Models were used in the manuscript "A novel design method for customized visual delimiting surveys for plant pests based on transects and scouting," by Barney P. Caton, Godshen R. Pallipparambil, and Hui Fang. This paper describes a novel approach for designing custom visual delimitation surveys, called ‘Delimitation via Transect Data and Scouting,’ or DTDS.
To evaluate the methodology, we identified published delimitation survey plans or results of activities for six species. One is a plan for a pest plant (or weeds) [which has no corresponding model]; the other five species have host plants. Using available information, we created customized DTDS plans for each species, and then compared the sampling effort required to complete the original and novel plans, for localized populations. We used simulations of each case study to compare the results under uncertainty and to evaluate outcomes when mapped spatially. See text for more information about published plans.
Overview. Models were built to emulate the survey conditions (e.g., areas, host densities, infestation rates) and survey plan specifications (areas and hosts inspected). Outputs were number of infected/infested hosts detected, by plan or scenario. We also estimated the inspection time required per host and the total time taken, again by plan or scenario.
Functions and parameters. Because of the number of models and survey plans created for evaluations, we cannot exhaustively present the functions or parameters used. Parameters came from the source or were standardized: only survey specifications differed in simulations, not situational details. For parameters with a single, mean estimate (e.g., trees per km2), in every case we added uncertainty by using lower and upper values that were ten percent different from the mean, and used a uniform distribution to sample values (i.e., every value equally likely). For example, if authors estimated 10,000 hosts per km2, the lower limit was 9,000 and the upper limit was 11,000.
Nearly all functions used were basic arithmetic, such as calculating infestation densities (no. per unit area), area sizes (e.g., π × R2), or widths. One exception was the binomial process. In this process, n independent, identical trials are run, each one with the same probability of success, p, producing some number of successes, s (Vose 2000): s = RiskBinomial(N, p).
This function was used, for example, to find the number of infested cells or hosts detected, where N was the number inspected and p was the infestation rate. We assumed perfect detection (i.e., sensitivity = 1.0) for simplicity.
General specifications. The models were all coded in spreadsheets and run using @Risk ver. 7.5.1 Professional Edition (Palisade Corporation, 31 Decker Road, Newfield, NY 14867), a Microsoft Excel add-in. Unless otherwise specified below, simulation settings were as follows: number of iterations = 100,000; sampling type = Latin Hypercube; and random seed = 101.
See the README for descriptions of each data file.
|Release Date|| |
|Spatial / Geographical Coverage Area|| |
POLYGON ((-168.3984375 -45.236217535866, -168.3984375 82.089708031507, 190.8984375 82.089708031507, 190.8984375 -45.236217535866))
Ag Data Commons
|Spatial / Geographical Coverage Location|| |
|Temporal Coverage|| |
January 11, 2023
|Contact Name|| |
Caton, Barney P.
|Public Access Level|| |
|Program Code|| |
005:051 - Department of Agriculture - Safeguarding and Emergency Preparedness/Response
|Bureau Code|| |
005:32 - Animal and Plant Health Inspection Service