区域土壤盐分时空变异性大，采用经典统计和地统计方法无法准确判断取样时间不规则、空间位置不一致的土壤盐分的时空变化趋势。本文以内蒙古河套灌区隆胜研究区68个监测点0~1.8 m土壤剖面4582个土壤盐分数据为基础，利用时空地统计方法分析区域土壤盐分时空变化特征，比较时空克里金较传统空间克里金插值的精度提升效果，并验证时空地统计方法在监测点减少50%情况下预测区域盐分时空动态的能力。结果表明：（1） 该研究区土壤盐分空间变异系数的变化范围是0.43~1.14，为中强变异，0~0.6 m根系层生育期积盐、非生育期脱盐，0.6~1.8 m土壤剖面生育期脱盐、非生育期积盐、农田土壤盐分有明显的季节性规律。（2） 和度量模型能较好拟合盐分时空经验半方差，各层土壤盐分预测值和观测值间的均方根误差RMSE均小于0.21 dS·m^-1，较传统空间克里金的RMSE小0.02~0.09 dS·m^-1。（3） 采用该方法在减少50%监测点情况下确定的土壤盐渍化分布与所有取样点确定的结果一致性较高，0~0.6 m和0.6~1.2 m土壤盐分面积间的相对误差MRE分别为-13.20%和-8.35%，RMSE为466.67 hm²和494.43 hm²，决定系数R²为0.79和0.72。时空克里金同时利用了土壤盐分时间和空间上的更多信息，实现了稀疏盐分监测点数据集土壤盐分时空动态的精确估计，可极大提高区域土壤盐分时空格局监测的效率。

For maintaining crop yield in salt-affected dry agricultural settings, monitoring and analyzing spatio-temporal dynamics of soil salinity over broad areas is crucial yet challenging due to its high variability. The most popular techniques for evaluating spatial distribution patterns and temporal trends are classical statistical analysis and traditional geostatistical analysis, but they are not suitable for accurately capturing spatio-temporal trends of soil salinity due to irregular sampling time and inconsistent spatial position during sampling time. Spatio-temporal Kriging is an extension of spatial geostatistics to space-time geostatistics and may overcome this problem effectively because its model covariance/variance is a function of both space and time. However, its application in spatio-temporal modeling and prediction of regional soil salinity is still unclear. Based on 4582 soil salinity data of 0-1.8 m soil profiles from 68 monitoring locations in the Longsheng study area of Hetao Irrigation District, Inner Mongolia, spatio-temporal variation characteristics of regional soil salinity using a spatio-temporal geostatistical method, and spatio-temporal Kriging interpolation accuracy was compared with traditional spatial Kriging interpolation. Furthermore, the ability of spatio-temporal Kriging to obtain regional soil salinity dynamics was verified using less than half of the original monitoring locations. The results showed that the spatial variation coefficient of soil salt in the study area ranged from 0.43 to 1.14, which was categorized as medium to strong variability. Regional averaged soil salinity dynamics had obvious seasonal variation characteristics, and the root zone (0-0.6 m) soil salinity accumulated in the crop growing season and desalted in the fallow season, while the deep soil salinity (0.6-1.8 m) was the opposite. The sum-metric model can fit the temporal and spatial experience semi-variance of soil salinity well, and the root mean square error (RMSE) between the predicted value and observed value of soil salinity in each layer was less than 0.21 dS·m^-1, which was 0.02-0.09 dS·m^-1 less than that of traditional spatial Kriging. The areas of different soil salinity determined by 32 sparse monitoring locations were in good agreement with those determined by all sampling sites, and the mean relative error between areas of different soil salinity for 0-0.6 m and 0.6-1.2 m were -13.20% and -8.35%, respectively. Similarly, the respective RMSE were 466.67 hm² and 494.43 hm² and the determination coefficient (R²) were 0.79 and 0.72, indicating that spatial distribution of soil salinity obtained by sparse monitoring locations is consistent with the results of all sampling locations. Spatio-temporal Kriging significantly improves the prediction accuracy of soil salinity compared with ordinary Kriging, since it uses more information on soil salinity in time and space. The accurate estimation of spatio-temporal dynamics of soil salinity in the data set of sparse monitoring points was realized, which can greatly improve the monitoring efficiency of the spatio-temporal pattern of soil salinity in the region.