Mann-Kendall (MK) trend test is a nonparametric statistical test. When the modified MK (MMK) test was used for serial trend test, it was considered to be stricter than the MK test considering the influence of self-correlation of time series. In this article, the MK method and the MMK method were combined for trend tests of water requirements (ET_c) for two main economic crops, i.e., cotton and sugar beet, at 41 stations in Xinjiang over two periods of 1961-1961 and 2015-2099. The influences of serial self-correlation of ET_c on trend test results were investigated, and the spatial distribution of trends in ET_c of each site and the whole Xinjiang were presented. The results showed that: (1) When the order of self-correlation j was greater than 0, the statistics obtained from the MMK test should be adopted. (2) Over 1961-2010, ET_c of both cotton and sugar beet had certain temporal-dependent structures. Over 2015-2099, ET_c of cotton had very significant long-range correlation, but ET_c of sugar beet hadn't any temporal dependence. (3) Over 1961-2010, which the trends in ETc of cotton at 37 sites decreased, of which the trends in ETc of cotton at 8 sites were significant. Over 2015-2099 for A2 scenario, trends in ET_c of cotton at all of the sites in Xinjiang decreased, of which trends in ET_c of cotton at 13 sites were significant, while trends in ET_c at 28 sites were not significant. For B2 scenarios, only at Luntai the trend in ET_c of cotton increased insignificantly, trends in ET_c of cotton at the remaining 40 sites decreased. Over 2015-2099 for A2 scenario, trends in ET_c of sugar beet at 13 sites increased, of which trends at 3 and 10 sites were significant and insignificant, respectively; trends in ET_c of sugar beet at the remaining 28 sites decreased, of which the trends were significantly and insignificant at half of the 28 sites. For B2 scenario, trends in ET_c of sugar beet at 9 sites increased. (4) Over 1961-2010, trends in ET_c of cotton and sugar beet in the whole Xinjiang decreased significant and insignificant, respectively. Over 2015-2099 for A2 and B2 scenarios, both trends in ET_c of cotton decreased significantly; trends in ET_c of sugar beet also decreased insignificantly for A2 scenario but significantly for B2 scenario. (5) For both cotton and sugar beet, Sen's slope b were related to the linear slope SL with high coefficients of determination with linear function, which reflected the high consistency of the two variables when describing the trend amplitudes (6) It is necessary to consider the effects of self-correlation for ET_c series when using MK test. (7) The synthetic effects of decreases in wind speed and solar radiation as well as the increases in precipitation and relative humidity on the restraining of ET_c surpassed the effects of increases in air temperatures in Xinjiang. In conclusion, decreases in the trends of ET_c series showed that there were reduced water demands of cotton and sugar beet in the future in Xinjiang, and the risk of drought also reduced. Trend variations in ET_c for cotton and sugar beet had important reference value for making the crop irrigation planning and drought disaster prevention under the background of climate change in the arid and semi-arid areas.