Use existing forecast infrastructure for generating forecasts with each individual contaminant and different combinations, in order to test contaminant models. It might help us isolate problems with fitting HCDs and metals, e.g. what k ranges are needed, what binning works best, etc under the assumption that the models are correct. In mocks and data we are testing those features while also being unsure if our models are describing well what is in the data, making some tests more complicated.
Use existing forecast infrastructure for generating forecasts with each individual contaminant and different combinations, in order to test contaminant models. It might help us isolate problems with fitting HCDs and metals, e.g. what k ranges are needed, what binning works best, etc under the assumption that the models are correct. In mocks and data we are testing those features while also being unsure if our models are describing well what is in the data, making some tests more complicated.