Floating Percentile Model and supporting functions to inform and improve sediment benchmark development
RFPM()bibentry
'RFPM' is an open-source implementation of the floating percentile model (FPM), which was originally developed by Avocet (2003) using Visual Basic for Applications and distributed to US Pacific Northwest regulatory agencies as a Microsoft Excel-based tool. 'RFPM' was developed independently by Claire Detering and Brian Church with support from John Toll and others at Windward Environmental LLC.
The purpose of the FPM is to generate aquatic toxicity-based sediment quality benchmarks for management of contaminated freshwater sediment sites. These benchmarks are intended to act as classification thresholds, meaning that an exceedance of benchmarks would imply that toxicity (as categorically defined) is likely in the sediment sample. The FPM has been used at sites in the US Pacific Northwest for many years, particularly after being published by the Washington State Department of Ecology in 2011.
The primary function in 'RFPM' is FPM, which runs the FPM algorithm on a data.frame object that includes concentrations of
chemicals in sediment as well as a logical toxicity classification column called "Hit". Example datasets are provided. The output of FPM
includes a set of sediment quality benchmarks for chemicals with significantly higher concentrations when Hit == TRUE
than when Hit == FALSE. Plots comparing the Hit == TRUE and Hit == FALSE data can also
be generated as a diagnostic tool. Supplemental functions (e.g., optimFPM) can help to optimize FPM inputs (resulting in more accurate benchmarks) or evaluate
the relative importance of each chemical among the FPM benchmarks (i.e., chemVI).
For 'RFPM', the FPM algorithm has been changed from the original Avocet (2003) model. Key changes are as follows:
A decision tree is implemented to select statistically appropriate hypothesis tests for chemical selection. This can be overridden if the
original Excel-based tool method is desired; see ?chemSig for details. The chemical selection and FPM algorithm have been integrated into FPM, though chemical selection can still be run separately, if desired.
The iterative looping of the FPM algorithm over multiple false negative limits was
not included in FPM. We find this functionality of the Excel-based tool to be confusing and unnecessary.
Instead, we believe the results generated for different false negative limits should be independently generated rather than dependent on prior model runs.
Thus, FPM allows for multiple false negative limits to be input in a seqeuential but independent manner, resulting in a data.frame of benchmarks with one
row per false negative limit. In R terminology, FPM has been vectorized.
Functions were developed to optimize the overall reliability of FPM sediment quality benchmarks. Optimization can either maximize the overall
reliability of the benchmarks, or it can reduce the difference between false positive and false negative predictions of toxicity. The latter
might be useful as a compromize between over- and under-conservatism of benchmarks.
The functions are optimFPM and cvFPM. Both functions can be used
to optimize the false negative limit, and optimFPM can also be used to optimize the type-I error rate (alpha), which drives the chemical selection
aspect of FPM. optimFPM and cvFPM also provide graphical representations of the optimization process,
which helps to provide context for false negative limit and alpha selections.
New functions are also available to quickly calculate chemical "variable importance" statistics using chemVI. These statistics
can inform the user about the influence of specific chemicals over the set of FPM benchmarks and, for example,
whether certain benchmarks can be ignored without a significant loss of predictive ability.
Avocet. 2003. Development of freshwater sediment quality values for use in Washington State. Phase II report: Development and recommendation of SQVs for freshwater sediments in Washington State. Publication No. 03-09-088. Prepared for Washington Department of Ecology. Avocet Consulting, Kenmore, WA. Ecology. 2011. Development of benthic SQVs for freshwater sediments in Washington, Oregon, and Idaho. Publication no. 11-09-054. Toxics Cleanup Program, Washington State Department of Ecology, Olympia, WA.