In order to find the mechanism of soil erosion control and crop yield increase in ridge-furrow rainwater harvesting production in semi-arid regions of China, a modified SCS-CN model was used to estimate the runoff for ridges with 3 widths (30, 45, and 60 cm) and 3 different materials (soil crust, biodegradable film, and plastic film) based on the rainfall and runoff from 2012 to 2016. The parameters of CN, λ, and α in the modified SCS-CN model were calibrated according to the rainfall and runoff from 2012 to 2013, and the applicability of the modified SCS-CN model to rainwater harvesting ridges was evaluated using rainfall and runoff from 2014 to 2016. The results indicated that for ridges mulched with soil crust, biodegradable film, and plastic film, CN in the modified SCS-CN model ranged from 82.0 to 82.9, from 97.9 to 98.8, and from 98.3 to 99.9, respectively; The parameter of λ ranged from 0.045 to 0.071, from 0.251 to 0.327, and from 0.189 to 0.213, respectively; The parameter of α ranged from 2.35 to 2.89, from 4.21 to 4.82, and from 3.32 to 3.99, respectively; Mean relative error ranged from 1.36% to 3.41%, from 1.21% to 3.42%, and from 2.48% to 5.42%, respectively; the Nash-Sutcliffe efficiency ranged from 0.97 to 0.98, from 0.97 to 0.98, and from 0.96 to 0.98, respectively. Effects of classification of daily rainfall were not obvious on the parameters of CN, λ, and α in the modified SCS-CN model. The difference of CN, λ, and α between 3 different materials was greatly larger than that between 3 ridge widths. The Nash-Sutcliffe efficiency was higher than 0.95 and the relative error was lower than 10%. The results indicated that the modified SCS-CN model was applicable to runoff estimation for rainwater harvesting ridges and the model performance was acceptable. These findings may be helpful in controlling soil erosion and increasing rainwater use efficiency.