Yu Jen Chung Department of Marine
Science, Naval Academy, Kaohsiung, Taiwan fc8001020@gmail.com
Yu Ru Chen Institute of Ocean
Technology and Marine Affairs, National Cheng Kung
University, Tainan, Taiwan tp6m3bj6@gmail.com
Laurence Z.H. Chuang Institute of Ocean
Technology and Marine Affairs, National Cheng Kung
University, Tainan, Taiwan zsuhsin@mail.ncku.edu.
tw
Yiing Jang Yang Institute of Oceanography, National Taiwan
University, Taipei, Taiwan yjyang67@ntu.edu.tw
Li Guang Leu Department of Marine
Science, Naval Academy, Kaohsiung, Taiwan lgleu@mail.cna.edu.tw
Abstract—High-frequency (HF) ground wave radar’s over-the-horizon property can effectively overcome the restraints posed by the curvature of the earth. HF radar transmits low-power radar waves and its detection distance is up to several hundred kilometers. However, radar in HF bands may be easily contaminated by noise, which has temporal and spatial variability, and ionospheric interference, limiting system performance. Therefore, understanding the regional properties of noise environmental is critical. This study investigates the spatial-temporal properties of noise in long-term radar data and then analyzes the relation between noise and ionospheric indices.
A comparison of the time series of ionosphere and noise indices indicates a diurnal fluctuation property. These results through cross correlation show the consistency of SUAO and HABN HF radar sites and a dramatic change of correlation coefficients over time delay. Most of the strongest negative correlations at both radar sites were found with lead time of noise index about time delay between 2 to 4 h during test period.
Keywords—HF radar; noise index; ionosphere index I. INTRODUCTION
High-frequency (HF) ground wave radar transmits low-power radar waves. Its detection distance is up to several hundred kilometers with little attenuation. Since the sea surface is an excellent conductor, HF radar’s over-the-horizon property can effectively overcome the limits caused by the curvature of the earth [1, 2]. HF radar has many advantages, such as low cost, active detection, and near-real-time performance for various applications, especially in large-area monitoring [3, 4].
However, radar in HF bands may be easily contaminated by noise, which has temporal and spatial variability, and ionospheric interference, limiting system performance. Taiwan is located in an equatorial ionization anomaly area, where the influence of the ionosphere is stronger, and thus understanding the properties of noise and interference is necessary.
Various filtering techniques that depend on noise distribution and amplitude are used to eliminate environmental noise. A background for different kind noise environments in detection algorithm can be difficult to construct and has no unique procedure. Therefore, understanding the spatial-temporal properties of environmental noise using regional analysis is critical. This study investigates the spatial-temporal
properties of a noise environment in long-term radar data, and then determines the appropriate index for measuring ionospheric clutter. Finally, a correlation analysis, including auto- and cross-correlation, is applied to the long-term radar data to determine the relations between these time series.
II. HFRADAR AND IONOSPHERIC CLUTTER A. HF radar system
The SeaSonde long-range system, uses a three-element receive antenna and the multiple signal classification (MUSIC) algorithm to determine a target’s bearing, and operates at 4.3-5.4 MHz to achieve an average range of 100-220 km with a range resolution of 3-12 km [5]. The range series and Doppler spectra of this system are obtained using the first and second fast Fourier transforms of the time series data collected using its Acquisition program.
Fig. 1. SUAO and HABN, located on the east coast of Taiwan, are two long-range SeaSonde HF radar sites for ocean current measurements (http://www-codar.oc.ntu.edu.tw).
SUAO and HABN, located on the east coast of Taiwan, are two long-range SeaSonde HF radar sites with central frequency 4.4 MHz for ocean current measurements, as shown in Figure 1 [6]. The data from these two long-range SeaSonde HF radar sites for the year 2013 were used to investigate the local properties of environmental noise. For both radar sites, noise index estimation was used to identify unique regional
characteristics. The noise index, which is used to express the spatial-temporal distribution of environmental noise, was developed in this study to understand noise distribution.
B. HF radar data
The Doppler spectra of the SeaSonde HF radar system is referred to as the range-Doppler (R-D) spectra or R-D map, as shown in Figure 2. Data collected by HF radar contains various signal components, including first-order sea echoes, second-order continuum, vessel echoes, and radar interference, etc.
The dominant first-order sea echoes, due to Bragg resonance [7], make it possible to estimate ocean current information. The raw R-D spectra, not averaged by CSPro, contain 6 spectra series, including the auto-spectra and cross-spectra series of three antennas. In order to accurately estimate the noise distribution, the auto-spectra of the monopole antenna were used to measure the noise in this study. Figure 2 shows the SeaSonde R-D spectra measured at night on October 29, 2013.
Fig. 2. SeaSonde R-D spectra measured at nigh on October 29, 2013.
C. Ionospheric clutter
The ionosphere, a region of the Earth’s upper atmosphere that influences radio propagation to distant places, is ionized by solar radiation. The formation of the ionosphere is due to the dissociation of the atmosphere, which forms positive ions and free electrons under ultraviolet light. The maximum density of free electrons is in the mid-level, and forms D, E, and F layers (tens to hundreds kilometers high). The ionosphere is the plasma form of ionized gas, and the F layer has the highest density and the largest scale of plasma [8]. The ionosphere has several types of variation, including solar cycle, diurnal variation, seasonal variation, geographical variation, and temporal variation. At night, ionization in the E and D layers of the ionosphere is extremely weak; the F layer is the only significant layer of the ionosphere. In the daytime, the D, E, and F layers become much more ionized, creating an additional, weaker region known as the F1 layer. The F2 layer is present in the daytime and at night, and is an important region for electromagnetic wave refraction [8].
HF radar, a surface wave system, illuminates a certain area with a vertically polarized electromagnetic wave along the sea surface. However, part of the electromagnetic waves emitted by HF radar may be directed upward towards the atmosphere, resulting in ionospheric clutter due to its reflection by the ionosphere. Since Taiwan is located in an equatorial ionization anomaly area, the ionosphere affects it severely, and thus understanding ionospheric variations is critical.
The D layer of the ionosphere absorbs the electromagnetic wave, preventing the echo from returning to the sea surface
during the daytime. At night, when the D layer is not present, the radar echo is reflected from the F layer. This kind of clutter, similar to noise caused by the sea state, may affect system performance at night. Second-order ionospheric clutter, one of three types of ionospheric clutter, commonly occurs with the condition of occurrence of Bragg resonance. This condition requires ocean waves that have wavelength that is now equal to that of the radar wavelength [9].
III. METHODOLOGY A. Noise index
In order the assess HF radar noise, the noise index is computed by averaging the energy in the monopole auto-spectra over Doppler bins that are assumed to contain only noise. In an R-D spectrum, there are various signals between first-order Bragg lines. Most signals have intensities starting to remain low level from the position of the theoretical second-order Bragg lines. Using noise estimation methods for similar systems [10, 11], the proper average regions were determined from the frequencies of the second-order Bragg lines to the wings of each side. The noise level was estimated at each range cell of a specific spectrum, as shown in Figure 3. The coherent integration time was 1,024 s, so there were 3 or 4 R-D spectra within 1 h.
Fig. 3. Average regions of noise level estimation for specific range cell [12].
B. Ionosphere Index
The I95 index provides information about the influence of differential ionospheric biases as experienced by users of differential GPS positioning [13, 14]. Thus, based on the ionospheric model coefficients [13], the I95 ionospheric index was chosen to assess the ionospheric clutter in this study. The index values of I95, which use a 95% margin of all ∆I values, are computed based on the differential ionospheric biases in the directions south-north (∆ILAT) and west-east (∆ILON) per hour.
This information can be combined as [15]:
∆𝐼𝐼 = �∆𝐼𝐼𝐿𝐿𝐿𝐿𝐿𝐿2 + ∆𝐼𝐼𝐿𝐿𝐿𝐿𝐿𝐿2 (1)
C. Cross-correlation of the time series
Cross-correlation, a measure of similarity of two similar series as a function of the shift of one relative to the other, is commonly used in signal processing. In this section, the cross-correlation of the time series of noise and ionospheric index
-0.4 -0.2 0.0 0.2 0.4
Doppler Frequency (Hz)
Power Density (dBm)
Positive 1stOrder Bragg Peak Negative 1stOrder
Bragg Peak Zero Doppler
Frequency
Positive Average region Negative Average
region
values is analyzed to measure their similarity. Let x(t) and y(t) be functions of time and τ be the time delay. Then, the cross-correlation 𝑅𝑅𝜏𝜏 is defined as:
𝑅𝑅𝜏𝜏 = ∫ 𝑥𝑥(𝑡𝑡)𝑦𝑦(𝑡𝑡 + 𝜏𝜏)𝑑𝑑𝑡𝑡−∞∞ (2)
Therefore, the time delay τ can be determining from the cross-correlation analysis of two series.
IV. RESULTS AND DISCUSSION
The spatial-temporal analysis of regional noise and the results of correlation analysis for the long-term radar data are described below.
Fig. 4. Mean noise levels at SUAO and HABN HF radar sites in 2013 [12].
A. Distribution of HF noise environment
The above signal processing procedure was used the estimate the HF radar system noise environment from the noise index values. The outputs were plotted and used to evaluate the two HF radar sites. The mean hourly noise levels at the SUAO and HABN HF radar sites, shown in Figure 4, show that the noise levels at these sites remained at a high level from 9:00 to 21:00 , significantly decreased from 22:00 to 3:00, gradually increased from 4:00 to 8:00, and rapidly increased to a value of higher than -147 dBm at 9:00. It is worth mentioning that the minimum mean noise level was from 3:00 to 5:00. The maximum mean noise level was around 11:00 UTC (universal time coordinated), corresponding to a local time of 19:00, which is the start of nighttime in Taiwan.
B. Correlation analysis
The time series of the noise index outputs exhibits a diurnal cycle, which is related to temporal deviations in the ionospheric clutter driven principally by solar radiation. In this section, the noise index values and ionospheric index values are compared to explore their relationship and then correlation coefficients are evaluated to estimate the strength and direction of the linear relationship between the two types of index value.
Finally, cross-correlation analysis is used to determine the time delay between a noise level value and an I95 value.
Figure 5 shows a comparison between I95 and noise values at the SUAO HF radar site. There is a diurnal fluctuation and a negative correlation between the ionospheric index and noise level at a certain time period in December, 2013. A weak downhill linear relationship was determined from auto-correlation analysis. The three levels of I95 values denote different degrees of ionospheric influence. These levels are normal, median, and high activity, with values of 8, 4, and 2, respectively. Most I95 values are greater than the median activity in December, 2013. There are four peaks of the I95
values during the nighttime period in Figure 5. Their values remain around 30, which is much greater than the high activity value. At the HABN HF radar site, the noise level values conversely remain relative lower level at the same time, represents their diurnal properties.
Fig. 5. Comparison between I95 and noise level at SUAO HF radar site in certain period in December, 2013; the abscissa denotes the hour of the day; the ordinate denotes the noise values in dBm at right side and ionospheric index values at left side.
Although it appears that there is a clear relation between the noise index and the ionospheric index, these data are weakly negatively correlated, and thus it is reasonable to assume that there exists a factor of time displacement that influences the relationship. Adjusting the time delay of the time series of noise and ionospheric indices may thus improve the relation strength. Cross-correlation analysis was applied to confirm this assumption. Figures 6 and 7 show the cross-correlation results of SUAO and HABN HF radar sites with time delays of -6 to 6 h on December 9 to 12, 2013. The results show a dramatic change over time delay. The strongest negative correlation was found between 2 to 4 h, which means that the noise values lead the I95 values by about 2 to 4 h. Similar results were found for both sites.
C. Discussion
The ionosphere is mainly influenced by solar irradiation and solar periodical activity. The strength of solar radiation determines the variation of ionized gas concentration, and is affected from time, space, and solar activity. Due to the strong sunlight in the daytime, severe ionization leads to a high density of free electrons, which at the lower level of the atmosphere dissociate and form the D and E layers (including sporadic-E) [16]. The D layer, located at 50 to 100 km above sea level, is easily absorbing HF radar echoes, especially at noon. Theoretically, most backscattering energy of HF radar is absorbed by this layer. At night, there is lower production of
110 120 130 140 150 160 170 180 190
0
positive ions and free electrons, and thus the plasma density in the ionosphere decreases; the D or E layer may not clearly exist due to the dissociative recombination of ions and electrons.
Similarly, the F1 and F2 layers recombine to become the F layer at night. Although the F layer is relatively high above sea level, it has a higher density than D or E layer and easily backscatters HF radar echoes. The recombination in the D and E layers results in backscattering from HF radar through the F layer at night. Either the electromagnetic wave from the system itself or interference from other regions may easily increase the radar noise level due to the disappearance of the D layer.
Above results proved the diurnal property. Moreover, the time delay which noise leads ionospheric index between 2 to 4 h when the strongest negative correlations occurred may help estimate the ionospheric activity. Therefore, in the daytime, there is less interference from the ionosphere, making measurements more accurate, especially for vessel detection.
Fig. 6. Crosscorrelation results for SUAO HF radar site with time delay of -6 to -6 h on December 9 to 12, 2013.
Fig. 7. Crosscorrelation results for HABN HF radar site with time delay of -6 to -6 h on December 9 to 12, 2013.
V. CONCLUSION
Long-range SeaSonde HF radar data from two locations on the coast of Taiwan collected in 2013 were analyzed. The distribution of environmental noise at HABN and SUAO HF radar sites was characterized using the noise index, and then correlation analysis was applied to the noise index and ionospheric clutter values to determine their relation. A comparison of the time series of ionosphere and noise indices
indicates a diurnal fluctuation property and a weak negative correlation. The cross-correlation analysis results show the similarity of data at both sites and an undulating fluctuation of correlation coefficients over time delay. The strongest negative correlations about time delay between 2 to 4 h at SUAO and HABN HF radar sites occurred when time of noise leads that of ionospheric index. These results may be used to assess the effects of the ionosphere and help improve the accuracy of present applications.
ACKNOWLEDGMENT
This work was funded by the Ministry of Science and Technology under grants MOST 105-2221-E-006-137 and MOST 106-2623-E-002-002-D. The authors would like to thank the National Land Surveying and Mapping Center, Ministry of the Interior for providing ionosphere data.
REFERENCES
[1] Headrick, J.M and Skolnik, M.I., “Over-the-Horizon in the HF Band”, Proceeding of IEEE, Vol. 62, pp. 664-673 (1974).
[2] Skolnik, M.I., Introduction to Radar Systems, McGraw-Hill, 3rd Edition (2001).
[3] Chung, Y.J., Chuang, L.Z., and Yang, W.C., “Feasibility studies of ship detections using SeaSonde HF radar”, proceeding of the IEEE Geoscience and Remote Sensing Symposium, pp. 2892-2895 (2013).
[4] Chuang, L. Z., Chung, Y. J., and Tang, S. T., “A Simple Ship Echo Identification Procedure With SeaSonde HF Radar”, IEEE Geoscience and Remote Sensing Letters, Vol. 12, Iss. 12, pp. 2491-2495 (2015).
[5] CODAR (online). Retrieved July 15, 2015b, from http://support.
codar.com.
[6] Fang, Y.C., Wang, J., Yang, Y. J., Mau J.C., and Huang, M.J.,
“Preliminary Results of CODAR Surface Current Observations Northeast of Taiwan”, International Workshop on Operating System for Marine Environment Monitoring and Forecasting, pp. 59-77 (2011).
[7] Crombie, D.D., “Doppler Spectrum of Sea Echo at 13.56 Mc/s”, Nature, Vol. 175, pp.681-682 (1955).
[8] Davies, K., Ionospheric Radio Waves, Blaisdell Publishing Company, Waltham (1969).
[9] Ponsford, A.M. and J. Wang. A Review of High Frequency Surface Wave Radar for Detection and Tracking of Ships. Turkish Journal of Electrical Engineering & Computer Sciences, Vol. 18, Iss. 3, pp. 409-428 (2010).
[10] Cosoli, S., Bolzon, G. and Mazzoldi, A., “A Real-Time and Offline Quality Control Methodology for SeaSonde High-Frequency Radar Currents”, Journal of Atmospheric and Oceanic Technology, Vol. 29, Iss. 9, pp. 1313-1328 (2012).
[11] Whelan, C., Barrick, D., Emery, B. and Washburn, L., HF Radar Calibration with Automatic Identification System Ships of Opportunity Phase II Final Report (2011).
[12] Chung, Y. J. Echo Identification of Vessel Activities Using HF Radar and Relevant Threat Evaluation, doctoral dissertation, National Cheng Kung University, Tainan (2015).
[13] Wanninger, L., “The Performance of Virtual Reference Stations in Active Geodetic GPS-Networks under Solar Maximum Condition”, Proceeding of ION GPS 99, Nashville TN, pp. 1419-1427 (1999).
[14] Wanninger L., “Ionospheric disturbance indices for RTK and Network RTK positioning”, Proceedings of ION GNSS 2004, Long Beach, California, pp. 1849-2854 (2004).
[15] Chen, X., Landau, H., Vollath, U., “New Tools for Network RTK Integrity Monitoring”, Proceedings of IONGPS/ GNSS 2003, pp.
1355-1361 (2003).
[16] Davies, R. M. & Smith, E. K. Sporadic E at VHF in the U.S.A., Proc.
IRE, 47, 762 (1959).