Abstract: The saturated permeability, k_s is a key parameter in solving geotechnical engineering problems such as stability analyses of slopes and earth-retaining structures during rainfall. Precisely determining this quantity for soils in the field is difficult. Hence, this paper suggests a simple approach for estimating the saturated permeability of loess in the field using a new, modified, double-ring infiltrometer. We use this device, which can perform infiltration tests accurately and effectively, to study the infiltration characteristics of loess. The results of these tests show that water infiltration into uniform, dry loess can be considered to divide the soil into a saturated, wetting, and dry-soil zone. Moreover, the depth of the saturated zone is approximately equal to that of the wetting zone, and the degree of saturation in the wetting zone is greater than 70%. The soil-moisture profile derived from the infiltration tests is similar to that given by the Green-Ampt model. The field tests also indicate the double-ring infiltrometer method based on the conventional Green-Ampt model (Specification of soil test, SL237-042-1999) provided larger k_s values of the loess. To determine the saturated permeability more reliably, we suggest a modified approach herein, which is based on the Green-Ampt model and employs the soil-water characteristic curves (SWCC) along with the main wetting path. The concept of the modified approach is to redefine the meanings of the key parameters of the Green-Ampt model (the matric suction head at the wetting front, the wetted depth, and the infiltration rate). We take the matric suction head at the wetting front to be that corresponding to the initial soil moisture content; its value can either be measured with tensiometers in the experimental field or else determined by the SWCC along with the main wetting path. The wetted depth during the infiltration test period is taken to be that observed by the soil moisture sensors. In the Green-Ampt model, we also adopt the average infiltration rate in the early infiltration stage rather than the steady-state infiltration rate. From the results of our infiltration tests and laboratory permeability tests, we find this approach to be a suitable method for determining the saturated permeability of loess in the field.