This paper describes the methodology used to eliminate environmental effects to a time series of short distance(24 m) leveling across a geological fault in order to detect potential precursors of earthquakes from the residual time series.First,linear trends and periodic signals are removed from the observations with the help of a recursive least squares method.Then,a controlled autoregression model is used to predict the enviromnental contribution due to precipitation and evaporation.The difference between the leveling data and these predictions are considered to be due to endogenic crustal processes.Applying this methodology to leveling data of approximately 13 years at Niukouyu crustal deformation station near Beijing reveals an anomalous deformation before the Tangshan earthquake (M7.8),which has its epicenter about 180 km from the survey site.However,in the time period considered,another anomaly occurred which cannot be related directly to a seismic event. This paper describes the methodology used to eliminate environmental effects to a time series of short distance (24 m) leveling across a geological fault in order to detect potential precursors of earthquakes from the residual time series. First, linear trends and periodic signals are removed from the observations with the help of a recursive least squares method. Chen, a controlled autoregression model is used to predict the enviromnental contribution due to precipitation and evaporation. The difference between the leveling data and these predictions are to be due due to endogenic crustal processes. Applying this methodology to leveling data of approximately 13 years at Niukouyu crustal deformation station near Beijing reveals anomalous deformation before the Tangshan earthquake (M7.8), which has its epicenter about 180 km from the survey site. Still, in the time period considered , another anomaly occurred which can not be related directly to a seismic event.