GPS Carrier-to-Noise Density Prediction Using Regression Trees

The carrier-to-noise density measured by global positioning system (GPS) receivers can be used as an indicator of signal quality and to predict the receivers' performance. It is the ratio of the carrier power to the noise power per unit bandwidth, expressed in decibel-Hertz (dB-Hz). The traditional method of estimating the carrier-to-noise density involves computing the narrow-band to wide-band power ratio of the received GPS signal in real-time. In this work, we propose three empirical models to predict the carrier-to-noise density of GPS signals based on commonly available spatial and temporal parameters, such as time of day, satellite pseudorandom noise code, azimuth, and elevation of the observed GPS satellites. Two models were created by finding the second-order polynomial best-fit, while the third model was created using bagged regression trees. Three years of observations from a GPS receiver at Sharjah (25.2827° N, 55.4621° E), United Arab Emirates, have been used in to create these three models. Among the three proposed models, the bagged regression trees model offered the best performance in terms of root-mean-squared error, mean absolute error, and coefficient of determination.