Measurements of length of day using the Global Positioning System

Length-of-day (LOD) estimates from the seven Global Positioning System (GPS) analysis centers of the International GPS Service for Geodynamics have been compared to values derived from very long baseline interferometry (VLBI) for a recent 16-month period. All GPS time series show significant LOD biases which vary widely among the centers. Within individual series the LOD errors show time-dependent correlations which are sometimes large and periodic. Clear correlations between ostensibly independent analyses are also evident. In the best case the GPS LOD errors, after bias removal, approach Gaussian with an intrinsic scatter estimated to be as small as similar to 21 mu s/d and a correlation time constant of perhaps 0.75 day. Integration of such data to determine variations in UT1 will have approximately random walk errors which grow as the square root of the integration time. For the current best GPS performance, UTI errors exceed those of daily 1-hour VLBI observations after integration for similar to 3 days. Assuming the stability of LOD biases can be reliably controlled, GPS-derived UT1 can be useful for near real time applications where otherwise extrapolations for several days from the most current VLBI data can be inaccurate by up to similar to 1 ms.