Bridging Real-Time Precise Point Positioning in Natural Hazard Warning Systems during Outages of MADOCA Corrections

Abstract

Real-time Precise Point Positioning (RT-PPP) is the primary positioning method used in natural hazard warning systems (NHWS), e.g. for monitoring tsunami and earthquakes. The Japanese Multi-GNSS Advanced Demonstration tool for Orbit and Clock Analysis (MADOCA) is a promising service that enables RT-PPP. Currently it includes GPS, GLONASS and QZSS orbits and clock corrections in addition to code biases. However, one concern for continuous hazard monitoring by using RT PPP is the severe decline of positioning accuracy if a discontinuity in receiving these corrections occur, for instance due to a temporary user modem failure. In this paper, we present a method that can sustain RT PPP with 3D accuracy less than 20 cm when such a break takes place. For short outages less than 30 minutes we predict MADOCA orbits using a Holt-Winters’ autoregressive model, and for longer outages up to 2 hrs, the most recent International GNSS Service (IGS) ultra-rapid orbits can be used for GPS observations. Moreover, the clock corrections are predicted as a time series using a joint quadratic polynomial and sinusoidal model. The best regression period to estimate the required model parameters is discussed based on autocorrelation analysis of the corrections. The time lengths of the sinusoidal terms are estimated from analysis of the data in the frequency-domain. The prediction model parameters are estimated sequentially using a sliding time window with short intervals to reduce the computational load. Evaluation of the proposed method is performed at a site resembling a NHWS station and positioning accuracy were compared for the cases when using the original corrections and when using the predicted corrections for 1 hr, assuming that within this period the outage can be fixed. The experimental results proved validity of the presented approach where positioning accuracy of 20 cm was maintained during the prediction period.