smoothed curves until you reach your desired standard deviation (y axis value), then the corresponding x axis value represents the estimated number of samples/sites needed to achieve that desired error rate. Please note that the number of samples/sites to select is based on simulations and is only approximate. Generally, after the curves flatten, there is little estimated improvement in the error. If you must pick one sample size/number of sites to use across all three parameters, best practice is to choose the maximum result from among the three curves. If the number of prior sites is small or there are few unsampled oyster reefs in the managed area, the curves might not decrease for one or more of the parameters. This pattern is indicative of statistical estimation issues, so it is recommended to only consider the parameter curves that show the expected pattern for sampling guidance.

1. If there are sites that should be removed from consideration, enter those in a .csv file with OBJECTID, lon and lat as headers. You can open a dialog box to select your file using the “Browse” button in panel 1.

2. You can also select ranges of depths or reef areas to include using the buttons in panels 2a and 2b. Beware that too stringent selection criteria could drastically reduce or even eliminate remaining sites.

3. Click on the button in panel 3 and drag the slider to the number of sites you would like to generate. Clicking “Submit” will close the dialog and compute the sites. Choosing more than you will need can be helpful, in case some sites turn out to be inaccessible. If you need to change the number of sites after already generating a set, do not just generate the remainder as a second set; instead, re-generate the entire set of sites using the new total number.

4. Check the box next to GRTS sample sites in the layer selection panel to view your sites on the map.

5. Use panel 5 at the bottom left to export the GRTS sample sites to one of three formats: csv, kml or gpx. The file has OBJECTID, lon and lat. The order of the reefs is randomized so if you need to eliminate reefs you can remove them from the top or bottom of the list.