Saline soil as one of important land resource reserves, it should be properly used and effectively improved to bring into play the resource potential and increase the integral production capacity. It's not only significantly great to protect future food security and the ecological environment in China, but also one of the important ways to carry out sustainable agricultural development in China. The rational use and management of saline soil improvement must be built on the right soil water and salt dynamic prediction. Research on soil water and salt dynamics, acquiring the distribution and movement of water and salt in the soil, dynamic analysis of the impact of soil water and salt, and establishing forecast model of soil water- salt dynamics is of great significance in the soil water-salt transport mechanism research and playing an important role in promoting the knowledge of soil salinization evolution and controlling soil secondary salinization. In this paper,aiming at the primary problems existed in salt-affected soil improvement in North China Plain and coastal region in North Jiangsu Province, with the support of long-term soil water salt dynamics observational data in Fengqiu county, Henan province and Rudong county, Jiangsu province, on the basis of analyzing soil water salt regime and their influencing factors in the representative saline soil areas, time series analysis, artificial neural network and support vector machines are used to simulate and forecast the dynamics of soil water and salt in two regions above mentioned, and then soil water salt dynamic forecasting models suitable for different salt-affected regions are proposed in the condition of comparative analysis of soil water and salt in dynamic forecasting models. Major content and results are as follows:1. The soil water salt regime and relationship between it and its factor are expounded based on long-term observation data of soil water and salt dynamics in Fengqiu County.The large-scale soil column simulation experiments with the treatments of the combination of typical soil textures and different groundwater levels in North China Plain indicate: Rainfall and evaporation are the leading factors of soil water and salt regime, and rainfall has pulse to raise soil moisture content and reduce soil salinity in the profile;Groundwater table and the thickness of clay sandwich affect the whole level of soil water and salt dynamics,the obstruction of clay interlayer from salinizing the soil gets stronger as it approaches groundwater table; Vegetation has an effect on soil water and salt dynamic with time and distribution in the space, crop root absorbs water and nutrients, diving to speed up the supply of soil water, thereby affecting the distribution of soil water and salt in the profile.2. On the basis of long-term soil moisture, salinity and temperature monitoring field experiment in Rudong county, North Jiangsu province, the local soil water and salt dynamics and other factors impacts on it are well elucidated.Soil moisture content increases gradually as soil depth increasing, and it fluctuates more frequently and its fluctuation amplitude is much wider while soil depth is more close to soil surface in field monitoring experimental site in Rudong County. As the groundwater level is shallow and the groundwater salinity is high, the distribution of soil salinity in profile is characteristic of the profile type that soil salt gathers in the topsoil. Soil salt content in surface soil at monitoring site fluctuates much frequently during the experiment period because of highly frequent alteration of rainfall and evaporation. There are two major influencing factors that affects on soil water and salt dynamics, the one is meteorological factor, rainfall and evaporation play dominant role in influencing the direction and the intensity of local soil water and salt, while rainfall is greater than evaporation during a certain time period ,it implies salt elution in soil profile, and when evaporation rate is greater than rainfall during other time period, it means salt gather to soil surface; the other is fluctuation of groundwater table, the dynamic change of it affects soil water and salt moving down and up in soil profile, the average groundwater table is about 1m ,this may be the primary cause of soil profile type that soil salt gathers in the topsoil.3.Time series analysis, artificial neural networks and least squares support vector machines are used to simulate and forecast soil water and salt monitoring data in Fengqiu county, Henan province, and at the same time the comparison among three methods is made.Modeling analysis of soil column water and salt data in the basement, Fengqiu county using ARIMA, artificial neural network and least-squares support vector machine shows that: there are different best models corresponding to soil salt and water at different layers, and the best models strongly depends on the data itself; In the best soil water salt dynamic time series forecasting models, BP neural network model accounts for the biggest proportion, the LS-SVM model follows, and ARIMA model takes up the minimum proportion. But with regard to these three methods, when the sample is not large, the least-squares SVM model has a wider adaptability and generalization, and BP neural network models rely on the experiences.4. Modeling analysis and forecasting are used to soil water and salt data in Rudong county, Jiangsu province with the support of the three methods--time series analysis, artificial neural networks and least squares support vector machines, at the same time the comparison among three methods is made.In the short-term forecasting of soil water content, salinity, and soil temperature at the field experimental site in coastal region of North Jiangsu province, artificial neural networks and LS-SVM model that based on nonlinear mathematical theory show much better than ARIMA model because of artificial neural network and LS-SVM could simulate non-linear feature of short-term time series very well. With regard to these three models, LS-SVM model has a wider adaptability and generalization, model establishment is very simple, and model parameters could be automatically searched for the optimal approximations. However, when the sample size is larger, model computation speed decline significantly, it needs considering to use artificial neural networks or ARIMA model instead.5. On the basis of comparison of three methods on soil water and salt modeling, the best soil water and salt forecasting model suitable for different salt-affected region are proposed.Soil water and salt time series in different soil horizons have different best forecasting models, and the type of model strongly rely on the feature of soil water and salt time series themselves. The simple ARIMA models are likely to achieve very good forecast results when soil water and salt time series at different soil depths indicate certain linear trend or periodicity. While soil water and salt time series at different soil depths has the characteristic of nonlinearity, artificial neural network and LS-SVM models show much better than ARIMA models. LS-SVM model and supported by ARIMA model is the first choice for soil water and salt time series modeling in Fengqiu county, Henan province, and artificial neural network is comparatively ideal choice for short-term soil water and salt time series forecasting in Rudong county, Jiangsu province.