Stochastic simulation of water drainage at the field scale and its application to irrigation management

To estimate water drainage at the field scale, a frequently adopted approach is using a deterministic model with field-averaged hydraulic parameters. However, spatial variability of soil hydraulic properties in the field is a potential source of error. This study evaluated the effect of spatial variability of surface soil saturated hydraulic conductivity (Ks) on water drainage. A conditional simulation (CS) method was used to generate a random field of surface soil Ks based on 117 observed values in the study area. The random field of surface soil Ks was then coupled with a dynamic soil water movement model (HYDRUS-1D). Water drainage during a period of 3 months was stochastically simulated with a total water input of 354 mm (including 270 mm of irrigation). Accumulated drainage beyond 2 m soil depth ranged from 23.7 to 64.7 mm, which accounted for 8.8–24.0% of the irrigation input. In addition, the accumulated drainage was also calculated using the measured Ks data and Ks estimated by an ordinary kriging (OK) method. Results obtained from the three methods showed that the accumulated quantities of water drainage obtained by the CS method agreed well with those from measured Ks data, while the water drainage range was narrowed by the OK method because of its apparent ‘smoothing effect’. The effect of spatial variability of surface soil Ks on water drainage was demonstrated by the three methods and their results were all better than a traditional method that did not consider the spatial variability of surface soil Ks. An irrigation schedule was finally determined using the CS method. When the irrigation input was controlled between 190 and 200 mm, the schedule saved about 747 m3 of water in a 1 hm2 field, accounting for about 28% of the traditional irrigation applied, and the mean accumulated quantity of water drainage was only 2.3 mm, far lower than the 58.9 mm generated by the traditional method.