Radio Science

A method for automatic scaling of sporadic E layers from ionograms

Authors


Abstract

[1] A method for automatic scaling of the maximum frequency and virtual height of a sporadic E layer is presented. A set of ionograms recorded at the ionospheric observatory of Gibilmanna was used to test the performance of the algorithm. The test was performed by comparing the data obtained automatically with the values scaled by an operator.

1. Introduction

[2] Recently, the Istituto Nazionale di Geofisica e Vulcanologia (INGV) developed a low-power (less than 200 W) advanced ionospheric sounder called AIS-INGV [Zuccheretti et al., 2003; Bianchi et al., 2003]. This ionosonde was installed at the ionospheric stations of Gibilmanna, Rome and Terra Nova Bay, in Antarctica. Together with the ionosonde, Autoscala, a computer program for the automatic scaling of critical frequency foF2 and MUF(3000)F2 from ionograms was developed [Scotto and Pezzopane, 2002; Pezzopane and Scotto, 2004, 2005].

[3] The aim of this work is to present a procedure to identify the presence of a sporadic E (Es) layer in the ionogram estimating its maximum frequency ftEs and the associated virtual height hEs. This procedure is designed for application as an extension of the current version of Autoscala.

[4] The Es layers are thin structures (usually about 1 km thick) in the E region having enhanced electron density. Sporadic E layers may be very flat and uniform or may form clouds of electrons 2–100 km in size and moving horizontally at 20–130 m s−1. From a fixed observation point on the Earth's surface, Es layers can last from a few minutes to several hours. Unlike other ionospheric parameters such as the critical frequency of the F2 region, foEs values do not cluster round a mean; sporadic E is often not present at all and at other times foEs reaches values many times greater than its mean [Whitehead, 1989]. For this reason the word “sporadic” must be considered in the sense that the Es events are unpredictable but they are statistically very common at midlatitude in late spring and early summer.

[5] An Es event sensibly affects radio communication systems. In the HF range it can cause the maximum usable frequency (MUF) for reflection in the E layer to be higher than the corresponding MUF for the F layer. Therefore the reliable automatic real time monitoring of ftEs is desirable for space weather applications. In addition hEs can be a useful parameter to distinguish different Es layers formed by the downward propagation of wind shear convergent nodes associated with tide [Haldoupis et al., 2006].

[6] On ionograms the Es layer appears as a strong reflection that can occur over a range of heights from about 90 to 120 km or more. The maximum frequency reflected can be greater than the critical frequency of any of the normal layers. An Es layer is classified according to the URSI standard in 11 morphological types [Piggot and Rawer, 1972]. The routine described in this work is designed to scale ftEs and hEs for all these Es types except for the retardation type (identified by the letter r) typical of the auroral zone. In the case of multiple Es layers the routine was thought unable to scale the layer having the maximum frequency by type as suggested by the URSI standard [Piggot and Rawer, 1972]. However, with regard to this issue no conclusion can be reached because in the data set considered there are few ionograms showing multiple Es layers.

2. Automatic Scaling Method for Sporadic E Maximum Frequency and Virtual Height

[7] The routine for automatic scaling of the Es layer was developed along similar lines to the routine currently used for foF2 and MUF(3000)F2 [Scotto and Pezzopane, 2002, Pezzopane and Scotto, 2004, 2005]. This routine is based on image recognition techniques giving as output ftEs whose meaning depends on the antenna system of the ionosonde. If only the ordinary component is recorded the output corresponds to foEs, while if only the extraordinary component is recorded the output corresponds to fxEs. For the ionograms considered in this study both components were recorded. For such ionograms the routine described in this paper limits itself to give as output ftEs but is unable to specify this value as foEs or fxEs.

[8] The technique relies on a set of curves having the typical shape of the Es layer. The mentioned set is defined setting appropriate bounds for the height and the frequency. In particular curves having maximum frequency that does not exceed the modeled critical frequency foE of the normal E region are not considered. The shape of each curve is defined by several parameters p1, p2,…pn. For each curve the local contrast C(p1, p2,…pn) with the recorded ionogram is calculated with allowance made for both the number of matched points and their amplitude. The curve having the maximum value of C is then selected. If this value of C is greater than a fixed threshold Ct the selected curve is considered as representative of the Es trace. The value ftEs is thus obtained as the maximum frequency of the curve together with the associated height hEs. On the contrary if Ct is not exceeded then the routine assumes the Es trace is not present on the ionogram.

3. Comparison With the Manual Method

[9] The test was performed using a set of 1421 ionograms recorded in July 2004 and January 2005 by the ionosonde AIS-INGV installed at the ionospheric station of Gibilmanna. This data set was divided into two subsets. The subset Y, containing the ionograms for which the operator observed an Es trace, and subset N containing the ionograms for which the operator did not observe any Es trace.

[10] For each subset we considered (1) the number of ionograms for which the routine detected the Es trace scaling the maximum frequency and virtual height and (2) the number of ionograms for which the routine did not detect the Es trace.

[11] Figure 1 shows an ionogram for which the operator observed the Es layer and scaled ftEs: in this case the routine detected the Es and scaled ftEs correctly. Figure 2 is an ionogram for which the Es trace presents some discontinuities. In this case the operator scaled ftEs (5.1 MHz), and this value was also correctly scaled by the routine.

Figure 1.

Ionogram having a clear Es trace. In this case the routine properly scaled ftEs as 6.9 MHz (in green, the trace identified by the software) in close agreement with the value given by an experienced operator.

Figure 2.

Ionogram for which the Es trace has some discontinuities. In this case the routine scaled ftEs as 5.1 MHz (in green, the trace identified by the software) in close agreement with the value given by an experienced operator.

[12] The main cause of error is related to the presence of noise in the recorded ionogram. Figure 3, a case of an ionogram for which both the operator and the routine detected the Es layer, but the automatically scaled value is not acceptable. In this case the signal at 8.8 MHz was incorrectly recognized by the routine as the last part of the Es trace. At the same time the operator considered such an echo as noise giving a value of 6.9 MHz for ftEs. Figure 4 shows an ionogram for which the software failed by not detecting an Es layer that is actually present. This ionogram belongs to the 33 cases reported on Tables 1a and 1b for which the above mentioned threshold Ct was not exceeded, although an Es trace was observable.

Figure 3.

Ionogram for which the routine failed, giving an unacceptable automatically scaled value. In this case the signal at 8.8 MHz was incorrectly recognized by the routine as the last part of the Es trace (in green, the trace identified by the software), while the operator considered such an echo as noise and gave a value of 6.9 MHz for ftEs.

Figure 4.

Ionogram for which the software failed by not detecting an Es layer that is actually present.

Table 1a. Behavior of the Es Scaling Routine for ftEsa
OperatorAutoscala
Es DetectedEs Not Detected
  • a

    The test was carried out on the ionograms recorded at Gibilmanna in July 2004 and January 2005.

Es observed540 acceptable, 36 not acceptable33
Es not observed-812
Table 1b. Behavior of the Es Scaling Routine for hEsa
OperatorAutoscala
Es DetectedEs Not Detected
  • a

    The test was carried out on the ionograms recorded at Gibilmanna in July 2004 and January 2005.

Es observed572 acceptable, 4 not acceptable33
Es not observed-812

[13] The results of the data analysis for ftEs are reported in Table 1a and can be summarized as follows.

[14] 1. Among the 609 ionograms for which the operator observed an Es layer the routine correctly identified 576 cases and failed in 33 cases.

[15] 2. With reference to the ionograms in which the presence of an Es layer was assumed both by Autoscala and by the operator, ftEs was acceptably scaled in a high percentage of cases (540 out of 576, equal to 93.8%) (in this work a value is considered acceptable if within ±0.5 MHz of the value obtained by the operator); in this case the results are also presented in the form of a histogram in Figure 5.

Figure 5.

Differences (δ = automatic-manual) between the values of ftEs for ionograms for which both the INGV software and the operator identified an Es layer. Out of 576 cases the results were: for 484 cases, −0.1 MHz ≤ δ ≤ 0.1 MHz; for 23 cases, 0.1 MHz < δ ≤ 0.3 MHz; for 12 cases, −0.3 MHz ≤ δ < −0.1 MHz; for 15 cases, 0.3 MHz < δ ≤ 0.5 MHz; for 6 cases, −0.5 MHz ≤ δ < −0.3 MHz; for 32 cases, δ > 0.5 MHz; for 4 cases, δ < −0.5 MHz.

[16] 3. When the Es layer is not observed in the ionograms, the routine never wrongly detected the presence of an Es layer.

[17] Analogous results for hEs are shown in Table 1b. These results highlight the good reliability of the autoscaled values of hEs once the routine correctly detected the presence of an Es layer. In this case a value is considered acceptable if within ±5 km of the value obtained by the operator.

4. Conclusions

[18] The results reported in Tables 1a and 1b show that the routine for automatic scaling of Es layers can reliably identify the cases in which there is an Es layer. The results reported in Table 1a indicate that acceptable values of ftEs are obtained in 540 out of 576 cases, or 93.8% of cases, for the ionograms for which both the software and the operator observed an Es layer.

[19] For the same ionograms, the results reported in Table 1b indicate that acceptable values of hEs are obtained in 572 out of 576 cases, equal to 99.3% of cases.

[20] For these reasons we can conclude that the INGV routine for automatic scaling of Es layers performs well and can be added to Autoscala. It is possible to see on the Internet the real time ionograms recorded and autoscaled by the AIS-INGV/Autoscala system installed in the station of Gibilmanna by connecting to the Web site http://ionos.ingv.it/Gibilmanna/latest.html.

Ancillary