Generalized from long-term (1945-1989) observational data on water temperature at stations of the Hydrometeorological Service network that describe reservoirs of different types in northwestern Russia, empirical relationships of the thermal regime on various geographical factors are established (lake morphometry, geographical latitude and altitude above sea level, residence time). We have demonstrated that for stable thermal stratification to emerge, a lake must have a certain combination of geometric parameters. Regional boundaries between epi-, meta- and hypothermal types of lakes have been quantified depending on the area and maximal depth of the lakes. Analysis of the results has shown that the temperature in the upper 5 m layer of water mainly depends on the latitude, whereas at a depth of more than 10 m it is more significantly affected by the lake morphometry, inflow and outflow of rivers. Modeled curves describing average annual daily surface temperature values for lakes of different size and depth were received with use of the 6-parameter regression function. All parameters are dimensional and have a clear physical interpretation. This feature favorably distinguishes the function from previously published regression models.