It is significant to analysis the spatial and temporal variation of soil nutrients for precision agriculture especially in large-scale farms. For the data size of soil nutrients grows once after sampling which mostly by the frequency of one year or months, to discover the changing trends of exact nutrient would be instructive for the fertilization in the future. In this study, theories of GIS and geostatistics were used to characterize the spatial and temporal variability of soil nutrients in paddy rice fields in the Erdaohe farm of Heilongjiang Province, China, which located in the north of Daxing'an Mountains, has an area of nearly 36.1 million hectares for paddy rice planting. The soil samples, collected from 2009 to 2013 once a year, were sampled based on the spatial distribution of paddy rice fields, counting as 651 in 2009, 1488 in 2010, 954 in 2011, 483 in 2012, and 471 in 2013. These samples were analyzed for pH, soil organic matter (SOM), available nitrogen (AN), available phosphorus (AP), and available potassium (AK). In these measurement results, value of pH is not very variable among four years, ranging from 4.6 to 6.4 in 2009, 5.1 to 6.5 in 2010, 5.98 to 6.4 in 2011, 4.88 to 6.52 in 2012, and 4.94 to 6.21 in 2013, and the coefficient of variation (C.V. %) were 4.77, 3.73, 3.60, 4.42, 3.11 from 2009 to 2013, all of which had a weak spatial variability. Besides, the other four nutrients had a medium spatial variability, with the highest one of AK. On the other hand, for the general trend, spatial variation of soil organic matter (SOM) increased from 2009 to 2013, and decreased the rest AN, AP, AK. After calculating and comparing the spatial and temporal variation in whole farm area, variations between management regions is the second research point. To reach this aim, we chose interpolation method of kriging to generate grids for every soil test data. Among these five factors, only data of soil pH fit the normal distribution, the other four factors for several years need to be transformed for better results. Take the interpolation of AN as an example, Box-Cox transformation parameters were chosen as 0.1 in 2009, and Log transformation was used in 2012 and 2013, the temporal geographic maps revealed that in 2009, region I had the highest level of AN, and the next year most regions has the same level of AN, except for region VIII which is lower than the other six regions. In 2011, region II, IV, VI had a higher level than region I, III, V, VIII, while the next year kept the same except for region V becoming higher. Based on this study, conclusion acquired is that from 2009 to 2013, the spatio-temporal variations decreased in soil pH, AN, AP, AK, and increased in SOM. Moreover, according to the comparison of interpolation results, these five soil properties in Erdaohe farm remained not very stable in the past four years, which could implicate important significance in future research with consideration of fertilization, rice yield and other factors.