Typhoon Melor and ionospheric weather in the Asian sector: A case study



[1] The Space Environment Group of the National Institute of Information and Communications Technology, Tokyo, Japan, operates four ionosondes at Okinawa, Yamagawa, Kokubunjii, and Wakkanai in the Asian sector. Okinawa is located at the lowest latitude and lies at the northern edge of the northern equatorial anomaly, while Wakkanai is located at the higher latitudes of the midlatitude ionosphere. For this study, ionograms obtained from the Internet are analyzed using the automated Expert System for Ionogram Reduction (ESIR). An anomalous 3 day foF2enhancement observed by the Wakkanai ionosonde from 9 to 11 October 2009 forms the basis for this study. The scientific question being addressed pertains to the remarkably quiescent geomagnetic activity experienced during the extended solar minimum between cycles 23 and 24 that enables a search for the ionospheric response to weather in the lower atmosphere. The analysis of the ionograms from these four stations using the proprietary ESIR technique provides an extended database of electron density profiles that describes the ionospheric variability as a function of latitude, local time, and season. In addition, independent observations of the ionospheric TEC used by the USU Global Assimilation of Ionospheric Measurements (GAIM) model verify the anomalous ionospheric behavior as well as establishing its extent. Typical solar minimum conditions were seen during this study, with geomagnetic activity restricted to well-characterized corotating interaction region (CIR) events. After eliminating geomagnetic and solar disturbances as drivers of the October 2009 anomaly, the presence of Typhoon Melor is suggested as a possible source mechanism for the ionospheric anomaly.

1. Introduction

[2] The connection between severe weather in the troposphere and effects in the F region has a long and controversial history. The first possible observation of a hurricane generating an ionospheric response was made by Bauer [1958]. His evidence was based on an increase in foF2, measured by an ionosonde in response to approaching storms. Later, tropospheric gravity waves and ionospheric motions were observed by an HF radar in response to the passage of Hurricane Eloise [Hung and Kao, 1978], and Bertin et al. [1978]stated that medium-scale gravity wave energy from the troposphere could approach that provided by solar UV and EUV on occasion. More recently,Bishop et al. [2006] presented the results of multiple ionospheric measurements as tropical storm Odette passed within 600 km of the Arecibo Observatory. In this case, the Arecibo Incoherent Scatter Radar (18.3°N, 66.75°W) observed large velocity variations, and the ionosonde located at Ramey (18.5°N, 67.2°W) also observed intense midlatitude spread F; GPS occultations within the storm path confirmed the presence of irregularities through the presence of strong F region scintillations.

[3] This study presents ionospheric observations made in the Asian sector, specifically in the vicinity of Japan, at a time when Typhoon Melor was passing by northern Japan. Ionosonde observations identified a specific 2 to 3 day enhanced foF2 anomaly over Japan during this time. Further evidence of this enhanced foF2 anomaly is found in GPS TEC observations. These results, in the absence of normal geomagnetic or solar disturbances, are qualitatively compared to the location and path of Typhoon Melor.

2. Ionosonde Observations

[4] The Space Environment Group of the National Institute of Information and Communications Technology (NICT) in Tokyo, Japan, operates a network of four ionosondes that provides ionograms to the scientific community. Figure 1 indicates the locations of these four ionosondes and Table 1provides their coordinates. The ionograms generated by these four instruments were captured in real time as 2-D web page images with a cadence of 15 min at Space Environment Corporation (SEC). The real-time acquisition resulted in occasional gaps due to server downtime or other network interruptions. The images were then automatically processed to identify the main ionospheric layers, i.e., the E, F1, and F2 layers, and also the electron density profile (EDP).

Figure 1.

Location of four ionosondes: OK426, YG431, TO536, and WK546.

Table 1. Japanese Ionosondes
NameIdentifierGeographic LatitudeGeographic Longitude

[5] The ionogram images were analyzed with the Expert System for Ionogram Reduction (ESIR) software package developed by Space Environment Corporation [Rice et al., 2009]. The ESIR analysis software is independent of ionosonde hardware and uses a patented pattern recognition technology [Sojka et al., 2009]. ESIR performs additional analysis to ensure the robustness of the eventual electron profile and profile parameters by rejecting ionograms which have inadequate traces or excessive contamination from sporadic E or spread F. If a credible virtual height frequency profile can be created, the ESIR software uses the POLAN ionogram analysis routine [Titheridge, 1985] to obtain a true height frequency profile and an electron density profile. The application of ESIR to 2-D ionogram images is challenging due to the limited frequency, height, and dynamic range resolution provided by the images, as well as the lack of polarization (O/X trace) data. The ability to analyze such images is important due to the vast historical ionogram archives that exist on film and on paper; conversion of these ionograms to useful digital parameters would provide access to decades of ionospheric observations. The study presented here uses the results of the first version of this 2D image analysis capability, together with hourly scaled values from NICT (available online athttp://seg-web.nict.go.jp/e-sw/download/iono-monthly/index.html) for comparison during key time periods.

[6] The period of interest focuses on 3 days: 9, 10, and 11 October 2009 (days 282, 283 and 284). The year 2009 itself was noteworthy for being part of the extended solar minimum period between solar cycles 23 and 24. During the year, all four ionosonde station observations were consistent with a quiet solar minimum climatology, almost devoid of short-term weather.Figure 2 provides an 11 day glimpse of these quiet conditions. Automated ESIR foF2results from the four station ionograms are shown for the 11 days prior to the period in question. Gaps in the traces are due to either rejection of unsuitable ionograms by ESIR, or a failure to obtain real-time ionograms due to instrument, server, or processing outages. On inspection, 2 days, 268 and 271, have somewhat enhanced daytime foF2values at the most equatorward station OK426; otherwise, all four stations show a similar day-night diurnal modulation (local nighttime is approximately 0800–2100 UT in this sector, based on the absence of the daytime E layer in the ionograms). The most poleward station, WK546 at 45.2°N, shows only a consistent diurnal variation during this period.

Figure 2.

Diurnal trend for postequinox solar minimum at the four ionosondes.

[7] All four ionosondes' nighttime minimum foF2 measurements are the same at about 3 MHz ± 0.5 MHz. However, the daytime peak foF2shows a measurable increase from the highest latitude (WK546) at 6 ± 0.5 MHz to the lowest latitude (OK426) at 9 ± 1.0 MHz. This trend is consistent with the expectation that the lower midlatitude station OK426 is near the northern equatorial anomaly where the ionospheric density is larger than that at midlatitudes. The day-to-day variability of the lowest-latitude foF2is also largest, again associated with the expected day-to-day variability in the equatorial anomaly eastward electric field [Scherliess and Fejer, 1999].

[8] Figure 3 shows the ESIR foF2 results for the 10 day period 280–289. At first glance, this seems to be even less variable than the data in Figure 2. However, the WK546 nighttime foF2 values are not uniform during this period. The nighttime WK546 foF2 values range from about 3.0 to 4.0 MHz in Figure 2 and also for most of the days in Figure 3; the exception are days 282 and 283, when foF2 increases to a nighttime peak of 5 MHz. This sustained 2 day nighttime enhancement is sufficiently significant that it cannot be attributed to instrumental problems or statistical variability.

Figure 3.

Quiet conditions continue, but Wakkanai (WK546) shows changes on days 283–284.

[9] Figure 4 explores this trend in more detail. In Figure 4a, the stack plot corresponds to an enlarged section of Figure 3 with a limited foF2 frequency scale. In this stack plot, the WK546 nighttime enhancement is more evident. Figure 4b shows the WK546 foF2 data, this time including hourly values from the World Data Center (WDC) scaling to fill in gaps. Both ESIR and WDC data demonstrate the enhanced nighttime foF2 (dark lines) compared to day (lighter lines). The greatest night enhancement occurs between 1000 and 1800 UT on day 282; it diminishes over the next 2 days, and returns to a typical value by day 285. Table 2 lists the WK546 range of foF2 and its corresponding NmF2 values for each night between day 281 and 285, based on one standard deviation from the mean between 0800 and 2100 UT. The percent difference between the densities and those of the “normal” day 281 are shown. The day 281 nighttime NmF2 values range from 0.9 × 1011 to 1.7 × 1011 m−3; however, on day 282 (9 October 2009), the maximum nighttime density increased to 4.0 × 1011 m−3. Indeed, this nighttime density is only 7% lower than the following day's daytime average NmF2 of 4.34 × 1011 m−3. To maintain the nighttime density at these levels requires either a significant change in wind/electric field, or a production source. This requirement will be discussed later.

Figure 4a.

Expanded view of the active period for stations WK546 (first panel), TO536 (second panel), YG431 (third panel), and OK426 (fourth panel).

Figure 4b.

Expanded view of the active period WK546, including ESIR and WDC values with night values emphasized.

Table 2. WK546 Night Ranges of foF2 and NmF2
DayMinimum ValuesMaximum Values
foF2 (MHz)NmF2 (m−3)NmF2 (%)foF2 (MHz)NmF2 (m−3)NmF2 (%)
2812.79.2 × 10101003.71.7 × 1011100
2824.42.4 × 10112595.74.0 × 1011233
2833.61.7 × 10111794.82.9 × 1011166
2843.11.2 × 10111314.22.2 × 1011126
2853.11.2 × 10111283.61.6 × 101192

[10] It could be argued that TO536 south of WK546 does see a nighttime foF2 enhancement on day 282, while the YG431 and OK426 exhibit only a slight enhancement, beginning on day 282 and extending several days beyond. In Figure 4, the TO536 nighttime foF2 on day 282 is somewhat increased from levels of 3 ± 0.5 MHz to values of 3.6 ± 0.5 MHz, but by day 283 the nighttime value has returned to the quiet day level.

3. Independent Observational Evidence

[11] Ground-based GPS ionospheric total electron content (TEC) measurements were made during this period from a global distribution of locations, including the Asian-Japanese sector. These observations of slant path TEC were assimilated into the Utah State University (USU) Global Assimilation of Ionospheric Measurements Gauss Markov (GAIM-GM) model. The GAIM-GM model is a physics-based ionospheric model that uses a Kalman filter procedure to assimilate slant TEC measurements [Scherliess et al., 2004, 2006; Schunk et al., 2004; Sojka et al., 2007]. For the study in question, the GAIM-GM model was running in real time using only slant TEC ground-based observations as data being assimilated.

[12] Figure 5shows the GAIM-GM output over a 10 day period which includes the day 282 period of interest. The modeled vertical TEC over the location of the WK546 ionosonde (Figure 5, top) and over the TO536 ionosonde location (Figure 5, bottom) are shown. In both panels, the period of interest, days 282 to 284, shows elevated nighttime TEC in addition to slight increases in the daytime TEC. These enhancements coincide with those found described for the ionosondes in section 2. Figure 5 (top) has a strong nighttime TEC enhancement on day 282 which decreases over the next 2 days to quiet nighttime values. This enhancement on day 282 from 3 TECU to over 6 TECU is a factor of 2 or more, which is consistent with the WK546 ionosonde's nighttime NmF2enhancement at this location. At the lower-latitude location TO536, the TEC increase is present on day 282 but less pronounced, amounting to about 50%. In comparing the GAIM-GM TEC and the ionosonde observations, it is important to note that the ionosonde provides a very local measurement while the GAIM-GM values are based on a 3-D voxel whose latitude and longitude extents are 4.67° and 15°, respectively, in this sector. Furthermore, the GAIM-GM assimilation procedure has its own correlation scales of comparable size. To evaluate this spreading effect, the GAIM-GM observations are presented as differences with respect to the quiet period prior to day 282. Also instead of TEC, the GAIM-GM values for theF region peak density are used.

Figure 5.

USU GAIM-GM modeled TEC for the ionosonde sites.

[13] Figure 6 shows the NmF2 difference in density at WK546 for the period of interest. A dashed line indicates the zero difference and provides a reference. In Figure 6, the rapid rise and slower recovery over 3 days is evident. Using the same differencing procedure, the analysis is repeated for locations north, south, east, and west of WK546. Figure 7 shows this grid of differences in NmF2 over an 8 day period. Each panel is labeled with the specific location as a geographic latitude and east longitude. In Figure 7 (top), locations are 10° poleward of WK546 and show only a weak signature during the day 282 to 284 period of interest. Figure 7 (middle), 45.2 N 141.8E, shows the strong nighttime enhancement seen in the WK546 ionogram data.

Figure 6.

NmF2 difference in density (cm−3) at Wakkanai (WK546).

Figure 7.

Differences in NmF2 (cm−3) for locations north, south, east, and west of WK546 (45.2, 141.8) over an 8 day period.

[14] In contrast, 10° equatorward (Figure 7, bottom) shows a strong but different type of enhancement. In all cases, the NmF2 enhancement begins on day 282, as did the ionosonde event. In the bottom row, though, the enhancements have a highly structured diurnal dependence. Panel 35.2 N 141.8E (near TO536) and panel 35.2 N 161.8E have dramatic spikes centered on 0 UT, corresponding to local dawn. In the GAIM NmF2 data and in the TO536 ionograms, these spikes correspond to rather subtle features: more rapid rates of increase in the dawn F2 density after day 282 than on the baseline days. It is not clear how these changes in the morning F2growth rate relate to the change in nighttime density observed to the north at WK546; follow-up studies will be needed to determine the full extents of these phenomena.

4. Locating a Driver Mechanism

[15] The usual suspects for producing impulsive modification in the F region ionosphere are geomagnetic storms and substorms. The WK546 event lasted from 2 to 3 days, mainly in the form of nighttime density enhancement. The common signature of a geomagnetic storm in the F region is typically observed over large areas as a negative phase with deep depletions and/or a positive phase with strong enhancements. However, during this extended period of solar minimum, geomagnetic activity was particularly mild, and no ionospheric storm signatures were seen by the ionosondes during the period of interest. Figure 8 (top) shows the geomagnetic Kp record over a period of 40 days, centered on day 280. There is no credible geomagnetic activity around the beginning of the day 282 event. Indeed, from day 282 through day 284, only Kp values of 0, 0+, and 1 are present. On day 284, Kp reached 3 for the first time in the prior 12 days.

Figure 8.

Kp, F10.7, and F10.7A values during the period of interest.

[16] Other than ionospheric storms, unusual electron density enhancements have been reported in the Asian sector, but these observations typically cover a much larger geographical area than observed in this case by the ionosonde chain or indicated by the GAIM-GM model. For example,Maruyama [2006]reported a strong TEC enhancement extending to 45°N in the evening, but attributed it to storm-enhanced density (SED). The lack of any geomagnetic or ionospheric storm signatures in the period of interest, and the absence of indications of equatorward TEC enhancement would exclude such a SED event. (The equatorward enhancements inFigure 7are rate-of-change differences, not absolute density increases, as explained in the previous section.)

[17] Another reported electron density enhancement, the midlatitude summer nighttime anomaly (MSNA) described by Lin et al. [2010], provides a northern counterpart to the well-known Weddell Sea Anomaly (WSA). However, MSNA observations of TEC enhancement in northern Japan are most pronounced in early summer (April–June) and have not been seen after August. The MSNA enhancement in the Japan sector discussed by Lin et al. has a generally poleward progression before local midnight, and the enhancement should be observed by other sites such as TO536 before midnight. Overall, the geographical distribution of the enhancement observed by WK546 is much more localized than the MSNA.

[18] Possible solar sources, i.e., EUV, are similarly unlikely. Figure 8 also shows the solar 10.7 cm radio flux index and its 80 day average. Both are singularly devoid of variability that could generate a 3 day event. Direct nighttime enhancement would also not be associated with a solar ionization source.

[19] The next set of possible candidates for the anomaly would be atmospheric sources. Noting that this event has a rapid rise on day 282 with a subsequent slower decay over 2 to 3 days, it is unlikely that gravity waves, tides, or planetary waves in the ionosphere are directly responsible. Furthermore, the affected region, estimated from GAIM-GM results, is quite localized, with a latitudinal extent of about 20° and a longitudinal extent of perhaps 40°. Recently, both the stratospheric and ionospheric communities have uncovered a strong correlation between the northern hemisphere polar stratospheric warmings and subsequent modification in the equatorial ionosphere [Chau et al., 2009]. However, the northern stratospheric warming occurs in January. To date, the equivalent F region effects for the southern stratospheric warmings are being investigated but appear to be less significant (B. Fejer, private communication, 2010).

[20] During the same period, detailed analysis of ionosonde observations at Jicamarca, Peru showed no event occurring in the atmosphere similar to that observed over Japan [Eccles et al., 2011]. Given that the ionosonde observations suggest that the event is localized to the WK546 latitudes, and the GAIM-GM assimilation of GPS TEC data also indicates that the event is localized in this sector, a more local source is needed.Figure 9 shows the trajectory of Typhoon Melor; the track passes near TO536 on 8 October (day 281), and weakens as it moves off the coast on day 282. The discussion of Typhoon Melor presented at http://en.wikipedia.org/wiki/Typhoon_Melor_(2009) provides a description of the storm's life cycle. Typhoon Melor was classified as a “super typhoon” and its path over northern Japan was unusual because it traveled relatively slowly (B. Anderson, private communication, 2009).

Figure 9.

Track of Typhoon Melor. Points are at 6 h intervals, and circle size is proportional to the wind speed of the storm. The storm passed near TO536 on 8 October (day 281).

[21] Perhaps more relevant to the ionospheric observations over northern Japan is that this storm did not simply dissipate in the cold North Pacific. On 11 October, its remnants combined with an extratropical storm forming near Alaska and moved south to cause flooding on the west coast of the United States. This storm was one of the most intense October storms recorded in California (http://en.wikipedia.org/wiki/October_2009_North_American_storm_complex). Thus the typhoon carried a considerable amount of energy north and east into the North Pacific after its landfall in Japan. Some of this energy could be transferred to the upper atmosphere through gravity waves, as observed by Hung and Kao [1978].

5. Discussion and Summary

[22] The intriguing coincidence between an ionosonde located in northern Japan experiencing a very unusual three-night density enhancement and similar GAIM-GM perturbations at a time when a major typhoon was plaguing the same area encourages a discussion on severe tropospheric weather events coupling into the F region. The work ofBishop et al. [2006] also provides evidence for this; specifically, they showed that the Arecibo Incoherent Scatter Radar observed very dynamic, strong F region plasma drifts when tropical storm Odette passed Arecibo. This plasma drift observation provides a connection as to a possible cause of nighttime F region density enhancement. If a “localized” atmospheric hot spot led to a similar region of heating in the thermosphere with associated vertical winds, then a nighttime F region maintenance mechanism could be produced.

[23] In the case of Typhoon Melor, its longevity as it moved from Japan into the North Pacific leads to the conjecture that the combination of the warm storm and cold ocean waters could produce strong penetrative convection above the storm, which is effective for launching atmospheric gravity waves into the middle and upper atmosphere [Stull, 1976]. In modeling gravity wave structures above mesoscale thunderstorms, Pasko et al. [1997] found that gravity wave heating in the mesosphere on a time scale of about 10 h could compete with integral daily heating produced by solar radiation. It is thus plausible that, in the absence of geomagnetic storm drivers, this strong, slow moving tropospheric storm over cold North Pacific waters delivered sufficient energy via upward propagating gravity waves to affect nighttime ionospheric densities in a limited region, and affected layer formation rates over a larger area.

[24] This study presents observational evidence that, beginning on 9 October 2009, a major F region nighttime density enhancement event was initiated. This event was only present over a latitudinal region of ±10° in the northern part of Japan, and the F region recovery lasted from 2 to 3 days. No magnetic or solar activity was present that could drive this localized event. However, Typhoon Melor moved past northern Japan at this very time, bringing significant amounts of energy from the warm equatorial regions into the cold North Pacific environment. Curiously, in the days prior to landfall, Typhoon Melor was much more powerful and its path brought it near the ionosondes OK426 and YG431, but neither instrument registered any unusual F region variability. Thus one might speculate that the injection of energy into the colder northern atmosphere and enhanced convection over the North Pacific were crucial to producing the F region disturbances. More research is required to explain how this typhoon scenario could produce the F region weather event.

[25] This study shows the usefulness of modeling with assimilated data as a supplement to direct observations of this anomaly. The GAIM-GM data used here utilized GPS TEC measurements, and not the ionosonde observations, providing an independent description of the event. Currently, GAIM-GM results are being examined for evidence of ionospheric disturbances associated with recent major tropospheric storms in the American sector in conjunction with other observations. It is hoped that such efforts will lead to improved understanding of connections between tropospheric and ionospheric weather events.


[26] Space Environment Corporation carried out this work with support from the National Science Foundation (ATM-0521487) and a Small Business Innovative Research contract from the Air Force Research Laboratories, Hanscom AFB contract FA8718-04-C-0002. GAIM-GM development was supported by a grant from the Office of Naval Research (N00014-09-0292) and DoD funding through Northrup Grumman (subcontract 7500055856). The authors would like to thank Bruce Anderson for helpful discussions on the tropospheric weather associated with typhoons and Robert Hunsucker for atmospheric wave references