With climate change, it is becoming more challenging for water-limited cities to sustain historic watering levels in urban parks, leading park managers to consider changes to park designs. But whether and to what extent the public value parks that deviate from conventional designs featuring large areas of irrigated lawn remains uncertain. We use a choice experiment to assess public preferences for different park groundcovers in Perth, Australia. With a scale-adjusted latent class model, we identify optimal groundcover compositions for four preference classes. We find that while having some watered grass in urban parks is important, the public are also accepting of non-irrigated alternatives. Incorporating at least 40% native vegetation cover can increase the utility the public derives from parks and conserve water. Park managers also have a degree of flexibility in designing parks that vary from the optimal groundcover composition but that still deliver near-optimal benefits to communities.