Input–output balances of potassium (K) in agricultural fields have been calculated using five long-term field experiments located on varying parent materials (from loamy sand to clay) in South and Central Sweden. Each experiment consisted of variable K fertilizer regimes and two separate arable crop rotation systems. The inclusion of a grass/clover ley and regular additions of farmyard manure was used to simulate a livestock system (System I), although no actual grazing occurred. The second rotation (System II) also included oil seed and sugar beet in addition to cereals. In System I crop residues were removed while for the grass free System II all crop residues were incorporated and no farm yard manure was added. Experimental plots were fertilized with either 125 or 150 kg N ha^–1 yr^–1, and received K fertilizer at one of four rates: zero K, replacement of K removed by the previous crop, replacement of K removed by the previous crop + 40 kg K ha^–1 yr^–1, or replacement of K removed by the previous crop + 80 kg K ha^–1 yr^–1. Exchangeable potassium (K_ex) concentration in the top soil (0–20 cm) was analyzed every year, using an extractant of 0.1 M ammonium lactate and 0.4 M acetic acid (pH 3.75). Potassium concentration in the crop was analyzed after every harvest. Measurements of K inputs and outputs were calculated for each rotation over a 40 year-period (kg ha^–1 yr^–1). The results of K field balances for individual treatments will be described in relation to changes in topsoil K_ex for different soil types and in particular the sustainability of K delivery to grass ley in organic and low input cropping systems. The K balances followed the expected general trend for individual K fertilizer treatments. Plots that received zero fertilizer K showed negative K balances which ranged from 29 to 64 kg ha^–1 yr^–1 in rotation System I, compared to 10 to 26 kg ha^–1 yr^–1 for System II. The K replacement treatments were all close to being in balance, whereas the two highest K-fertilizing rates produced a surplus of K. Differences in the measured response on K_ex were apparent between soil types. In one sandy loam soil, the balance of non K-fertilized plots within System I gradually became more negative with time during the experimental period. This increase in net off-take of K from the soil was significant (p<0.01, r² = 0.58), and indicated increasing release of K by chemical weathering of soil minerals, or depletion of exchangeable K, or both. Despite 40 years of negative balances for plots not fertilized with K, no significant depletion of K_ex was found for the clay or sandy loam soils. Furthermore, the presumably small leaching of K made replacement of harvested K sufficient to maintain K_ex. In the sandy loams there was even a significant increase of K_ex in the topsoil, presumably due to release of K by weathering, or by biocycling of K taken up from deeper soil horizons. However, the loamy sand became depleted in K_ex, for the plots that received either zero K or straight replacement of harvested K. At this site, the highest K fertilizer application rate was needed to maintain the level of K_ex unchanged. For the non-fertilized plots, only those sited on clay rich soils were able to maintain a concentration of ~2 % K in grass ley (dry weight); on the sandy loams and the loamy sand, herbage concentrations were generally less than 2 % in the zero-K treatments. This provides evidence for the requirement of K fertilizer, to ensure an adequate K concentration in herbage harvested from certain soils.