switch from rlnorm to rnorm for K~Kpred in linear, binned models
aappling-usgs opened this issue · 1 comments
aappling-usgs commented
when K has a universal prior (pool_K600='none') then sure, make it a lognormal. because streams vary in their Ks lognormally.
when K has a site-wide prior ('normal') then i could see normal or lognormal.
when K is a linear or piecewise linear function of Q ('linear','binned') then it should probably be normally distributed around K_pred, because K_pred already can rise exponentially with discharge, and we just don't expect that much variation (certainly not extreme variation) between daily values of K on days with the same Q.
aappling-usgs commented