Geoelectromagnetic Hazards and Impacts on American Communication and Power Systems • Watts Up With That?

HT/Yooper

^{Research Article}

Jeffrey J. Love, E. Joshua Rigler, Michael D. Hartinger, Greg M. Lucas, Anna Kelbert, Paul A. Bedrosian

First published: 22 June 2023

https://doi.org/10.1029/2022SW003379

Abstract

An analysis is made of geophysical records of the 24 March 1940, magnetic storm and related reports of interference on long-line communication and power systems across the contiguous United States and, to a lesser extent, Canada. Most long-line system interference occurred during local daytime, after the second of two storm sudden commencements and during the early part of the storm’s main phase. The high degree of system interference experienced during this storm is inferred to have been due to unusually large-amplitude and unusually rapid geomagnetic field variation, possibly driven by interacting interplanetary coronal-mass ejections. Geomagnetic field variation, in turn, induced geoelectric fields in the electrically conducting solid Earth, establishing large potential differences (voltages) between grounding points at communication depots and transformer substations connected by long transmission lines. It is shown that March 1940 storm-time communication- and power-system interference was primarily experienced over regions of high electromagnetic surface impedance, mainly in the upper Midwest and eastern United States. Potential differences measured on several grounded long lines during the storm exceeded 1-min resolution voltages that would have been induced by the March 1989 storm. In some places, voltages exceeded American electric-power-industry benchmarks. It is concluded that the March 1940 magnetic storm was unusually effective at inducing geoelectric fields. Although modern communication systems are now much less dependent on long electrically conducting transmission lines, modern electric-power-transmission systems are more dependent on such lines, and they, thus, might experience interference with the future occurrence of a storm as effective as that of March 1940.

Key Points

• Extreme geomagnetic field variation realized during the March 1940 storm might have resulted from the interaction of ICMEs
• Long-line interference in the U.S. occurred during local daytime and in the upper Midwest and East, where surface impedance is high
• Voltages measured on grounded long lines during the storm exceeded 1-min voltages that would have been induced by the March 1989 storm

Plain Language Summary

On 24 March 1940, a pair of concentrated, and possibly interacting, bursts of solar wind forced an intense magnetic storm on Earth. Geomagnetic field variation during the storm induced high-amplitude geoelectric fields in the solid Earth’s conducting interior. These geoelectric fields drove uncontrolled currents in grounded long-wire communication- and electricity-power-transmission systems in the United States and Canada, causing significant operational interference in those systems. This interference was primarily experienced in the upper Midwest and the eastern United States, and many incidents of interference were reported in the popular press. Voltages monitored on several lines were greater than studies estimate would have occurred during the great storm of March 1989. In terms of its impact on communication and power systems, the March 1940 magnetic storm was one of the most significant ever experienced by the United States. Modern communication systems are less dependent on long electrically conducting transmission lines. On the other hand, modern electric-power-transmission systems are more dependent on such lines, and they, thus, might experience interference with the future occurrence of a storm like that of March 1940.

1 Introduction

The front page of the New York Times (1940a) declared that, on 24 March 1940, something akin to a “sun-spot tornado” had disrupted long-line communication transmission systems. The Washington Post said that an “invisible sun-spot storm” had interfered with radio, telephone, and telegraph communication systems across “half the world,” including trans-Atlantic communication systems (Smith, 1940). The headline of the Boston Daily Globe (1940) used more modern terminology, announcing that the United States had been hit by a “magnetic storm.” As a result of the storm, numerous protective breakers had to be replaced on American communication systems (Germaine, 1940; Ireland, 1940). In other parts of the world, interference was reported on telegraph systems, for example, in Australia (e.g., Morning Herald, 1940), in Great Britain (e.g., Daily Telegraph, 1940), and in India (e.g., Rangaswami & Basu, 1940). Interference was also reported by electricity utility companies, including alternating current waveform “distortion,” “reactive power,” “tripping out” of transformers, and “blown transformer fuses” (Davidson, 1940). John A. Fleming, then Director of the Department of Terrestrial Magnetism, Carnegie Institution of Washington, noted the quality that distinguished the March 1940 storm was the “unusual violence” of its geomagnetic “fluctuation” (Fleming, 1940, his p. 476, 480). These qualities, in his opinion, likely meant that the March 1940 storm was “the greatest magnetic disturbance which has ever been recorded” up to that point in time (Fleming, 1940, his p. 480). Harold W. Newton of Greenwich Observatory offered a slightly more qualified summary, describing the March 1940 storm as “remarkable for its agitation” and “one of the big storms of (the) century” (Newton, 1940, his p. 130).

Although these effects of space weather were experienced eight solar cycles ago, well before the space age, scientists and engineers had a pretty good, if qualitative, understanding of their causes. For example, scientists knew that the magnetic storm of March 1940 was caused by the “emanation” of a neutral stream of electrically charged particles from a group of sunspots (e.g., Nicholson, 1940), but they did not understand the mechanism that caused this emanation. Scientists knew that, after passage through interplanetary space and arrival at the Earth, the action of this stream of particles on the Earth’s magnetic field caused the storm (Chapman & Ferraro, 1930), and they knew that geomagnetic disturbance during the storm was concentrated beneath electric currents flowing in the ionosphere (e.g., Fleming, 1940; McNish, 1940; Nicholson, 1940). Storm-time geomagnetic field variation induced geoelectric fields in Earth’s electrically conducting interior. At the surface, potential differences drove quasi-direct geomagnetically induced currents (GICs) between the grounding points of long wires and cables that interfered with communication- and electric-power-transmission systems. And engineers knew, even in 1940, that this interference tended to be in regions where subsurface rock is generally electrically resistive (e.g., Corr, 1935; Germaine, 1940; Ireland, 1940). In particular, engineers appreciated that telegraph and telephone systems were vulnerable to the effects of magnetic storms (e.g., Gish, 1936), but, in 1940, engineers believed that electric-power-transmission systems were not especially vulnerable to magnetic storms (e.g., Fleming, 1940). These days, attention is usually focussed on the interference that storm-induced geoelectric fields can bring to long-line power-transmission systems (e.g., Kappenman, 2004; Molinski, 2002). In retrospect, it is, therefore, noteworthy that the March 1940 storm interfered with both communication and power systems.

Modern understanding of solar-terrestrial physics and space-weather effects on engineered systems is firmly based on physical principles (e.g., Abda et al., 2020; Gaunt, 2016; Kelbert, 2020; Pilipenko, 2021), but it remains a challenge to estimate the impact that future intense storms might have on electric-power systems (e.g., Pulkkinen et al., 2017; Thomson et al., 2010). Important lessons can be learned from the analysis of past magnetic storms (e.g., Boteler, 2019; Hapgood, 2019; Knipp et al., 2021; Lanzerotti, 2017; Love et al., 2022). In a recent report, Hayakawa et al. (2022) summarize the natural phenomena that were the storm of 24 March 1940. Our focus, here, is on the interpretation of ground-level geoelectromagnetic variation brought by the March 1940 storm and the deleterious impact it had on long-line communication- and power-system interference experienced across North America, and, especially, the contiguous United States (CONUS). Results inform modern estimates of storm-induced geoelectric hazards (e.g., Lucas et al., 2020), the vulnerability of power systems to these hazards (e.g., Pennington et al., 2021; Piccinelli & Krausmann, 2014), and risk (e.g., Baker et al., 2014; Eastwood et al., 2017; Oughton et al., 2019).

2 Power and Communication Systems

The magnetic storm of March 1940 arrived at a time of significant change in American communication systems. The use of the telegraph, which had flourished since the middle of the 19th century, was, by 1940, declining as the telephone was becoming an increasingly important form of communication (e.g., Fischer, 1992; John, 2010). The Mann-Elkins Act of 1910 and the Transportation Act of 1920 required telegraph and telephone companies to share transmission lines as “common carriers” in exchange for reasonable fees (e.g., Huber & Kellogg, 1999; John, 2010). This helped to reduce redundancy in communication-transmission services. The Federal Communications Commission (FCC), established in 1934 with passage of the Communications Act, regulated and broadly promoted universal access to telecommunication services. In 1940, the American telegraph services industry was dominated by the Western Union Corporation, the American Telephone and Telegraph Corporation (AT&T), and the Postal Telegraph & Cable Corporation. AT&T and its subsidiary companies formed the Bell System which was the primary FCC–regulated telephone service. These companies contributed substantially to the interconnected long-wire communication system that enabled the transmission of long-distance telegrams and phone calls, and the transmission of television programs across CONUS (e.g., Jewett, 1940; Mueller, 1997), Figure 1.

In 1940 America, electric-power-transmission systems were much less extensive than communication systems, Figure 2. Power companies served regional, state, and, sometimes, neighboring-state communities, but a national-scale power grid, a familiar concept today, did not exist back then. There were relatively few grid interconnections; and very long high-voltage transmission lines were not yet especially common (e.g., Cohn, 2017). In 1940, American urban centers and most city homes had electricity, but most farms and many rural areas did not. However, all of this was in the process of changing. With significant support from the Rural Electrification Administration and the Rural Electrification Act of 1936 (e.g., Tobey, 1997), the portion of farms in the United States with electricity increased from 10% in 1930 to 33% in 1940 (Pate, 1975, his Chapter S, Tables 108–119). From 1930 to 1940, generating capacity in the United States modestly increased, from 41,153 to 50,962 MW, although, over the same interval of time, the number of generating stations slightly decreased, from 4,043 to 3,918 (Pate, 1975, his Chapter S, Tables 86–94). From these statistics, we can infer that generating stations, still small by modern standards, were getting larger, and transmission distances between generating stations and load centers, still short by modern standards, were increasing. In 1940, the electric-power industry was only beginning to take advantage of economies of scale that were already being exploited by the communications industry.

Shortly after the March 1940 storm began, Captain N. H. Heck of the Coast and Geodetic Survey (CGS) remarked “It may not be the worst storm we’ve known, but such storms seem worse than ever now, because there’s so much new electrical equipment everywhere to be put out of order” (Times Dispatch, 1940). In the years and decades that followed, even more new equipment would be put “everywhere,” although with different effect. As the national communication system was expanding, its technology was evolving to rely on a diversity of transmission technologies, microwave transmission stations, optical cables, and satellites (e.g., O’Neill, 1985). Although connected in terms of signals, modern communication systems do not much rely on long-wire electrical conductors. As a result, American communication transmission systems are, today, much less vulnerable to magnetic storms than they were in 1940. On the other hand, in 1940, American power transmission systems were expanding. This continued in the subsequent decades. From 1940 to 1950, the proportion of American farms with electricity increased from 33% to 78% (Pate, 1975, his Chapter S, Tables 108–119), and generation capacity increased from 50,962 to 82,850 MW, but the number of generating stations decreased from 3,918 to 3,867 (Pate, 1975, his Chapter S, Tables 86–94). Such progress was enabled by larger generating stations, greater grid interconnection, and longer transmission lines (e.g., Edison Electric Institute, 1973, their Table 49B). Today, virtually all American farms are served with electricity, transmission lines are sometimes very long, and CONUS has, essentially, three giant electric-power interconnections (e.g., Cohn, 2017). Since longer lines mean greater integrated storm-time geopotentials between grounding points, these developments may have contributed to an increase in the vulnerability of CONUS electric-power systems to magnetic storms (e.g., Barnes et al., 1991; Kappenman, 2004; Liu et al., 2010; Piccinelli & Krausmann, 2014; Slothower & Albertson, 1967). In this light, then, we understand why modern interest in storm-induced geoelectric fields is focused on impacts on electric-power systems rather than on communication systems.

3 Transmission-System Ground Connections

Early telegraph and telephone systems were often based on the transmission of voltage signals on single wires connecting transmitting and receiving stations (e.g., Magnusson et al., 2001, their Appendix D). Ground connections at each station allowed current to return through the Earth, and, outside of periods of geomagnetic disturbance, grounding provided a means for (nearly) nulling out the transmission wire’s baseline voltage. Closed-circuit or “fully metallic” systems—with current return coming through a second wire—offered a less noisy means of signal transmission (e.g., Fagan, 1975, his p. 203). By 1940, metallic systems had replaced most single-line systems (Germaine, 1940; Ireland, 1940; McNish, 1940). Metallic systems were not, however, entirely ungrounded, as shown in Figure 3. If lightning or system malfunction resulted in significant potential differences between long transmission wires, current would “break” across protector blocks and flow into the Earth. But with storm induction of significant potential differences in the Earth (significant “geopotentials”), the blocks would operate in the opposite way, with current flowing from the Earth and into equipment. Ireland (1940) notes that, for AT&T telephone systems in place at the time of the March 1940 storm, the average protector broke at 450 V, although some broke with potential differences as low as 350 V and others as high as 575 V. Ireland (1940, his p. 191) reports that on 24 March 1940, protector blocks broke on some (unspecified) lines that were only 20–25 miles (32–40 km) long, indicating geopotentials of >350 V and geoelectric fields of >8.75 V/km.

4 Long-Line Anomalies and Geopotentials

We use a list of 74 “anomalies” representing reports of operational interference at CONUS communication-network stations during the 24 March 1940, magnetic storm (Love & Murphy, 2022); 72 are reported by Germaine (1940); 2 are reported by Ireland (1940); and 2 are reported by both Fleming (1940) and McNish (1940). All of these instances of interference are attributed by these authors to Earth currents induced by the storm. Storm-induced system voltages on long lines connecting communication depots were sufficient to trigger the protective blocks discussed in Section 3. Germaine notes that, as a result of the storm, nearly 800 protectors in just the state of Wisconsin required replacement. Germaine does not give occurrence times for the anomalies he reports, although he notes that severe communication-system interference occurred on 24 March 1940, between 10:00 and 16:00 Eastern Standard Time (EST = Universal Time (UT)−05:00). Both Germaine (1940, his Figure 2) and Ireland (1940, his Figures 7 and 8) show (incomplete) time series of geopotential measured on a grounded AT&T line between Minneapolis, Minnesota, and Fargo, North Dakota; anomalous geopotential commenced at about 9:00 EST, with amplitudes of >300 V lasting for perhaps 10 min, and instantaneous amplitudes exceeding 400 V (even outside the long gap in the time series). Assuming a straight line distance of 344 km between these cities, a potential of >400 V corresponds to a geoelectric field of >1.16 V/km. Ireland provides a vivid description (but without a data plot) of the measurement of potential on a New England Telephone and Telegraph line between Boston, Massachusetts, and Keene, New Hampshire. Anomalous potentials commenced at about 09:00 EST (Ireland, 1940, his p. 193), and by 10:00 EST, the geopotential exceeded the 125-V limit of the measuring voltmeter. This motivated the insertion of a second voltmeter in series with the first. At 11:00 EST, the geopotential exceeded 300 V, and a third voltmeter was inserted in series with the others. At 11:30 EST, the geopotential exceeded 450 V, and a fourth voltmeter was inserted. At 11:35 EST, the 600 V limit of the four voltmeters was exceeded. With no additional voltmeters at hand, engineers estimated (by extrapolation) that the potential might have attained 800 V for “about 4 or 5 s.” Given the 119-km distance between Boston and Keene, this geopotential corresponds to a geoelectric field of 6.72 V/km. Fleming and McNish report anomalous geopotentials on a Western Union line between New York City and Binghamton, New York, commencing at about 09:00 EST. Judging from their plots, reproduced here as Figure 4, there were periods, early on in the storm, when amplitudes of >150 V lasted for perhaps 10 min. Later on, the sensitivity of the voltmeter was reduced (and the peak-to-peak recording range increased). Although the recorded trace is not clear during this time, amplitudes of several hundred volts lasted for perhaps 10 min. Fleming and McNish note a maximum of >800 V, after about 12:00 EST lasting “about 30 s” (McNish, 1940, his p. 363). Given the 222-km distance between New York City and Binghamton, a geopotential of >800 V corresponds to a straight-line geoelectric field of >3.60 V/km. The anomalies for the March 1940 storm reported in journal articles are given in Supporting Information S1.

To supplement our list of operational communication-system anomalies, we searched newspaper articles for reports of telegraph and telephone-system interference during the 24 March 1940, storm specific for a city or town (not general reports of national-scale interference). We found 110 such reports. The formats of many of the newspaper articles are similar—reproducing content supplied by the Associated Press news service, together with local details of the storm’s impact. It is these local details that interest us. Qualitative descriptions range from “wire trouble” for the local bureaus of the Western Union and Postal Telegraph companies (e.g., Philadelphia Inquirer, 1940; Times Dispatch, 1940), to “wrecked” communication lines (Green Bay Press-Gazette, 1940), to “interruption” of radio programs that would normally have been “piped” across the country on telephone lines and then broadcasted locally (e.g., Minneapolis Star, 1940). As dramatic as such descriptions sound, quantitative details are more useful for our analysis. Of the newspaper reports, 57 of the 110 give the start time of the interference, and 53 of the reports (all from CONUS) give the end time of the interference; these instances of interference were primarily experienced from 10:00 to 17:00 EST, March 24. We found three newspaper articles that mention storm-induced geopotentials on long lines. On a line of unstated ownership between St. Louis, Missouri, and Springfield, Missouri, >750 V was measured (New York Times, 1940b). For a straight-line distance of 314 km, this potential corresponds to a geoelectric field of >2.39 V/km. On a line possibly owned by the Western Union between Washington, DC, and Cumberland, West Virginia, 400 V was measured (Smith, 1940). For a straight-line distance of 354 km, this potential corresponds to a geoelectric field of 1.13 V/km. The North American communication-system anomalies for the March 1940 storm we have identified from newspapers are included in Supporting Information S1.

Our list includes 23 operational anomalies for North American electric-power systems, all of which are attributed to storm-induced geoelectric fields on 24 March 1940, (Love & Murphy, 2022). These are reported, with some specificity, by Davidson (1940, 1941), of the Consolidated Edison Company of New York on behalf of 22 power companies in both CONUS and Canada; the anomalies are variously described as “voltage dips,” “tripping” of transformer banks, “reactive power,” etc. 10 of the 23 anomalies are given with start and end times. In most cases, the name of an associated electrical substation or generating station is given, sufficient to identify site latitudes and longitudes using Department of Homeland Security records and internet searches; in other cases, only the city or town name is given. As already noted, power-system interference during the March 1940 storm was not serious enough to interrupt electricity transmission service (Davidson, 1940, 1941). Other than the measured geopotentials reported for long lines connecting specific grounded endpoints, the long-line anomalies for the March 1940 storm are, essentially, qualitative. Furthermore, each anomaly at each identified station is related to interference on long lines that connected one or more other stations, but, in most other cases, we do not know the system connectivity for the reported interference. The North American power-system anomalies for the March 1940 storm are given in Supporting Information S1.

In addition to the geopotential measurements discussed above, we know of two measurements on relatively short lines. Geopotential variation was measured on a 43-km-long AT&T line between Netcong, New Jersey, and Stroudsburg, Pennsylvania (Durkee, 1940, his Figure 2); the amplitude reached >100 V, for a straight-line geoelectric field of >2.33 V/km. Geopotential was also measured on a 76-km-long line between Butte, Montana and Helena, Montana (Montana Standard, 1940), 150 V, for a straight-line geoelectric field of 1.97 V/km. In Section 11, we compare measured long-line potentials with statistical estimates of extreme-value geopotentials, such as would have been realized during the storm of March 1989. Those estimates rely on local electromagnetic transfer functions obtained from magnetotelluric measurements acquired at survey sites across CONUS with a nominal 70-km spacing. But Murphy et al. (2021) find that, in regions with spatially complicated geology, which are often associated with spatially complicated surface impednace, accurate estimates of geopotentials require magnetotelluric surveying with site spacing closer than the length of the corresponding long line. In this respect, the Netcong-Stroudsburg and Butte-Helena lines were short in comparison to the site spacing of the survey data we use. Therefore, we do not consider these short lines in our comparisons of the March 1940 measured geopotentials with the March 1989 estimates.

A few anomalies are reported with start times that precede the 08:49 EST time of the first sudden commencement of the 24 March storm, discussed in Section 7. For example, vigorous geopotential variation on the Netcong-Stroudsburg line is said to have begun at “about” 08:42 EST (Durkee, 1940, his p. 306). Such “vigorous variation” could, plausibly, have actually begun 7 min later. We infer that the analog geopotential recording system that was used by AT&T provided somewhat inaccurate time stamps. Additionally, in Racine, Wisconsin, the local Western Union office could receive “no messages” from Chicago from 01:00 to 12:00 Central Standard Time (CST) or 02:00 to 13:00 EST on 24 March (Journal Times, 1940). The start time, here, is hours earlier than the first sudden commencement on 24 March. A possible interpretation is that the reported start time is simply erroneous; perhaps, instead of 01:00 CST, what was meant was 10:00 CST? We do not attempt to adjust any of these seemingly problematic time stamps. They do not affect any of our interpretations.

5 Analog Magnetic Observatory Data

Since the middle of the 19th century, ground-based magnetometer observatories around the world have monitored geomagnetic field variation (e.g., Barraclough et al., 1992; Love, 2008). In 1940, CGS operated two magnetic observatories within CONUS: the Cheltenham (CLH), Maryland, observatory (1940 geomagnetic latitude: 49.90°N) (e.g., Love & Lewis, 2021) and the Tucson (TUC), Arizona, observatory (1940 geomagnetic latitude: 40.19°N) (e.g., Love et al., 2017). The Dominion Observatory, an institution of the Canadian federal government, operated two magnetic observatories: the Agincourt (AGN), Ontario observatory (54.87°N), and the Meanook, Alberta, observatory (e.g., Lam, 2011; Newitt & Coles, 2007). At these facilities, geomagnetic field variation was recorded automatically using analog magnetometer systems (e.g., Schröder & Wiederkehr, 2000): For each geomagnetic vector component, a beam of light was projected onto a small mirror attached to a magnetized needle suspended on a fiber. The orientation of the needles would change in response to geomagnetic field variation. Resultant deflections of the light beams were recorded as wiggly time-series traces on photographic paper mounted on a cylinder that rotated once per 24 hr. Each day, an observatory worker would remove the paper “magnetogram” from the cylinder, develop it with chemicals, and measure the amplitudes of the traces to obtain a sequence of hourly values for each vector component.

Before the era of electronic digital data, hourly analog observatory data were published in tables in yearbooks. The CGS, however, appears to have never published yearbooks for 1940, likely due to other priorities brought by the World War. Still, by a route we do not know, CLH and TUC hourly average values ended up being archived in computer-readable form at the World Data Centers (WDCs), although, as with other observatory data, without the metadata given in yearbooks. Hourly values are useful for some applications, but they do not represent the rapid storm-time geomagnetic field variation of relevance to studies of induced geoelectric fields. For that, raw analog magnetograms are more useful. Unfortunately, most of the historical CGS magnetograms were lost years ago. Fortunately, reproductions of CLH and TUC magnetograms for the March 1940 storm can be found in Fleming (1940); we show a copy of the CLH magnetograms in Figure 5. The CLH traces are readable before the storm’s first sudden commencement at 08:49 EST on 24 March and for about an hour after that (Ludy, 1940). But during the storm’s main phase, geomagnetic field variation was extremely rapid and of very high amplitude; the light beams were only faintly recorded on the photographic paper, making that part of the CLH magnetogram “unsatisfactory” for detailed use, as noted by McNish (1940). The TUC magnetograms were similarly unsatisfactory (Hershberger, 1940). From this, we understand why the WDC values for TUC have a 5-hr data gap from 10:00 to 14:00 EST on 24 March. On the other hand, the WDC digital archive of hourly CLH values appears to record the entirety of the 24 March storm, but given the reports by Fleming and Ludy, we are suspicious of their integrity; we suspect that they are either crude estimates or some sort of filler values; here, a yearbook would likely have been informative. We have not seen copies of the AGN magnetograms, but we note that the yearbook for AGN records several gaps in hourly values on 24 March and 25 March (Jackson & Ross, 1964). The WDC digital archive for another CGS observatory, that at San Juan (SJG), Puerto Rico, contains several identical consecutive values in horizontal intensity on 24 March. We suspect that these are filler values representing problematic recording; again, a yearbook would likely have been informative. Outside of North America, the March 1940 storm was successfully recorded at low-latitude observatories in Cape Town (CTO), South Africa (Magnetic Observatory, University of Cape Town, 1944), Alibag (ABG), India (Rangaswami & Basu, 1940), and in Kakioka (KAK), Japan, Figure 6. At mid-latitudes, field variation at Watheroo, Australia was too “violent” to be satisfactorily recorded (Parkinson, 1940), but geomagnetic field variation was successfully recorded at Niemegk, Germany (e.g., Fleming, 1941, Figure 17).

6 Geomagnetic Indices

Geomagnetic activity indices are important products derived from observatory magnetometer time series (e.g., Mayaud, 1980; Rangarajan, 1989). Among the oldest and most frequently used are local K indices (e.g., Menvielle & Berthelier, 1991). In detail, K values are integers that are roughly proportional to the logarithm of the minimum-to-maximum range of irregular geomagnetic field variation realized over 3-hr intervals of time at individual mid-latitude observatories, with K = 0 for variation of only a few nanoteslas (very calm geomagnetic conditions) up to K = 9 for a variational range greater than several hundred nanoteslas; note that at K = 9, the index is effectively saturated since greater values are not accommodated in the definition of the index (Bartels et al., 1940). The exact variational range assigned for individual K values is set mainly on the basis of geomagnetic latitude. Historically, K indices were developed to facilitate inter-comparison of geomagnetic field variation from one observatory to another. In Figure 7a, we show K time sequences from the CLH and AGN observatories for the March 1940 storm (Johnston, 1941; Johnston & Heck, 1941). Both CLH and AGN attained K = 9 twice during the storm, although we infer that these values from these observatories are rough estimates for the most active periods of the March 1940 storm, when CLH and AGN disturbance was so severe as to make detailed use of the analog magnetograms “unsatisfactory.” In comparison, at the Fredericksburg (FRD), Virginia, observatory, which replaced CLH in 1956, geomagnetic variation attained K = 9 four times during the March 1989 storm and five times during the October 2003 Halloween storm.

The storm-time disturbance index Dst is calculated from a weighted average of horizontal-component geomagnetic disturbance measured at several low-latitude observatories (Sugiura & Kamei, 1991). Dst is useful for identifying the primary phases of storm development, and it is often interpreted in terms of the strength of the magnetospheric ring current. With main-phase strengthening of this current and the corresponding decrease in low-latitude, horizontal-component geomagnetic field strength, Dst decreases from its pre-storm, near-zero value. Minimum Dst is often used as a measure of the absolute strength of a magnetic storm (e.g., Gonzalez et al., 1994). The standard version of Dst is provided as a service of the World Data Center for Geomagnetism, Kyoto et al. (2015), Kyoto WDC; their version of the index covers the years 1957-present. Recently, Hayakawa et al. (2022) used historical yearbook records and magnetograms from four low-latitude observatories to construct a Dst time sequence for the March 1940 storm; we use their Dst time sequence, Figure 7b. They estimate that the storm’s greatest strength occurred at 15:00–16:00 EST when Dst reached its minimum value of −389 nT; see Figure 7b. This value is comparable to that of the October 2003 Halloween storm; minimum Dst = −383 nT, but less than that of the March 1989 storm; minimum Dst = −589 nT (e.g., Allen et al., 1989). A statistical analysis of extreme Dst values suggests that a storm with minimum Dst < − 389 nT occurs, on average, about once per 1 or 1.5 solar cycles (e.g., Love, 2020, his Figure 7a; Love, 2021, his Figure 2).

So, at least in terms of the local K and global Dst magnetic indices, the March 1940 storm can certainly be described as “big,” but not so big as to be considered rare. We recall that Fleming described the March 1940 storm as being one of “unusual violence” and as having rapid geomagnetic “fluctuation” (Fleming, 1940, his p. 476, 480). In terms of just the minimum-to-maximum range of variation, such as measured by K-indices, Fleming noted that the March 1940 storm was “not much greater” than that of April 1938. Indeed, we appreciate that magnetic storms are complicated phenomena, that some storms can be extreme in some ways, but not so extreme in other ways. Standard geomagnetic indices, by their definition and low time resolution, are relatively insensitive measures of rapid geomagnetic field variation that can efficiently induce the geoelectric fields that interfere with grounded long-line systems. Such interference across CONUS is, perhaps, the most notable characteristic of the March 1940 storm. We return to this issue in Section 11, where we compare geopotentials measured on long lines during the March 1940 storm with geopotentials that would have been induced on similar lines during the March 1989 magnetic storm.

7 A Narrative of Events

From 19 March to 1 April, and during the declining phase of solar cycle 17, a complex sunspot group crossed the solar disk (e.g., Jones, 1955). This active region was the source of numerous solar flares, including, notably, a cluster of flares that erupted on 22 March between 20:33 and 21:28 EST, and a brilliant flare that erupted on 23 March at 06:24 EST (e.g., d’Azambuja, 1940; Jones, 1955; Newton, 1940), inducing a solar-flare effect perturbation in dayside ground-based magnetograms (e.g., Jones, 1955). The great magnetic storm of 24 March 1940, was preceded by another storm, one of modest intensity. It began at 01:17 EST on 23 March, with a sudden commencement, seen in magnetograms from low-latitude observatories as a step-like positive enhancement in horizontal intensity (Hershberger, 1940; Ludy, 1940; Parkinson, 1940; Rangaswami & Basu, 1940). A storm initial-phase then followed, a duration of time characterized by positive Dst, something that is usually caused by solar-wind pressure with northward interplanetary magnetic field (IMF), which is not optimal for field-line connection onto the magnetopause. In this case, the initial-phase was from 01:00 to 04:00 EST, 23 March. This was followed by a storm main phase and a decline in Dst, which soon reached a minimum Dst = −72 nT from 17:00 to 18:00 EST, 23 March.

A pair of sudden commencements heralded the great magnetic storm of 24 March 1940. The first of these might have been due to a coronal mass ejected at the time of the 20:33–21:28 EST 22 March cluster of flares; Hayakawa et al. (2022) assign an ejection time of about 21:00 EST 22 March. At the Earth, the first sudden commencement of 24 March occurred at 08:49 EST. At low-latitude observatories around the world, an increase in horizontal-component intensity was recorded, for example, as 73 nT at KAK (Araki, 2014). Accepting Hayakawa et al.’s inferred ejection time for the storm’s first interplanetary coronal-mass ejection (ICME), its Sun-Earth transit time was 35.82 hr for an average speed of 1,160 km/hr, or 0.0279 astronomical units/hr (AU/hr). Following the sudden commencement, an initial-phase then ensued, with Dst positive for two consecutive hours, 42 nT from 09:00 to 10:00 EST and 45 nT from 10:00 to 11:00 EST. An initial-phase was clearly recorded in KAK horizontal-component magnetograms, Figure 7, attaining a temporary maximum at about 09:30 EST; a similar initial-phase is not seen in the CTO magnetogram (Araki, 2014). From 07:00 to 10:00 EST on 24 March, and inclusive of the first sudden commencement, AGN and CLH K attained values of 7, Figure 7a. From 10:00 to 13:00 EST, AGN K attained a value of 8, CLH K attained a value of 9, Figure 7a. Starting at about 10:00 EST, horizontal-component geomagnetic disturbance at CLH was so rapid that the photographic paper scarcely recorded the variation (Fleming, 1940, his figure on p. 477); CLH K-values are likely estimates based simply on the extreme levels of disturbance. Starting at about 10:00 EST, large-amplitude variation with periods of about 15 min was recorded in horizontal-component KAK and CTO magnetograms, but these were only faintly recorded in the ABG magnetogram.

After the first commencement, at about 09:00 EST, variation in measured geopotential was noted on the New York City to Binghamton (Fleming, 1940; McNish, 1940), the Minneapolis to Fargo (Ireland, 1940), and the Boston to Keene (Ireland, 1940) lines. At 09:30 EST, telegraphs used by the New York Central Railroad were “hampered” (Daily News, 1940). Starting at 10:00 EST, when horizontal-component geomagnetic disturbance at CLH was so great as to scarcely be recorded, communication-system interference was widespread and significant. For example, the city of Cincinnati, Ohio, was described as having been “bombarded” by the Sun’s electricity, with significant interference caused on telegraph and telephone systems, making service “at times, impossible” (Cincinnati Post, 1940). Starting at 10:00 EST, in Madison, Wisconsin, a local Western Union manager reported that fuses “blew out” continuously (Wisconsin State Journal, 1940). Also starting at 10:00 EST, in Fitchburg, Massachusetts, the local Western Union system was “disrupted” due to “burnt-out relay equipment.” More globally, starting at 10:20 EST, Associated Press communication links on transatlantic cables between the United States and Europe were “interrupted” (Boston Daily Globe, 1940). Starting at 10:30 EST, additional communication-system interference was widespread and significant. In New York City, for example, teletype machines, used for transmitting news service reports and business information, suddenly went “haywire,” delivering garbled text messages, but because 24 March 1940, was Easter Sunday, the impact on many businesses was less than it would have been on a normal working day (New York Times, 1940b). Starting at 10:30 EST, in Minneapolis, 250 of 481 Northwestern Bell Telephone toll circuits were “out of service,” and land-line “pipe” transmission of radio programs was interrupted (Minneapolis Star, 1940). Also starting at 10:30 EST, the Associated Press reported that their “285,000-mile network of leased lines” was “out of use” (Smith, 1940).

At about this time, the decline in horizontal-component intensity in the KAK magnetogram suggests that the storm was entering a main phase, something usually associated with a turning of the IMF southward, but, then, the arrival at Earth of the storm’s two ICMEs caused a second sudden commencement at 10:38 EST on 24 March. Notably, this was one of the largest commencements in magnetometer history, with a positive increase in horizontal-component intensity of 164 nT at CTO (Araki, 2014), >273 nT at KAK (Araki, 2014), 321 nT at ABG (Rangaswami & Basu, 1940), and, incredibly, >425 nT was possibly recorded at TUC (the report is anecdotal) (Hershberger, 1940). These amplitudes are truly unusual; less than 5% of the commencements recorded at KAK have amplitudes greater than 50 nT (Araki, 2014). This commencement was possibly due to a coronal mass ejected at the time of the 06:24 EST, 23 March brilliant flare (e.g., Hayakawa et al., 2022). Accepting this inferred ejection time for the storm’s second ICME, its Sun-Earth transit time was 28.23 hr, for an average transit speed of 1,472 km/s or 0.0354 AU/hr (e.g., Hayakawa et al., 2022)—more than two standard deviations higher than the average ICME speed at 1 AU (e.g., Yurchyshyn et al., 2005). Still, such speeds are not unusual for great storms. The August 1972 storm, for example, was due to three ICMEs, the last one traveling 0.0685 AU/hr (e.g., Cliver et al., 1990). In contrast to the first commencement, the second commencement was recorded as a spike in magnetograms from low-latitude observatories, no initial-phase transpired. Instead, the storm entered into a main phase with a rapid decline in low-latitude horizontal-component intensity due to intensification of the equatorial ring current. From 10:00 to 13:00 EST on 24 March, and inclusive of the second sudden commencement, AGN K attained a value of 8, CLH K attained a value of 9, Figure 7a. Only minutes after the second commencement, at 10:44 EST, the first interference for electric-power systems, with the Niagara Hudson Power Company in Lockport, western New York state experiencing alternating current waveform distortion (Davidson, 1940). Independently, and quickly thereafter, at 10:45 EST, the Northern States Power Company in Minneapolis, Minnesota, reported a voltage surge, and at 10:50, Central Maine Power reported tripping of transformer banks (Davidson, 1940). After the second commencement, additional interference was caused on CONUS communication systems.

From 11:00 to 12:00 EST, the storm’s main phase began, and Dst declined to −143 nT. At 11:00 EST, Postal Telegraph in Washington D.C. reported “paralysis” on their lines (Smith, 1940). By 11:35 EST, storm-induced geopotential on the New England Telephone and Telegraph, Boston to Keene line possibly reached 800 V (Ireland, 1940). The short duration of this geopotential peak, only “about 4 or 5 s” (Ireland, 1940, his p. 193), indicates rapid inducing geomagnetic field variation. At 11:45 EST, New England Telephone and Telegraph reported a “bad interruption” of service at Brattleboro, Vermont (Brattleboro Reformer, 1940). At 11:48 EST, simultaneous to within a minute, three separate instances of interference were experienced on electric-power systems (e.g., Fleming, 1940): The Philadelphia Electric Company, Pennsylvania, reported a transformer tripping (Davidson, 1940; Philadelphia Inquirer, 1940), as did the Ontario Hydro Electric Power Commission; the Consolidated Edison system of New York reported the first of a series of voltage disturbances (Davidson, 1940). Shortly after all of that, from 11:50 to 11:54 EST, the Northern States Power system reported their “most severe” voltage surges (Davidson, 1940). From 12:00 to 13:00 EST, Dst declined to −252 nT, and it was, at about this time, that geopotential measured on the Western Union New York City to Binghamton line reached a peak of >800 V (Fleming, 1940; McNish, 1940) for “about 30 s” (McNish, 1940, his p. 363). From 13:00 to 14:00 EST, Dst declined further to −299 nT. Using a simple parameterization of the latitudinal radius of the auroral oval in terms of Dst (e.g., Milan et al., 2009), we infer that, during the period when Dst < − 300 nT, the southern edge of the auroral oval would have descended to New York City (1940 geomagnetic latitude: 51.99°N), suggesting that this North American metropolis was exposed to severe levels of geomagnetic disturbance generated by a main-phase auroral electrojet.

During the storm’s main phase, from 13:00 to 16:00 EST, AGN K increased to 9, but CLH K declined to 8, Figure 7a. At 14:00 EST, the Associated Press reported that their entire “10,000-mile wirephoto network” was “out of order” (New York Times, 1940b; Smith, 1940). Dst continued to decline, reaching its deepest value from 15:00 to 16:00 EST at −389 nT. For an hour or two more and into the storm’s final recovery, interference continued on some long-line systems, but after this, we found no additional reports of interference for either communication or power systems (e.g., New York Times, 1940a). During the local night of 24–25 March, aurorae were witnessed in many American skies (e.g., Morning Examiner, 1940).

8 Interacting Coronal-Mass Ejections

From Section 7, we note that the two sudden commencements on 24 March, both of high amplitude, occurred relatively close in time, just 1.82 hr apart. In light of this observation, we hypothesize, as others have before us, that ICME interaction might have been partly responsible for the “violent” ground-level geomagnetic disturbance realized during the 24 March 1940, storm (e.g., Hayakawa et al., 2022; Lefèvre et al., 2016). Such interaction is known to result in highly geoeffective ICMEs (e.g., Burlaga et al., 1987; Hess & Zhang, 2017; Lugaz et al., 2017; Scolini et al., 2020). The interaction hypothesis is partly supported by a straightforward calculation. First, consider that, at 1 AU, the typical radial dimension of a CME cloud of ejecta is approximately 0.15 AU (Burlaga et al., 1990; Klein & Burlaga, 1982), and that the typical thickness of an ICME sheath, the standoff distance between the leading edge of the ICME mass and the shockwave that it is generating, is approximately 0.10 AU (Russell & Mulligan, 2002). Then, consider the ejection times of the two ICMEs that arrived at the Earth on 24 March, given in Section 7; by the time of the storm’s first commencement, the two ICMEs of 24 March were separated by only 0.0646 AU. From this, we infer that the separation distance between the 24 March storm’s two ICMEs was likely less than the radial dimension of the plasma cloud at 1 AU. This suggests that, by the time the 24 March ICMEs arrived at the Earth, the shock of the second ICME was compressing the backside of the first CME cloud, concentrating its turbulent plasma and magnetic field, perhaps by a factor of (0.10 + 0.15)/0.0646 ≈ 4—simple evidence supporting the hypothesis that the ICMEs were interacting. In the reference frame of the solar wind, radial compression would have shortened the radial autocorrelation length scale in the turbulent structures of the solar wind and IMF. Then, as these compressed structures passed the Earth, from the reference frame of the Earth, the shortening of the autocorrelation timescale forced the magnetospheric-ionospheric system into a state of rapid time dependence. We further speculate that the 23 March storm preconditioned interplanetary space (e.g., Temmer et al., 2017), and also the magnetosphere (e.g., Kamide et al., 1998; Kozyra et al., 2002), for the severe disturbance that followed on 24 March.

Comparisons should be made with other great magnetic storms having multiple sudden commencements. For example, the March 1989 storm was preceded by two sudden commencements (e.g., Boteler, 2019; Nagatsuma et al., 2015). These were caused by interacting ICMEs (e.g., Boteler, 2019; Nagatsuma et al., 2015), although because the sudden commencements were separated in time by 6.27 hr, the interaction was possibly not quite as great as it was for the March 1940 storm. The August 1972 storm (Knipp et al., 2018) was preceded by three sudden commencements, all on the same day. The first two were separated in time by only 1.02 hr, so we might infer that the related ICMEs were strongly interacting, but the second and the third commencement (which initiated the storm) were separated by a much longer 18.57 hr. These differences in ICME arrival times might have influenced the ensuing evolution of each storm, but any specific effects related to consecutive sudden commencements are, here, purely qualitative supposition.

9 Local Time, Latitude, Storm Phase

From Figure 7c, we note that almost all North American long-line system interference caused by the March 1940 storm occurred during local daytime (near noon). Most communication-system interference came soon after the second sudden commencement on 24 March, and most of this was in CONUS. We note, furthermore, that the first report of interference for North American power systems came only a few minutes after the second sudden commencement, and this was in CONUS. The first report of interference on power systems in Canada came more than an hour later. It was early in the storm’s main phase that maximum geopotentials were measured on long lines in eastern CONUS. For comparison, the March 1989 storm induced severe power-system interference across North America, but, before CONUS, interference was experienced in Canada where it caused the collapse of the Hydro-Québec system just minutes after the storm’s second sudden commencement and after local midnight. Most interference on CONUS power systems caused by the March 1989 storm came during storm main phase and across a range of local times, during both daytime and nighttime (e.g., Boteler, 2019, his Table A1; Love et al., 2022, their Figure 5). Both the Carrington Event of September 1859 and the great storm of May 1921 induced interference on telegraph systems around the world during storm main phase and across a range of local times (e.g., Boteler, 2006; Love et al., 2019).

In totality, these reports of long-line system interference induced by individual great storms do not clearly indicate a pattern in local time and storm phase. Neither do integrated studies of many storms, mostly of modest intensity. Shi et al. (2022) analyze 1-s resolution geoelectric time series collected in the upper Midwest during a magnetotelluric survey. Geoelectric amplitudes >0.50 V/km occurred most frequently during local nighttime (Shi et al., 2022, their Figure 6a) but across different storm phases (Shi et al., 2022, their Table 2). A study of many years of 10-s resolution GIC data collected at auroral latitudes finds that amplitudes tend to be greatest during substorms and at local nighttime (e.g., Hajra, 2022), although some GIC events were specifically identified during local daytime (e.g., Tsurutani & Hajra, 2021). Other studies of the geomagnetic field variation, measured with 1-min and 10-s resolutions at auroral latitudes, find that rates of change, across all storm phases, are greatest at local pre-noon and pre-midnight (e.g., Juusola et al., 2015; Schillings et al., 2022; Viljanen et al., 2001). A different study of 1-min-resolution magnetometer data from a globally distributed set of observatories finds that extreme-value disturbance, again, across all storm phases, tends to be greatest at auroral latitudes, pre-midnight (e.g., Rogers et al., 2020, their Figure 5). And yet a different study of 1-min-resolution magnetometer data finds that extreme-value disturbance, again, across all storm phases, tends to be greatest at mid-latitudes from local morning to local midnight (Freeman et al., 2019, their Figure 6).

Since our focus is on the March 1940 storm and ground-level geoelectromagnetic variation that brought interference to long-line systems across North America, and, especially, CONUS, and since this interference occurred during local daytime, the nightside substorm paradigm is not very relevant in our interpretations. Having said that, dayside reconnection and dayside electrojets, which we speculate played roles in the March 1940 storm, can be understood as the dayside equivalent of the nightside substorm (e.g., Glassmeier & Heppner, 1992). This suggests opportunities for research. Even while research is concentrated on nightside ground-level geomagnetic disturbance that can be generated by substorms (e.g., Engebretson et al., 2020; Vorobev et al., 2019; Zou et al., 2022) and the GICs that they can induce (e.g., Hajra, 2022), the March 1940 storm serves as a reminder of the need for research on “dayside substorms” (e.g., Gromova et al., 2016; Le & Russell, 1993; Tsurutani & Hajra, 2021), which might, themselves, contribute significantly to the induction of GICs, possibly across mid-latitudes during extremely intense solar-wind conditions.

10 Geographic Organization of Impacts

In Figure 8a, we show a map of the locations of the CONUS March 1940 communication-system anomalies. These are far from uniformly distributed in geography. We note, in particular, the concentration of anomalies in eastern CONUS. We also note the scatter of anomalies across the upper Midwest. In Figure 8b, we show the locations of newspaper reports of local telegraph and telephone-system interference. And, in Figure 8c, we show the locations of power-system anomalies for the March 1940 storm. Although we do not expect long-line communication- and power-system anomalies to be identically distributed, it is important to recognize that the maps in Figures 8a and 8c show consistencies—anomalies of both types are concentrated in eastern CONUS and in parts of the upper Midwest, and these anomalies are rare elsewhere. We also do not expect the (mostly anecdotal) newspaper reports of communication-system anomalies and the anomalies reported in formal publications (authored by communication company representatives) to be identically distributed—the similarities we see in Figures 8a and 8b suggest that the two data sets are fairly representative of March 1940 interference, even though some differences are to be expected. Love et al. (2022) show that the geographic density of historical power-system anomalies in CONUS is disproportionately greater than the geographic density of power-system substations, suggesting that another factor is responsible for the concentration of anomalies in the upper Midwest and East.

The geographic expression of long-line system interference during the March 1940 storm was strongly influenced by the Earth’s surface impedance. To see this, we use results from long-period (10–10,000 s) magnetotelluric surveys across CONUS (Schultz, 2010). At each survey site, simultaneous geomagnetic and geoelectric field variation is parameterized in terms of a frequency-dependent magnetotelluric transfer function—a surface-impedance tensor divided by the permeability of free space (e.g., Chave and Jones, 2012; Simpson & Bahr, 2005; Unsworth, 2007). In Figure 9, we show a map of transfer-function amplitude at survey sites with a nominal spacing of 70 km and for sinusoidal variation at (a) 120 s and (b) 600 s; each ellipse at each site shows transfer-function amplitude as a function of the azimuth of the induced geoelectric field (e.g., Berdichevsky & Dmitriev, 2008, their Equation 1.91). Notably, transfer-function amplitude is high over regions characterized by (electrically resistive) crystalline igneous and metamorphic rock (e.g., Bally et al., 1989), for example, in the upper Midwest (e.g., Bedrosian, 2016; Yang et al., 2015). Transfer-function amplitude is high and polarized in eastern CONUS. Transfer-function amplitude is low and not particularly polarized over (electrically conductive) sedimentary basins, such as in Michigan and eastern Montana.

Important correlations are seen between Figures 8 and 9. During the March 1940 storm, long-line system interference was primarily experienced in the upper Midwest and the East, where, generally, surface impedance is high. Conversely, during this storm, long-line interference was rare in other parts of CONUS, where, generally, surface impedance is low. It is worth noting that, even in 1940, engineers were aware that, if only in general qualitative terms, long-line system interference tended to be high (low) in areas where subsurface rock is electrically resistive (conductive) (Germaine, 1940, his Figure 1; Ireland, 1940, his Figure 5). Indeed, even years earlier, engineers had noted similar correlations from projects dedicated to monitoring geopotentials on grounded long lines (e.g., Corr, 1935, her Figure 1). Similar geographic patterns of long-line system interference have been realized across CONUS during other great storms. Love et al. (2022) show that CONUS power-system interference during the March 1989 storm was experienced in the upper Midwest and the East, again, where surface impedance is high. Additionally, Love et al. (2022) convolve maps of geomagnetic field variation derived from digital magnetometer data with magnetotelluric transfer functions to estimate storm-time geoelectric field amplitudes. They show that power-system interference during the March 1989 storm was experienced when and where geoelectric fields were of high amplitude. The lack of useful magnetometer data across CONUS, Section 5, precludes a similar analysis for the March 1940 storm. The Carrington Event of September 1859 induced GICs and substantial operational interference on telegraph systems in Boston, Massachusetts, and Philadelphia, Pennsylvania (e.g., Boteler, 2006). The great storm of May 1921 induced GICs on telegraph lines that started fires at two and possibly three different railroad stations in New York state (e.g., Hapgood, 2019; Love et al., 2019). The August 1972 storm induced power-system interference across the upper Midwest and East (Albertson and Thorson, Jr., 1974).

Given that, generally speaking, Canada is located beneath the geomagnetically active auroral zone, and that the highly populated provinces of Ontario and Québec are located over the (high surface impedance) Superior Craton (e.g., Hill et al., 2021), we are surprised that, despite searching, we have found few reports of the location, timing, and nature of communication-system interference in Canada. Indeed, many Canadian newspaper reports of the storm focus on what happened in the United States (e.g., Ottawa Journal, 1940; Windsor Star, 1940). While recognizing that storm-time geomagnetic disturbance is usually greater at Canadian latitudes than it is at CONUS latitudes, we might expect that Canadian communication companies had, very early on, experienced greater storm-induced interference than CONUS communication companies. Recalling, now, from Section 3, the importance of long-line grounding, we speculate that, as Canadian telegraph transmission was being transitioned from single-wire systems to metallic systems, to avoid frequent interference, the Canadian companies might have set their protector blocks to break-point voltages higher than those in CONUS. This would have allowed Canadian communication systems to avoid much of the interference that was experienced across CONUS during the March 1940 storm.

11 Comparisons With the March 1989 Storm

Next, we examine the geopotentials measured during the March 1940 storm on the five CONUS long lines shown on Figure 10 and discussed in Section 4. What do they tell us about the intensity of the storm? And how, for example, do they compare with geopotentials that would be induced on similar lines by the most intense storm of the space age, that of March 1989 (e.g., Boteler, 2019)? Following methods outlined by Lucas et al. (2020) and Love et al. (2022), essentially, 1-min-resolution map sequences of March-1989 CONUS geomagnetic variation are constructed by geographic interpolation of digital horizontal-component magnetometer time series acquired at North American magnetic observatories operated by the U.S. Geological Survey and Natural Resources Canada. These maps are convolved with magnetotelluric transfer functions, Figure 9, in the vicinity of each long line to obtain 1-min-resolution geoelectric field time series at each survey site. Projections of the geoelectric fields are integrated along the length of each long line to obtain 1-min-resolution geopotential sequences. For each line, we identify the maximum geopotential attained during local daytime during the March 1989 storm. In Figure 10, we list both the maximum geopotentials measured on the long lines during the March 1940 storm and our estimates for the March 1989 storm. The geopotential measured on the line between Washington, DC, and Cumberland, West Virginia, at 400 V (Smith, 1940), is rather comparable to the 246 V we estimated would have been induced during the March 1989 storm. Otherwise, the March 1940 geopotentials are all significantly higher than our March 1989 estimates; they are, on average, about a factor of >9 times higher.

One might hypothesize that differences in temporal resolution contribute to these differences. Our geopotential estimates for March 1989 have 1-min-resolution, whereas the measured geopotentials are, essentially, instantaneous spot samples taken from (apparently) rapidly varying time series. In Section 4, we note that, during the storm, geopotential on the Minneapolis to Fargo line exceeded 400 V (for an average electric field of 1.16 V/km) at several instances, but there were also periods, some of perhaps 10 min, when the potential exceeded 300 V (0.87 V/km) (Germaine, 1940, his Figure 2; Ireland, 1940, his Figures 7 and 8). It is important to recognize that 300 V exceeds the maximum 157 V we estimate for the March 1989 1-min-resolution geopotential on this line. During the March 1940 storm, geopotential on the Boston to Keene line exceeded the 600 V limit of a series of four voltmeters (5.04 V/km). Unfortunately, the temporal resolution of the 600 V value is not specified and cannot readily be inferred, but the value of 800 V (6.72 V/km) was estimated to have lasted for only “about 4 or 5 s” Ireland (1940, his p. 193). Although both values, 600 and 800 V, far exceed the 47 V we estimate for the March 1989 geopotential for this line, this comparison does not account for possible sub-minute variation in the measured potential. During the March 1940 storm, geopotential on the New York City to Binghamton line appears to have exceeded several hundred volts for periods of about 10 min (Fleming, 1940, p. 479; McNish, 1940, p. 362), and >800 V (>3.60 V/km) was estimated to have lasted for “about 30 s” (McNish, 1940, his p. 363). These values exceed the 76 V we estimate for the March 1989 geopotential for this line. Independent studies of data recording recent magnetic storms find that peak 1-min-resolution geoelectric amplitudes are typically 10%–30% lower than peak 1-s-resolution amplitudes (Grawe et al., 2018; Love et al., 2018; Shi et al., 2022). On the basis of all of this evidence, we understand that rapid (sub-minute) geoelectric field variation might partially account for our estimated one-minute-resolution March 1989 geopotentials falling below those measured on long lines during the March 1940 storm.

12 Comparisons With Industry Benchmarks

The geopotentials measured during the March 1940 storm on the five CONUS long lines listed on Figure 10 should be compared with the 100-year geoelectric benchmark used by the North American Electric Reliability Corporation (NERC) (2016) in projects for checking the exposure and vulnerability of power systems to extreme-event geoelectric fields. Using NERC’s formulas, using the 1940 geomagnetic latitude for the mid-point between Boston and Keene, and taking the highest equivalent surface impedance available from among the various CONUS physiographic regions given in the NERC report, we obtain <4.68 V/km as the NERC 100-yr geoelectric benchmark for a Northeast metropolis like Boston and vicinity. Since we have used the highest possible NERC impedance parameter, the value of <4.68 V/km should be regarded as an upper limit. Still, this upper limit is lower than the geoelectric field of 6.72 V/km estimated from measured geopotential on the Boston to Keene line during the March 1940 storm. We recall, furthermore, from Section 3, that Ireland (1940) estimates, on the basis of blown protector blocks, that geoelectric fields of >8.75 V/km were not uncommon across affected areas of CONUS during the storm. All of this indicates that the NERC 100-year geoelectric benchmark was exceeded across parts of CONUS during the March 1940 storm.

13 Conclusions

When the magnetic storm of March 1940 arrived, American communication-transmission systems were continental in scale, whereas electric-power-transmission systems were still mostly localized and regional in scale, Section 2. Both communication- and power-transmission systems relied on long grounded, electrically conducting transmission lines. In subsequent years and decades, communication technology would evolve to become much less dependent on such lines, while power transmission would grow in scale and become more dependent on them. These different trajectories would contribute to communication systems becoming less vulnerable to magnetic-storm-induced geoelectric fields, and power systems becoming more vulnerable to induced geoelectric fields. As summarized in Sections 3 and 4, the magnetic storm of March 1940 was, in terms of the interference it brought to long-line communication and power systems, one of the most significant ever experienced by the United States. Storm-time geomagnetic field variation was of great amplitude, but, just as importantly, it was also of great rapidity. Geomagnetic variation overwhelmed the recording capabilities of analog magnetometer systems in operation at the time at observatories across North America, Section 5. Standard magnetic indices, by virtue of their definition and limited time resolution, Section 6, do not entirely represent the storm’s actual variational “violence.” A time-ordered narrative of events, Section 7, reveals that most long-line communication- and power-system interference across CONUS was experienced during local daytime, after the second of two storm sudden commencements, and during the early part of the storm’s main phase. The narrative suggests that interacting ICMEs, Section 8, played an important role in forcing geospace into a heightened level of disturbance. The narrative also suggests that reconnection at the magnetopause and “dayside substorms” drove intense ionospheric electrojet currents across dayside mid-latitudes, Section 9. Long-line system interference was geographically organized, with greatest (least) interference experienced across regions of CONUS characterized by high (low) surface impedance, Section 10, most notably in the upper Midwest and eastern United States. Measured potential differences on grounded long lines, Section 11, exceeded 1-min-resolution potentials that we estimate would have been realized during the March 1989 storm. In some places, geoelectric amplitudes induced during the March 1940 storm, Section 12, exceeded American electric-power-industry benchmarks.

Acknowledgments

We thank K. A. Lewis, R. D. Gold, B. S. Murphy, and B. R. Shiro for reviewing a draft manuscript. The CONUS magnetotelluric survey was supported by the National Science Foundation (NSF) EarthScope project (2006–2018) (Schultz et al., 2019–2020; Williams et al., 2010); NASA (2019) (Schultz et al., 2019–2020), and, since 2020, the U.S. Geological Survey (USGS) has been supporting the survey’s completion (Schultz et al., 2020–2023). Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. government.