Mount Etna despite being an extremely active volcano which, during the last 400 years, has produced many lava flow flank eruptions has rarely threatened or damaged populated areas. The reconstruction of the temporal evolution of potentially hazardous flank eruptions represents a useful contribution to reducing the impact of future eruptions by and analyzing actions to be taken for protecting sensitive areas. In this work, we quantitatively reconstructed the evolution of the 1981 lava flow field of Mt Etna, which threatened the town of Randazzo. This reconstruction was used to evaluate the cumulated volume, the time averaged discharge rate trend and to estimate its maximum value. The analysis was conducted by comparing pre- and post-eruption topographic surfaces, extracted by processing historical photogrammetric data sets and by utilizing the eruption chronology to establish the lava flow front positions at different times. An unusually high discharge rate (for Etna) of 640 m3/s was obtained, which corresponds well with the very fast advance rate observed for the main lava flow. A comparison with other volcanoes, presenting high discharge rate, was proposed for finding a clue to unveil the 1981 Etna eruptive mechanism. A model was presented to explain the high discharge rate, which includes an additional contribution to the lava discharge caused by the interception of a shallow magma reservoir by a dike rising from depth and the subsequent emptying of the reservoir.
 The most destructive flank eruption of Mount Etna volcano occurred in 1669, when nine villages and a large part of the city of Catania were destroyed [Corsaro et al., 1996]. More recently, the eruptions of 1923, 1928, 1971, 1981 and 1991–92 have threatened inhabited areas, although only that of 1928 produced significant damages. The reconstruction of the temporal evolution of past lava fields represents a useful contribution for evaluating the hazard of future eruptions and allows better planning of mitigation measures aimed at reducing their impact on inhabited areas. The knowledge of the lava flow emplacement history gave us the opportunity to analyze the feasibility and the effectiveness of building a barrier system for controlling and delaying the lava flow advance toward sensitive areas, also considering different barrier configurations. This analysis can be performed by simulating the mitigation measures trough predictive numerical codes for lava flow propagation which provide potential inundation scenarios. This investigation has already been performed on the 2001 Etna lava flows that endangered the Rifugio Sapienza tourist facilities [Scifoni et al., 2010]. Basaltic lava flows are phenomena that usually leave enough time to prepare mitigation actions, as happened during the 1983, 1992 and 2001 Etna eruptions, even if in some cases it was not possible to significantly reduce the damages. However, in few cases eruptions can produce fast moving lava flows that do not allow any mitigation action and may cause major damages and fatalities. Major damages occurred happened during the 1923 [Ponte, 1923], the 1928 [Duncan et al., 1996] and the 1981 [Global Volcanism Program, 1981] eruptions when flow advance rates of up to several hundred of meters per hour were observed. The reasons for such different behavior is not yet fully investigated despite the fact that fast moving flows represent the major hazard linked to Etna's effusive activity. Consequently, understanding the characteristics of these eruptions would improve the capability of forecasting their occurrence and more effectively planning mitigation actions.
 The 1981 Etna eruption had a significant impact on population because it threatened the town of Randazzo (Figure 1), and the worry was that it might be an event similar to what happened in 1928 when lava flows destroyed the town of Mascali. During the first few hours of the 1981 eruption, the main lava flow cut the “Circumetnea” railway, the main road “S.S. 120,” the railway “FF.SS. Taormina – Randazzo” and the main road “S.P. Randazzo – Moio.” Due to lava expanding on an almost flat area many houses and farmland properties were destroyed. The front of the main flow reached, only 40 h after the beginning of the eruption, the bed of the Alcantara River situated about 10 km away from the vent.
 In this work we quantitatively reconstruct the evolution, in the time and space of the 1981 lava flow field. The final volumes were obtained by comparing the pre- and post-eruption Digital Elevation Models (DEMs) and orthophotos extracted using digital photogrammetry. Volumes of discharged lava were estimated at different times during the eruption from the final volume and by plotting the lava fronts on the orthophotos on the basis of the chronicles reported on scientific papers and newspapers. The obtained volumes allowed us to calculate the trend of the Time-Averaged Discharge Rate (TADR) and its maximum value, which represent the key parameters for hazard evaluation. TADR is obtained by measuring the volume emplaced over a known time interval [Harris et al., 2007]. The analysis here presented shows that the 1981 eruption was an extraordinary event characterized by a high discharge rate. The comparison between the 1981 Etna flow and lava flows from other volcanoes presenting similar discharge rate values supported the formulation of a model for the 1981 eruptive mechanism, as below described. The proposed conceptual model for understanding the mechanism of the 1981 Etna eruption represents an advance on the previous studies of this eruption because it merges information on the underground dike intrusion, derived by the previous geophysical analyses, and the new quantitative information on the discharge rate trend, highlighting the eruption peculiarity and proposing a new model for Etna eruptive mechanism.
 The main achievement of this work is the outline of a complex interaction between the dike that fed the eruption and a small shallow reservoir located inside the volcano edifice. Because the formation of shallow magmatic reservoirs is common in basaltic volcanoes, the mechanism proposed for the 1981 eruption can be applied to understanding similar eruptions which produce fast moving and dangerous lava flows. On basaltic volcanoes, this model could represent a contribution for discerning a simple dike-fed eruption from a complex multisource-fed one such as the 1981 eruption.
 After the end of the 1979 flank eruption, vigorous activity resumed at Etna in April 1980. The volcano was characterized by both discontinuous strombolian activity and lava fountain episodes at the South East Crater (SEC), as well as at Voragine (VOR) and Bocca Nuova (BN) summit craters (Figure 2). On 1st September, the activity shifted to the North East Crater (NEC) forming a strong lava fountain eruptive episode characterized by the emission of two lava flows 3–4 km-long. Early the next day the activity waned but another similar paroxysmal episode occurred at the same crater (NEC) on 6th September and was characterized by both violent explosive activity and lava overflow for about 10 h. Vigorous strombolian activity resumed again on 5th February at the same crater and lava poured out from its base forming three lobes that traveled about 2 km up to 7th February. Ash emissions from the VOR were noticed both at the end of the NEC activity and during the last days of the month, when the SEC also showed similar activity. During the first half of March, ash emission together with ejection of spatter were observed from the VOR and, sporadically, from the NEC [Villari, 1983].
 The 1981 flank eruption was preceded by a swarm of local earthquakes, which started on the morning of 16th March. Figure 2 shows an evolution map of the eruptive fissures, the evolution of the lava flows as well as the infrastructures and toponyms reported in the following. On 17th March at 1:37 P.M. (local time) a NS trending eruptive fissure (F1a), opened at 2595 m a.s.l. on the northern slope of the volcano. The F1a initial activity was characterized by lava fountaining, at the same time a sudden emission of ash from the BN was observed. From 1:37 P.M. to 5:21 P.M. another three eruptive fissures opened downslope, two (F1b and F1d) followed a NW trend, while one (F1c) showed a WNW strike [Villari, 1983]. Short-lived lava flows were emitted along the F1 fissures in correspondence to the four distinct segments (F1a, F1b, F1c, and F1d). At 6:55 P.M. a new eruptive fissure (F2a) opened to the East of Mt Spagnolo between 1800 and 1350 m a.s.l. From the lowermost section of this fissure, between 1400 and 1350 m a.s.l., a large lava flow (L2a) started and, between 8:00 P.M. and midnight of 17th March, it rapidly inundated the Circumetnea railway and the main road (S.S. 120), reaching an elevation of 730 m a.s.l. at about 5 km from the vent. By 18th March at 9:00 A.M. the main lava flow had advanced an additional 1 km, covering the railway Taormina – Randazzo and the main road (S.P. Randazzo – Moio). At 10:00 P.M. on the same day, additional eruptive fissures (F2b and F3) opened at lower altitudes (between 1350 and 1310 m a.s.l. and between 1227 and 1117 m a.s.l., respectively) producing lava flows (L2b and L3) that slowly advanced toward Randazzo. The L2a flow slowed reaching the Alcantara riverbed (600 m a.s.l.) at 11:00 A.M. on 19th March. During this day the L2a flow remained confined within the Alcantara riverbed and continued to be fed, thickening its frontal portion but it not further advanced. On midday of 20th March the lava outpouring from F2a stopped and the main lava flow L2a reached its final length of 10 km. At the same time, around noon, a small spatter cone built up above the F3 emission point as a consequence of a weak explosive activity that accompanied the slow advancing of the only active lava flow (L3). The eruptive activity from F3 fissure continued with variable intensity until the end of 23rd March when the lava front stopped at 926 m a.s.l.
3. Available Topographic Data
 Pre- and post-eruption surfaces were analyzed together with post–eruption orthophotos in order to precisely define the limits of the 1981 lava flows and evaluate their final volumes. The pre-eruption DEM was extracted by photogrammetric photos acquired during an aerial survey performed in 1978 at a flight altitude of 3000 m (scale 1:20.000). The data processing was carried out using a digital photogrammetric workstation which allows the generation of DEMs and orthophotos through a semi–automatic procedure [LH Systems LLC, 1999]. Stereo pairs obtained from 13 digitalized aerial photos configured in 4 different strips, formed a photogrammetric block that enabled the coverage of the whole area of interest (Table 1). Ground Control Points (GCPs) were identified and measured on recent high resolution DEMs and orthophotos in areas not covered by lava flows. GCPs were used for the orientation procedure in conjunction with a set of Photographic Control Points (PCPs) measured on a photogrammetric data set [Kraus, 1997]. Aerial photos collected during a photogrammetric survey performed in 2004 (Table 1) were processed to extract post–eruption topography. In order to obtain a complete coverage of the lava field below 2000 m a.s.l., it was necessary to integrate the 2004 DEM with a DEM obtained by interpolating a 1:10.000 contour map updated to 1999 [Coltelli et al., 2007].
Table 1. Characteristic of Photogrammetric Data
1978 (Pre-eruption Topography)
2004 (Post-eruption Topography)
Mean image scale
Ground pixel dimension (cm)
 In order to assess the accuracy of the 1978 and 2004 DEMs, terrestrial laser scanner and GPS surveys were carried out in 2008 and in 2009 on selected test areas. GPS control points, distributed both inside and outside the lava flow area were measured and post-processed using four GPS stations of the INGV (Istituto Nazionale di Geofisica e Vulcanologia) permanent network as reference points. The elevation of each GPS point was compared with that extracted at the same location from the pre and post-eruption DEMs. The comparisons produced an average difference of about 0.3 m for the post-eruption DEM and of about 0.9 m for the pre-eruption DEM.
 The laser scanner survey was performed on a portion of the L3 lava flow front to collect independent and sufficiently accurate data to verify the quality of the post-eruption photogrammetric DEM. The acquired data were geo-referenced using artificial targets positioned with GPS receivers. The height differences between the laser survey and the post eruption photogrammetric DEM are characterized by a mean and a standard deviation of 1.75 m and 0.52 m, respectively.
 Giving that the three comparative analyses, i.e., the two performed between pre- and post-eruption photogrammetric DEMs and GPS points, and the third between the post-eruption DEM and the laser scanner survey, gave similar values of mean and standard deviation, we can conclude that the pre- and post-eruption surfaces have similar accuracy. Therefore the differences observed between pre- and post-eruption DEMs on areas unaffected by lava flows (1.3 ± 1.0 m) are adopted to quantify the error associated with the thickness measurements used for estimating lava volumes as described in the following chapter.
4. Quantitative Reconstruction of the Lava Flow Volumes and TADR Estimation
 The vertical differences between the pre and post-eruption DEMs were considered only inside the flow limits, excluding the internal areas not covered by new lava, and enabled the preparation of the residual map, which shows the distribution of lava thicknesses (Figure 3). Unfortunately, the simple difference between the two surfaces does not provide a correct estimate of the volume because the DEM are not exactly corresponding to the situation soon before and after the event. The upper portion of the L2a flow destroyed mature and dense forests thus the height of the vegetation should be considered in the volume calculation. Similarly, the lower portion of the L2a was in part modified as a consequence of refurbishment of the area. Field measurements were performed for measuring, on the unchanged areas, the lava thickness to be assigned to the adjacent modified zones.
 In order to estimate the volumes (V) of the lava flows we compared the pre- and post-eruption DEM using the following expression:
where Δx = 5 m is the linear dimension of the square cells and Δzij is the height variation between the 1978 and the 2004 DEMs that represents the lava thickness. Since Δx is constant V can also be evaluated as:
where ni is the number of cells inside the lava flow limits while np is the number of cells covering the flow perimeter and = is the average lava thickness.
 The variance associated with the lava volume is calculated by applying the variance propagation law to (2):
where (σ) = 1 m is the vertical accuracy of the lava thickness evaluated as the standard deviation of the terrain residuals, that is the height variations between the two DEMs evaluated in unchanged areas while (σΔx) = 0.5 m is a horizontal accuracy evaluated as twice the orthophoto resolution. Since the horizontal error is only due to the drawing of the lava flow field limits, the cells inside the lava flow can be considered as having zero error thus ni · σΔx = 0, therefore the third and fourth terms of equation (3) are equal to zero and the standard deviation of the volume V can be obtained as follows:
This procedure for estimating the volume uncertainties does not take into account the systematic errors since the co-registration procedure described before should have reduced the biases between the two DEMs.
 This methodology was applied to evaluating the volume of both the lava and the pyroclastic products emplaced during the 1981 Etna eruption from each eruptive fissure (Table 2). The total volume is 22.75 × 106 m3 with a relative error of 24% estimated following the above reported procedure. The total volume of the lava emplaced divided by the total time of the eruption (142 h) allowed us to calculate an eruption rate of 44.5 m3/s.
Table 2. Opening Time, Name, Trend and Length of the 1981 Etna Eruptive Fissures: Name, Length, Area, Total Volume of Lava and Maximum Advance Rate of Lava Flows Emitted From the Different Fissuresa
Fissure Trend (deg)
Max. Advance Rate (m/s)
In the column for eruption deposits L, P1 and P2a indicate lava flows, pyroclastic fall deposit and spatter ramparts, respectively.
17th March 1981, from 1.37 P.M.
17th March 1981, from 6.55 P.M.
18th March 1981, from 10.00 P.M.
 The temporal evolution of the main (L2a) and the lowermost flows (L2b and L3) were reconstructed in order to estimate their discharge rate trends during the eruption. This reconstruction was based on the information on front position, recovered from the event chronologies in paragraph 2 and allowed to map the lava advancement (Figure 2). Sufficient data were not available for defining the lava flow limits at regular time intervals. The partial and cumulative volumes of L2a, L2b and L3 (Figure 4a, right axis) were reconstructed by splitting the final lava thickness through the flow field limits drown at each time step (Figure 2) and measuring the volume only inside the corresponding area. The temporal evolution of the eruption and the analysis of final lava thickness indicate that the flows were emplaced mostly as single units and that the super-imposed lava units can be considered negligible, except for the L2a lava fronts in the Alcantara riverbed. The TADRs were then obtained by dividing the partial volumes by the corresponding time spans (gray bars in Figure 4a, left axis). The volume of the spatter rampart (0.55 × 106m3) built up above F2a was included and proportionally distributed in the calculus of the main-flow partial volumes and then in the TADR.
 A continuous trend was then interpolated through to an exponential fit of the 1981 TADR values (Figure 4a). Such trend shows a rapid increase followed by a slow decline (waxing-waning behavior of Wadge ) typical of many basaltic eruptions and has a maximum value of 640 m3/s.
5. Analysis of the Results
 Several authors have previously estimated the total volume, the averaged effusion rate (corresponding to our TADR) on a portion of the eruption and the eruption rate of the 1981 Etna eruption. Romano and Vaccaro  reported a total volume of 30 × 106 m3 and the eruption rate of 58 m3/s. Guest et al. , who studied the flow field development of the 1981 eruption, reported a total volume of 20 × 106 m3 and an average effusion rate of 128 m3/s in the first 40 h. Del Negro et al.  roughly estimated the average effusion rate during the first 24 h reporting that it might have approached 300 m3/s, a large value for Etna eruptions.
 Despite the fact that our estimate of the 1981 total volume and of the eruption rate are not very different from the published data, through the analysis of the TADR trend it was possible to estimate a maximum value of 640 m3/s that was reached in a very short time which is quite unusual for known Etna eruptions. If we consider for comparison the 2001 eruption [Coltelli et al., 2007], it is clear that, even though the shape of the TADR trend of its main flow (Figure 4b) resembles very well that of the 1981 eruption, the maximum values are very different, reaching 32 m3/s and 640 m3/s, respectively. The high TADR values observed for the 1981 eruption requires an explanation, then we conducted a comparative analysis among a number of basaltic eruptions from Etna and other volcanoes, for which effusion rate values are available in literature.
 The eruption rate is available for 60 Etna flank eruptions occurring between 1607 and 2008 [Behncke et al., 2005; Branca and Del Carlo, 2004, 2005; Romano and Sturiale, 1982; Smethurst et al., 2009; Tanguy et al., 2007]. These values are plotted against the corresponding durations and volumes (Figure 5) and show that the 1981 eruption with a value of 44.5 m3/s, is located at the high end of the Etna's eruptions plot. However, the eruption rate cannot be considered as a significant parameter for assessing the intensity of an eruption because it does not distinguish events characterized by high rates and short durations. A second comparison has been made by considering a reduced data set including only those events for which the maximum effusion rate is known (Table 3 and Figure 6). As shown in Figure 6 the 1981 eruption was very different from other Etna flank eruptions. A similar comparison has been performed on a set of well-documented basaltic eruptions from other volcanoes (Kilauea, Mauna Loa, Nyiragongo and Piton de la Fournaise) that had high effusion rates (Table 4 and Figure 6). The 1981 eruption appears to be more similar to these eruptions than to those of Etna therefore their eruptive mechanisms associated are discussed below with the aim of finding a clue to explaining the 1981 Etna eruption.
Table 3. Volume, Duration and Maximum Effusion Rate for Etna Eruptions
 Kilauea and Mauna Loa eruptions showed high effusion rates which are associated with strong lava fountaining [Rowland and Walker, 1990]. On Kilauea, the Mauna Ulu lava fountains were typically hundreds of meters high up to 540 m [Rowland and Walker, 1990] while those produced during four short Puu Oo episodes in 1984 and in 1985 were between 352 and 441 m high [Heliker and Mattox, 2003]. On Mauna Loa, huge effusion rates were related to the development of very long curtains of fire associated with broad lava fountaining tens of meters high [Rowland and Walker, 1990].
 The Nyiragongo eruptions of 1977 and 2002 exhibited very fast advancing lava flows resulting from the emptying of a lava lake located in the large summit crater of the volcano. In the 1977 eruption an extremely fluid, fast-moving (up to 60 km/h) lava flow drained the summit lava lake and covered several villages in a very short time [Tazieff, 1977]. During the 2002 eruption, several fissures opened on the S and NW flanks of the volcano, the upper fissures drained the summit lava lake whereas the lower fissures were supplied from a dike rising directly from the shallow plumbing system, forming lava flows that destroyed part of the city of Goma [Allard et al., 2002]. The maximum effusion rates (Table 4) for the 1977 and upper fissures of 2002 eruptions can be estimated by the lava volumes reported in literature (21 × 106 m3 [Tazieff, 1977] and 14 × 106 m3 [Favalli et al., 2006], respectively),even if the duration of the lava effusion is more uncertain due to the scarce available information. Tazieff  reports less than 1 h for the 1977 eruption, whereas the upper fissures of the 2002 eruption were active for at least 2 h [Allard et al., 2002]. These combined volumes and durations provide maximum effusion rates of about 5833 and 1944 m3/s for the 1977 and 2002 eruptions, respectively. As these values are among the highest ever observed for a lava flow and because of uncertainties in volume and duration, they can only be considered a rough estimate.
 An huge effusion rate was also observed during the April 2007 eruption of Piton de la Fournaise, when the Dolomieu crater at the summit collapsed [Coppola et al., 2009]. This eruption started with a dike propagating from the shallow plumbing system below the Dolomieu crater toward the lower flank, where lava flows were discharged at about 65 m3/s. The fast magma outpouring caused a sudden decrease in pressure inside the shallow reservoir that induced the collapse of the summit roof. The rock column collapsed into the reservoir and acted as a piston, increasing its internal pressure and causing the rapid drainage [Peltier et al., 2009]. The maximum effusion rate during this paroxysmal phase has been estimated to be more then 200 m3/s [Staudacher et al., 2009] which is quite a large value for the effusive eruption of this volcano that generally shows low effusion rates (from < 2 m3/s) for summit eruptions and up to about 20 m3/s for the initial phases of flank eruptions [Coppola et al., 2009].
 During the 1981 Etna eruption lava fountaining from F1 fissure, 100–200 m high [Global Volcanism Program, 1981], formed short-lived lava flows that accounted for only 3% of the total volume and had an average discharge rate of 90 m3/s. The main fissure, F2, presented only minor explosive activity. A curtain of lava, indicated by the large spatter rampart along this fissure, was formed by low (meters to about ten meters high) lava fountains, as proved by the lack of pyroclastic deposits that conversely are present on the eastern side of F1,a,b,d fissure, while the most voluminous lava flow occurred (87% of the total volume associated with the maximum discharge rate of 640 m3/s). Finally, F3 fissure produced mild strombolian activity and a minor lava flow (6% of the total volume at an average discharge rate of 3 m3/s). Therefore the magnitude of the maximum discharge rate of the 1981 Etna eruption cannot be associated with an eruptive mechanism analogous to those of the Kilauea and Mauna Loa eruptions, which are characterized by strong lava fountains of gas-rich basaltic magma that generate lava flows. On the contrary, a mechanism similar to that of Nyiragongo and Piton de la Fournaise eruptions, where the extraordinary effusion rates are clearly associated with sudden pressure changes in shallow plumbing systems or superficial reservoirs, can be inferred to explain the evolution of the 1981 eruption.
6. Discussion on the Eruptive Mechanism
 Given the peculiarity of the 1981 eruption, the analysis presented in this work on the discharge rate trend provides useful insights for investigating its eruptive mechanism. A variety of hypotheses for the mechanism of this eruption have been proposed in literature. Sanderson et al.  modeled the dike feeding the eruption using precise leveling and gravity data showing that, between August and September 1980, the magma rose from depth, filling a SSE-NNW trending fissure zone. In March 1981 a long eruptive fissure opened on the NNW flank due to a deeper radial intrusion of magma, as testified by the gravity change measured. Scott , on geochemical and petrographic basis, suggested that the 1981 eruption radially drained a hybrid magma that was the result of mixing residual 1979 magma with fresh magma during the dike filling from September 1980. Bonaccorso  re-analyzed the leveling data set of Sanderson et al.  and compared it with EDM data acquired between October 1979 and May 1982. He proposed a double-source model consisting of two tensile dislocations; the first associated with deeper magma injected at depth and the second related to magma ascent in the summit area that activated the eruptive fissure. Carbone et al.  applied a parametric inversion analysis of previous data (microgravity, leveling and EDM) and showed that when a two-tensile crack model is used [Bonaccorso, 1999] the observed vertical and horizontal ground deformations are underestimated. Therefore they suggested that the ascending magma filled a network of pre-existing interconnected fractures, allowing mass redistribution with no evident deformation. They also suggested that the significant effusion rates could be related to a low viscosity magma.
 The previously described models do not take into account the cause of both the peculiarly rapid evolution of the eruption and the exceptional value of its maximum discharge rate, except for the hint on low viscosity magma given by Carbone et al. . Indeed they envisaged a complex intrusive mechanism that drove the eruption, suggesting that up to two magma sources may have fed the eruption.
 As already outlined, the TADR trend resembles the general shape of a typical Etna eruption, except for the maximum value (640 m3/s). Such a value cannot be linked to the typical processes involved in some strong basaltic lava flow eruptions, i.e., vigorous magma injection of a gas-rich basaltic magma into an over pressurized dike [Rowland and Walker, 1990] or low viscosity magma. The former mechanism should be excluded because high lava fountaining was not observed during the main phase of the eruption. The latter mechanism is unlikely because the chemical composition of the 1981 lava is a hawaiite [Scott, 1983] similarly to other Etna eruptions [Romano and Vaccaro, 1986] that did not have high discharge rate. We suggest that, the magnitude of the maximum discharge rate can be ascribed to rapid drainage of a shallow or superficial reservoir and we identify four different phases below described (Figure 7) to explain the phenomena observed during the 1981 eruption.
 Phase 1. A long pre-eruptive phase was characterized by the ascent of magma inside the shallow plumbing system, as indicated by the strombolian activity and periodic ash emissions observed at the summit craters since the spring of 1980. In particular, three paroxysmal explosive episodes occurred from the NEC on 1st and 6th September 1980 and 5th February 1981 [Sanderson et al., 1983]. These explosive events were accompanied by abundant overflows of lava that represent a clear evidence of new magma rising from depth. We suggest that this magma filled the shallow part of the plumbing system, building a small magma reservoir inside the volcanic pile. The complete filling this reservoir caused the uprush of magma and a series of short-lived summit lava flow eruptions.
 Phase 2. The 1981 eruption started with the opening of fissures F1a, F1b, F1c and F1d, at progressively lower altitude, accompanied by lava fountains about 200 m high [Global Volcanism Program, 1981], as evidenced by tephra deposits dispersed to the west of F1a, F1b and F1d. Short-lived lava flows were also emitted. The volume of these lava flows and of the pyroclastic deposits is estimated to be 1.62 × 106 m3 (only 7% of the total volume) corresponding to a TADR of about 90 m3/s. The opening of these fissures can be related to a dike intrusion from the deeper part of the plumbing system located around 3 km below the sea level [Chiarabba et al., 2004] and the consequent arrival of a gas-rich magma that was responsible for the explosive activity. To estimate the magma driving pressure (Γ) at the F1 fractures we used the cubic law [Santini et al., 2011] i.e., the equation used to calculate fluid flow rate (Q) between two parallel plates:
where L and w are fissure length and opening, respectively and η is the viscosity. We consider the estimated TADR of F1 = 90 m3/s, η = 135 Pa x s (evaluated as below described), a fissure width w1 = ∼1 m [Bonaccorso, 1999; Santini et al., 2011] and a length L1 = 734 m which is the sum of the F1a,b,d lengths. These fissures represents the intersection with the surface of the deep-seated dike because they erupted as lava fountaining, while the F1c was not considered because it has a different orientation and emitted only lava flow, therefore it represents a local and very shallow propagation of the F1b fissure. The lava viscosity η = 135 Pa x s was evaluated by considering the viscosity law of a hydrous Etna basalt [Giordano and Dingwell, 2003], a water content of 0.3 wt. % [Scott, 1983] and an temperature T = 1100°C, which represents a plausible value of pre-eruption temperature of a magma flowing into the dike since most of measured eruptive temperature of Etna lava flows range between 1080 and 1095°C [Pompilio et al., 1998]. Using these values, and the cubic law equation, we evaluate a driving pressure Γ1 = ∼0.15 MPa/km, necessary for generating the observed TADR of 90 m3/s.
 Phase 3. The main phase of the eruption began with the opening of F2 fissure, from which a large lava flow was emitted. A Spatter-rampart forming explosive activity was also observed, however the explosivity index (E), which is the percentage of the total volume that is pyroclastic material, is low (E = 2.8) during this phase with respect to Phase 2 (E = 33.3). This observation suggests that F2a was, at least partially, fed by a gas-depleted magma that had resided for a certain period in an open reservoir into the shallow portion of the plumbing system. This is in agreement with petrological analyses [Scott, 1983]. Therefore we suggest that during Phase 3 the previously intruded dike (Phase 2) have expanded the surrounding wall rocks, intercepting and draining the shallow magma reservoir previously re-filled (Phase 1).
 The cubic law equation was used to estimate the driving pressure Γ2 = ∼0.44 MPa/km necessary for generating from the F2a fissure a discharge rate as high as 640 m3/s, considering a length L2 = ∼1760 m and width w2 = ∼1 m [Bonaccorso, 1999; Santini et al., 2011]. Since Γ1 is lower than Γ2 the dike intrusion from the deeper part of the plumbing system alone was not sufficient for explaining the F2 maximum discharge rate. The discrepancy between the two values of the driving pressures can be seen as the key factor to explain the eruptive mechanism of the 1981 Etna eruption. An overpressure in the magma reservoir of about 4.41 MPa was determined using the relationship described by Wadge :
where ρm (magma density) = 2650 kg/m3, ρv (volcanic pile density) = 2400 kg/m3 [Wadge, 1977] and h is the depth of the magma reservoir that we hypothesized to be at 1800 m below the summit, in consideration of the elevation of the main vent. This overpressure produces, over a distance of about 9 km, measured between the reservoir and F2, a driving pressure of 0.49 MPa/km. By inserting this value in the cubic law and the dimension of F2 a discharge rate of 710 m3/s that has the same order of magnitude of the measured value. The inflow of magma from the reservoir into the dike led to a sudden pressure rising which triggered a horizontal propagation of the dike. This resulted in the opening of a shallow NNW crack along which the F2a and then F2b fissures emerged at lower elevation than F1. Such a rising pressure in the shallow dike caused the high discharge rate of 640 m3/s from F2. Thus the significant volume of erupted lava and of the spatter rampart (19.83 × 106 m3), which account for the 87% of the total volume, can be associated to the described draining process.
 Phase 4. The dike continued to propagate and the opening of another fissure (F3) produced a rapid pressure fall in the plumbing system. This caused, for a feedback mechanism, the collapse of the summit craters as evidenced by continuous and strong ash emission observed during the eruption [Villari, 1983; Global Volcanism Program, 1981]. During this final phase (from 20th March to the end of the eruption) the F3 fissure produced effusive and low explosive activities. Lava flowing toward the village of Randazzo progressively slowed and stopped after 3 days. The low volume of this lava flow and the very low TADR (3 m3/s) suggest that F3 vent was fed during the waxing phase of the dike and was no longer supplied by the shallow magma reservoir. In fact, a weak explosive activity (hornito-forming spatter ejection) occurred at the lower vent evidencing that the dike emptying was driven only by residual magma degassing.
7. Concluding Remarks
 In this work we analyzed the temporal evolution of the 1981 Etna eruption after having evaluated the cumulated volumes and the TADR trend throughout the whole eruption. Final volume, eruption rate, as well as maximum discharge rate, were compared with the same quantities recorded for past Etna eruptions. The analysis showed that the 1981 eruption is located at the high end of the data set formed by the eruptions of the last 400 years if we compare the values of total lava volumes and eruption rates. On the other hand, if we compare the maximum effusion rates (available only for the most recent eruptions) the 1981 event is dissimilar from the others because it had a very high value (640 m3/s) which has never been recorded in recent Etna activity. The analysis was extended to other volcanoes presenting comparable high values, i.e., Kilauea, Mauna Loa, Nyiragongo and Piton de la Fournaise. The eruptive mechanisms of these volcanoes are different from those usually associated with Etna flank eruptions and thus provided useful insights for the interpretation of the behavior of the 1981 eruption. More specifically, the analysis of some recent eruptions of these volcanoes suggested that the rapid evolution of the 1981 main flow and its huge discharge rate can be related to a sudden emptying of a shallow magma reservoir interacting with a dike intrusion from the deeper part of the plumbing system.
 A four-step model was proposed to reconstruct the phenomena observed during the eruption. In the first step a shallow magma reservoir was emplaced into the volcano edifice, then at the eruption onset a dike from depth fed a strong explosive activity at the highest fissures (F1) that vanished when more fissures (F2) opened at lower elevation marking the beginning of the third step. The change of the eruptive style toward a very high lava discharge was due to the intercepting and draining of the shallow magma reservoir. This mechanism is supported by an estimation of the driving pressure from the reservoir emptying which is sufficient to producing a discharge rate higher than the 640 m3/s evaluated in this work.
 The complex interaction between an eruption feeder dike and a small shallow reservoir, that often built up within basaltic volcanic edifices, proposed for the 1981 eruption can be applied to understand similar lava flow eruptions which produce fast moving and dangerous lava flows. Finally, the reconstruction here presented, improves the knowledge of the 1981 eruption characterized by very fast moving lava flows, which are uncommon within Etna eruptions despite they represent a serious concern for the safety of potentially inundated areas. Future hazard evaluations should take into account that an eruption characterized by a rapid evolution can be foreseen if, thanks to improvements in monitoring systems, signs of the formation of shallow magma reservoirs inside the volcanic edifice are detected.
 We are grateful to Michele Dragoni and Paolo Baldi, who provided helpful comments and suggestions. This work was partially supported by the Project V3-LAVA 2007–2009 INGV-DPC agreement.