When estimating a true population parameter from survey data, there is always some level of uncertainty as a result of imperfect detection, imperfect observation, spatial and temporal variation, and sampling error. In this report, we discuss the sources of uncertainty in New Zealand braided river bird counts. We use Monte Carlo simulations to illustrate the effect of different survey designs on uncertainty in counts for two species of bird: wrybill (Anarhynchus frontalis) and black-fronted tern (Sterna albostriata). The simulations were based on observed counts from previous river surveys. In general, larger annual changes can be detected with less uncertainty than smaller changes. Additional survey effort, e.g. replicate counts within a year, replicate sections within a river or replicate surveys over sequential years, will reduce uncertainty.