Normalization of model checklist for nightly observations

[jmtmp]

Hi,

If I understand well, the model, at final step, checks predicted species with historical ebird dataset for given lat/lon and current week of the year. Only when the occurence in the historical dataset is higher than defined threshold, prediction is accepted.

Furthermore I understand that historical eBird dataset incluses significantly higher number of daily observations than nightly observations. In order to identify owls and other nocturnal birds, the threshold in the model has to be set to very low probability, which has bad side efffect that the model misidentifies some rare daily birds.

I suppose and I also quickly checked that eBird dataset includes full timestamp of any (majority) of observations. So my question is, would not be possible to normalize the model checklist for daily/nigtly observations in some way or at least test the prediction against lat/lon & week & night/day?

Thanks,
Martin