Abstract: In order to overcome the problem that the determination of non-scale interval and fractal dimensional computation were affected by human factors, the new method of scale transformation was presented according to the principles of notable regression effect, the smallest residual sum of squares and weak autocorrelation of residual error, which providing reference for the determination of study scale. The inhomogeneous distribution of the measuring stations of irrigation experiment network in Henan province was described and its fractal characterization and the dimensional resolution were analyzed by means of GIS. The results showed that the station network distribution in Henan province had different scale domain, the fractal dimension of the 110～190, 200～310, 320～410 and 420～490 km range was 1.768, 1.317, 0.763 and 0.283, respectively, and their correspondingly dimensional resolutions were respectively 0.232, 0.683, 1.237 and 1.717. It had been indicated that the detectability of phenomena by the current network was highest in 110~ 190 km scale and did not describe well the crop water requirement in the scale of less than 100km. In scale of more than 420km, the current structure of network appeared redundance and some measuring stations should be reduced.