Effects of long-term fertilization on balance, forms and use efficiency of potassium in oasis grey desert soil under a typical crop rotation system as affected by fertilization treatment were studied in an arid area. The long term fertilization experiment was designed to have 9 treatments, i.e., CK (No fertilization), N, NP, NK, PK, NPK, NPKS (NPK plus straw), NPKM (NPK plus organic manure) and 1.5NPKM (1.5 folds of NPKM). Results show that the balanced fertilizer treatments were much higher than the unbalanced fertilizer treatments in yield (except the NP). However, Treatment NP was an exception, and did not differ much from the balanced fertilizer treatments in yield (p>0.05), indicating that the grey desert soil is not in deficit of potassium. Budgeting of soil potassium revealed that only Treatment NPKS was gaining in K, while all the others were losing, with K deficit reaching as high as 268 ~ 2 996 kg hm^-2. Among all the K treatments, Treatments NPKM and 1.5NPKM were the highest in apparent use efficiency of K, reaching up to 81.2% and 38.9%, respectively, much higher than all the others (p<0.05), indicating that addition of manure may improve potassium use efficiency. After the long-term fertilization experiment, potassium varied significantly in form in all the treatments. Treatments NPKM and 1.5NPKM were much higher than all the other treatments（p <0.05）in non-exchangeable K, non-specific absorbable K, water soluble K and exchangeable K, demonstrating the importance of addition of organic manure or straw to NPK in maintaining potassium fertility in grey desert soil. In short, the conventional fertilization method fails to meet K demand of the crops, leading to growing K deficiency in farmlands of grey desert soil. More attention should be given to application of more potassium fertilizer or addition of organic manure or incorporation of crop straw in the farmlands of grey desert soil.