TABLE 3:

Factors for consideration in the design of experiments for validation of bioavailability models

Factors to consider	Comments	Caveats
Experimental design	• Capture full concentration–response range to calculate toxicity estimates • Use appropriate number of treatment replicates (Green 1979; ASTM, OECD, USEPA standardized methods) • Consider positive and negative controls (ASTM, OECD, USEPA standardized methods) • Replicate experiment over time to demonstrate reproducibility • Stipulate test acceptability criteria (ASTM, OECD, USEPA standardized methods) • Use accepted and defined statistical methods (Green 1979; ASTM, OECD, USEPA standardized methods)	• The number of treatments can be determined based on the type of study that has been chosen (e.g., field versus lab) and based on the resources available • Regardless of the model, predicted ECx should optimally be the midrange of the test concentrations used • Test conditions should be tightly controlled • Test acceptability criteria exist for well-established protocols (e.g., Organisation for Economic Co-operation and Development 2004; Elonen 2018), but caution must be taken when testing with novel species for which such an extensive database or well-established protocol does not exist
Selection of species/taxa	• Consider how model was developed and the geographical location to which the model will be validated, including desirability of or requirement for taxonomic representation from that location • Validate model with species/taxa that the model was intended to represent • Determine if the model is to be extrapolated to relevant species/taxa, if it is predictive of ecologically relevant species/taxa (e.g., Schlekat et al. 2010; De Schamphelaere et al. 2014), and if it is to be applied to a range of taxa (microalgae, invertebrates, vertebrates) • Use standard characteristics of test organism (e.g., test starts with daphnid neonates <24 h old) • Consider sensitivity of the testing laboratory’s strain of the species to be tested, relative to the sensitivity of the strains in the data set used to parameterize/calibrate the model	• Taxonomic diversity should be considered (i.e., if the model is based on one species or averaged across a taxonomic grouping [e.g., invertebrates]) • Although validating with species/taxa the model was developed for is a good check, these taxa might not be available to the user or relevant to the geographic region of interest • Caution must be taken when extrapolating models from one taxonomic group to another (Van Sprang et al. 2009; Organisation for Economic Co-operation and Development 2016) • Unusually low or high relative sensitivity of the strain used in the validation test might lead to an erroneous conclusion about model bias
Exposure conditions	• Use exploratory runs of the model during the experimental design phase to help determine appropriate water chemistry • Consider ecologically relevant exposure conditions, water chemistry within the bounds of concern to the user and the geographic range of interest, exposure water flow (static, renewal, or flow-through), and implications for chemical equilibrium • Measure routine water chemistry parameters (e.g., temperature, pH, conductivity, hardness, alkalinity, DO, DOC) • Ensure measurement of appropriate metal form or species • Use appropriate lighting (both intensity and light cycle)	• Caution must be taken if validating the model outside of the water chemistry parameter ranges for which it was originally developed
Media collection	• Decide whether using natural or synthetic dilution water • Follow guidance for best practice of natural water collection and synthetic water preparation	• In natural waters, consider concentrations of metals and speciation-modifying factors (e.g., pH, hardness, natural DOC, and ligands) • Although synthetic waters allow tight control of physical–chemical parameters, they do not necessarily represent waters of interest
Endpoints	• Ensure appropriate endpoints for the validation objective (e.g., ECx for reproduction in daphnids) • Consider the endpoints predicted by the model	• Use caution when extrapolating to endpoints the model was not developed to predict

ASTM = American Society for Testing and Materials; DO = dissolved oxygen; DOC = dissolved organic carbon; ECx= x% effect concentration; OECD = Organisation for Economic Co-operation and Development; USEPA = United States Environmental Protection Agency.