Salinity in semi-arid shallow lakes tends to vary inversely with water level: fresh when full and more saline when dry. Despite this well recognized co-variation and its use in numerical water and salt budget models or in paleo-salinity studies there is no accepted form that the empirical relationship should take or a physical explanation for its shape and why it varies between lakes. Using data from a long-term study of 152 shallow lakes across southwest Australia, variously impacted by long-term climate drying and anthropogenic salinization we show that empirically salinity varies with lake water level as a piecewise power-law with up to four distinct segments. A simple mathematical model is developed which explains how piecewise relationships emerge in these lakes, indicating that segments are variously controlled by salt precipitation, vertical water fluxes (evaporation and rainfall), lake outflow hydrology and lastly equilibrium with inflows. Exponents of segments are shown to depend on universal scaling relationships for lake bathymetry and lake-stage discharge relationships. Using this knowledge, we were further able to quantify salinization in data poor lakes and distinguish impacts from long-term climate drying from secondary salinization, both of which were contributing to widespread increases in salinity levels. This study shows that salinity - water level data is underappreciated for its ability to qualitatively and quantitatively describe lake hydrology and disentangle causes of rising salinity in shallow lakes.