Effect of sporadic E clouds on GPS radio occultation signals

GEOPHYSICAL RESEARCH LETTERS, VOL. 37,, doi:10.1029/2010gl044561, 2010 Effect of sporadic E clouds on GPS radio occultation signals Z. Zeng 1 and S. Sokolovskiy 1 Received 1 July 2010; revised 9 August 2010; accepted 12 August 2010; published 29 September 2010. [1] In this study the effects of sporadic E (Es) clouds on GPS radio occultation (RO) signals are investigated by modeling wave propagation and analyzing observational GPS RO data. It has been shown that when the Es clouds are aligned with the propagation direction, they cause defocusing of the GPS RO signal, accompanied by scintillation above and below the defocusing region (due to the interference of direct and refracted radio waves). These effects result in specific U shape structures in the amplitude of the GPS RO signals. When Es clouds are tilted with respect to the propagation direction, the effects reduce and disappear with the increase of the tilt angle. The U shape structures are clearly identified in the observed GPS RO signals mainly at tangent point (TP) heights 90 120 km, but also at much lower TP heights. The latter indicates that some Es clouds are tilted with respect to the local horizon (this has also been shown in other studies). The distributions of the observed U shape structures in latitude, local time, tilt angle, and vertical thickness of the cloud are evaluated in this study based on one month of COSMIC (Constellation Observing System for Meteorology, Ionosphere and Climate) RO data in July 2009. Citation: Zeng, Z., and S. Sokolovskiy (2010), Effect of sporadic E clouds on GPS radio occultation signals, Geophys. Res. Lett., 37,, doi:10.1029/2010gl044561. 1. Introduction [2] The sporadic E (Es) layer manifests itself as a nonuniform, wavy layer, or multiple layers, or a composition of irregular elongated clouds of intense ionization within the lower E region. The appearance of the Es clouds has well pronounced seasonal and latitudinal dependencies [Wu et al., 2005] and less pronounced dependency on local time. It is commonly accepted that the Es layer is formed by tidaldriven wind shear that compresses the metallic ions into a thin layer [Whitehead, 1989]. Several mechanisms can be responsible for the quasi periodic (QP) irregularities of the Es layer. Woodman et al. [1991] explain the QP irregularities by gravity waves. Tsunoda et al. [1994] considered polarization electric field produced by gravity wave modulation to explain some effects observed in the QP Es structures. Larsen [2000] considers Kelvin Helmholtz (K H) instability responsible for the QP structure of the Es layer. Bernhardt [2002], by numerical modeling of the K H instability, obtained both QP structure and splitting of the Es layer into two layers. Cosgrove and Tsunoda [2004] consider coupling between E and F regions of the ionosphere to deduce characteristics of instability in both regions. A review of the mechanisms 1 COSMIC, University Corporation for Atmospheric Research, Boulder, Colorado, USA. Copyright 2010 by the American Geophysical Union. 0094 8276/10/2010GL044561 of formation of irregular Es layer is given by Mathews [1998]. [3] The Es clouds have much higher electron density than the regular E layer, and they affect radio wave propagation in the HF and lower frequency bands. When the Es clouds are aligned with the propagation direction, they also affect propagation in the VHF and UHF bands. In particular, they affect the GPS radio occultation (RO) signals used for remote sensing of the atmosphere from low Earth orbiting (LEO) satellites. Over the past decades, the Es clouds were extensively studied by different methods, including ionosondes, incoherent scatter radars and backscatter radars operating from the ground [Whitehead, 1989; Houminer et al., 1996; Mathews, 1998]. Also, two dimensional images of the Es layer were obtained by radio beacon tomography [Bernhardt et al., 2005] and by radio induced aurora [Djuth et al., 1999; Kagan et al., 2000; Bernhardt et al. 2003]. Recently GPS RO has been used for studies of the Es clouds based on the fluctuation of RO signals [Hocke et al., 2001]. Wu et al. [2005] evaluated global Es morphology based on Challenging Minisatellite Payload (CHAMP) data; Arras et al. [2009] evaluated the Es tidal signature based on CHAMP, Gravity Recovery and Climate Experiment (GRACE) and Constellation Observing System for Meteorology, Ionosphere and Climate (COSMIC) data. [4] In this study, we model the propagation of GPS RO signals through Es clouds and demonstrate that their effects result in specific U shape structures in the amplitude of RO signals, which have not been reported in other studies. These structures are identified in COSMIC RO signals. We present examples of the modeled and observed structures. By applying an algorithm for identification of these structures for one month of COSMIC data, we evaluate the distributions of the observed Es cloud events in latitude and local time (consistent with results from previous studies); in the observational height (this allows evaluation of the distribution of the Es cloud tilts with respect to local horizon); and in the vertical thickness of the cloud. 2. Modeling of the Effects of Es Clouds on Propagation of the GPS RO Signals [5] Despite numerous studies there is still some controversy about the structure of Es layer. Most commonly, Es layer is considered to be composed of separate elongated clouds. Some studies report multiple layers or location of the clouds at multiple dedicated heights. Radar studies from the ground report concaved shape of the clouds. In the GPS RO, the Es layer often causes strong scintillation of RO signals. Although most intensive scintillation is commonly observed at tangent point (TP) heights h of about 90 120 km (where Es clouds are located), in many cases scintillation is also observed at lower h, i.e. is caused by irregularities located far away from TP. In order to explain scintillation caused by Es 1of5

Figure 1. study. Layout of the model of Es cloud used in this clouds, Wu et al. [2005] considered a model of horizontally elongated, vertically modulated ionized structures. While such structures may exist, their effect on propagation is significantly reduced with increasing distance from TP. In order to produce significant scintillation, the elongated Es cloud located far away from TP must be aligned with the propagation direction and not with the local horizon (i.e. the cloud must be tilted). The tilts can result from the modulation of the height of Es layer by gravity waves [Woodman et al., 1991] and K H instability [Larsen, 2000; Bernhardt, 2002]. The evidence of Es cloud tilts and their estimates, ranging from several deg to about 20 deg, are available from some studies [From and Whitehead, 1978; Parkinson and Dyson, 1998; Chu and Yang, 2009]. As for the scintillation, it can be caused by multipath propagation behind an elongated refractivity structure without the vertical modulation of the refractivity inside the structure. [6] In this study we consider a simple model of the Es cloud that has the horizontal extension L and the vertical distribution of electron density N e (z) =N e0 W H (z,dh), where W H (z,dh) is Hanning window of the width DH. The cloud may be aligned or tilted with respect to the propagation direction. The angle a denotes the tilt of the cloud in the vertical plane coinciding with the propagation direction (hereafter referred to as occultation plane), as this is schematically shown in Figure 1. The refractivity inside the cloud is negative N ffi 40N e / f 2 (where N e is in el/cm 3 and f is the carrier frequency in MHz). When the Es cloud is aligned with the propagation direction, its central part results in the defocusing of radio waves, while the edge parts (where N e (z) has opposite convexity) result in the focusing. When the Es cloud is tilted by an angle a > D H / L, it does not affect propagation (in terms of bending of rays) between heights z 1 and z 2,as shown in Figure 1. For modeling of the wave propagation, we apply the multiple phase screen method [Sokolovskiy, 2001], using the incident plane wave, and a distance of 3,000 km between the Es cloud and the receiver (this is consistent with the GPS RO observations from a LEO at 800 km orbit height). [7] Figure 2 shows amplitudes of the modeled RO signals after propagation through the model of Es cloud with different parameters, N e0, L, DH and a (indicated in the Figure 2 caption). For a = 0 (Figure 2, top), the amplitude in the shadow of the cloud is reduced due to defocusing. The spikes due to focusing and scintillation (caused by the interference of direct and refracted waves) occur above and below the shadow region. Extension of the regions of strong scintillation depends on the maximal bending angle which is proportional to N e0 and L, and inverse proportional to DH. These effects, resulting in U shape amplitude structures, are seen in Figure 2 (top). We tested other distributions N e (z), in particular, resulting in smaller focusing above and below the shadow region, but the U shape structures remain due to the scintillation caused by multipath. In reality, layers or clouds are never strictly plain but have some curvature. In order to produce noticeable propagation effect, the section of the layer satisfying the condition: the distance between the arc and the chord is smaller than the thickness of the layer, must have sufficient horizontal extension. Modeling of the propagation through a continuous wavy layer (not shown here) results in multiple U shape structures corresponding to the regions of the layer aligned with the propagation direction. When the thickness of the Es cloud DH is comparable or larger than the Fresnel zone ( 1 km for the GPS observations from LEO), the extension of the region of reduced amplitude is consistent with DH and thus can be used for evaluating DH. When DH reduces below 1 km (case F), the extension of the region of reduced amplitude does not reduce but remains at about 1 km due to diffraction (the scintillations around this region are interpreted as the diffraction pattern). Tilt of the Es cloud by an angle a > DH /L, substantially reduces the propagation effects (Figure 2, bottom). The extension of the region where the Es cloud does not affect propagation (z 2 z 1 in Figure 1) increases with increasing a. 3. Observations of the Effects of Es Clouds on the GPS RO Signals [8] The modeled structures, such as those shown in Figure 2 (top), are clearly seen in the observed GPS L1 RO Figure 2. Modeled effects of the Es clouds on the amplitude of GPS RO signals for different parameters of the model (the modeled clouds are shifted in h by 20 km for better visualization). (top) a =0(A F); N e0 =6 10 5 el/cm 3 (A, B), 3 10 5 el/cm 3 (C F); L = 100 km (A, C); 50 km (B, D); 25 km (E); 10 km (F); DH =4km(A C), 2 km (D), 1 km (E), 0.5 km (F). (bottom) N e0, L, DH from the case (A) in Figure 2 (top) and a = 1 deg (A), 3 deg (B), 5 deg (C), 7 deg (D), 9 deg (E). 2of5

Figure 3. The amplitudes of COSMIC RO signals showing the U shape structures similar to those modeled in Figure 2. signals (the L2 signal, tracked in the semi codeless mode in COSMIC receivers, becomes very noisy and practically unusable in the presence of scintillations induced by Es clouds). Figure 3 shows examples of L1 amplitude (C/A signal to noise ratio (SNR)) from COSMIC observations in July 2009. The structures similar to that modeled in Figure 2 are seen and observed at different TP heights h. Those observed at h well below the height of the Es layer, 90 120 km, apparently correspond to tilted Es clouds, approximately aligned with the direction from GPS to receiver. For h < 20 km the effect of the neutral atmosphere on amplitude may become significant and observations of the Es clouds are less reliable. This limits the observations of the Es cloud tilts projected on the occultation plane by approximately a < 9 10 deg. [9] For statistical analysis we use the following criteria to identify the effects of Es clouds (or Es events) in the observational data. The amplitude A of the GPS L1 RO signal (C/A SNR sampled at 50 Hz) is represented as the function of h and the A(h) is smoothed by the sliding averaging with 1 and 20 km windows, A(h) and A(h). The Es event is identified if the following conditions are satisfied: (i) A(h) < 0.6 A(h), in this case the interval (h1, h2) where A(h) < 0.95A(h) is associated with the thickness of the cloud DH; (ii) the standard deviation of A(h) calculated in 1 km interval at the center of the interval (h1, h2) (or in the interval (h1, h2), if h2 h1 < 1 km) is smaller than both standard deviations of A(h) calculated in the intervals (h1, h1 1km) and (h2, h2 + 1km), and at least three times smaller than any one of them. [10] Figure 4 shows distributions of the Es events (identified with the above described criteria) in latitude and h (Figure 4a) and in latitude and local time (Figure 4b) in July 2009. Both distributions agree well with results of previous studies [Wu et al., 2005; Arras et al., 2009]. Figure 4c shows distribution of the Es events in h (in terms of the numbers of events in 1 km bins). The Es events observed at h < 90 120 km show that the Es clouds are aligned with the propagation direction, i.e., tilted with respect to local horizon (possible mechanisms of the tilts are mentioned in the section 2). Figure 4d shows distribution of the Es events in the thicknesses of the Es cloud DH estimated as explained above (in terms of the numbers of events in 0.1 km bins). Solid and dashed lines correspond to 80 km < h < 125 km and h < 80 km. The position of the maximum of the distributions DHmax 1.5 km is consistent with the estimates of the thickness of the Es clouds, 0.6 2 km, by Houminer et al. [1996]. The difference of the distributions at DH > DHmax may indicate, on average, smaller thickness of the tilted than of the horizontal Es clouds. This could be related to effect of the gravity wave on a spherical layer, which modulates not only the height but also the thickness of the layer [Woodman et al., 1991], however this may need further validation. [11] The used method of detecting Es clouds is based on identifying specific structures in RO signals and, generally, the results depend on tunable parameters used in the identification. Testing shows that changing the parameters within reasonable limits (based on the modeled structures in Figure 2) affects the total number of the identified Es events, but does not significantly affect the distributions of the Es events in latitude, local time and h (Figures 4a 4c). However, the distribution of the thicknesses of Es clouds (Figure 4d) is more affected. In particular, the structure of this distribution for DH < DHmax is affected by the smoothing window which was chosen in accordance with the Fresnel s scale 1 km. With the window reduced beyond the Fresnel s scale, this distribution will be affected by diffraction effects (see discussion of the simulations in section 2). In any case, this distribution may not reliably reproduce the true distribution of the thicknesses for DH < 1 km. [12] The distribution in Figure 4c can be treated in terms of the probability density wh of observing the Es event at a height h. The wh depends on the probability density wb of an Es cloud to be tilted at an angle b with respect to local horizon. Though solving for wb from wh is an under determined problem, some useful information about wb can be obtained Figure 4. (a d) Distributions of the Es cloud events in COSMIC data from July 2009. See text for details. In Figure 4d, left and right Y axes correspond to solid and dashed lines. 3 of 5

extensions or less regular structures or larger curvature and thus produce less effects on RO signals. Figure 5. Distributions w h (and w b ) modeled for z e = 105 km, H e = 40 km. Other parameters: Db =3.75deg, C =0.05(solidline);Db =3.75deg,C = 0 (dashed line); Db = 7.5 deg, C = 0 (dot dashed line). by modeling w b under certain reasonable assumptions. The angle a, which characterizes the tilt of an Es cloud pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi in the occultation plane, is related to b by: tan a =tanb / 1 þ tan 2 where is the rotation angle of the Es cloud around the local vertical, thus a b. Under the assumption of equal probability of an Es cloud to be characterized by an angle, the probability density w a to observe the tilt a in the occultation plane is: w ðþ Z =2 1 þ tan 2 w ðþpffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffid ð1þ tan 2 tan 2 The probability density w h is then expressed through the integral along the line of sight: w h ðhþ Z 1 h rdz w z ðþw z ½arccosðp=rÞŠpffiffiffiffiffiffiffiffiffiffiffiffiffiffi r 2 p 2 where p = r e + h, r = r e + z (r e is the Earth s radius) and w z is the probability density of an Es cloud to be located at a height z. For modeling w h by equations (1) and (2) we use the following models: w z (z) W H (z z E, H E ) where W H is the Hanning function, z E and H E are median height and thickness of the E layer, and w b (b) exp[ (b/db) 2 ]+C. Figure5 shows w h modeled for different parameters z E, H E, Db and C as indicated in the caption. A distinctive feature of the distribution in Figure 4c is a rather slow decay well below the E layer, i.e., for h <60 80 km. It is difficult to model such slow decay with only Gaussian distribution w b without an offset C. In Figure 5, the solid line, which results in better modeling of w h at h <60 80 km, corresponds to C = 0.05, while two dashed lines correspond to C = 0 and different Db (the distributions w b are shown in the internal panel). This indicates that the distribution of the Es cloud tilts is changing rather slowly for the tilt angles greater than 5 10 deg. Most clouds have tilts within 5 10 deg. This is a smaller number than 10 20 deg found from ionograms by Paul [1990]. This may be related to that the Es clouds with large tilts have less ð2þ 4. Conclusions [13] In this study, by modeling wave propagation of GPS signals received in LEO, we found that Es clouds result in specific U shape structures in amplitude of the received RO signals. Such structures are identified in the GPS RO observations. By detecting the U shape amplitude structures in COSMIC RO signals from July 2009, we estimated their distributions in latitude, local time and TP height. The first two distributions are consistent with the distributions of the Es clouds obtained in other studies. We have used the distribution of the observed U shape structures in TP height for evaluating the distribution of the tilts of the Es clouds, approximated by Gaussian function and additive constant in the interval of tilt angles between 0 and 9 10 deg. The effect found and investigated in this study provides a tool for monitoring Es clouds and studying their morphology by the use of GPS RO observations. [14] Acknowledgments. This study was supported by the National Science Foundation under Cooperative Agreement 0918398, CSA ATM 0939962. The authors are grateful to anonymous referees for their useful comments. References Arras, C., C. Jacobi, and J. Wickert (2009), Semidiurnal tidal signature in sporadic E occurrence rates derived from GPS radio occultation measurements at higher midlatitudes, Ann. Geophys., 27, 2555 2563, doi:10.5194/angeo-27-2555-2009. Bernhardt, P. A. (2002), The modulation of sporadic E layers by Kelwin Helmholtz billows in the neutral atmosphere, J. Atmos. Sol. Terr. Phys., 64, 1487 1504, doi:10.1016/s1364-6826(02)00086-x. Bernhardt, P. A., N. A. Gondarenko, P. N. Guzdar, F. T. Djuth, C. A. Tepley, M. P. Sulze, S. L. Ossakow, and D. L. Newman (2003), Using radio induced aurara to measure the horizontal structure of ion layers in the mesosphere and lower thermosphere, J. Geophys. Res., 108(A9), 1336, doi:10.1029/2002ja009712. Bernhardt, P. A., C. A. Selcher, C. L. Siefring, M. Wilkens, C. Compton, M. Yamamoto, S. Fukao, T. Ono, M. Wakabayashi, and H. Mori (2005), Radio tomographic imaging of sporadic E layers during SEEK2, Ann. Geophys., 23, 2357 2368, doi:10.5194/angeo-23-2357-2005. Chu, Y. H., and K. F. Yang (2009), Reconstruction of spatial structure of thin layer in sporadic E region by using VHF coherent scatter radar, Radio Sci., 44, RS5003, doi:10.1029/2008rs003911. Cosgrove, R. B., and R. T. Tsunoda (2004), Instability of the E F coupled nighttime midlatitude ionosphere, J. Geophys. Res., 109, A04305, doi:10.1029/2003ja010243. Djuth, F. T., et al. (1999), Large airglow enhancements via wave plasma interactions in sporadic E, Geophys. Res. Lett., 26, 1557 1560, doi:10.1029/1999gl900296. From, W. R., and J. D. Whitehead (1978), On the peculiar shape of sporadic E clouds, J. Atmos. Terr. Phys., 40, 1025 1028, doi:10.1016/ 0021-9169(78)90006-5. Hocke, K., K. Igarashi, M. Nakamura, P. Wilkinson, J. Wu, A. Pavelyev, and J. Wickert (2001), Global sounding of sporadic E layers by GPS/ MET radio occultation experiment, J. Atmos. Sol. Terr. Phys., 63, 1973 1980, doi:10.1016/s1364-6826(01)00063-3. Houminer, Z., C. J. Russell, P. L. Dyson, and J. A. Bennett (1996), Study of sporadic E clouds by backscatter radar, Ann. Geophys., 14, 1060 1065. Kagan, L. M., M. C. Kelley, F. Garcia, P. A. Bernhardt, F. T. Djuth, M. P. Sulzer, and C. A. Tepley (2000), The structure of electromagnetic waveinduced 557.7 nm emission associated with a sporadic E event over Arecibo, Phys. Rev. Lett., 85, 218 221, doi:10.1103/physrev- Lett.85.218. Larsen, M. F. (2000), A shear instability seeding mechanism for quasiperiodic radar echoes, J. Geophys. Res., 105, 24,931 24,940, doi:10.1029/ 1999JA000290. 4of5

Mathews, J. D. (1998), Sporadic E: Current views and recent progress, J. Atmos. Sol. Terr. Phys., 60, 413 435, doi:10.1016/s1364-6826(97) 00043-6. Parkinson, M. L., and P. L. Dyson (1998), Measurements of mid latitude E region, sporadic E, and TID related drifts using HF Doppler sorted interferometry, J. Atmos. Sol. Terr. Phys., 60, 509 522, doi:10.1016/ S1364-6826(97)00058-8. Paul, A. K. (1990), On the variability of sporadic E, Radio Sci., 25, 49 60, doi:10.1029/rs025i001p00049. Sokolovskiy, S. (2001), Modeling and inverting radio occultation signals in the moist troposphere, Radio Sci., 36, 441 458, doi:10.1029/ 1999RS002273. Tsunoda, R. T., S. Fukao, and M. Yamamoto (1994), On the origin of quasi periodic radar backscatter from midlatitude sporadic E, Radio Sci., 29, 349 365, doi:10.1029/93rs01511. Whitehead, J. D. (1989), Recent work on mid latitude and equatorial sporadic E, J. Atmos. Terr. Phys., 51, 401 424, doi:10.1016/0021-9169(89)90122-0. Woodman, R. F., M. Yamamoto, and S. Fukao (1991), Gravity wave modulation of gradient drift instabilities in mid latitude sporadic E irregularities, Geophys. Res. Lett., 18, 1197 1200, doi:10.1029/91gl01159. Wu, D. L., C. O. Ao, G. A. Hajj, M. de la Torre Juarez, and A. J. Mannucci (2005), Sporadic E morphology from GPS CHAMP radio occultation, J. Geophys. Res., 110, A01306, doi:10.1029/2004ja010701. S. Sokolovskiy and Z. Zeng, COSMIC, University Corporation for Atmospheric Research, PO Box 3000, Boulder, CO 80301, USA. (zzeng@ucar.edu) 5of5