A quantitative hydrogeomorphic approach was applied to an extensive survey of temporary wetlands in the Doñana National Park (SW Spain) in search for quantitative thresholds for wetland classification. Twenty freshwater ponds on the aeolian sand mantle and 46 sites on silty-clay substrate, located in the southern marshland, were surveyed during the heaviest rainy period so far recorded (October 1995-September 1997. On average, temporary ponds showed higher water depth, longer flooding period, lower conductivity (\0.5 mS cm -1 ), lower pH (6.7), lower phosphate concentration (0.4 lM) and a more balanced proportion of Ca 2+ /Na + than temporary marshes. During floods, marshland sites exhibited higher water transparency, pH (9.5), alkalinity (3.5 meq l -1 ), conductivity (8.2 mS cm -1 ), phosphate concentration (0.7 lM), Na + and Ca 2+ concentrations (97.2 and 3.5 meq l -1 , respectively) than ponds. Study sites were significantly segregated (ANOSIM test: R = 0.88, P \ 0.01, n = 92) in relation to water depth and conductivity. A conductivity of 1.6 mS cm -1 is proposed as a threshold between marshland sites and ponds during floods. Marshland sites were further segregated into two groups (ANOSIM test: R = 0.777, P \ 0.01, n = 23) according to the Na + /Ca 2+ ratio (in meq l -1 ) at a threshold value of 25. An ordination by PCA showed that five variables grouped 81.4% of the total variance in two axes. The first PCA axis (60.7% of variance) separated temporary wetlands into ponds and marshland sites according to variables related to substrate and hydrology (Na + /Ca 2+ ratio, conductivity, water depth and flooding period). Other variables (e.g., water transparency, alkalinity, pH, submersed macrophyte biomass, phosphate, nitrate and planktonic chlorophyll concentrations) did not produce a significant segregation between marshland and pond sites during floods. Further discrimination within each wetland type was thus not achieved.