Rediscovering the observations of solar prominences from 1906 to 1957 recorded at the Madrid Astronomical Observatory

– The Madrid Astronomical Observatory implemented a solar observation program from 1876 to 1986. In addition to sunspots, the observers at this observatory recorded other solar features such as promi-nences. In this work, we have consulted the documentary sources of the Madrid Astronomical Observatory (the information is not digitally available), digitized the records of the observers on the annual number of prominences, and constructed a homogeneous series of the total and hemispheric annual number of promi-nences with heights of 25 00 and more for the period 1906 – 1957. To evaluate the quality of the data and assess their potential, we have compared the Madrid prominence series with the number of prominences recorded by the Astronomical Observatory of the University of Coimbra and other time series such as the sunspot number index, solar radio ﬂ ux at 10.7 cm, and sunspot areas. We have also analyzed the hemispheric prominence numbers and the asymmetry index. We obtained the strongest correlation between Madrid and Coimbra prominence series ( r = 0.7), whereas the correlations between the Madrid prominence series and the other solar activity time series are similar ( r (cid:2) 0.6). In addition, we found that the correlation coef ﬁ cient between the Madrid prominence series and the sunspot number is lower than that from the Coimbra prominence series and the sunspot number. We suggest that these differences are a consequence of the way prominences were counted in the Madrid Astronomical Observatory.


Introduction
Prominences are structures in the solar atmosphere appearing like arcs on or above the solar limb (Parenti, 2014). The first prominence observations were made during solar eclipses. For example, during the eclipse of 1860, the observations performed in Spain by Warren de la Rue from Rivabellosa, and by Angelo Secchi and José Monserrat from Desierto de las Palmas (using even photography) confirmed that prominences are a solar feature and their origin is not in the Moon or optic effects (de la Rue, 1862;Secchi, 1875;Foukal, 2004). Since then, spectroscopy started to develop and the first systematic observations of prominences were recorded at astronomical observatories (Bocchino, 1933;McIntosh, 1979;Rusin et al., 1988;Chatterjee et al., 2020;Carrasco et al., 2021; see also the chapter 4 of the monograph by Vaquero and Vázquez, 2009). One of these observatories where prominences were observed was the Madrid Astronomical Observatory.
The Madrid Astronomical Observatory carried out a solar observation program from 1876 to 1986 spanning from the last part of Solar Cycle 11 to the first months of Solar Cycle 22 (Ruiz-Castell, 2008;Aparicio et al., 2014). Sunspots were mainly recorded in this period at the observatory although there are some gaps in the observation series. Moreover, several sunspot catalogs were published by the observatory: the Aguilar catalog covering observations from 1914 to 1920 (Lefèvre et al., 2016), the Carrasco catalog including records for the period 1931-1933 (Aparicio et al., 2022a), and the "modern" catalog with observations for 1952-1986(Aparicio et al., 2018. In addition to the sunspot observations, other solar features of the Sun were observed at the Madrid Astronomical Observatory such as prominences and chromospheric faculae (flocculi) (López Arroyo, 2004). In particular, observations of solar prominences were recorded in this observatory during the 20th century and summarized in annual tables for the period 1906-1957(Jiménez Landi, 1913Gullón, 1950;López Arroyo, 2004).
This work is the first step to exploring the prominent observations made at the Madrid Astronomical Observatory. In Section 2, we present the Madrid prominence observations, the instruments used, the responsible observers, and the retrieved data. We include an explanation and an analysis of the homogenized series we have constructed in Section 3 and the main conclusions of this work in Section 4. Daily observations and summaries including annual data on different parameters that characterize the solar prominences were published in yearbooks and bulletins of the observatory (see Appendix). We note that there is no digital version of the prominence data recorded at the Madrid Astronomical Observatory.
In addition to the daily number of prominences, measurements of the base, height, and intensity (the latter from 1908 onwards) of the prominences were recorded. From 1907 onward, a specific drawing was made for each prominence (López Arroyo, 2004). Note that only a few examples of the drawings of the prominences recorded in Madrid were published in the yearbooks and bulletins. Figure 1 shows one of these examples including a prominence observed at the Madrid Astronomical Observatory on 11 September 1935.
Since it is not possible to have an actual idea of a prominence only from its base, height, and intensity, the International Astronomical Union agreed that the projected area of the prominences should also be recorded (Gullón, 1950). Thus, this parameter was also systematically measured from 1925 onward at the Madrid Astronomical Observatory. Since the projected areas of the prominences were not recorded in Madrid before 1925, Jiménez Landi (1925) measured it from the original observations for the period 1910-1920. Note that no observations were performed for the period 1921-1924 due to unfavorable circumstances, as mentioned by Gullón (1950). The observations continued until November 1936, when they stopped due to the Spanish Civil War (1936)(1937)(1938)(1939). Then, they were performed in Valencia during the period 1937-1938 and resumed in Madrid in September 1939. The last interruption, due to a reconstruction of the pavilion of the observation tower, spanned 1942 and part of 1943. Finally, the prominence observations continued until 1957.

Instrumentation
The instrument used to carry out the first prominence observations was a Grubb spectroscope composed of four 60°prisms and two semi-prims with a configuration that produced a scattering equivalent to that obtained with ten prisms (Jiménez Landi, 1913). The spectroscope was mounted on a Grubb equatorial telescope with a 20 cm aperture and 3 m focal length. In 1925, the spectroscope was substituted by one of the Zeiss brand. In 1936, the Zeiss spectroscope was moved to Valencia and was mounted on a Zeiss equatorial telescope with a 15 cm aperture and 2.2 m focal length. In September 1939, the instrument was returned to Madrid to resume the observations at the Madrid Astronomical Observatory.

Observers
Several astronomers were responsible for the prominence observation program. It was started by Francisco Íñiguez in February 1906. From 1907to 1921 in charge of the observations. After that, the observations were interrupted until 1925. Then, they were resumed by Pedro Jiménez Landi. Upon his retirement in 1932, he was substituted by Enrique Gullón. During the stage in Valencia (from the end of 1936-1938), Rafael Carrasco and Mariano Martín Lorón performed the prominence observations. Finally, in 1954, Manuel López Arroyo assisted Enrique Gullón (Gullón and López Arroyo, 1955).

Data
In this work, two documentary sources including data on prominences observed at the Madrid Astronomical Observatory have been used: (1) yearbooks and bulletins of the observatory (see Appendix) and (2) the Madrid prominence series by Gullón (1950). We remind that prominence drawings were made by the observers of the Madrid Astronomical Observatory. However, they have not been considered for this work because, inter alia, only a few examples are available in the publications of the observatory. As a future task, it would be of interest to locate (and later digitize) all these original prominence drawings made in Madrid.  The total and hemispheric prominence observations performed at the Madrid Astronomical Observatory for the period 1906-1957 were published in yearbooks and volumes of bulletins of the observatory (see Appendix). Daily records and summaries with monthly and annual values were shown in tables in those documentary sources available only in the printed version. We have extracted and combined the records from all those sources to build the annual prominence number series recorded in Madrid in digital format. Table 1 lists the total and hemispheric number of prominences per year included in the summary tables published by the Madrid Astronomical Observatory in the yearbooks and bulletins for the period 1906-1957. We emphasize that the annual number of prominences in Table 1 does not constitute a homogeneous series because:

Yearbooks and bulletins
(1) in 13 of the 47 years (1906-1910, 1918-1920, 1925, 1928, 1930, 1932 and 1938), not all the prominences observed at the Madrid Astronomical Observatory were taken into account for these statistics (some eruptive prominences and other type of prominences named "clouds" by the observers, due to their shapes as clouds, were not considered in this case) (Aguilar, 1910), (2)  We note that the value of the total frequency (ft) shown in Table 1 is not equal to the sum of the values of the frequency for the northern (fn) and southern (fs) hemispheres in some years (1906,1926,1935,1936,1938,1940,1943,(1946)(1947)(1948)(1949)(1950)(1951)(1952)(1953). This can be due to typos in the summary tables published by the observatory or the way to round values of frequencies by the observers.    The annual number of prominences considered by Gullón (1950) was slightly larger than that recorded in the summary tables published in the yearbooks and bulletins of the Madrid Astronomical Observatory (Table 1) in a few years (1906-1910, 1918-1920, 1925, 1928, 1930, 1933, and 1938). This is because Gullón (1950) took into account all prominences recorded in the original observations without discarding eruptive and "cloud" prominences. We note that information on how many eruptive/cloud prominences were ignored is not included in the documentary sources.
On the other hand, prominences considering any height were only listed at the Madrid Astronomical Observatory from 1906 to 1911. While in 34 of the 47 years of the observational period, astronomers of this observatory recorded prominences with heights of 25 00 and more. To construct an annual homogenized prominence number series of prominences with heights of Table 2. Data of the homogenized series of the annual number of prominences with heights of 25 00 and more of the Madrid Astronomical Observatory obtained in this work for the period 1906-1957. The columns represent: (i) the year of the observations, (ii) the number of observation days in the corresponding year, (iii) the number of prominences observed in the corresponding year, and frequency of the (iv) total, (v) north, and (vi) south number of prominences (that is, the total, north, and south number of prominences divided by the number of observation days).  were not recorded (Gullón, 1950). For it, Gullón calculated the average percentage of prominences with heights between 25 00 and 30 00 with respect to the number of prominences of 25 00 and more for the years 1906-1912, 1931-1933, 1935-1936, 1941, and 1944-1949

Homogenization of the Madrid prominence number series
We have constructed a homogenized series of total and hemispheric annual prominence numbers from Madrid data for the period 1906-1957 using the values for prominences with heights of 25 00 and more by Gullón (1950) for the period 1906-1949 and those recorded in the yearbooks and bulletins of the observatory for the period 1950-1957, which followed the methodology by Gullón (1950). This homogenized series of the total and hemispheric annual number of prominences with heights of 25 00 and more recorded at the Madrid Astronomical Observatory for the period 1906-1957 is included in Table 2 (the digitized version is available as Supplementary Material of this work).
The total number of observation days for the period 1906-1957 was 6383. This represents a temporal coverage of 37.0% discarding the years 1921-1924 and 1942 when no observations were made at the observatory. Thus, the annual mean of the number of observation days is around 136. The number of observation days as well as the daily mean of the total and hemispheric number of the homogenized series of prominences observed at the Madrid Astronomical Observatory are shown in Figure 2. Note that the sum of the number of prominences with heights of 25 00 and more of the Madrid homogenized series (column "N" in Table 2) is 30308 of which 28,519 were observed and 1789 were calculated by Gullón (1950) applying the corrections indicated above. In addition, the sum of the values of the frequency for the northern (fn in Table 2) and southern (fs) hemisphere is not equal to the total frequency (ft) in some years (1906, 1914, 1935-1936, 1938-1940, 1943, 1946-1953). As in Table 1, this can be due to typos in the summary tables published by the observatory or the way to round the values of the frequencies by the observers. Also, there is a change in the way to present the values by the Madrid Astronomical Observatory since values corresponding to the observations for the period 1906-1949 were provided with two decimals and from 1950 with only one decimal.
One can see that the homogenized series of the Madrid prominence number (MPS) follows the 11-yr solar cycle (Fig. 2). The strongest solar cycle according to the MPS for the period 1906-1957 is Solar Cycle 15 (maximum in 1917) with a daily mean in the number of prominences around 10 and the weakest one is Solar Cycle 14 (maximum in 1908) with around 6. Note that the strongest cycle for that period according to the sunspot number was Solar Cycle 18 and the weakest one was Solar Cycle 14 (this in agreement with the Madrid prominence number series).
The temporal homogeneity of the MPS for the period 1906-1957 has been evaluated using the Standard Normal Homogeneity Test (Alexandersson, 1986). This test is useful to detect inhomogeneities in the series, which could appear because of, for example, changes in the instruments used to observe or the way to count, in this case, prominences. According to this test, the annual series of the total number of prominences and the total, north, and south frequencies are homogeneous at a 95% significant level.

Comparison with other indices of solar activity
We compared the annual MPS with other indicators of solar activity. First, we show the comparison between MPS and the prominence series computed by Carrasco & Vaquero (2022) from the observations made at the Astronomical Observatory of the University of Coimbra (CPS) for the period 1929-1944 (Fig. 3).
We note some differences in the solar cycle behavior between Madrid and Coimbra series. The number of prominences recorded in Coimbra is significantly larger than that in MPS. Moreover, the minimum of Solar Cycle 17 was one year later in MPS (1933) with respect to CPS (1932), the maximum of that cycle was one year before (1936 in MPS), and the minimum of Solar Cycle 18 could be two years later (it was in 1942 according to CPS, whereas it was in 1944 in MPS although we note that there is no data in MPS in 1942).
The best linear fit between MPS and CPS is ( Fig. 4): CPS = (1.6 ± 0.4) Á MPS + (0.9 ± 1.9), with r = 0.71 and p-value = 0.003. Furthermore, the correlation coefficient between the annual values of the sunspot number index and the CPS for the period 1929-1944 is r = 0.86 with p-value < 0.001, whereas it is r = 0.70 and p-value = 0.003 between the sunspot number and MPS regarding that same observation period of Coimbra.
We suggest that the differences between both series may be due to the way to count prominences in both observatories. The annual MPS only includes prominences with heights of 25 00 and more while all prominences were counted in CPS regardless of their height. Part of the discrepancy may be due to the fact that the annual calculations shown in Figure 3 for each prominence series were not made using common observation days in both observatories. In our future work, we will search for, digitize and analyze the daily observations of prominences of the Madrid Astronomical Observatory to try to understand these differences.
We also compared the annual MPS with other indicators of solar activity (Figs. 3 and 4). For example, the best linear fit between MPS and the annual sunspot number (version 2, Taking into account this low value and the fact that, due to repair works on the dome where the telescope was located, the observations of prominences were not evenly distributed throughout the year 1957 (Gullón and López Arroyo, 1958), the observations may not represent accurately solar activity in the aforementioned year.
We computed the asymmetry index following the definition used in recent works such as Veronig et al. (2021) and Aparicio et al. (2022b): D = P n À P s , where D is the absolute asymmetry, P n and P s is the number of the prominences observed in the northern and southern hemisphere according to the MPS. This index is represented in Figure 5.
One can see that the southern hemisphere was dominant in the declining phase of Solar Cycle 14 (À0.13) and in Solar A.J.P. Aparicio et al.: J. Space Weather Space Clim. 2023, 13, 5 Cycle 15 (À0.64), whereas the northern hemisphere dominated in Solar Cycles 16 (2.63), 17 (0.21), 18 (0.69), and the rising phase of 19 (2.20). Note that the previous values in parentheses are the sum of the annual values of the asymmetry index in each solar cycle. We note that Aparicio et al. (2022b) also concluded that the sunspot numbers in the northern hemisphere were dominant in Solar Cycles 17, 18, and 19. Furthermore, the maximum value of the northern hemisphere according to the daily mean of the prominence number was larger than that of the southern hemisphere in Solar Cycles 15-17, whereas it is the opposite in Solar Cycles 14, 18, and 19.

Conclusions
This work is the first step in the study of the prominence observations performed at the Madrid Astronomical Observatory. Here, we present a digitized version of a homogenized series of the total and hemispheric annual number of prominences with heights of 25 00 and more, which covers the period 1906-1957. The values were extracted from documentary sources in printed versions, i.e., the yearbooks and bulletins of the observatory and Gullón (1950), corrections were applied in some years of the period 1906-1938, and the homogeneity of the series was checked. This digitized version is available as Supplementary Material for this work.
We have made an analysis of this annual series explaining, in addition, the methodology and instruments used to carry out the prominence observations at the Madrid Astronomical Observatory. We have clarified that the methodology to count prominences in Madrid changed over time.
We have compared the MPS with sunspot number values provided by SILSO, the solar radio flux at 10.7 cm, the sunspot area series by Mandal et al. (2020), and the prominence series by Carrasco and Vaquero (2022) from the Astronomical Observatory of the University of Coimbra. The highest correlation was found between the prominence series of Madrid and Coimbra (r = 0.71). However, we found that the correlation coefficient between the MPS and the sunspot number is significantly lower than those between the sunspot number and the other solar activity time series. In particular, the correlation coefficient between Coimbra series (1929)(1930)(1931)(1932)(1933)(1934)(1935)(1936)(1937)(1938)(1939)(1940)(1941)(1942)(1943)(1944) and the sunspot number is r = 0.86 (it is r = 0.7 between Madrid and the sunspot number when the Coimbra study period is considered). These differences, as well as those detected in the time of maxima and minima between the different series, may be due to the way the prominences were counted in Madrid since only prominences with heights of 25 00 and more were considered. Our analysis of the asymmetry index from the MPS showed that the southern hemisphere was dominant in the last part of Solar Cycle 14 and Solar Cycle 15, and the northern hemisphere dominated from Solar Cycle 16 to the rising phase of 19, when MPS finished.
This work is a continuation of our efforts to analyze solar activity using observations carried out at the Madrid Astronomical Observatory. Studies reconstructing the total and hemispheric sunspot number series from Madrid data  and analyzing and providing machine-readable versions of different sunspot catalogs recorded at this observatory have already been published. In the future, more data from solar prominences and other solar features observed at the Madrid Astronomical Observatory shall be made available.