Abstract River water quality is the outcome of multiple processes and factors, which vary spatially and temporally. In this study, the key spatial and temporal scales and the most important environmental factors explaining river water quality at these scales were analysed by generalized additive models (GAMs). Water quality was studied through total phosphorus (median = 61.9 µg l⁻¹) and nitrogen (1388.1 µg l⁻¹), pH (6.7) and water colour (143.3 mg Pt l⁻¹). Environmental factors covered variables from land use/cover, climate and other landscape characteristics. The spatial scales used were the closest 50, 100, 200, 500 and 1000 m buffer zones around the river channel, in addition to the entire catchment area. Temporality was studied through the entire year and four periods, which were determined by the natural variation in discharge. In the comparison of spatial scales of environmental factors, the variation in phosphorus, nitrogen and water colour was best explained using environmental data from the broadest scale, the entire catchment. In contrast, the variation in pH was best explained using data from the closest (50 m) buffer zone, the riparian area. In modelling temporal scales, the variation in water quality variables was best explained during discharge maximum periods and when the environmental data covering the entire year were considered. Nutrients were related specifically to agriculture, water colour to lake percentage and pH to pastures. The results showed the suitability of GAMs in water quality studies.