Assessing the health of any ecosystem requires the measurement of a number of indicators. For an assessment of ecosystem health to have real value, it must not only address status, but also evaluate the risk of that status changing to a less healthy state. Although numerous indicators of ecosystem health have been proposed (Rapport et al 1985; National Research Council 1993), few have been tested to date. Such testing must incorporate evaluation of the indicator's sensitivity so as to provide unambiguous and reliable values for the assessment of ecosystem status. Indicator sensitivity must also be known so as to enable an evaluation of the risk of change in the health status of the ecosystem.