In a system in which fertilization is recommended, diagnosis of soil K availability and the establishment of critical levels are made difficult by the possibility of a contribution of non-exchangeable forms of K for plant nutrition. Due to its magnitude, this contribution is well diagnosed in long term experiments and in those which compare fertilization systems with positive and negative balances in terms of replacement of the K extracted by plants. The objective of this study was to evaluate K availability in a Hapludalf under fertilization for sixteen years with the addition of K doses. The study was undertaken in an experiment set up in 1991 and carried out until 2007 in the experimental area of the Soil Department of the Federal University of Santa Maria (Universidade Federal de Santa Maria - UFSM), in Santa Maria (RS), Brazil. The soil was a Typic Hapludalf submitted to four doses of K (0, 60, 120 and 180 kg ha-1 K2O) and subdivided in the second year, when 60 kg ha-1 of K2O were reapplied in the subplots in 0, 1, 2 and 3 times. As of the fifth year, the procedure was repeated. Grain yield above ground dry matter and total K content contained in the plant tissue were evaluated. Soil samples were collected, oven dried, ground, passed through a sieve and submitted to exchangeable K analysis by the Mehlich-1 extractor; non-exchangeable K by boiling HNO3 1 mol L-1 and total K by HF digestion. Potassium fertilization guidelines should foresee the establishment of a critical level as of which the recommended dose should accompany crop needs, which coincides with the quantity exported by the grain, without there being the need for the creation of broad ranges of K availability to predict K fertilization. In adopting the K fertilization recommendations proposed in this manner, there will not be K translocation in the soil profile.