Topographic complexity is a key component of habitat, which has been linked to increased species richness in many ecological communities. It can be measured in various ways and it is unclear whether these different measurements are mutually comparable when they relate to plant species richness at different spatial scales. Using a densely sampled set of observations for Rhododendrons (406 species and 13,126 georeferenced records) as a test case, we calculated eight topographic complexity indices from a 250-m resolution digital elevation model and examined their correlations with Rhododendron species richness in China at seven spatial scales: grain sizes 0.05°, 0.1°, 0.25°, 0.5°, 1.0°, 1.5°, and 2.0°. Our results showed that the eight topographic complexity indices were moderately to highly correlated with each other, and the relations between each pair of indices decreased with increasing grain size. However, with an increase in grain size, there was a higher correlation between topographic complexity indices and Rhododendron species richness. At finer scales (i.e. grain size ≤ 1°), the standard deviation of elevation and range of elevation had significantly stronger correlations with Rhododendron species richness than other topographic complexity indices. Our findings indicate that different topographic complexity indices may have positive correlations with plant species richness. Moreover, the topographic complexity–species richness associations could be scale-dependent. In our case, the correlations between topographic complexity and Rhododendron species richness tended to be stronger at coarse-grained macro-habitat scales. We therefore suggest that topographic complexity index may serve as good proxy for studying the pattern of plant species richness at continental to global levels. However, choosing among topographic complexity indices must be undertaken with caution because these indices respond differently to grain sizes.