Abstract: ‍ ‍Wind power， a clean energy， has gained interest worldwide. The performance of airborne surveillance radars may be affected by the reflected echoes of wind farms. Short coherent processing intervals （CPIs） cause problems such as a reduced spectral resolution of the radar echoes and unclear fretting characteristics of wind farms， and the performance of the clutter suppression method is affected. In this study， the effect of modulation on the spectrum of radar echoes and the influence of ground clutter are considered， and the sparse characteristics of transient radar echoes in the frequency domain are used. A feature recovery method of wind farm clutter for airborne radar， based on Augmented Lagrangian Method and Variable Splitting （ALM-VS）， is proposed. First， multiple signal components are obtained by sliding the window. Second， the split-augmented Lagrangian method is used to iteratively solve the optimal sparse representation coefficients of each component， and each signal component can be recovered. The complete data can be intensively reconstructed by using the inverse sliding window to improve the spectral resolution and micro-motion characteristics of the wind farm radar echoes. Based on this， a combined method of morphological component analysis （MCA） and low rank matrix optimization （LRMO） is used to suppress non-stationary dynamic wind farm clutter for the short CPI airborne surveillance radar. The experimental results show that the feature recovery method based on ALM-VS can improve the spectral resolution of the radar echoes for short CPI airborne surveillance and subsequently enhance its micro-motion characteristics. Wind farm clutter can be suppressed if the target and clutter are either in the same or different distance units. Under the same conditions， the efficiency of the proposed ALM-VS method can be increased by 85.6% compared with the Iterative Adaptive Approach （IAA） method.