1. The value of measuring ecosystem functions in regular monitoring programs is increasingly being recognised as a potent tool for assessing river health. We measured the response of ecosystem metabolism, organic matter decomposition and strength loss, and invertebrate community composition across a gradient of catchment impairment defined by upstream landuse stress in two New Zealand streams. This was performed to determine if there were consistent responses among contrasting functional and structural indicators. 2. Rates of gross primary production (GPP) and ecosystem respiration (ER) ranged from 0.1 to 7.0 gO2 m−2 day−1 and from 0.34 to 16.5 gO2 m−2 day−1 respectively. Rates of GPP were variable across the landuse stress gradient, whereas ER increased linearly with the highest rates at the most impacted sites. Production/respiration (P/R) and net ecosystem metabolism (NEM) indicated that sites at the low and high ends of the stress gradient were heterotrophic with respiration rates presumably relying on organic matter from upstream sources, adjacent land or point sources. Sites with moderate impairment were predominantly autotrophic. 3. Declines in the tensile strength of the cotton strips showed no response across part of the gradient, but a strong response among the most impaired sites. The rate of mass loss of wooden sticks (Betula platyphylla Sukaczev) changed from a linear response to a U-shaped response across the impairment gradient after water temperature compensation, whereas leaf breakdown at a subset of sites suggested a linear loss in mass per degree-day. Three macroinvertebrate metrics describing the composition of the invertebrate community and its sensitivity to pollution showed similar linear inverse responses to the landuse stress gradient. 4. The first axis of a redundancy analysis indicated an association between landuse stress and various measures of water quality, and wooden stick mass loss, the invertebrate metric % EPT [percentage of macroinvertebrate taxa belonging to the Ephemeroptera, Plecoptera and Trichoptera (excluding Hydroptilidae] taxa, P/R and NEM, supporting the utility of these structural and functional metrics for assessing degree of landuse stress. The second axis was more strongly associated with catchment size, ER and GPP which suggests that these indicators were responding to differences in stream size. 5. Our results suggest that nonlinear responses to catchment impairment need to be considered when interpreting measurements of ecosystem function. Functional indicators could be useful for detecting relatively subtle changes where the slope of the response curve is maximised and measurements at the low and high ends of the impairment gradient are roughly equivalent. Such responses may be particularly valuable for detecting early signs of degradation at high quality sites, allowing management responses to be initiated before the degradation becomes too advanced, or for detecting initial moves away from degraded states during the early stages of restoration. Close links between structural and functional indices of river health across an impairment gradient are not necessarily expected or desirable if the aim is to minimise redundancy among ecological indicators.