We present a 30 year long data set of satellite-derived time-averaged lava discharge rates (TADR) for Mount Etna volcano (Sicily, Italy), spanning 1980–2010 and comprising 1792 measurements during 23 eruptions. We use this to classify eruptions on the basis of magnitude and intensity, as well as the shape of the TADR time series which characterizes each effusive event. We find that while 1983–1993 was characterized by less frequent but longer-duration effusive eruptions at lower TADRs, 2000–2010 was characterized by more frequent eruptions of shorter duration and higher TADRs. However, roughly the same lava volume was erupted during both of these 11 year long periods, so that the volumetric output was linear over the entire 30 year period, increasing at a rate of 0.8 m3 s−1 between 1980 and 2010. The cumulative volume record can be extended back in time using data available in the literature. This allows us to assess Etna's output history over 5 centuries and to place the current trend in historical context. We find that output has been stable at this rate since 1971. At this time, the output rate changed from a low discharge rate phase, which had characterized the period 1759 to 1970, to a high discharge rate phase. This new phase had the same output rate as the high discharge rate phase that characterized the period 1610–1669. The 1610–1669 phase ended with the most voluminous eruption of historic times.
 For Mt. Etna (Sicily, Italy), Wadge and Guest  showed that, between 1971 and 1981, although being erupted as part of 37 discrete effusive eruptions, output was marked by a linear decadal-scale trend, with lava being erupted at an average rate of 0.7 m3 s−1. This trend also characterized the 1980s and 1990s, an average output rate of around 1 m3 s−1 being apparent in the erupted volume data for 1975–1995 given by Allard . However, a change in eruption style was witnessed during the flank eruption of July-August 2001. This event was marked by an anomalous degree of explosivity, as well as by an eccentric eruption of a fast rising, volatile-rich batch of amphibole-bearing magma which contrasted with activity styles witnessed during the preceding 30 years [Calvari and the Whole Scientific Staff of INGV-Sezione di Catania, 2001; Behncke and Neri, 2003a; Corsaro et al., 2007]. This 2001 eruption was followed by a further eccentric eruption in 2002, during which explosivity was again remarkably high [Spampinato et al., 2008]. This has led some to suggest that the supply and eruption dynamics at Etna were changing, with the 2001 event initiating this change [Allard et al., 2006; Behncke and Neri, 2003a; Clocchiatti et al., 2004].
 Such a hypothesis can be tested using long-term discharge rate records. For Etna, such a record exists, and is represented by the archive of thermal data available from the satellite-borne sensor, the Advanced Very High Resolution Radiometer (AVHRR). This data set now spans 30 years, and dates back to the launch of the first AVHRR sensor on 13 October 1978 [Cracknell, 1997]. These data allow extraction of time-averaged lava discharge rates following the methodology of Harris et al. . Such measurements can be made at a minimum temporal resolution of 12 h, the return period of the NOAA satellites on which AVHRR is flown. For the period 1980 to 2010, we have identified 1792 cloud-free images spanning 23 eruptions at Etna, yielding a data set capable of testing the hypothesis and detailing any variation in the rate of output over time.
2. Method and Data
Harris et al.  presented a method for obtaining time-averaged lava discharge rate (TADR) from satellite thermal infrared data. This was initially tested using AVHRR data collected during Etna's 1991–1993 eruption. The method reduces to the application of an empirical relation that allows TADR to be obtained from a measure of active lava area [Wright et al., 2001]. The derived discharge rate then relates to the time-averaged flux required to spread the lava over a given area. The method uses one band of data collected in the thermal infrared, typically between 10 μm and 12 μm, and applies a simple mixture model to the pixel-integrated spectral radiance to extract lava area (RTIR). To do this, the pixel is assumed to contain a mixture of lava-free land and active lava so that, if the temperatures of the two components are assumed, the pixel portion occupied by the two components can be calculated using the pixel-integrated radiance, which is written:
where Rlava and Rland are the spectral radiances for the lava and lava-free land surfaces, and plava and pland are the pixel portions occupied by these two components. Note that plava and pland must sum to one, so that pland can also be written 1 – plava. Now, if we can assume values for Rlava and Rland, we can calculate the pixel portion occupied by active lava from:
 Following Harris et al. , the radiance of the lava-free land can be obtained from surrounding, active lava-free pixels, and the lava surface radiance can be assumed over a range of radiances (obtained from the equivalent temperature range). The result yields a range of pixel portions occupied by the lava, ranging from a maximum value obtained with the maximum assumed lava temperature, to a minimum value obtained with the minimum assumed lava temperature. Multiplying the pixel portions by pixel area, and summing for all hot pixels containing active lava, gives the active lava area.
in which x is the coefficient that defines a positive, linear relation between TADR and A. Two relations are required, one for the maximum estimated flow area (Amax) and one for the minimum estimated flow area (Amin). This yields a range of TADRs, between which the actual value usually lies. The relation is best set through best fitting through calibration using ground-measured values [Harris et al., 2010]. For effusive eruptions at Etna between 1980 and 1999, the relation labeled “old” in Table 1 provided good fits with field measured values when used with Amax obtained assuming a typical lava surface temperature of 100°C and Amin assuming 500°C [Harris et al., 1997, 2000, 2007]. However, best fitting using TADRs derived from LIDAR data collected simultaneously with satellite passes during Etna's 2006 eruption yielded a different best fit relation for this eruption; this relation is labeled “new” in Table 1. For this new case Amax obtained assuming a typical lava surface temperature of 100°C and Amin assuming 600°C provided the best fit with the LIDAR data [Harris et al., 2010]. For each eruption we thus apply a range of conversion cases, and use the case which provides the best fit between satellite-data-derived and field-measured TADR. Based on best fitting using ground-based data acquired during Etna's 1999 eruptions at the SE Crater and Bocca Nuova, we also applied a modified version of the old relation, as given in Table 1. This “modified old” relation uses with Amax obtained assuming a typical lava surface temperature of 100°C and Amin assuming 1000°C [Harris and Neri, 2002].
Table 1. New, Old, and Modified Old Relations Used to Convert Active Lava Area to TADR at Etna Using AVHRR and MODIS Dataa
Temperature Condition (°C)
The second column gives the temperature condition used to apply the mixture model of equation (2) and hence to supply the required area value for each relation.
TADRmin (m3 s−1) = 2.5 × 10−6 (m s−1) Amax (m2)
TADRmax (m3 s−1) = 43 × 10−6 (m s−1) Amin (m2)
TADRmin (m3 s−1) = 5.5 × 10−6 (m s−1) Amax (m2)
TADRmax (m3 s−1) = 150 × 10−6 (m s−1) Amin (m2)
TADRmin (m3 s−1) = 1.7 × 10−6 (m s−1) Amax (m2)
TADRmax (m3 s−1) = 162 × 10−6 (m s−1) Amin (m2)
 We apply these conversions to (1) generate TADR time series for each eruption for which satellite data are available during the period 1980–2010 and (2) assess, through comparison with independent field measurements of TADR, which best fit relation applies to each eruption (as done in Figures 1–3). We use these data to volumetrically characterize each eruption in terms of the TADR trend, as well as the maximum and mean TADRs that feed each eruption. By integrating the TADRs for each eruption (using trapezium rule) we also provide an estimate for the volume erupted during each eruption, as well as the mean output rate (MOR) for each eruption, this being the total volume erupted divided by the eruption duration [Harris et al., 2007]. We can thus use our data set to define each effusive eruption in terms of intensity (maximum TADR) and magnitude (total volume erupted).
 To apply the methodology described above we used all AVHRR data obtained by the University of Dundee (Dundee, Scotland) receiving station between 1980 and 2010. Following Harris et al. , we used the band 4 (10.3–11.3 μm) pixel-integrated temperatures, where all data were calibrated and corrected for nonlinearity effects [e.g., Brown et al., 1993; Weinreb et al., 1990; Kidwell, 1995] by the NERC based at the Plymouth Marine Laboratory (Plymouth, UK). Data were then corrected for atmospheric and emissivity effects, and pixel areas were calculated as a function of the sensor instantaneous field of view and scan angle. Data between 10 and 12 microns need to be corrected for atmospheric upwelling radiance (Rup), atmospheric transmissivity (τ) and surface emissivity following (ɛ):
Rsat being the at-satellite radiance. We estimated Rup and τ as a function of scan angle using the MODTRAN atmospheric code, where typical values for a surface at an altitude of 0 m (at nadir and using a U.S. Standard atmosphere), averaged between 10 and 12 μm, are 1.79 × 10−5 W cm−2 sr−1μm−1 and 0.96, respectively. Emissivity averaged over the same wave band is also 0.96 (obtained from reflectance spectra for a sample of Etna 'a'a).
 For the period 2000–2010, we also used data from band 31 (10.78–11.28 μm) of the Moderate Resolution Imaging Spectroradiometer (MODIS) [Salomonson et al., 1989; Barnes et al., 1998]. This sensor was first launched in 1999, and provides data with a similar spatial and temporal resolution to the AVHRR, and at a similar wavelength. These data were processed in the same way. Here, the full database is, for the first time, collated in one place (see auxiliary material). While the pre-1993 portion of this database was processed by Harris , all post-1999 data were processed by Hawaii Institute of Geophysics and Planetology staff during various eruption crises and transmitted to Istituto Nazionale di Geosfisica e Vulcanologia-Sezione di Catania, as well as Centro Nazionale Terremoti (Rome), during those crises [Harris et al., 2007]. All data have been reprocessed here to check for correct application of TADR conversion coefficients (see auxiliary material).
 The main effusive events that occurred at Etna between 1980 and 2010 are listed in Table 2, along with their location, duration and number of available AVHRR and MODIS images. We see that there were around 36 effusive events, of which 25 lasted more than 1 day. During the period, AVHRR and MODIS data were available for 23 events, the remaining events being too short to capture, given the 12 h separation of AVHRR and MODIS passes, or because of cloud cover. For these 23 events, 1792 images were available of which 509 were from MODIS. We see that a total of 16 events comprise data sets of more than 10 images, allowing TADR time series to be generated for all of the main effusive events, with the exception of those occurring during 1997–1998, for which we have no satellite data (see Table 2). We also collated all available lava flow volume data for the period 1971–2010, as given in Appendix A. This allowed us to fill all of the gaps in the satellite-derived record.
Table 2. Main Effusive Eruptions at Etna, 1980–2010a
 The satellite-derived TADR time series for each effusive eruption are given in Figures 1–3. The full database on which these plots and following treatments are based are given, for each major effusive eruption listed in Table 3, as data files in the auxiliary material.
Table 3. Volumetric Characteristics for the Main (Greater Than 2 Days Long) Effusive Events Spanning 1983 Through 2010a
3.1. Classification of Eruptions on the Basis of Satellite-Derived TADR Trends
 Using our satellite-derived TADR time series we define four types of effusive event on the basis of characteristic trends in the TADR time evolution during each eruption. Type I and II trends were defined by Harris et al.  using the 1980–1996 portion of this data set and involve the following trends:
 1. Type I trends are characterized by a rapid waxing phase, followed by a longer waning phase, and are consistent with tapping of a pressurized volume [Wadge, 1981]. In Type I eruptions ∼50% of the volume is typically erupted in the first ∼25% of the total eruption time. Five eruptions show Type I trends, these being the eruptions of 1983, 1986–1987, 1991–1993, 2002–2003 and 2008–2009 (Figure 4a).
 2. Type II trends are characterized by relatively stable and low TADRs that span 0.1 and 5 m3 s−1, with ∼50% of the volume erupted in ∼50% of the total eruption time. Type II trends tend to be associated with summit eruptions. They are thus consistent with nonpressurized overflow from the summit craters. Four eruptions show Type II trends, these being the eruptions of 1984, 1985, 1996 and 1999's SE Crater eruption (Figure 4b).
 In new data processed for the post-2000 period we find two other trends, Types III and IV. Of these two trend types, the Type III trend has been described, and explained, by Steffke et al.  through comparison of satellite-derived TADRs and ground-measured SO2 fluxes for the 2004–2005 and 2006 eruptions; the Type IV trend has been examined by Harris and Neri , whereby:
 3. Type III trends are characterized by TADRs that increase with time and are associated with three eruptions, those of the SE Crater in 2001, plus those of 2004–2005 and August-December 2006 (Figure 4c). This trend likely results from the ascent of a magma batch which pushes a volume of degassed magma ahead of it, and is consistent with an imbalance observed between TADRs and SO2 emission rates during the opening phase of the 2004–2005 eruption [Steffke et al., 2011]. Eruption of the degassed volume contributes to the opening, low TADR, phase; arrival of the ascending batch contributes to the high TADR terminating phase.
 4. Type IV trends are extremely pulse-like. Peaks in such trends are of high amplitude and attain values of 10s to 100s of meters cubed per second. Such trends likely relate to the rise and eruption of a series of rapidly ascending magma batches. The first eruption of this type was the October-November 1999 eruption of the Bocca Nuova. This event was marked by 11 discrete paroxysmal events during which intense strombolian and lava fountain activity fed vigorous channelized 'a'a lava flows at TADRs of up to 120 m3 s−1 [Harris and Neri, 2002]. Two other eruptions show this Type IV trend, those being the July-August south flank eruption of 2001, plus the July 2006 eruption (Figure 4d).
3.2. Time Distribution of Effusive Event Types
 If we examine the distribution of events by Type through time we find that all of the large (>107 m3) volume eruptions prior to 1999 involved eruptions with Type I or II trends (Table 3), as already described by Harris et al. . Note, though, that we reclassify Etna's February-November 1999 eruption. This eruption was originally classified as Type I by Harris et al. . This assignment was made on the basis of a limited data set. However, processing of the full AVHRR data set available for this eruption reveal a Type II, slowly oscillating, trend as also found in the field-derived TADRs of Calvari et al. . In contrast, the Bocca Nuova eruption that occurred in the same year, during October and November, had a strongly pulsating character, as detailed by Harris and Neri , meaning that we here classify this eruption as Type IV. We also note that the opening, high TADR, phase of the 1983 eruption was missing from the AVHRR-derived data set due to cloud cover. The opening phase was, however, recorded by Frazzetta and Romano  whose field data clearly show this eruption to be of Type I character. Thus, Frazzetta and Romano's  field-derived TADR for this opening phase have been added to the plot in Figure 4 to justify its classification as Type I.
 Type III and IV trends dominate from 1999 onward (Table 3). The first event in this series, the January-July 2001 eruption of the SE Crater, shows a series of oscillations which, as detailed by Lautze et al. , increase in amplitude toward the July-August 2001 flank eruption. These oscillations lend a Type III character to the January-July 2001 eruption. Such Type III behavior is then also apparent during the July 2006 eruption of the SE Crater. Of the remaining post-2001 eruptions, both the eruptions of 2004–2005 and August-December 2006 show TADRs that build toward the end of the eruption, giving them also a Type III trend. The remaining two eruptions of this period are the 2002–2003 and 2008–2009 eruptions, which both have Type I trends.
 It is worth stressing that, although Type II and IV trends may look similar, the two classes are distinguished by the range of variability during pulses that interrupt the otherwise flat trends. For Type II trends the pulses generally have ranges of between ∼1 and 7 m3 s−1, whereas pulses in Type IV trends have ranges of up to 100 m3 s−1. As noted above, the TADR levels are also quite different between Type II and IV events.
 Finally, we have classified the eruptions based on the dominant or primary trend. Several eruptions have secondary trends overprinted on the primary trends. For example, while all Type III trends appear to have Type IV trends overprinted, the Type IV trends of the October-November 1999 and July-August 2001 eruptions also show a secondary declining trend with time, suggestive of a Type I influence. Note too, that the 2002–2003 eruption has been classified as Type I. This is mainly due to the contribution of high TADR flows during the opening phase on the northeast flank. Once these flows shut down, Type II trends are apparent for the following south flank emission.
 Calibration factors that need to be used to provide the best fit to independent ground-based TADR estimates, as given in Table 1, produce a grouped data set, with each grouping requiring a different calibration factor to achieve best fit. Each grouping also has distinct and characteristic maximum recorded TADR (Table 3), as well as eruption frequencies, durations, volumes and MORs (Table 4). This allows us to divide our data into two 11 year long periods on the basis of the volumetric character of the eruptions occurring within each period. These two periods are separated by a third period of low volumetric output, with the two high-volume periods spanning 1983–1993 (Period 1) and 2000–2010 (Period 3).
Table 4. Volumetric Characteristics for the Two Main Periods of Effusive Events Identified Between 1980 and 2010a
Volume (×106 m)
Mean Output Rate (m3 s−1)
Period 3b is the same as period 3 but without the two long-duration SE Crater eruptions that began and ended the period. Volumes are obtained from integrating the midpoint TADR value though time. For each period, mean output rate has been calculated by dividing the total volume emplaced by all eruptions during each period by the total duration of all activity during the same period. This is given on the “total” line below each period. Also given is the mean MOR for individual eruptions within each period (given on the “mean” line below each period).
Total (number = 5)
Total (number = 7)
Total (number = 5)
3.3.1. Period 1: 1983–1993
 The first period was marked by four eruptions each involving the output of 107 m3 of lava or more. Durations ranged between 120 and 471 days (mean = 204 days) to emplace lava volumes of between 18 and 183 × 106 m3 (mean = 80 × 106 m3) at mean output rates of 1.7 to 9.1 m3 s−1, with a mean of 4.6 m3 s−1 (Period 1, Table 4). Maximum TADRs were between 4 and 35 m3 s−1, with an average of between 2 and 12 m3 s−1 (Table 4).
 Best fits with field data, revealed that the “old relation” applied to this series of eruptions, as already detailed by Harris et al. . For example, applying the “old relation” to the 1991–1993 data set and integrating the TADRs through time yields an erupted volume of 183 ± 72 × 106 m3, compared with a volume measured using Electronic Distance Measurement by Stevens et al.  of 185 ± 22 × 106 m3.
3.3.2. Period 2: 1994–1999
 Following the eruption of 1991–1993, the most voluminous effusive event of the last 3 centuries [Calvari et al., 1994], there was a three year hiatus in effusive activity until 1996 (Appendix A). This period was followed by dominantly intracrater summit activity [Neri et al., 2005; Allard et al., 2006], and then by the two phases of the 1999 eruption (from SE Crater and Bocca Nuova [Calvari et al., 2003; Harris and Neri, 2002]). This period involved eight effusive events and was volumetrically transitional between Periods 1 and 3, with the 1999 SE Crater and Bocca Nuova eruptions requiring application of the “modified old relation.” The Bocca Nuova eruption also had a particularly high maximum TADR and was the first non-Type I or II event of our series (Table 3). During this 6.75 year long period, 36 × 106 m3 of lava were erupted at a TADR of 0.2 m3 s−1. Rothery et al.  showed an increasing volcanic radiance trend between 1996 and the end of 1999. This would convert to an increasing TADR trend, which builds to culminate in the high-intensity events experienced by the Bocca Nuova in 1999.
3.3.3. Period 3: 2000–2010
 The third period was marked by seven eruptions involving the output of 107 m3 of lava or more, with a total erupted volume similar to that of Period 1. However, although the frequency of eruption was higher in Period 3, eruption durations tended to be shorter, with a mean duration of 138 days, compared with a mean of 204 days for Period 1 (Table 4). Volumetrically, the character of the Period 3 eruptions also contrasts with those of Period 1 in that (1) emplaced lava volumes during individual eruptions were smaller, being between 2 and 68 × 106 m3 (mean = 39 × 106 m3) during Period 3, compared with a range of 18 to 183 × 106 m3 (mean = 80 × 106 m3) during Period 1 (Table 4); (2) the shorter durations for Period 3 eruptions meant that MORs were typically higher than during Period 1 eruptions, Period 3 MORs having a mean of 5.9 m3 s−1 compared with 4.6 m3 s−1 during Period 1 (Table 4); and (3) maximum TADRs were higher in Period 3 than during Period 1, being between 20 and 58 m3 s−1, with an average of between 3 and 24 m3 s−1 during Period 3, compared with a range of 4 and 35 m3 s−1 and averages of between 2 and 11.5 m3 s−1 during Period 1 (Table 3).
 These differences are accentuated if we remove the two long-duration SE Crater eruptions that began and ended the period, as done in Table 4 (Period 2b).
 Best fits with field data, from the sources indicated in Table 2 and given in the auxiliary material, also revealed that the “old relation” no longer applied to the second period eruptions, with the “new relation” instead applying to all eruptions in this series. For example, applying the “old relation” to the first three months of the 2004–2005 data set and integrating the TADRs through time yields an erupted volume of ∼15 × 106 m3, compared with a ground-measured volume of between 18.5 and 32 × 106 m3 [Burton et al., 2005]. However, use of the “new relation” yields a value ∼28 × 106 m3, a value more-or-less central to the field-measured volume. Likewise, applying the “old relation” to the TADRs for the lava flow field of the Lower Fissure System at 2100 m a.s.l. (the LFS1 flow field of Coltelli et al. ) active during the 2001 south flank eruption, and integrating the TADRs through time, yields an erupted volume of 13 ± 8 × 106 m3. This is low when compared with the volume of 21.4 × 106 m3 acquired from subtraction of digital elevation models generated during the eruption by Coltelli et al. . Instead, use of the “new relation” yields a value of 22.2 ± 6.1 × 106 m3, again in excellent agreement with the target value.
 Good fits between ground-measured volumes and volumes obtained from integrating up to 656 satellite-derived TADRs collected over eruptions lasting up to 500 days indicate that the satellite-based conversion is valid and can be trusted. In the case of the 2008–2009 eruption, for example, any systematic error in our TADRs would yield a large error in the final volume estimate obtained from integrating 656 TADRs over 420 days. However, the satellite-data-derived volume of 68 × 106 m3 is identical to the ground measured volume, and backs up our good fits between the ground-measured and satellite-TADR-derived volumes for the 1991–1993, 2001 and 2002–2005 lava flow fields. These data have now been collected over 30 years and used to track and assess effusion rates during individual eruptions [Harris et al., 1997; Harris and Neri, 2002; Lautze et al., 2004; Harris et al., 2007; Vicari et al., 2009; Harris et al., 2010]. Here, these data are collated allowing them to be used as a resource to statistically define and track Etna's longer-term volumetric output behavior.
 Our analysis of this database shows a change in the eruptive style, defined by volumetric properties, following the 2001 eruption (as seen in Tables 3 and 4). We find that, while the period 1983–1993 was characterized by less frequent, but longer-duration effusive eruptions at lower TADRs, 2000–2010 was characterized by more frequent effusive eruptions of shorter duration and higher TADRs. Another important difference is the mean lava volume of eruptions, the mean volume for eruptions between 1983 and 1993 being a factor of two higher than those for eruptions between 2000 and 2010, being 80 × 106 m3 for the former period and 39 × 106 m3 for the latter (Table 4). This change begins with the activity of 2001, suggesting that this year marked a change in the volumetric style of effusion. This change in activity style has already been explored in several previous studies which point to the increased explosivity and different chemistry of the erupted products during 2001 and 2002–2003 [Behncke and Neri, 2003a; Clocchiatti et al., 2004; Métrich et al., 2004; Corsaro et al., 2007]. After 2001, a new relation also has to be applied to our satellite data to derive TADRs that fit well with ground-based data. This new relation, as well as the old relation, is given in Table 1. If we compare the two relations we see that, to obtain the same areal coverage as was achieved using the old relation, a higher TADR is required in the new relation. This suggests a rheological change in the erupted lava, where an increase in viscosity and/or yield strength would force a decrease in the area over which a given volume of lava could spread at any given TADR. That all eruptions from 2001 onward required this new relation indicates that all lavas erupted since 2001 were subject to these new conditions. In addition, the time variation in TADR witnessed during each eruption changed from 2001 onward, with their Type III and IV trends being more consistent with batch-like ascent, as opposed to the pressurized or overflow-fed release of central conduit magma that generated the Type I and II trends of the first, 1983–1993, period. Together, these changes are consistent with more efficient ascent of batches from depth during 2001–2010, to erupt lava with a modified rheology and output rate when compared with the 1983–1993 period. This change may have been aided by the opening up of surface pathways for magma ascending from the deep source, to result in pathways which bypassed the central conduit and generated eccentric eruptions, as in 2001 and 2002–2003. This has been argued to be facilitated by processes such as flank slip [Acocella et al., 2003; Neri et al., 2005; Solaro et al., 2010; Ruch et al., 2010].
 However, in terms of long-term output nothing changed over the 1980–2010 period. If we plot the cumulative lava volumes for the two 11 year long periods spanning 1983–1993 and 2000–2010 we witness identical slopes (Figure 5a), meaning that the time-averaged output over the two periods was the same. As a result, the effused volume built at a steady, time-averaged rate of between 0.7 and 0.9 m3 s−1 which was stable through 2001. Thus, output remained steady even through the major change in style and location of surface emissions that occurred in 2001. Even the movement of Etna's eastern flank [Ruch et al., 2010; Solaro et al., 2010], responsible for the triggering of the 2004–2005 flank eruption [Burton et al., 2005; Neri and Acocella, 2006], did not produce any response in the long-term supply, as plotted in Figure 5. All that changed was the way in which this volume was partitioned between, and during, individual effusive events, as well as the spreading properties of the erupted lava.
 The steady output rate recorded between 1980 and 2010 has a time-averaged average value of 0.8 m3 s−1. This is similar to the value of 0.7 m3 s−1 obtained by Wadge and Guest  for the period 1973 to 1980. We can thus confirm that this trend has remained constant at least since 1971, as is apparent if we plot our data together with that of Wadge and Guest , as done in Figure 5b. Going further back still, we can use the data of Wadge et al.  to compare the 1971–2010 trend with those trends in operation since 1550, as done in Figure 6. The period prior to 1971 has already been examined by Wadge et al.  who pointed to two main segments in the cumulative volume plot: (1) a segment spanning 1610 to 1669 characterized by a relatively high output rate of 0.8 m3 s−1 and (2) a segment spanning 1759 to 1959 characterized by lower output rates of 0.2 m3 s−1. We note that the Wadge et al.  estimates for 1669–1971 were largely based on flank eruptive volumes. Instead, post-1971 volumes are much closer to a complete flank + summit budget. As discussed by Wadge and Guest  the flank-only rate for 1759–1970 was 0.17 m3 s−1, rising to 0.39 m3 s−1 if the 1971–1981 flank/summit ratio is used. Given this uncertainty, a better value for the second segment would thus be 0.2 ± 0.1 m3 s−1.
 It is noteworthy that, since 1971, we have returned to a trend identical to that witnessed prior to the 1669 eruption. This new trend has now been stable over 40 years. Indeed, if we consider the period 1971–2010 we find that 960 × 106 m3 of lava was erupted onto the flanks of the volcano in 40 years, with the volume being fairly evenly partitioned between these four decades (Table 5) to define a linear trend in cumulative output, which increases at a time-averaged rate of 0.8 m3 s−1 (Figure 5).
Table 5. Erupted Lava Volumes and Durations of Effusive Activity at Etna During the Period 1970 to 2010 Broken Out by Decadea
 During the high output rate phase of 1610–1669, eruptions were dominated by flank events [Wadge et al., 1975]. This phase was terminated by the effusive eruption of 1669. The 1669 eruption was the largest historical eruption recorded at Etna in terms of magnitude and intensity, with ∼1 km3 of lava being erupted in about four months of eruption (MOR ≈ 100 m3 s−1), to build a ∼17.5 km long tube-fed lava flow field that reached the sea south of Catania, destroying part of the city [Corsaro et al., 1996; Crisci et al., 2003]. This eruption signaled the end of an eruptive cycle that had lasted a few centuries and that had erupted plagioclase-rich lava, resulting from crystallization of a magma stored at shallow levels within the volcanic pile [Guest and Duncan, 1981; Corsaro et al., 1996]. After the 1669 eruption, emptying of the shallow system resulted in a caldera collapse of the summit area [Guest, 1973]. The volcanic activity that followed was significantly different in terms of both composition and volumes of lava erupted [Wadge et al., 1975; Tanguy, 1980; Guest and Duncan, 1981; Clocchiatti et al., 1988; Hughes et al., 1990; Condomines et al., 1995; Corsaro et al., 1996]. At the same time, the style of eruptive activity changed, with the new period showing an increasing number of summit eruptions [Wadge et al., 1975], as well as the lower time-averaged output rates apparent in Figure 6.
 The transition back to a high output rate period was heralded by the 1971 eruption [Wadge and Guest, 1981], followed a period of reorganization of the summit area, and coincided with the formation of the SE Crater. Guest  maps the development of a fissure that cut the summit with a NW-SE trend in 1956. This fissure eventually joined the Bocca Nuova and SE Craters, with the Bocca Nuova opening in 1968 at the NW end of the fissure, and the SE Crater forming at the SE end in 1971. The opening of the SE Crater coincides with the beginning of the modern, high output rate period. As noted by Branca and Del Carlo , this period was marked by an increase in summit activity, but also by voluminous flank eruptions and a high number of events from the newly formed SE Crater. Volumetrically, our data show that 542 × 106 m3 of the 779 × 106 m3 of lava erupted between 1980 and 2010 (or 70% of the total lava erupted) was associated with flank activity. It is worth noting that this figure does not change if we take into account also the volumes erupted by short-lived fire fountain events (Appendix B). In addition, 21 of the 38 (55%) of the effusive events listed in Table 2 were focused on the SE Crater, with all nine events since 2003 being centered on the SE Crater. Behncke and Neri [2003b] also point to the incidence of more frequent and more voluminous summit and flank eruptions since 1950, and remark that [Behncke and Neri, 2003b, p. 475] “if this trend continues, the activity of Etna might become similar to that of the 17th Century.” We can confirm that this is indeed the case, and the 17th century trend has now been in place for 40 years.
 If we consider the period 1971–2010, Etna experienced effusive activity on 3735 days, or 25% of the time (Table 5). Through this time there was variation in terms of the frequency and duration of effusive activity, but erupted volumes on a decadal scale were typically of the order of 300 × 106 m3 to keep the mean output rate between 0.6 and 0.9 m3 s−1; thereby maintaining a typical rate of 0.8 m3 s−1 (Table 5). This compares with the rate of 0.7 m3 s−1 obtained by Wadge and Guest  for the period 1971–1981, and 0.8 m3 s−1 for the period 1610–1669.
 The 30 year long satellite-derived discharge rate data set allows us to define the decadal volumetric behavior of the system, as well as to detail the volumetric trends and character of individual eruptions. This shows that, although the decadal-scale output remained stable between 1980 and 2010, the style with which this volume was erupted changed, with the 2001 eruption triggering a change in style: in 2001 eruption durations, syneruption volumetric trends, peak discharge rates and mean output rates all changed, as did the relation between discharge rate and flow area. Thus, although trends in long-term output appear stable and resilient, events such as that of 2001 can modify the manner in which this persistent flux is erupted. This suggests that the output rate is controlled by supply from the deep system, and cannot be changed by processes occurring in the shallow system; shallow system processes just serve to modify the way in which the flux is erupted. Thus, to change Etna's output rate, some profound change must occur in the deep system. Examining the cumulative output trends for the last 4 centuries show that such changes occurred in 1669 and 1971.
 We can conclude that Etna has displayed a remarkably steady output rate at least during the last 40 years. Similar steady state behavior, consistent with a constant flux of magma into and out of a crustal reservoir connected to the surface by an open conduit, has been detected at other volcanoes such as Stromboli [Bertagnini et al., 2003; Armienti et al., 2007; Calvari et al., 2011] and Montserrat's Soufriére Hills [Ryan et al., 2010; Wadge et al., 2010], and for shorter periods at Kilauea [Mattox et al., 1993] and Shishaldin [Petersen and McNutt, 2007]. This points to the resilience of output at such persistently active effusive (and extrusive) systems. Given that Etna is characterized by such steady supply and output rates we can propose that, unless the supply from deep system changes then, in the next decade, another 300 × 106 m3 of lava will be erupted onto Etna's flanks. Uncertainty arises, however, in projecting how superficial effects, such as flank spreading, will serve to determine the way in which this volume is distributed across the decade. Will it be erupted in a small number of longer, lower TADR events (as during 1983–1993), or will it be erupted in a larger number of shorter, higher TADR events (as during 2000–2010)? If the output trend of the volcano remains similar to that observed post-2001, we can expect short-duration effusive phases with high discharge rates. The key question is, though, when will this current phase of high output end, and how will that end be marked?
 Volumes, durations and mean output rates for effusive activity at Etna during the period 1970–2010 can be collated using the published data listed at the end of this appendix, plus the data generated by us using the satellite data. The resulting collation is given in Table A1. The types of eruptive event included in this collation are discussed in Appendix B, where we note that volumes erupted during explosive events are not included. Nor do we include the short-lived fountaining events that occurred between 1996 and 2001, although the events grouped within the 1990 eruption tabulated in Table A1 are actually four short fountaining events (as broken out in Table 2) lasting 0.375 h, 0.46 h, 0.375 h and 0.2 h.
Table A1. Volumes, Durations, and Mean Output Rates for Effusive Activity at Etna During the Period 1970–2010a
 Satellite-derived volumetric data presented in this paper cover a large proportion of the effusive products emplaced by Etna's eruptions during the past 30 years. These include almost all voluminous flank and summit effusive eruptions lasting more than 6 days. Appendix A lists 20 effusive events lasting greater than 6 days between 1980 and 2010, of which we have satellite data for 17. Just three are missing due to gaps in the data archive or cloud problems, these being the 10 day long eruption of September-October 1989 SE crater and E flank eruption, and the January-June 2000 eruption. We use volumetric measurements from the literature to fill in these missing volumes; we also use literature-derived volumes to fill in the data for effusive events lasting less than 6 days (as given in Appendix A).
 We find that the long-duration (greater than 6 days long) events account for 95 percent of the total volume erupted by effusive events (i.e., 737 × 106 m3 of the 775 × 106 m3 of lava erupted). Of this, 86 percent (666.5 × 106 m3) has been captured by our satellite data sets. Our analysis considers the volumetric nature of these large volume effusive events.
 Two types of eruptive event are not covered in the Appendix A listing, and thus in our analysis: (1) voluminous tephra emission over extended periods (as occurred during the effusive eruptions of 2001 and 2002–2003) and (2) short-lived episodes of lava fountaining that erupt lava and tephra at high intensity, but over relatively short time periods (typically just a few hours in duration), as occurred between 1996 and 2001.
 While the volume of tephra emitted during each of these event types cannot be quantified using the methods applied here, lava fountaining episodes are too brief to be captured by the satellites that this study relies on.
 To check the underestimate in total erupted volume from not considering these events, we collated available literature data for them. These give the following additional volumes emplaced during tephra emissions and lava fountaining between 1996 and 2010: (1) Erupted volume during 105 fountain events spanning 1996–2001 was 90 × 106 m3 (of which 72 × 106 m3 was lava) [Behncke et al., 2006]. (2) Erupted volume of pyroclastics during the July-August 2001 eruption was 5–10 × 106 m3 [Behncke and Neri, 2003b]. (3) Erupted volume of pyroclastics during 2002–2003 was 4.4 ± 0.6 × 1010 kg (converts to 16.3 ± 0.2 × 106 m3 dense rock equivalent using a rock density of 2700 kg m−3) [Andronico et al., 2008a]. (4) Erupted volume during the 4–5 September 2007 SEC Crater fountain event was 2–4 × 106 m3 (clastogenic lava flow) + 3.9–4.9 × 105 m3 = 2.4–4.5 × 106 m3 [Andronico et al., 2008b].
 For the period 1980–2010, inclusion of these volumes adds 1.17 × 108 m3 to the total of 7.75 × 108 m3, meaning that the Appendix A collation of volumes associated with the main effusive events includes around 87 percent of the total erupted volume (8.92 × 108 m3) during the three decades.
 For the period 2001–2010, inclusion of these volumes adds 2.73 × 107 m3 to the total of 2.76 × 108 m3 that we estimate was erupted during major summit and flank effusive events during this period (as given in Table 4), for a total erupted volume of 3.03 × 108 m3. Thus, around 90% of the total erupted volume over this period was recorded by our satellite-based data and was partitioned into flank or summit lava flow fields. Addition of these “missing” volumes increases the time mean output over this 9 year period from 0.97 m3 s−1 to 1.07 m3 s−1, thereby strengthening the trends, and conclusions, that we draw from our effused lava volume data set.
 We thank the three reviewers for their encouraging comments and helpful suggestions, as well as all University of Hawaii students who helped with AVHRR data processing, including John Bailey (1999 and 2001 data), Nicole Lautze (2001 data), and Lucas Moxey (2002–2003 data). All other data were processed by A.H. (pre-2000 data) and A.S. (post-2003 data). This contribution is in support of the LMV-based (PI: Franck Donnadieu) TerMex-MYSTRALS project “Contribution à l'évaluation des risques associés aux activités éruptives majeures de l'Etna: approche multidisciplinaire des processus et précurseurs.” The AVHRR data and processing support provided by the NERC-NEODAAS group at the Plymouth Marine Laboratory (Plymouth, UK; http://www.neodaas.ac.uk) made this work possible. Their continued work and collaboration are gratefully acknowledged.