Setting minimum head capsule abundance and taxa deletion criteria in chironomid-based inference models

Criteria for removing training set lakes and taxa in chironomid-based infer ence models, due to low abundances, have largely been ad hoc. We used an an oxia inference model and a hypolimnetic oxygen model from south-central Ont ario to determine what effect subfossil head capsule abundance and taxa del etion criteria have on fossil inference statistics. Results from six traini ng set lakes suggest that a minimum abundance of 40-50 head capsules is suf ficient for use in inference models, however more diverse samples likely re quire more than 50 head capsules. Taxa deletion criteria substantially impr oved the predictive ability of inference models (lowered the root mean squa red error of prediction (RMSEP)). The common practice of including taxa wit h only greater than or equal to 2% abundance in at least two lakes was one of the deletion criteria that much improved inference models. Similar delet ion criteria, such as greater than or equal to 2% in at least 3 lakes and g reater than or equal to 3% in at least 1 lake, produced comparable improvem ents (up to 18% reduction in RMSEP).