Midlatitude sporadic E episodes viewed by L band split spectrum InSAR

Transcription

1 FULL PAPER Open Access Midltitude spordic E episodes viewed y L nd split spectrum InSAR Msto Furuy 1*, Tkto Suzuki 1, Jun Med 2 nd Kosuke Heki 1 Astrct Spordic-E (Es) is lyer of ioniztion tht irregulrly ppers within the E region of the ionosphere nd is known to generte n unusul propgtion of very high frequency wves over long distnces. The detiled sptil structure of Es remins uncler due to the limited sptil resolution in the conventionl ionosonde oservtions. We detect midltitude Es y interferometric synthetic perture rdr (InSAR), which cn clrify the sptil structure of Es with unprecedented resolution. Moreover, we use the rnge split-spectrum method (SSM) to seprte dispersive nd nondispersive components in the InSAR imge. While InSAR SSM lrgely succeeds in decomposing into dispersive nd nondispersive signls, our results indicte tht smll-scle dispersive signls due to the totl electron content nomlies re ccompnied y nondispersive signls with similr sptil scle t the sme loctions. We lso exmine the effects of higher-order terms in the refrctive index for dispersive medi. Both of these detected Es episodes indicte tht smller-scle dispersive effects originte from higher-order effects. We interpret tht the smller-scle nondispersive signls could indicte the emergence of nitric oxide (NO) generted y the rections of metls, Mg nd Fe, with nitric oxide ion (NO + ) during the Es. Keywords: Ionosphere, Spordic-E, Totl electron content, Interferometric synthetic perture rdr, Split-spectrum method, Dispersive medi, Higher-order refrctive index, Nondispersive medi Introduction While the spordic-e (Es) lyer of the ionosphere hs een ttrcting rod reserch interest since the 1930s (e.g., Whitehed 1989; Mthews 1998; Hldoupis 2011), there still remin lrge uncertinties in the dynmics of the Es. Recently, glol nvigtion stellite system (GNSS) receiver dt hve een used to derive the totl electron content (TEC) long the microwve propgtion pth etween the GNSS stellite nd receiver. Tking dvntge of the dense GNSS receiver network dt in Jpn, Med nd Heki (2014, 2015) derived the TEC nomlies ssocited with midltitude Es episodes in Jpn nd demonstrted the detiled morphology nd dynmics of Es. In terms of the sptil resolution in the oservtion of Es, however, stellite-sed InSAR imging is more dvntgeous thn the GNSS network. Med et l. (2016) first succeeded in demonstrting n *Correspondence: furuy@sci.hokudi.c.jp 1 Deprtment of Erth nd Plnetry Sciences, Hokkido University, N10W8, Kit ku, Spporo, Hokkido , Jpn Full list of uthor informtion is ville t the end of the rticle Es episode in Jpn using oth GNSS TEC nd n InSAR imge derived from the Advnced Lnd Oservtion Stellite/Phsed Arry L-nd Synthetic Aperture Rdr (ALOS/PALSAR). ALOS ws lunched in 2006 y the Jpn Aerospce Explortion Agency (JAXA). Another dvntge of InSAR imging is tht no receivers hve to e deployed on the imging re, wheres the temporl resolution of InSAR is seriously limited y the stellite s recurrence intervl, which is 46 dys in the cse of ALOS/PALSAR. Med et l. (2016) ttriuted the phse nomlies in the InSAR imge to the Es episode, given the nerly identicl loction of the phse nomlies in the InSAR dt nd those derived from the GNSS TEC dt. We should note, however, tht GNSS TEC is physiclly distinct from the InSAR phse nomly, ecuse the InSAR phse includes the dispersive signl due to TEC nd the nondispersive phse dely tht hs een mostly ttriuted to polr molecules in the troposphere. In contrst to the dul frequency mesurement y GNSS, SAR imging hs een performed with single crrier frequency, nd no The Author(s) This rticle is distriuted under the terms of the Cretive Commons Attriution 4.0 Interntionl License ( which permits unrestricted use, distriution, nd reproduction in ny medium, provided you give pproprite credit to the originl uthor(s) nd the source, provide link to the Cretive Commons license, nd indicte if chnges were mde.

2 Pge 2 of 10 opertionl corrections for the ionosphere hve een conducted to dte. However, it hs een demonstrted tht the rnge split-spectrum method (SSM) could virtully perform dul frequency imging, therey llowing for the possile opertionl correction of the dispersive effects (Brcic et l. 2010; Rosen et l. 2010; Gom et l. 2016). Our first ojective is to report the results of our ppliction of the SSM to two Es episodes in Jpn: the first episode is the one reported y Med et l. (2016), nd the second one is new Es episode detected y ALOS2/ PALSAR2 lunched in 2014, follow-on mission of JAXA s ALOS/PALSAR. The second ojective is to exmine the impct of higher-order terms in the refrctive index on the estimtes of dispersive signls, which is motivted y the results sed only on the conventionl first-order refrction index. Theory nd processing methods Genertion of single look complex imge for SSM While SAR imging is performed with single crrier frequency, f 0, the ctul wveform of ech rdr pulse is frequency modulted with its vrile instntneous frequency s liner function of time. Therefore, it hs finite ndwidth, B w, on the order of MHz in the frequency domin, which controls the rnge resolution (Fig. 1, left). The ctul single look complex (SLC) imge is se-nded, nd the effect of the crrier c f 0 B w /2 f 0 B w /2 f 0 f 0 +B w /2 B w H f H f 0 f 0 +B w /2 s(t) = T e j2 f 0 sinc[ B w (t )] s H (t) = T H e j2 f H e j2 ( f H f 0 ) sinc[ B w H (t )] s H (t) = T H e j2 f H sinc[ B w H (t )] f f H f H + B H H B H w /2 w /2 Fig. 1 (Left) originl rnge spectrum with its center frequency f 0 nd ndwidth B w. (Right) corresponding se-nded SLC dt from point trget t R. Here, T nd τ indicte the pulse length nd the dely time 2R/c from trget, respectively; c is the speed-of-light. The sinc function is defined s sinc(x) = sin(πx)/πx. (Left) ndpssfiltered rnge spectrum with new higher centrl frequency f H nd ndwidth B H w. (Right) corresponding ndpss SLC dt from point trget t R. T H indictes the reduced pulse length. c (Left) sended rnge spectrum with new higher centrl frequency f H nd ndwidth B H w. (Right) corresponding se-nded SLC dt from point trget t R with new center frequency f H frequency is numericlly included in the phse vlue of ech slnt rnge pixel (Fig. 1, right). A key component of SSM is the splitting of the originl ndwidth into two su-nds with new lower nd higher crrier frequencies (Fig. 1), followed y stndrd InSAR processing t ech new crrier frequency (Brcic et l. 2010; Rosen et l. 2010; Gom et l. 2016). As such, SSM is performed t the expense of rnge resolution, wheres the zimuth resolution remins the sme s in the originl SAR imge. While Gom et l. (2016) hve detiled the processing steps, we ddress the procedure for SSM, pying prticulr ttention to the genertion of two new SLC imges from the originl SLC imge. We first pply two ndpss filters long the rnge xes of the SLC imge, so we cn split the originl B w into two nrrower nds; Fig. 1 illustrtes the ndpss filtering t higher frequency. The ndpss SLC imge, however, is not se-nded nd includes the modulted phse tht corresponds to the devition of the new crrier frequency from the originl crrier frequency (Fig. 1, right). We, therefore, need to demodulte the ndpss SLC, so it is se-nded with the new center frequency, f H (Fig. 1c). We perform the demodultion in the time (rnge) domin, multiplying the conjugte of the demodulted phse t ech slnt rnge. We repet this procedure for the new lower frequency, f L. When we hve new SLC imges with new higher nd lower crrier frequencies, we cn perform the stndrd processing of the differentil InSAR. Estimtion of dispersive phses The refrctive index of dispersive medi explicitly depends on the crrier frequency, which llows for seprting the ionospheric dispersive phse from the nondispersive phse, such s tht due to tropospheric propgtion dely nd ground displcement signls. Following the phse-sed GNSS rnge mesurement (Syndergrd 2002; Kim nd Tinin 2007; Hernández-Pjres et l. 2014), we write the frequency dependence of the InSAR phse Δφ s follows: f f 0 φ = φ Nondisp + φ TEC f 0 f f0 2 + φ Mg f 2 + φ Bend f0 3 f 3 where φ Nondisp, φ TEC, φ Mg, nd φ Bend indicte the contriution from the nondispersive phse, firstorder TEC, second-order TEC, nd the third-order ryending phse, respectively, nd f 0 is the originl crrier frequency. The second-order TEC term φ Mg indictes the Frdy rottion effect y the geomgnetic field nd is often dued the geomgnetic effect. Although (1)

3 Pge 3 of 10 the finl term, φ Bend, includes the third-order TEC, higher-order geomgnetic, geometric ry-ending, nd ry-ending due to the excess pth dely nd trnsverse heterogeneity of TEC (Hernández-Pjres et l. 2014), we cll it the ending term for simplicity ecuse the former two contriutions would e much smller thn the more dominnt terms in φ TEC nd φ Mg. In the literture regrding high-precision GNSS positioning, the secondorder geomgnetic nd the third-order ending effects re estimted to e orders of mgnitude smller thn the first-order TEC, wheres these higher-order terms re currently tken into ccount in the opertionl GNSS positioning (e.g., Syndergrd 2002; Kedr et l. 2003). Dul frequency phse dt for oth GNSS nd InSAR re used to simply solve for the first two terms of Eq. (1) neglecting higher-order dispersive terms. Eqution (13) for SSM in Gom et l. (2016) is lso derived under this ssumption. In contrst to the GNSS phse, the InSAR phse indictes the difference etween two cquisition epochs, nd the dominnt first-order term cn e significntly reduced in the InSAR phse. Becuse the effect of Es is included in either one of the two imge cquisitions, nd Es indictes n nomlously dense ionized lyer in the E region of the ionosphere, the higher-order dispersive terms my hve significnt impct on the InSAR phse nd my e reveled. In order to solve for the effects of the higher-order terms in ddition to the two conventionl terms ove, we my split the originl ndwidth into four nds to mtch the numer of oservtions to those of the unknowns. However, the elements of the mtrix ecome numericlly close to ech other. Under the ville ndwidth nd in the presence of phse noises, the condition numer for the resulting mtrix ecomes uncceptly lrge, preventing the estimtion of the higher-order terms. While estimting four unknowns from two InSAR dt sets t the two frequencies is n under-determined prolem, we cn derive the so-clled minimum-norm estimtes in such cse (e.g., Menke 2012). Below, we compre the conventionl estimtes with the minimum-norm estimtes. Another reson to consider the effect of higher-order terms in the dispersive phse is tht our results sed on the conventionl scheme do not necessrily result in distinct seprtion of dispersive nd nondispersive components s shown elow. Results Es episode on June 28, 2009 We first pply SSM to the Es event reported y Med et l. (2016). Detils of the ALOS/PALSAR dt sets, the new ndwidth, nd frequencies re shown in Tle 1. All the SAR dt in this study re cquired long descending pth, which psses during the locl dytime. Figure 2, indictes the unwrpped differentil interferogrms for the higher nd lower frequencies, respectively. The processing strtegy for the InSAR dt is siclly the sme s tht in Med et l. (2016) with the exception of the new center frequency nd nrrower ndwidth in the rnge xis. Although the nerly one-third ndwidth mkes the rnge resolution nerly three times corser thn the originl, we hve performed multi-looking with 50 looks in rnge nd 80 looks in zimuth, with pixel resolution of ~ 250 m. We do not discuss ny signls with sizes smller thn ~ 1 km. We used JAXA s precision orit dt nd did not re-estimte, nd we did not remove ny long-wvelength residul phse trend. Figure 2, is quite similr to the originl interferogrm in Med et l. (2016), suggesting tht the sptil resolution is sufficient for confirming the structure of the Es. Figure 3, shows the estimted nondispersive nd dispersive components using Eq. (13) of Gom et l. (2016) tht tkes into ccount only the first-order dispersion. We do not pply ny sptil filters in Fig. 3,, nd the estimted results re shown. Although we originlly expected tht the dispersive signl would dominte over the nondispersive signl, we oserve significnt nondispersive signls tht re prtly correlted with dispersive signls. Long-wvelength phse trends rise from the est to west oth in the dispersive nd nondispersive signls, while the EW phse grdient in ech signl is the inverse. In ddition, we notice smll-scle phse nomlies in oth the dispersive nd nondispersive signls, which re correlted with ech Tle 1 Detils of the L-nd InSAR dt sets Dtes (YYYYMMDD) Pth-frmes New high nd low frequencies (GHz) New ndwidth (MHz) B perp (m) Results , , Figures 2, 3 nd , Figures 6, 7 nd , 1410, , Additionl file 1: Fig. S , 1410, , Additionl file 1: Fig. S2 Occurrence of Es

4 Pge 4 of 10 H GHz (B w =9.33MHz) L GHz (B w =9.33MHz) cm in slnt-rnge Fig. 2 Unwrpped differentil interferogrm with new higher frequency nd new lower frequency. The ndwidth is reduced to 9.33 MHz from the originl 28 MHz for ALOS/PALSAR fine em single (FBS) polriztion mode. Detils of the InSAR dt re shown in Tle 1 other in terms of their loctions. A distinct seprtion etween dispersive nd nondispersive components does not seem to e successfully performed s ws performed t nother loction (Additionl file 1) nd s shown y Gom et l. (2016). Figure 4 gives the minimum-norm estimtes of nondispersive (Fig. 4), first-order TEC (Fig. 4), secondorder geomgnetic (Fig. 4c), nd third-order ending effects (Fig. 4d). The sum of the three dispersive terms is shown in Fig. 4e. Compring the two estimtes of nondispersive effects in Figs. 3 nd 4, the long-wvelength phse slopes cross the imge re significntly reduced in the minimum-norm estimte in Fig. 4. The smll-scle phse nomlies, however, remin in oth Figs. 3 nd 4, whose loctions re correlted with those of dispersive signls s noted elow. The three dispersive signls indicte interesting chrcteristics in their sptil distriutions. The ENE-WSW-trending phse dvnce ptches ner the costline re commonly oserved, suggesting tht the first-order term lone cnnot cpture the entire dispersive signls. We lso notice tht the smll-scle phse nomlies re more significnt in the higher-order geomgnetic nd ending effects; we do not pply ny smoothing filters in Fig. 4. Compring the conventionl solutions (Fig. 3, ) with the minimum-norm solution (Fig. 4, e), the overll phse nomlies for nondispersive (Figs. 3 nd 4) nd dispersive effect (Figs. 3 nd 4e) re lrgely consistent, wheres the long-wvelength EW phse trends re sent in the minimum-norm solutions in contrst to the conventionl solutions. Es episode on My 25, 2016 Next, we descrie the second detection of Es y GNSS nd InSAR, using ALOS2/PALSAR2 dt. To serch for the Es episode in Jpn, we first exmined the ionosonde dt t Kokuunji, Wkkni, nd Ymgw in Jpn, nd chose the dtes when the criticl frequencies of Es (foes) were higher thn 15 MHz during the locl dytime from My to August in 2015 nd 2016, s Es is known to e frequent in the locl dytime during the summer. The ionosonde oservtions were opertionlly performed y Ntionl Institute of Communictions nd Technology (NICT), Jpn. Second, we serched the ALOS-2/PAL- SAR-2 dt sets for oservtion res, dtes, nd times mtching the ionosonde dt ove s closely s possile. Third, we generted GNSS TEC mp whose res, dtes, nd times were the sme s the ove, nd if Es

5 Pge 5 of 10 Non-disp (Conventionl) Disp (Conventionl) cm in slnt-rnge Fig. 3 The estimted nondispersive component. The estimted dispersive component. Both, re derived on the ssumption of the firstorder TEC term lone in the refrctive index for dispersive medi ws confirmed in the GNSS TEC mp, we generted n interferogrm. The GNSS TEC mp in Fig. 5 indictes the presence of EW-trending TEC nomlies t 3:10 UTC on My 25, 2016, nd Fig. 5, c shows the time series of slnt TEC derived from stellite 31 t the sttion 0388 nd 0400, respectively. We detected the phse nomlies in the pir of Ferury 17, 2016 (Mster), nd My 25, 2016 (Slve), long trck from Okym to Kgw. Detils of the ALOS2/PALSAR2 dt sets re shown in Tle 1. Figure 6, indictes the derived high- nd low-frequency interferogrms derived y SSM. We first show our estimtes of nondispersive (Fig. 7) nd dispersive components (Fig. 7), which re derived on the ssumption of the first-order TEC effect lone for the dispersive component. In contrst to the cses in Figs. 3 nd 4, the dispersive phse nomlies re much greter thn the nondispersive, notly in the estern res. However, we cnnot ttriute ll of the dispersive signls to the Es episode, ecuse it might indicte the TEC in the F region nd could e lso due to the TEC during the other cquisition dte. However, the EW-trending steep phse chnges in Fig. 7 re presumly ttriutle to the Es on My 26, 2016, in light of the GNSS TEC mp in Fig. 5. It seems tht the ALOS2/PALSAR2 dt could not cover the centrl prt ut the estern edge of the TEC nomlies. Although nondispersive phses re nerly flt over the entire region, there gin rise loclized nd steep phse chnges in Fig. 7, which re closely correlted with those in the dispersive phse of Fig. 7. This is similr to the oservtions in the previous cse (Figs. 3 nd 4). The minimum-norm estimtes of ech term re shown in Fig. 8. The nondispersive signls in Fig. 8 re mostly flt s in Fig. 7, nd the loclized signls ner the phse jump still remin, while the mplitude of the loclized signls is reduced. The three dispersive signls in Fig. 8 d revel sptil chrcteristics similr to those in Figs. 4 in terms of sptil scles. Lrge-scle signls re notle in the lower-order terms (Fig. 8, c), while smllscle signls re dominnt in the higher-order terms (Fig. 8c, d). Compring the conventionl estimtes with the minimum-norm estimtes, oth the conventionl dispersive nd the summed dispersive signls re lrgely consistent (Figs. 7 nd 8e).

6 Pge 6 of 10 Non-disp (Min.norm) TEC c Mg cm in slnt-rnge d Bend e TEC + Mg + Bend cm in slnt-rnge cm in slnt-rnge Fig. 4 Minimum-norm estimtes of nondispersive, first-order TEC, c second-order TEC (geomgnetic), nd d third-order effect derived from the ALOS/PALSAR interferogrms in Figs. 2,. The sum of the dispersive effects d is shown in e Discussion Origin of dispersive nd nondispersive signls t the sme loctions Regrdless of whether we consider the effects of higherorder terms in the refrctive index, we hve identified smll-scle phse nomlies on the order of ~ 5 km or less in oth dispersive nd nondispersive components t the sme loctions. The dispersive phses originte in the free electrons ssocited with the Es, while the nondispersive phses t the sme loctions with comprle mgnitude re unexpected ut my indicte some physicl processes in the ionosphere. It is unlikely tht tropospheric nondispersive signls pper t the sme loctions ecuse of the huge differences in the height. Wht does the presence of nondispersive phse suggest? The origin of refrctivity is polriztion (e.g., Feynmn et l. 1963; Ch. 31 in Vol. 1, Ch. 32 in Vol. 2). The polriztion etween positive ions nd free electrons is responsile for the refrctivity, while the dispersive refrctivity in the ionosphere is due to the oscilltion of free electrons

7 Pge 7 of 10 c Fig. 5 GNSS TEC nomly mp t 3:10 UT on My 25, 2016, derived from stellites 23, 26, nd 31. The imged re y ALOS2 is shown with red rectngle. Slnt TEC nomly time series t the GEONET site 0388 derived from stellite 31. Solid verticl line indictes the imge cquisition time y ALOS2. c Sme time series t the GEONET site 0400 y the externl electric field. Positive ions shielded y free electrons cnnot oscillte ecuse of their hevier mss nd hve no dispersive effect. In ddition, it is known tht nitric oxide ion (NO + ) is the most undnt ions in the E region, nd metllic ions such s Mg + nd Fe + re known to e minly sustining the Es (e.g., Mthews 1998). While the therml ioniztion of meteorites is responsile for the genertion of metllic ion, the following chemicl rections with NO + hve een suggested to e eqully importnt (Ngt nd Tohmtsu 1973; Feng et l. 2013): Mg + NO + = Mg + + NO Fe + NO + = Fe + + NO We speculte tht the coexistence of dispersive nd nondispersive phses my e due to the nitric oxide (NO) molecule ccording to these rections, ecuse NO is well-known polr molecule. Our oservtion of nondispersive phses my, therefore, vlidte the relity of the rections ove. In ddition, the Es ptches shown in this study re highly ionized with foes well exceeding 15 MHz. In these cses, there must e sustntil trnsport of metllic ions into the Es region. However, it is uncler how nd where such lrge mount of metllic ions is deposited (T. Yokoym, personl communiction, Novemer 2, 2015). In ddition to the primry ioniztion y solr UV, rections of Mg nd Fe with NO + my ply crucil role s secondry driver for ioniztion in the dytime E regionit is lso importnt tht highly ionized Es ptches occur prticulrly during dytime when NO + density is rich. The increse in NO density is lso suggested y the depletion of NO + density t the pek electron density ltitude of Es (Ngt nd Tohmtsu 1973; Roddy et l. 2004). However, in view of the reduced mgnitude of the smll-scle nondispersive signls in the minimum-norm estimtes (Figs. 4 nd 8), we cnnot yet quntittively evlute how mny NO molecules were generted nd need to perform more cse studies. Non uniqueness of the estimted dispersive nd nondispersive signls Although the conventionl scheme to seprte dispersive nd nondispersive phses solves for two liner equtions with two unknowns, we hve shown tht the two solutions y the conventionl scheme re ctully not unique nd tht it my e necessry to estimte higher-order terms for the refrctive index. Since the conventionl solutions shre some chrcteristics with those derived from the minimum-norm solutions, we prefer the ltter minimum-norm solutions, t lest in the cse of ALOS/PALSAR in Fig. 4, ecuse they do not leve long-wvelength phse slope in either the dispersive or nondispersive phses. The significnt reduction in the long-wvelength phse slopes might e cused y minimizing the norm of the solutions. However, it seems tht the higher-order terms in the refrctive index llow us to isolte the scle dependence of the dispersive signl on frequency; smll-scle

8 Pge 8 of cm in slnt-rnge H GHz (B w =11.9 MHz) L GHz (B w =11.9 MHz) Fig. 6 Unwrpped differentil interferogrm with the new higher frequency nd the new lower frequency. The ndwidth is reduced to 11.9 MHz from the originl 79 MHz for the ALOS2/PALSAR2 strip mp high-resolution (SM1) mode. Detils of the InSAR dt re shown in Tle cm in slnt-rnge Non-disp (Conventionl) Disp (Conventionl) Fig. 7 The estimted nondispersive component. The estimted dispersive component. Both, re derived on the ssumption of the firstorder TEC term lone in the refrctive index for dispersive medi dispersive signls re likely due to the higher-order terms. However, we do not clim tht it is lwys necessry to estimte the effects of the higher-order terms. The cse studies in Gom et l. (2016) nd our results in the Additionl file 1 indicte tht the conventionl two unknown scheme could successfully seprte the dispersive nd nondispersive signls, ecuse they do not seem to show ny correlted phses in the other signls. Although it is uncertin now why the conventionl scheme works in some cses, it is noticed in the Additionl file 1 dt tht the totl phse chnges due to the dispersive signls re much greter thn those we hve found in our Es oservtions. The first-order TEC effect seems to e intrinsiclly dominnt.

9 Pge 9 of cm in slnt-rnge Non-disp (Min. norm) TEC (Min. norm) c Mg (Min. norm) cm in slnt-rnge cm in slnt-rnge d Bend (Min. norm) e TEC+ Mg + Bend Fig. 8 Minimum-norm estimtes of nondispersive, first-order TEC, c second-order TEC (geomgnetic), nd d third-order effect derived from the ALOS2/PALSAR2 interferogrms in Fig. 6,. The sum of the dispersive effects d is shown in e Conclusion We exmined two Es episodes in Jpn using L-nd InSAR SSM. Both cses indicted the presence of dispersive nd nondispersive smll-scle signls t the sme loctions, the ltter of which might indicte the presence of nitric oxide (NO) molecules generted y the chemicl rections in the Es. While the stndrd InSAR SSM sed only on the first-order TEC term could lrgely seprte the dispersive nd nondispersive phses, we might hve to consider the effects of higher-order terms in the refrctive index, depending on the nture of the dispersive signls. Additionl file Additionl file 1 Two exmples of distinct seprtion into dispersive nd non-dispersive phses. Arevitions Es: spordic-e; VHF: very high frequency; InSAR: interferometric synthetic perture rdr; SSM: split spectrum method; TEC: totl electron content; GNSS: glol nvigtion stellite system; ALOS: dvnced lnd oservtion stellite; PALSAR: phsed rry L-nd synthetic perture rdr; JAXA: Jpn Aerospce Explortion Agency; SLC: single look complex; ENE: est north est; WSW: west south west; EW: est west; foes: criticl frequencies of Es; NICT: Ntionl Institute of Communictions nd Technology. Authors contriutions MF, TS, nd JM performed SAR imge processing, while TS, JM, nd KH derived GNSS TEC mps. TS nd JM found the Es episodes. MF initilly wrote the mnuscript. All uthors red nd pproved the finl mnuscript. Author detils 1 Deprtment of Erth nd Plnetry Sciences, Hokkido University, N10W8, Kit ku, Spporo, Hokkido , Jpn. 2 Hokkido University Lirry, N10W8, Kit ku, Spporo, Hokkido , Jpn. Acknowledgements PALSAR 1.0 nd PALSAR-2 level 1.1 dt in this study re shred mong PIXEL (PALSAR Interferometry Consortium to Study our Evolving Lnd Surfce) under coopertive reserch contrct with the Erthquke Reserch Insitute, University of Tokyo. The ownership of ALOS-2/PALSAR-2 dt elongs to JAXA. MF thnks Drs. Dominique Deruw nd Giorgio Gom for discussing the implementtion of the rnge SSM in the erlier stge of this work. We cknowledge two nonymous reviewers, whose comments were helpful in improving the originl mnuscript. Avilility of dt nd mterils We generted ll the presented InSAR imges y processing ALOS/PALSAR level 1.0 nd ALOS2/PALSAR2 level 1.1 dt, which cn e serched nd purchsed from either RESTEC ( or PASCO ( en.los-psco.com). We cn shre the originl level 1.0/1.1 dt upon request

10 Pge 10 of 10 within frmework of collortion with our group. We used the processing softwre developed y Gmm Remote Sensing ( nd the sending of nd-pssed SLC ws performed y our originlly developed code. Competing interests The uthors declre tht they hve no competing interests. Ethics pprovl nd consent to prticipte Not pplicle. Funding This study ws prtly funded y Joint Reserch Progrm (B), 2015-B-02, t the Erthquke Reserch Institute, the University of Tokyo, nd prtly y the Joint Reserch Progrm, Suishin 14, of the Jpn Arctic Reserch Network Center. Pulishers Note Springer Nture remins neutrl with regrd to jurisdictionl clims in pulished mps nd institutionl ffilitions. Received: 10 August 2017 Accepted: 18 Decemer 2017 References Brcic R, Prizzi A, Eineder M, Bmler R, Meyer F (2010) Estimtion nd compenstion of ionospheric dely for SAR Interferometry. IEEE IGARSS, Proc. Feng W, Mrsh DR, Chipperfield MP, Jnches D, Höffner J, Yi F, Plne JMC (2013) A glol tmospheric model of meteoric iron. J Geophys Res Atmos 118: Feynmn R, Leighton R, Snds M (1963) The Feynmn lectures on physics. Cliforni Institute of Technology Gom G, Prizzi A, De Zn F, Eineder M, Bmler R (2016) Towrd opertionl compenstion of ionospheric effects in SAR Interferogrms: the splitspectrum method. IEEE Trns Geosci Remote Sens 54(3): Hldoupis C (2011) A tutoril review on spordic E lyers. In: Adu MA, Pnchev D, Bhttchryy A (eds) Aeronomy of the erth s tmosphere nd ionosphere, IAGA specil sopron ook series 2. Springer Netherlnds, Dordrecht. Hernández-Pjres M, Argón-Ángel À, Defrigne P, Bergeot N, Prieto-Cerdeir R, Grcí-Rigo A (2014) Distriution nd mitigtion of higher-order ionospheric effects on precise GNSS processing. J Geophys Res Solid Erth 119: Kedr S, Hjj GA, Wilson BD, Heflin MB (2003) The effect of the second order GPS ionospheric correction on receiver positions. Geophys Res Lett 30: Kim BC, Tinin MV (2007) Contriution of ionospheric irregulrities to the error of dul-frequency GNSS positioning. J Geod 81: org/ /s Med J, Heki K (2014) Two-dimensionl oservtions of midltitude spordic E irregulrities with dense GPS rry in Jpn. Rdio Sci 49: Med J, Heki K (2015) Morphology nd dynmics of dytime mid-ltitude spordic-e ptches reveled y GPS totl electron content oservtions. Erth Plnets Spce 67:89. Med J, Suzuki T, Furuy M, Heki K (2016) Imging the midltitude spordic E plsm ptches with coordinted oservtion of spceorne InSAR nd GPS totl electron content. Geophys Res Lett 43: doi.org/ /2015gl Mthews JD (1998) Spordic E: current views nd recent progress. J Atmos Terr Phys 60(4): Menke W (2012) Geophysicl dt nlysis: discrete inverse theory. Acdemic Press, Cmridge Ngt T, Tohmtsu T (1973) Modern eronomy (in Jpnese). Shoko, Tokyo Roddy PA, Erle GD, Swenson CM, Crlson CG, Bullett TW (2004) Reltive concentrtions of moleculr nd metllic ions in midltitude intermedite nd spordic-e lyers. Geophys Res Lett 31:L org/ /2004gl Rosen P, Hensley S, Chen C (2010) Mesurement nd mitigtion of the ionosphere in L-nd Interferometric SAR dt. In: Proceedings of the IEEE Rdr Conference. Syndergrd S (2002) A new lgorithm for retrieving GPS rdio occulttion totl electron content. Geophys Res Lett. org/ /2001gl Whitehed JD (1989) Recent work on mid-ltitude nd equtoril spordic E. J Atmos Terr Phys 51: