Abstract: Time-lapse seismic traveltime tomography is an important approach for characterizing subsurface changes. However, the conventional adjoint-state method based on absolute-difference (AD) traveltime residuals is sensitive to source-origin time errors and near-surface statics. Under complex near-surface conditions, such common-mode traveltime perturbations can be easily mis-mapped into deep velocity anomalies, thereby compromising the recovery of weak time-lapse signals. To address this issue, this study investigates time-lapse traveltime tomography under complex near-surface statics errors and introduces a double-difference (DD) adjoint-state method. By constructing relative traveltime residuals between receiver pairs, the proposed method systematically reduces common-mode errors at the data level and thereby mitigates their influence on inversion gradients and model updates. First, two simple numerical experiments are designed to compare the gradient stability of the AD and DD methods under systematic traveltime perturbations. Then, for the time-lapse imaging problem, two models are considered: an ideal near-surface model and a complex model containing an air layer, undulating topography, and a near-surface low-velocity layer. Under both sequential and independent inversion schemes, the two methods are evaluated in terms of their ability to recover time-lapse velocity anomalies. The results show that both methods satisfactorily recover the time-lapse anomalies under ideal conditions. Under complex near-surface conditions, however, the DD method achieves more stable gradients and better anomaly recovery. Quantitative analysis indicates that, under the designed statics-error scenarios, the DD method reduces the RMSE within the anomaly region by approximately 35%–40% compared with the AD method. In addition, a single-survey field data test shows that, under the same inversion settings, the DD method provides clearer imaging and higher resolution in areas with drastic local velocity variations than the AD method. These results suggest that the DD adjoint-state method exhibits strong robustness against statics errors under complex near-surface conditions and provides a feasible approach for time-lapse traveltime tomography under complex surface conditions.