We investigate how spatial coherence and temporal variability of precipitation over the western United States changes with temporal aggregation and filtering. We begin by calculating principal components (PC) from finely gridded daily precipitation anomalies across the western United States. The daily data were aggregated to form 5-day, 10-day, 20-day, monthly, and seasonal anomalies and the empirical orthogonal functions (EOFs) were estimated. Our results suggest that the leading EOFs of western United States precipitation do not depend on the temporal aggregation of the data. To examine the sensitivity of the leading empirical orthogonal functions to temporal aggregation and filtering, we then filtered the data with the discrete wavelet transform (DWT) and repeated the EOF analysis for scales of 2 days to 10 years. It appears that western precipitation is strongly persistent with EOFs that do not depend on the temporal aggregation of the data or the timescale of the analysis from 4 days to 7 years. To explore the origin of this time invariance of the EOFs, we simulated synthetic data sets and demonstrated the importance of spatial correlation on the dominant patterns of variability. Precipitation anomalies appear to display a large degree of temporal persistence that is reflected in the similarity of spatial patterns of precipitation ranging from days to seasons. Simulations suggest that the spatial correlation between adjacent grid points is the dominant factor shaping the EOFs.