Here are examples of questions this tool can help to answer, types of results it can give, and other use cases.
Given known inspection protocols (sample size and selection method) and records containing the inspection outcomes, what is the approximate contamination rate of the inspected consignments? By simulating inspections using the same protocol and calibrating the contamination parameters so that the simulated inspection outcomes match the actual inspection outcomes, we can estimate the contamination rate present in the actual consignments.
Given assumed contamination rates and consignment sizes, what percentage of contaminates are not being intercepted by border inspections when using various inspection methods? This information could be useful for estimating propagule pressure for quarantine-significant pest species.
For example, is the inspection success rate higher when we inspect fewer consignments using hypergeometric random sampling, or when we inspect all consignments using convenience sampling?
Given a specific inspection protocol, what level of contamination needs to be present in the consignments for us to detect it? Additionally, how much pest needs to be present to raise alarms?
In the following example, we inspect two boxes of each consignment, each containing between 1 and 50 boxes. We run the simulation with different contamination rates and compare slippage rates.
| Contaminated boxes | Missed |
|---|---|
| 90% | 1% |
| 80% | 4% |
| 70% | 9% |
| 60% | 16% |
| 50% | 24% |
| 40% | 34% |
| 30% | 47% |
| 20% | 61% |
| 10% | 77% |
Using a given inspection protocol, would we successfully intercept a newly emerging pest? In this case, we would modify the parameters that control how contamination in consignments from certain origin countries are added based on another model projecting an emerging pest.
Datasets like this one can be generated in case synthetic data with certain properties or without privacy issues are needed:
| Date | Port | Origin | Flower | Action |
|---|---|---|---|---|
| 1 | RDU | Estonia | Gloriosa | RELEASE |
| 1 | Miami | Hawaii | Gladiolus | RELEASE |
| 2 | RDU | Argentina | Actinidia | RELEASE |
| 3 | Miami | Argentina | Gladiolus | RELEASE |
| 3 | RDU | Hawaii | Ananas | RELEASE |
| 4 | Miami | Hawaii | Acer | RELEASE |
| 5 | Miami | Taiwan | Gladiolus | PROHIBIT |
| 5 | RDU | Estonia | Aegilops | RELEASE |
Example workflow is included in Obtaining synthetic F280 records.
- Jupyter Notebooks, Python scripts, and Bash scripts included in the repository
- Obtaining synthetic F280 records
- Outputs
Next: Consignments