Analysis of Ionosphere Gradient Using Japan GEONET Data

Large spatial gradients in ionosphere delay are a potentially threatening error source for the Local Area Augmentation System (LAAS). Therefore a better understanding of the ionosphere behavior during strong magnetic storms is crucial for LAAS so that it can more accurately evaluate its integrity and availability during these events. In order to obtain spatially-dense information on severe ionosphere delays, we use data from the Japan GPS Earth Observation Network (GEONET). GEONET is a very dense network of dual-frequency GPS receivers with more than 1200 receivers distributed across Japan. This density provides a significant advantage to the investigation of local ionosphere behavior. There are two major difficulties with the analysis of the GEONET data. The first issue is calculating accurate ionosphere delay measurements from the raw data provided by GEONET. We use the carrier-phase measurements of L1 and L2 while applying a calibration method for the interfrequency bias (IFB) that always appears in the processing of the dual-frequency data. The second issue is to screen out “artificial anomalies” that are due to GEONET receiver errors. Since the GEONET database is very large, partial data loss or erroneous measurements are occasionally included in the raw data. This poor-quality data can result in large errors in ionosphere delay calculations, and as a consequence, the estimated delays may appear to be ionosphere anomalies in the analysis. Based on identifying the geometrical differences between real and artificial gradients, we developed an error-screening technique that automatically excludes artificial gradients and output only actual large gradients from the huge GEONET database. We investigated five strong magnetic storms using the GEONET data. We found several instances of ionosphere spatial gradients much lager than normal during these storms. However, we found no gradients extreme enough to lead to unacceptable user errors for LAAS. Through these analyses, we have demonstrated the effectiveness of our method as a tool to analyze ionosphere behavior based upon GEONET data. 1.0 INTRODUCTION The Local Area Augmentation System (LAAS) provides real-time differential GPS corrections to aircraft users to support precision-approach operations. Since the service area of LAAS Ground Facility (LGF) is limited to the range of a user aircraft’s approach, the LGF and its user aircraft are close enough to cancel out the ionosphere delays in the user’s measurements with the DGPS corrections. However, if an LGF and/or user are exposed to a drastically large ionosphere gradient, the large difference of their ionosphere delays can result in a significant position error for the user. Therefore large spatial gradients in ionosphere delay are regarded as a potentially-threatening error source for LAAS. Several examples of extremely large ionosphere gradients that could cause the significant user errors have been observed [1-4]. Datta-Barua et al. investigated the Wide Area Augmentation System (WAAS) “supertruth” data of April 6, 2000 and found an ionosphere gradient potentially as large as 316 mm/km [1]. Prior to this, the research that developed the ionosphere error model for LAAS estimated nominal ionosphere gradients at 3 ~ 5 mm/km (one-sigma) under active ionosphere conditions [5-6]. Therefore, the 316 mm/km gradient is more than 60 times the one-sigma nominal gradient. Triggered by this discovery, simulations for the purpose of quantitatively understanding the vulnerability of LAAS against the ionosphere anomalies have been conducted [7-9]. Luo et al. simulated the user position errors during hypothesized ionosphere anomalies based on an ionosphere gradient “threat space” model which explains the large ionosphere gradient with three parameters: slope of the ionosphere delay, width of the slope, and speed of the movement of the slope [8]. This work showed that anomalous gradients within the threat space model could create errors intolerable for LAAS and discussed the means to mitigate these cases by making them non-available in advance. Although a better understanding about the vulnerability of LAAS against the ionosphere anomaly has been obtained from these simulations, an important question still remains open: do the threat models used in the simulations appropriately approximate real ionosphere spatial anomalies? In order to evaluate the integrity of LAAS, the threat model must be conservative. However, if it is needlessly conservative, the evaluated availability will be unnecessarily low. Therefore, a model having appropriate and well-bounded parameters is crucial to achieving the maximum user benefits from LAAS. In order to obtain the best possible model, a good method for precisely analyzing actual ionosphere anomalies is needed. In this paper, we introduce a method to analyze ionosphere anomalies by using the data from a very-dense network of GPS receivers in Japan, GEONET [10]. GEONET consists of more than 1200 dual-frequency receivers distributed throughout Japan. This degree of receiver density provides us with visibility of ionosphere delays to within 10 km to 50 km resolution, which is a significantly better than can be done with WAAS and IGS/CORS data from the U.S. In order to effectively deal with the huge GEONET database, we have employed a “coarse-to-fine” approach. First, we use a network of 350 ~ 400 receivers and search for ionosphere gradients larger than a particular threshold. In this process, receivers exposed to the gradients are identified. Next, we examine the data from all receivers located close to the identified receivers and fully analyze the ionosphere gradients affecting them. In addition to sheer size, there are several difficulties in handling the GEONET data. The first difficulty is that, because only raw data is provided by GEONET, ionosphere delays must be calculated from this raw data. We use the carrier-phase measurements of L1 and L2 to calculate ionosphere delays. In general, combinations of dual frequency carrier-phase measurements provide us with less-noisy ionosphere delay. On the other hand, in order to calculate accurate delays, we need to calibrate the interfrequency biases (IFBs) of both receivers and satellites which come from hardware differences in the L1 and L2 signal paths. The IFB calibration problem is not a new problem in the GPS community, and several IFB-calibration method have been proposed [11-13]. We apply one of these methods to the GEONET data, which is the method introduced by Hansen [11]. Another problem is the treatment of abnormal GEOENT receiver behavior. Unfortunately, partial data loss and/or large receiver noise occur in the GEONET data on rare occasions. This poor-quality data results in large errors in the ionosphere delay calculation, and delays with large errors generally form large apparent gradients. In addition to this poor-quality data, imperfections in IFB calibration can cause ionosphere gradients to be significantly biased. These artificial gradients must be distinguished from real ionosphere anomalies. We have developed a noise-screening technique to do this which is based on the differing geometrical characteristics between artificially large gradients and actual large gradients. We have analyzed five strong magnetic storms in Japan using the GEONET data. Based on the results of Luo’s simulations [8], we have concluded that no ionosphere gradient that could cause hazardous LAAS user errors occurred in these storms. Through these analyses, our method has demonstrated its effectiveness as a tool for analyzing the local-area ionosphere anomalies. 2.0 LARGE GRADIENT SEARCH ALGORITHM In order to effectively handle the very large GEONET database, we developed a method based on a “coarse-to-fine” error analysis approach. In this section, we describe the algorithm for the “coarse” part. The objective of this algorithm is to search for ionosphere gradients larger than a threshold within the network of 350 ~ 400 receivers and identify the receivers exposed to the largest gradients. The main reason to use a subset of all GEONET receivers is to reduce the computation cost (processing time and memory limitations). This algorithm consists of two processes: ionosphere gradient calculation and noise screening. In order to clearly describe the functions of these processes, we explain each process with showing results of a data processing for an actual strong magnetic storm occurred on November 6, 2001. This is one of the magnetic storms we have analyzed. The maximum Kp index of this storm is 8, and time at which the earth is impacted is estimated to be about 09:00 Japan Standard Time (GMT + 9 hours). 2.1 IONOSPHERE GRADIENT CALCULATION Raw ionosphere delay estimates are calculated using the L1 and L2 carrier-phase measurements in GEONET data. As mentioned above, the primary difficulty of ionosphere delay calculation with dual-frequency data is dealing with the interfrequency bias (IFB) that appears in the model due to the combination of L1 and L2 measurements. The IFB is caused by hardware differences in the L1 and L2 signal paths and varies for each receiver and each satellite. For IFB calibration, we employed the method introduced by Hansen. The details of this method are presented in [11]. Here, we describe only the basics of the method. The equations in (1) express the general form of the carrier-phase measurements on L1 and L2.