Estimating drift velocity of polar cap patches with all-sky airglow imager at Resolute Bay, Canada

Authors


Abstract

[1] Highly sensitive all-sky airglow imager has been operative at Resolute Bay, Canada (74.73°N, 265.07°E; AACGM latitude 82.9°) since January 2005. We present, as a first result from the imager, an event of polar cap patches drifting anti-sunward during the southward IMF conditions. Magnitude and direction of patch drift velocities are computed with a temporal resolution of 2 min by using the newly developed patch-tracking algorithm based on 2D cross correlation analysis. It is well visualized that the patches change their moving speed and direction drastically in a short time scale (a few minutes). Speed of the patch is primally controlled by the IMF Bz. Dawn-dusk component of the patch drift velocities is well correlated with the IMF By in agreement with published By dependence of the nightside polar cap convection. However, response of the patch drift direction to the IMF By is found to be much slower (≈20 min) than that of the drift speed to the IMF Bz (almost instantaneous).

1. Introduction

[2] Polar cap patches are known as regions of F-region electron density enhancement seen in the polar cap. They are generated near the dayside cusp and then transported across the polar cap toward the nightside in the antisunward convection [Crowley, 1996]. It is widely believed that patches are transported along the streamline of the polar cap convection, because they typically drift antisunward during the southward IMF conditions. However, there has been no study estimating drift velocity of patches continuously except for the study made by Fukui et al. [1994]. Fukui et al. [1994] have derived speed and direction of the patch drift velocities from 630.0 nm all-sky images. An average velocity was determined by tracing the location of “the center of gravity of the patch” at the beginning and end of the patch measurement. Hence, only one velocity vector is obtained for each individual patch (temporal resolution is about 15–20 min). They compared derived velocity with the IMF in a statistical way and obtained overall agreement between speed (direction) of the patch drift velocities and the IMF Bz (By). However, temporal variations of the patch drift velocities were not investigated in detail. In reality motion of the patches, as imaged by the all-sky imager employed in this study, seems to change in a shorter temporal scale (a few minutes). An analysis with higher temporal resolution is needed to clarify how patches change their moving path in response to the IMF variations. In this paper, speed and direction of the patch motion were computed from the 630.0 nm all-sky images with more objective method based on the 2D cross correlation analysis. Temporal behavior of the patch drift velocities is discussed in terms of changes of the polar cap plasma convection in response to the IMF variations. The imager employed in the current study has started routine measurements since January 2005. This paper presents the first results from the imager.

2. Observation

[3] The all-sky airglow imager used in the current study was developed by the Solar-Terrestrial Environment Laboratory, Nagoya University, as a part of the Optical Mesosphere Thermosphere Imagers (OMTIs) [Shiokawa et al., 1999]. The imager at Resolute Bay (74.73°N, 265.07°E) has been operative since January 2005. In the present analysis, we use airglow images at a wavelength of 630.0 nm (OI, emission altitude ranges from 200 to 300 km) which are obtained every 2 min with an exposure time of 30 s. Background continuum emissions from the sky is sampled every 20 min at a wavelength of 572.5 nm, which are employed to derive the absolute intensity of the airglow lines [Shiokawa et al., 2000].

[4] The interval presented is on February 3, 2005, between 0230 and 0830 UT. Figures 1a–1b show variations of the IMF By and Bz monitored by the ACE spacecraft located upstream of the Earth (Xgsm = 219.7 Re). An approximate solar wind velocity of 510 km s−1 measured by the spacecraft gives a delay of 43 min between the observed IMF feature and their incidence on the dayside magnetopause. The time-series of the IMF have been shifted accordingly. This time delay can be verified by comparing the SuperDARN measurement of convection enhancement near the cusp (not shown) with the southward turnings of the IMF Bz at 0708 UT. The IMF Bz component was predominantly negative, except for brief excursions to positive near 0238 and 0700 UT. The IMF By changed its sign more frequently, varying back and forth between −7 and 7 nT.

Figure 1.

(a–b) The IMF By and Bz as measured by the ACE spacecraft, time shifted by 43 min for the solar wind propagation delay. (c) North-South keogram of the 630.0 nm all-sky images on February 3, 2005. (d–o) Sequence of all-sky images from 0524 to 0630 UT. Superimposed are the vectors of the patch drift velocities derived with the patch tracking algorithm.

[5] Figure 1c shows the North-South keogram of 630.0 nm airglow images. To investigate perturbations in airglow intensity, deviations from 1-hour running averages are colour scaled in units of Rayleigh. Structures of high airglow intensity were observed continuously throughout the interval as slanted traces moving from north to south. This north to south motion corresponds to the antisunward drift of patches in the polar cap. 19 individual patches in total were observed, which have been given the identifying letters A to S to aid discussion. After the passage of the Patch F, region of enhanced luminosity appears on the southern edge of the keogram, which is due to the poleward expansion of the auroral oval following a substorm onset. Around this time period, observation of patches is greatly disturbed by the much brighter aurora.

[6] Figures 1d–1o show the sequence of the maps of 630.0 nm airglow image picked up every 6 min intervals. The original all-sky images have been converted into the AACGM [Baker and Wing, 1989] coordinates by assuming the emission height of 250 km. Dotted circle represents 80° magnetic latitude. The patch J appeared on the north-western edge of the FOV as a cloud-like structure at 0524 UT, and then drifted toward the southeast. This feature moved almost off the FOV by 0600 UT. At the same time, patch L appeared on the north-eastern edge of the FOV. This patch drifted from the northeast toward the southwest, and moved almost off the south-western edge of the FOV by 0630 UT. It is worth noting that the drift direction of the two patches is different. The drift direction of the patch J is inclined to the dawn while that of the patch L to the dusk. This change of drift direction is probably due to the shift of the IMF By from negative to positive around 0535 UT seen in Figure 1a.

[7] To investigate temporal variations of the patch drift velocities in more quantitative way, we have developed an algorithm for automated determination of exact drifting path of patches based on 2D cross correlation analysis. Consecutive pairs of original images (256 × 256 pixels) were processed according to the following procedures. (1) 25 (5 × 5) reference points (center of the images used for the correlation analysis) are defined near the zenith of the FOV. (2) The image is divided into 80 × 80 pixel template windows by setting the reference point as a center. (3) Template window is shifted in the succeeding image and calculate cross correlation coefficients between template and all shifted windows. (4) Drift velocity is determined by using relative displacement vector from template window to shifted one where maximum cross correlation coefficient is obtained. In this procedure, the effects of lens distortion and variations of intensity with zenith angle are removed before calculating correlations.

[8] As can be seen in Figures 1d–1o, small scale structures associated with patches tend to change their luminosity frequently, probably according to their own logic. These small scale variations also contribute to the cross correlation analysis and may contaminate estimated velocities. Our method, however, makes a cross correlation between consecutive images of 80 × 80 pixels, which is large enough to cover the whole patch. Then the estimated vectors primally follow the bulk motion of the main body of patches. The drift velocities of patches determined with this algorithm are superimposed on Figures 1d–1o, where the vectors with low cross correlation coefficient (<0.5) are not presented. It is clearly found that the drift vector of the patch J is tilted to dawn slightly. On the contrary, the drift vector of the patch L is inclined somewhat to dusk. These automatically estimated drift vectors are well consistent with the actual bulk motion of the patches.

3. Discussion

[9] In order to clarify how patches change their moving velocity with time, time-series of the patch drift velocity were generated by averaging the vectors obtained at the 25 reference points. In Figure 2 the time-series of (a) speed and (b) direction (θ angle: angle of azimuth relative to the noon-midnight line) of the patch drift velocities are plotted as a solid line with circles. Positive (negative) θ angle corresponds to the drift velocity inclined to dusk (dawn). Colour of the circles gives coefficient of the cross correlation analysis in the velocity estimation process, which can be used as a proxy for accuracy of the velocity determination. Correlation coefficient is greater than 0.6 for most of the data points, suggesting that the current correlation analysis can track the bulk motion of polar patches and then the derived drift velocities are reliable.

Figure 2.

(a) Time-series of speed and (b) direction (θ: angle of azimuth relative to the noon-midnight line) of the patch drift velocities, compared with the Bz and By components of the IMF. In the Figure 2a, the IMF Bz is plotted with reversed vertical scale (right-hand side).

[10] Before moving on, we discuss briefly on an accuracy of the velocity estimation. Our algorithm highly depends on the presence of patches in the FOV of the imager. Arrows at the top of Figure 2 indicate times when the center of the patch was at the zenith of the imager. Patches were observed almost continuously except for the gap at the time of the auroral breakup (from 0402 to 0430 UT). In addition, there exists an interval when patch was absent in the FOV (from 0500 to 0510UT: between Patch H and I). During this interval, the patch drift velocities are also unavailable.

[11] The velocities plotted in Figure 2 were calculated assuming the emission height of 250 km. Millward et al. [1999] suggested that this is not always a secure assumption within the polar cap. They suggested that the emission altitude is likely to be nearer 300 km on the poleward of the cusp. We also estimated velocities assuming the emission height of 300 km and compared them with the original values. Consequently, temporal variation of the velocities was found to be the same although the speed of patches was about 1.2 times larger than the original values. Since we are interested in the temporal variation of the drift velocities in response to the IMF, we move on with the assumption of emission altitude at 250 km.

[12] It has been well established that the plasma flow in the polar cap depends on the orientation and strength of the IMF. The polar cap convection speed should be an increasing function of the IMF Bz. The IMF By also introduces a dawn-dusk component in the polar cap plasma flow. Here we simply assume that velocity of the apparent patch motion is due only to the background plasma convection and discuss how the polar cap convection changes in response to the IMF. Before moving on, time-lag between variation of the IMF at the spacecraft and that of the patch velocities must be examined. Figure 3 shows cross correlation coefficients between IMF and patch velocities as a function of time lag from the spacecraft (a: IMF Bz vs speed of patches, b: IMF By and θ angle). The horizontal dashed-lines show the 99% significance level derived from Student's t-test. The error bars show the standard error of the coefficient, given by ε = (1 − r2)/(n − 2)1/2, where r is the value of the coefficient and n is the number of data points from which it was determined. Correlation coefficient between IMF Bz and speed of patches has two peaks at 44 and 66 min while coefficient between IMF By and θ angle has a broad peak around 66 min. Coefficients at these peaks are well above the statistical significance level. Here we simply determined the delay by taking the lag with the highest correlation coefficient, which is 44 min for “IMF Bz vs speed” and 66 min for “IMF By vs θ angle”, respectively. In Figure 2, time-series of the IMF Bz and By are overplotted with these time delays as a line filled with grey.

Figure 3.

Cross correlation coefficient between IMF and patch velocities as a function of lag ((a) IMF Bz vs speed, (b) IMF By vs θ). The horizontal dotted-lines show the 99% significance level.

[13] As shown in Figure 2a, agreement between patch speed and IMF Bz is generally good with the estimated 44 min delay. For example, increases (decreases) of the patch drift speed between 0520 and 0710UT have one-to-one correspondence to the decreases (increases) of the IMF Bz. In particular, decrease of the drift speed around 0650UT coincides well with sudden northward excursion of the IMF. Figure 2b demonstrates that the θ angle turns to positive (negative), which corresponds to the drift velocity tilted to dusk (dawn), when the IMF By is positive (negative). Statistical models of the high-latitude plasma convections [e.g., Ruohoniemi and Greenwald, 2005] showed that the nightside polar cap flow has an inclination from dawn (dusk) to dusk (dawn) for the positive (negative) By conditions. That is, the drift direction of patches is well controlled by the IMF By in agreement with published By dependence of the polar cap convection.

[14] Now we turn to discuss response time of the patch velocities to the IMF variations, especially why the response of the patch drift direction to the IMF By is much slower than that of the patch speed to the IMF Bz. As mentioned before, solar wind delay from the spacecraft to the dayside magnetopause was estimated to be 43 min, which was verified by the SuperDARN observation near the cusp. This implies that response of the patch drift speed to the IMF Bz occurs almost instantaneously (44 min delay from the spacecraft) even on the nightside. It is also worth noting that there is no peak of correlation coefficient between θ angle and IMF By at 44 min lag. In the past literature, there exist two different ideas concerning the response of the high-latitude convection to the change of the IMF Bz: (i) delayed response due to the propagation of the convection change first initiated near the cusp [e.g., Etamadi et al., 1988], (ii) almost instantaneous response in the entire polar ionosphere by magnetosonic waves [e.g., Ruohoniemi and Greenwald, 1998]. Our observation suggests that speed of the polar cap convections as derived from patch motions responds instantaneously to the variations of the IMF Bz.

[15] In contrast to the instantaneous response of the patch speed to the IMF Bz, an additional 23 min (total 66 min from the spacecraft) delay is needed to give the best agreement between the θ angle and IMF By. This much slower response can be identified also in the cross correlation analysis between the patch speed and IMF Bz (Figure 3a) although correlation coefficient is smaller than that of the primary peak at 44 min. This implies that ionospheric response to the upstream IMF variations consists of instantaneous component and slower component (i.e., two steps). Recently, similar two stage (instantaneous and 10–20 min delayed) response of the polar ionosphere to the IMF changes was reported by several authors [Nishitani et al., 2002; Lu et al., 2002]. Nishitani et al. [2002] presented that at the first stage ionospheric convection enhances and the polar cap boundary expands only in the noon and midnight sectors. In contrast, it takes 10–20 min from the initial response for the dusk-side polar cap boundary to shift equatorward. They claimed that the first response is associated with the propagation of magnetosonic waves [Ruohoniemi and Greenwald, 1998] and the second one is due to the redistribution of the newly created open flux in the polar cap region (based on the model by Cowley and Lockwood [1992]). In our observation, the first response at 44 min is clearly consistent with the magnetosonic wave theory [e.g., Ruohoniemi and Greenwald, 1998]. We claim that the second response at 66 min is related to the large-scale deformation of the convection streamline due to the redistribution of the newly created open flux. Past studies of the ionospheric response time to the IMF variation focused their attention mainly on the relationship between IMF Bz changes and ionospheric variations. Much slower response of the drift direction to the IMF By implied by the current measurement will be examined in more detail in the future.

4. Summary and Conclusion

[16] We have started routine measurement of the 630.0 nm airglow at Resolute Bay, Canada with all-sky imager since January 2005. Airglow enhancements associated with polar cap patches were continuously observed on February 3, 2005. An algorithm for automated estimation of the patch drift velocities has been developed and applied to the sequence of the patch images. Consequently, speed and direction of the patch drift velocities were obtained with a temporal resolution of 2 min. Then, it is well visualized that polar cap patches change their speed and drifting direction drastically in close relation to the IMF variations. Speed of the patch was primally controlled by the IMF Bz. Direction of the patch drift velocities was well correlated with the IMF By in agreement with published By dependence of the nightside polar cap convection. However, response of the drift direction to the IMF By was found to be much slower (≈20 min delayed) than that of the drift speed to the IMF Bz (almost instantaneous). This suggests that the response of the polar cap convection to the IMF variation can be divided into two steps (instantaneous and 20 min delayed).

Acknowledgments

[17] We thank Y. Katoh, M. Satoh, and T. Katoh of the Solar-Terrestrial Environment Laboratory (STEL), Nagoya University, for their kind support of airglow imaging observations. This work is supported by a Grant-in-Aid for Scientific Research (16403007) of the Ministry of Education, Culture, Sports, Science and Technology of Japan and the Project 2 of the Geospace Research Center, STEL. The authors wish to thank N. Ness at Bartol Research Institute for access to data from MFI and SWE instruments onboard the ACE spacecraft. Special thanks also go to the people in the Narwhal Arctic Service at Resolute Bay for their kind and helpful support in operating the optical instrument.

Ancillary