Study On Spatial Variability Of The Rice Growth Informations

As the low utilization rate of the irrigation water, the shortage of water resources has been seriously wasted. Thus it is important to know crop transportation and evaporation for precise irrigation and efficient water management. The consumption of the rice irrigation water accounts for the largest proportion in Jiangsu Province. In this condition, to understand the spatial variability of the rice growth status and water resource, based on farmland spatial information processing technology research, this paper studied on spatial heterogeneity and the relativity on the growth, output and evapotranspiration of the rice in the test field of Water Conservancy Technology Promotion Station of Changshu City. And the research used evaluating indicator to compare the Original Kriging method with the radial basis function neural network (RBFNN) method. The main conclusion of this paper are as follows.(1) the traditional theory of statistic analysis showed that, under this research condition, the rice LAI reached a peak in heading period. The earlier and later stages of the rice were medium weak variation, the mid-terms were weak variation. Different growth stages of the rice height and the evapotranspiration based on a day were weak variation. The variation of the rice output was medium weak variation. Under the confidence level of 90% and the relative error of 5%, the reasonable sampling numbers of each variable were less than the actual numbers of samples, so the numbers of samples met to the research requirements. The LAI, height and output of the rice were normal distributions, sectional stages of the rice evapotranspiration were abnormal distributions. According to related system analysis of the rice growth informations, the results showed that at heading period, the LAI, height and output had the remarkable correlation in the large farmland.(2) After normal transformation or specific value treatment, the research used the geostatistics theory to analyze the spatial variability of the rice growth informations. The results showed that, the semi-variogram models of the rice growth informations were Spherical models or Gauss models, and the range were between 6 meter and 9 meter. Original Kriging could be used to forecast the rice LAI. The effectiveness on fitting the rice evapotranspiration by Original Kriging depended on the date. But the Original Kriging was not fit to forecast the rice height and output.(3) The research used the correlation index R2 and statistic I in the comparison of the models. The result indicated that the RBFNN models of the rice LAI and height were obviously excellent than the semi-variogram models. The precision of interpolation in the rice output and evapotranspiration using the RBFNN was slightly exact than the Kriging.