Motivated by the idea of total least squared method, a total error-based multiquadric method (MQ-T) has been developed to decrease the effect of both horizontal and vertical errors inherent in sample points on surface modeling. Two examples including a numerical test and a real-world example were, respectively, employed to test the robustness of MQ-T to sample errors. The numerical test indicates that when sample points are only subject to vertical errors, MQ-T has a similar performance to MQ. When sample points are subject to horizontal errors, MQ-T is more accurate than MQ. In the real-world example, MQ-T was used to construct DEMs with sample points collected by a total station instrument, and its accuracy was compared with those of classical interpolation methods including inverse distance weighting, ordinary Kriging (Kriging) and ANUDEM. Results indicate that with the decrease of sample density, the interpolation accuracies of all methods become lower. Regardless of sample density, MQ-T is always more accurate than the other methods. Yet, compared with Kriging, MQ-T has a peak-cutting problem.