In salt marshes, species richness changes along environmental, disturbance and productivity gradients forming a complex network of abiotic and biotic factors. On 2,691 plots along 121 transects, we sampled vegetation along the German mainland North Sea coast (13 regions) during 1987–1989. Applying regression tree analysis (RTA), we now used this large data set to analyse variance in species richness (SRich) in relation to 13 explanatory variables varying on different scales. SRich (mean, 4.9 per m2) was significantly correlated to most variables. Only six variables were included in our final model, together explaining 68.5 % of variance in SRich, in hierarchical order: moisture, salinity, evenness, nitrogen, region and elevation. Predominantly, SRich was limited by environmental heterogeneity (moisture, salinity and nitrogen, each explained approx. 50 % variance). SRich tended to be high on plots exhibiting a combination of low moisture, salinity and nitrogen values, with high evenness—and especially high in some regions when plots were lying high in relation to mean high tide. Grazing regimes did not affect SRich significantly. In conclusion, our model showed that SRich in the study area was predominantly explained on a small scale and less along large-scale gradients. RTA proved suitable to identify the set of variables that mainly explained variance in SRich. Our tree model improves the understanding of richness patterns in salt marshes and can be used to predict species richness for the study area. Furthermore, our data provide a reference to detect richness changes due to, for example, management changes or sea level rise.