A multi-decadal view of the heat and mass budget of a volcano in unrest: La Soufrière de Guadeloupe (French West Indies)
David E. Jessop, Séverine Moune, Roberto Moretti, Dominique Gibert, Jean-Christophe Komorowski, Vincent Robert, Michael J. Heap, Alexis Bosson, Magali Bonifacie, Sébastien Deroussi, Céline Dessert, Marina Rosas-Carbajal, Arnaud Lemarchand, Arnaud Burtin
AA multi-decadal view of the heat and mass budget of a volcanoin unrest: La Soufri`ere de Guadeloupe (French West Indies)
David E. Jessop S´everine Moune Roberto Moretti Dominique GibertJean-Christophe Komorowski Vincent Robert Michael J. HeapAlexis Bosson Magali Bonifacie S´ebastien Deroussi C´eline DessertMarina Rosas-Carbajal Arnaud Lemarchand Arnaud BurtinReceived: date / Accepted: date
Orcid IDs:David E. Jessop: 0000-0003-2382-219XS´everine Moune: 0000-0002-8485-0154Roberto Moretti: 0000-0003-2031-5192Jean-Christophe Komorowski: 0000-0002-6874-786XVincent Robert: 0000-0002-9016-7167Michael J. Heap: 0000-0002-4748-735XMagali Bonifacie:0000-0002-4797-043XMarina Rosas-Carbajal: 0000-0002-5393-0389
Abstract
Particularly in the presence of a hydrothermal system, many volcanoes output large quantitiesof heat through the transport of water from deep within the edifice to the surface. Thus, heatflux is a prime tool for evaluating volcanic activity and unrest. We review the volcanic unrest atLa Soufri`ere de Guadeloupe (French West Indies) using an airborne thermal camera survey, andin-situ measurements of temperature and flow rate through temperature probes, Pitot-tube andMultiGAS measurements. We deduce mass and heat fluxes for the fumarolic, ground and thermalspring outputs and follow these over a period spanning 2000–2020. Our results are compared withpublished data and we performed a retrospective analysis of the temporal variations in heat fluxover this period using the literature data.We find that the heat emitted by the volcano is 36 . ± . . ± . , this equates to a heatflux density of 627 ±
94 W / m , which is amongst the highest established for worldwide volcanoeswith hydrothermal systems, particularly for dome volcanoes. A major change at La Soufri`ere deGuadeloupe, however, is the development of a widespread region of ground heating at the summitwhere heat output has increased from 0 . ± . . ± . Introduction
Hydrothermal systems in active island-arc andesitic volcanoes are produced by the interaction of hotmagmatic fluids, essentially gaseous water, CO , H S and/or SO and HCl, produced by magmadegassing at depth with marine or meteoric water at shallower depths and the host-rock (Sigurdssonet al., 2015; Hedenquist and Lowenstern, 1994). Cooling through interaction with water (dissolution1 a r X i v : . [ phy s i c s . g e o - ph ] D ec nd/or absorption into deep ground waters and mixing with meteoric and sea-water) and the host-rockcauses chemical species to be reduced (Giggenbach, 1975; Moretti and Stefansson, 2020). Hence, thegeochemical profile of fluids discharged at the surface of a hydrothermal system is typically differentto that at depth. Hydrothermal systems can undergo sudden and catastrophic changes in behaviour.Two events in recent years, in particular, have highlighted the importance of understanding all aspectsof hydrothermal volcanoes and their hazardous behaviour: the September 2014 Ontake (Japan) andDecember 2019 Whakaari (White Island, New Zealand) eruptions, both of which resulted in the tragicloss of human life.The boiling of geothermal liquids liberates a fraction of the dissolved gases, which fractionate intothe vapour phase that ascends to the surface through steam-dominated fumaroles. Partial condensationof these vapours into ground waters may generate steam-heated waters likely to disperse laterallywhere they can further mix with external waters and discharge as thermal springs (Hedenquist andLowenstern, 1994; Sigurdsson et al., 2015). Therefore, significant amounts of heat are emitted as thesuper-heated steam, generated by these interactions, rises towards the surface through networks ofcracks, fissures and more porous rock within the edifice. The super-heated steam either condensesnear the surface or escapes to the atmosphere through fumaroles (Chiodini et al., 2001; Fischer andChiodini, 2015; Stimac et al., 2015). Heat emission can occur in several forms. First, where resistanceto flow is low (high permeability subsurface) the steam may reach the surface without condensing and,second, where resistance to flow is high (low permeability subsurface) the steam may condense nearthe surface. In the first scenario, the fumarolic output is high and significant amounts of heat andmass are transferred to the environment. In the second scenario, fumarolic output is correspondinglylower and heat is brought to the surface by forced convection and liberated to the environment byradiation and conduction (Harris, 2013; Gaudin et al., 2016). This leads to thermal anomalies (groundheating) and small, low-flux fumaroles typically distributed over quite large areas (cf. Aubert et al.,1984; Aubert, 1999; Harris and Stevenson, 1997; Harris and Maciejewski, 2000, for example). Inmany cases, ground heating far exceeds the fumarolic output in terms of energy transfer (Matsushimaet al., 2003; Mannini et al., 2019). Due to the high heat capacity of water, direct fumarolic degassingand diffuse small fumarole/soil degassing are generally the two major components of heat loss athydrothermal volcanic systems (Aubert, 1999; Chiodini et al., 2001). The final component of heattransfer in hydrothermal volcanic systems is through a network of thermal springs which typicallyappear along the flanks or base of the system, taking advantage of structural discontinuities. Thesesprings discharge water, initially heated by volcanic gases, that has either condensed deep within theedifice or nearer the surface when it has come into contact with the water table (Fischer and Chiodini,2015; Stimac et al., 2015).Whilst the relative importance of the different heat loss mechanisms will vary from volcano tovolcano and may vary in time at a given site, volcanic heat flow in general is indicative of (e.g. Hardee,1982; Lardy and Tabbagh, 1999; Harris et al., 2009): 1. The state and position of the magma body;2. The porosity/permeability of the edifice or dome; 3. The extent of infiltration of external waterinto the system. As such, spatio-temporal variations in heat flow are of particular importance forboth monitoring and fundamental research and allow us to greatly constrain numerical models of themagmatic and plumbing systems (Di Renzo et al., 2016).Hydrothermal systems play a fundamental role in providing and enhancing the physico-chemicalconditions that promote rock alteration, as well as the pressurisation of hydrothermal fluids. Theseprocesses act as strong forcing and triggering agents on the dynamics of volcanic activity by mechan-ically weakening edifice-forming volcanic rock (Pola et al., 2012; Wyering et al., 2014; Heap et al.,2015; Mordensky et al., 2019) and, therefore, promoting recurrent partial flank collapses (L´opez andWilliams, 1993; de Vries et al., 2000; Reid et al., 2001; Reid, 2004; John et al., 2008), as observed at LaSoufri`ere de Guadeloupe (Komorowski et al., 2005; Rosas-Carbajal et al., 2016). Escalating pressuri-sation of hydrothermal systems, as a result of permeability loss due to hydrothermal alteration, canalso lead to explosive activity (Heap et al., 2019) that can reach paroxysmal levels with non-magmaticlaterally-directed turbulent pyroclastic density currents or blasts (e.g. Bandaisan, Japan, in 1889).Hydrothermal alteration has also been observed to reduce the thermal conductivity and thermal diffu-sivity of andesite for a given porosity (Heap et al., 2020). Finally, the hydrothermal system is a strongmodulator of geophysical and geochemical signals of magmatic unrest and can generate a plethora of2on-magmatic unrest signals that render monitoring, as well as their interpretation and forecasting,very challenging (Pouget et al., 2015).In this paper which spans the past 20 years with particular emphasis on the 2010–2020 period, weconcentrate on the use of thermal measurements to infer the state of unrest of a major hydrothermalvolcanic system, that of La Soufri`ere de Guadeloupe (Lesser Antilles). We present the first studyfor this volcano that fully integrates measurements of all the heat sources over such a long period oftime. La Soufri`ere de Guadeloupe is a good target for such a study due to the wealth of geochemical,geological and geophysical data acquired on the volcano. As such, it is often considered a naturallaboratory representative of andesitic hydrothermal systems. Context
La Soufri`ere de Guadeloupe (16 . ° N, -61 . ° E, alt. 1467 m, hereby referred to as La Soufri`ere) isan andesitic dome volcano situated in the south of the Basse-Terre island of Guadeloupe (French WestIndies), which is part of the Lesser Antilles volcanic arc and is the most recent edifice of the GrandeD´ecouverte complex (445 ka). La Soufri`ere is amongst the most active and potentially deadly of thevolcanoes in the Lesser Antilles Arc (Komorowski et al., 2005). Hydrothermal activity is sustained bygas and heat transfer from a 6–7 km deep andesitic magma reservoir to shallower aquifers (Pichavantet al., 2018). Owing to an extensive hydrothermal system, La Soufri`ere has undergone a series of sixphreatic and hydrothermal explosive eruptions since the last major magmatic eruption in 1530 C.E.(Komorowski et al., 2005). The most recent, and probably most famous, eruption was in 1976–77(Feuillard et al., 1983; Hincks et al., 2014).The present edifice dates back at least 9150 years (Komorowski et al., 2005; Legendre, 2012),during which time several major magmatic eruptions have occurred, the latter in around 1530 C.E.,when the current dome was emplaced (Komorowski et al., 2005; Boudon et al., 2008). Since thislast magmatic event, there have been a number of phreatic and/or hydrothermal explosive eruptions.The last eruption occurred in 1976–77, following which the volcano became essentially dormant until1992 when seismic activity and steam emissions from summit fumaroles recommenced (OVSG-IPGP1999-2001 ; Zlotnicki et al., 1992; Komorowski et al., 2001, 2005). Summit degassing has graduallyincreased concomitantly with other observables (seismic, gas flux and concentration, ground and fu-marole temperatures, deformation, emissions of chlorine-rich acid gases), over the past ∼
30 years.This has included the appearance of two new high-flux fumaroles (Napol´eon Nord and Napol´eon Est,labelled NAPN and NPE on Fig. 1; OVSG-IPGP 2014-2016; Komorowski et al., 2005; Villemant et al.,2014; Moretti et al., 2020a), extensive zones of substantial surface heating and scalding of vegetation.Several fumarolic sites on the flanks characterised by a low state of activity since 1976, graduallyvanished. At the summit, Tarissan (TAS), Crat`ere Sud (CS), la Fente du Nord, Gouffre 56 (G56) andthe Lacroix fumaroles had all become inactive by 1984 (Komorowski et al., 2005; Boichu et al., 2011;Feuillard, 2011; Ruzi´e et al., 2013).An increase in activity in 2018 raised speculation that the volcano is in a state of growing unrestand is likely to undergo another eruptive episode in the near future (Moretti et al., 2020a). Until 2014,ground thermal anomalies and accompanying soil degassing had likely been limited to the areas directlysurrounding the major fumaroles, as well as the Faille de la Ty/Ravine Claire/Matylis structure (Fig. 1;OVSG-IPGP 2014-2020; Komorowski et al., 2005; Lesparre et al., 2012; Brothelande et al., 2014). Inrecent years, however, a number of thermal anomalies and altered zones have been observed such as atthe Zone Fumerolienne Napol´eon Nord (ZFNN) at the summit, delimited by NAPN, Crat`ere Dupuy(DUP) and TAS, adjoining the Breislack fault (BLK) and in the upper Matylis ravine (Fig. 1, OVSG-IPGP 2014-2020 and this work). Increasing fluxes and acidification of the water and gas rising upwithin the volcano has led to significant alteration and weakening of the edifice, leaving it vulnerableto flank collapse during even moderate seismic activity or extreme rainfall (Komorowski et al., 2005;Rosas-Carbajal et al., 2016).The summit vents are located near major fractures and fault zones, i.e. zones of high verticalpermeability (see Fig. 1 and Zlotnicki et al., 1992; Komorowski et al., 2005). These are likely to have Materials and methods
Aerial thermal surveys were carried out in 2010 and 2019, MultiGAS and Pitot-tube measurementshave been carried out monthly since 2017 and the thermal springs have been sampled monthly since2000. Our measurements are effectively contemporaneous even though sampling times differ betweendifferent methods and sites. We calculated errors on our estimations using standard error propagationformulae (Ku, 1966; Gibbings, 1986). Examples of how to apply these formulae and a table of relativestandard errors for all the parameters used in this study can be found in the Supplementary Material.
Ground thermal anomaly flux
We used airborne thermal imagery to measure the extent and distribution of thermal anomalies overthe entire volcano using an InfraTec VarioCam HD thermal camera (8–14 µ m) with 640 ×
480 pixelresolution. A 15 mm focal length lens (56 . × . ° FOV), gave an instantaneous field of view (IFOV)of 1 .
65 mrad. The distinguishable temperature difference between neighbouring pixels, NE∆ T , was0 . L sol and L atm in Fig. 3). Thus the true temperature of theground can be expressed as T = (cid:32) T − T − (1 − τ ) T g (cid:15)τ (cid:33) / , (1)where T cam is the brightness temperature seen by the camera, T atm is the brightness temperatureof the upper atmosphere, T g is the temperature of gases between the object and camera, τ is thetransmissivity of an atmospheric and volcanogenic gas mixture between camera and the ground and (cid:15) is the emissivity of the ground (Fig. 3).We converted at-camera (brightness) temperature to absolute temperature by applying Eq. 1. Fu-marole plumes and areas outside the region of interest were masked. We calculated τ using a radiative4ransfer model, with the surface-camera distance given by the georeferenced images and GPS locationof the camera (Kochanov et al., 2016; Berk and Hawes, 2017). We took the surface emissivity to beconstant for all the heated areas with (cid:15) = 0 .
95 in line with that found for other studies on andesiticsystems (Sekioka and Yuhara, 1974; Gaudin et al., 2016).We note that not all of the steam condenses before reaching the surface. Condensed liquids drainaway to be discharged elsewhere in the system (i.e. through thermal springs, in which case the heattransported is accounted for in the thermal springs heat budget) and any residual heat transferredto the ground, where it is accounted for in the soil heat budget. Hence we do not consider heattransported by condensed water here (cf. Gaudin et al., 2015). Our heat balance is thus (Sekioka andYuhara, 1974; Matsushima et al., 2003; Harris, 2013; Mannini et al., 2019) Q soil = Q soil , rad + Q soil , conv (2) Q soil , rad = A heated (cid:15) soil σ (cid:0) T − T (cid:1) (3) Q soil , conv = A heated h c ( T − T amb ) (4)where Q soil is the soil heat flux and subscripts rad and conv refer to radiative and conductive compo-nents of Q soil , respectively, T is the ground temperature, T atm is the ambient temperature and A heated is the heated area, (cid:15) soil is the soil emissivity. The heat transfer coefficient, h c , depends on severalfactors, particularly the local wind speed, w . We use the Schlichting-Neri model (Neri, 1998; Gaudinet al., 2013) h c = 1500 w ( z ) (1 .
89 + 1 .
62 log( z/z )) − . , (5)where z is the height above the surface and z is a measure of the surface roughness. Eq. 5 has beenshown to produce results that are consistent with the surface heat balance at La Soufri`ere (Gaudinet al., 2013), such that the heat conducted to the surface equals Q soil . We note that the surfaces on thevolcano where heat transfer occurs consist typically of centimetric blocks and thus we take z = 0 .
01 mas our roughness scale. We determined w from measurements at the Sanner weather station (cf. PitonSanner in Fig. 1) at the time of thermal image acquisition. The anemometer at Sanner is approximately2 m above ground level, so we take z = 2 m in our calculations. For wind speeds between 5–10 m / s, asseen on the 22 November, we find h c between 21.1 and 42 . / (m K). Considering error propagation,we estimate a relative standard error of about 10% on the radiative and convective flux measurements,and thus about 15% for the total flux.
Fumarole heat and mass fluxes
In-plume fumarole steam flux via MultiGAS traverses
The OVSG MultiGAS consists of an IR spectrometer for CO determination and electro-chemicalsensors for SO , H S and H . The atmospheric pressure is determined with the sensor installed on theCO spectrometer card. The MultiGAS also includes an externally-fitted relative humidity (RH) sensor(Galltec, range: 0–100% RH, accuracy: ± ° C, resolution:0 . ° C), to determine water vapour concentration (Moussallam et al., 2017). H O determination withthese external sensors reduced the risk of underestimating the measured water/gas ratios due to steamcondensation in the inlet. An onboard GPS receiver tracked the location of the instrument at 1 Hz.Data were visualised on an external tablet in real time. More detailed information about the OVSGMultiGAS, its design and performance characteristics can be found in Tamburello et al. (2019) andMoretti et al. (2020a,b).Fumarolic gas fluxes were determined for the three main vents that generate plumes (CS, TASand G56, Figure 1) following Allard et al. (2014) and Tamburello et al. (2019). The horizontal andvertical distributions of gas species in the plume cross-sections were measured a few meters downwindfrom the vents during orthogonal traverses on foot. Gas concentrations were measured at two differentheights (typically 0.9 and 2 m) as the volcanic gas plumes are generally flattened to the ground bystrong trade winds (2–14 m / s) and have a maximum height of ca. 3–4 m above the ground at eachmeasuring site with a maximum gas density centred at between ∼ . Observatoire Volcanologique et Simologique de la Guadeloupe fluxes are derived by multiplying the CO concentration integrated over the plume cross sectionwith the wind speed measured during the gas survey with a hand-held anemometer. We use CO as the volcanic marker as, due to its more conservative behaviour compared to H S and due to thefaster response of the IR CO sensor compared to the electro-chemical H S sensor, this way theMultiGAS is able to detect rapid concentration changes during plume transects. This avoids fluxunderestimations and leads to more accurate gas flux measurements (Tamburello et al., 2019). Dueto the high atmospheric background for H O and CO , our walking profiles start and end in pureatmospheric background in order to characterise and then subtract the ambient air composition fromour recorded data. Steam fluxes, ˙ m , are derived from the CO flux by multiplying it by the weight ratioof H O/CO . Steam flux estimates were possible only when water was successfully determined via theexternal RH sensor. It is important to note that some temporal variability of steam fluxes could be dueto: i) different ambient humidity and weather condition at the summit between field measurements;ii) occasional partial steam condensation on the external sensors. Indeed, particularly for tropicalvolcanoes such as La Soufri`ere, water vapour in the plume rapidly condenses upon contact with theatmosphere. However, this condensed water is not taken into account by the MultiGAS measurements.It has been shown that, in such tropical conditions, properly accounting for the condensed water addsapproximately 35% to the steam flux estimations (Gaudin et al., 2016), an increase which we considerin our analysis. Lastly, wind speed is the main source of error in quantifying volcanic gas fluxes,leading to typical standard errors on steam flux estimation of about 40%. At-vent fumarole fluxes via Pitot-tube measurements
Measurements of the steam exit speed at the vent of several fumaroles were made using a Pitot-tubeinstrument based around Freescale MPX2200AP and MPX2010DP temperature compensated pressuresensors that measured the dynamic pressure in the moving stream and ambient (stagnation) pressure.Pressure readings were taken at 3 .
75 Hz and the median of 10 measurements was recorded by anArduino Due. Uncertainty in the pressure readings was 3 . / s. From these values, the speed of a moving stream of gas, u , of density ρ wascalculated as (Massey and Ward-Smith, 1998) u = (cid:115) pρ (6)where ∆ p = p − p is the dynamic pressure, p is the stagnation pressure and p is the free-streampressure. Vent temperatures were simultaneously measured using a PT1000 resistance temperaturesensor with an instrumental error of ± ρ ( p, T ), using numerical codes based on IAPWS thermodynamiccalculations (Wagner and Pruß, 1993, 2002). Measurements were taken repeatedly at different pointsacross the vent in order to build up an idea of the velocity distribution. Typically 7 measurementswere taken and the median velocity from these measurements was used in the calculations that follow.From vent speed, we deduce the mass flux from the fumaroles which, as water vapour contributes upto 98% of the total mass (Allard et al., 2014; Tamburello et al., 2019; Moretti et al., 2020a, OVSG-IPGPbulletins 2017-2020;), is equivalent to the steam flux,˙ m = ρ ¯ uA ≈ ρ steam ( T )¯ uA (7)where ¯ u is the mean vent speed (equivalent in this case to median vent speed), A is the area of the vent.Whereas in Moretti et al. (2020a), vent area was estimated by eye by the Pitot-tube operator, here wecalculate A by analysing thermal images. We repeatedly took thermal images looking straight into thevents throughout the period when Pitot-tube measurements were made, from which we manually tracedaround the vent perimeter and, using the on-camera laser distance measurements, then converted tophysical area (i.e. in m ) via a pixel-to-physical length conversion as per Bombrun et al. (2018). Weestimated the relative standard error on mass flux measurements to be 10%.6 eat flux estimations The heat released through fumarolic activity is essentially due to cooling and condensation of the vol-canic steam. Fumarole heat flux can generally be decomposed into two contributing factors: radiationby the heated vent surface, Q rad , and the specific and latent heat carried by the gas phase, Q gas , sothat Q fumarole = Q rad + Q gas (Harris, 2013; Gaudin et al., 2016). Heat lost to the surroundings throughthe walls of the fumarole pipes is not considered as part of this heat budget, but are accounted forthrough the geothermal heating of the surrounding ground (Stevenson, 1993; Mannini et al., 2019), asshown in the previous section. Following Harris (2013); Allard et al. (2014); Gaudin et al. (2016) wewrite these as Q rad = A(cid:15)σ (cid:0) T − T (cid:1) (8) Q gas = ˙ m ( c p,v ( T ) ( T − T boil ) + L ( T ) + c p,l ( T boil − T amb )) (9)where (cid:15) is the ground emissivity, σ is the Stefan-Boltzmann constant, c p is the specific heat capacity , L ≈ / (kg K) the latent heat of condensation, T is the temperature of the steam, T boil ≈ . ° Cis the boiling temperature of water at the dome altitude and T amb ≈ ° C is the ambient temperatureat the summit. The subscripts v and l refer to the vapour and liquid states of water, respectively, with c p,v ≈ .
015 kJ / (kg K) and c p,l ≈ .
200 kJ / (kg K) for summit temperatures and pressures. During thesurvey period, T ranged from 96.9 and 108 . ° C for CS and has been measured in the water lake at TASto be approximately 97 . ° C (OVSG-IPGP 2016-2020). Since the G56 vent is in a c. 30 m deep cavitywithin the volcano, requiring specialised equipment to access, it is impractical to measure it directly.Thus we estimate that the temperature at the G56 vent is at the boiling temperature of water. Wenote that c p and L are functions of p and T and were solved for using similar numerical routines as fordensity. T boil is a function of pressure only and is also deduced from the IAPWS formulations (Wagnerand Pruß, 1993, 2002).Given the instrumental and measurement errors summarised in the text above, and using errorpropagation techniques (see Gibbings, 1986, for example), we estimated the standard errors on the fluxestimation using the Pitot-tube and MultiGAS instruments. In the case of the Pitot-tube instrument,the standard error in estimating Q fumarole is dominated by the mass flux and radiative flux terms and,overall, is around 10%. The standard error for our estimates based on MultiGAS measurements isdominated by the uncertainty in the mass flux measurements alone and so is about 40%. Thermal Springs
The nine thermal springs situated around the base of the current dome have been monitored regularlyby the OVSG since 1978 by manually measuring temperature and flow rate. The majority of siteshave been visited on a 1–3 month basis to take manual temperature readings as well as physico-chemical parameters such as pH and conductivity, and to take samples for future chemical analysis(Villemant et al., 2005, 2014). During these outings, and when it was possible, volumetric flow rate,˙ V , was deduced from the time taken to fill a container of known volume. This process was repeated6–10 times and we report here the mean value of these measurements. From this, we calculate themass flow rate, ˙ m spring = ρ ˙ V , which then allows us to calculate the heat flux as the sum of specific,evaporative and radiative heats, Q spring = Q spec + Q evap + Q rad ≈ Q spec = ˙ mc p,l ( T )( T − T amb ) . (10)We drop evapotransport, Q evap , and radiative heat losses, Q rad , in Eq. (10) as these contribute negli-gibly to the heat budget. The relative standard error on these measurements is about 5%. Results
Ground heat flux
We show our results from the analysis of the thermal images (Fig. 2) in Table 1. For each site withdetected thermal activity (summit, lower Ravine Matylis, Ravine Claire and FTY), we have determined7adiative and convective fluxes as well as the flux density, q i = Q i /A . As large fluxes can be observedby low intensity emissions over a large area, we also calculate the total heat flux density, q = ( Q rad + Q conv ) /A heated , as a metric for comparing intensity between sites (Table 1) with a relative standarderror of about 6%. At the summit we found the radiative flux to be 0 .
74 MW and the convective flux tobe 4 .
94 MW, with h c = 41 . ± . / (m K) (Eq. 5) and A heated = 14 070 m (areas were calculated bycounting the number of heated pixels, e.g. Fig. 2). These values were by far the largest in magnitudeof all the sites, and larger than the total heat fluxes for the other sites combined. This finding issupported by the heat flux density, which is considerably greater than any other site.We calculated h c = 37 . ± . / (m K) for wind speeds during acquisition of the images of theflank sites, which was used in the calculations for all sites. Owing to a relatively large emitting surfaceof 8010 m , the heat flux at Ravine Claire (RC) is second to the summit with radiative and convectivefluxes of 0 .
13 MW and 0 .
75 MW (Table 1), respectively. In the lower Matylis ravine, a strong thermalanomaly leads to high flux densities (42 and 238 W / m for radiation and convection, respectively),although a low heated area (1630 m ) keeps the overall fluxes low. We identified two sites alongFTY (FTY0 and FTY1 in Figs. 1 and 2) which have similar results for flux density and had a totalflux of 0 .
58 MW and a mean flux density of 227 W / m . We note that all these sites (Matylis, RC,FTY) are linked to the Ty N-SE and Galion N-S faults that cut the dome (Komorowski et al., 2005;Rosas-Carbajal et al., 2016). Fumarole heat and mass flux
The mass and heat fluxes are shown in Fig. 4 a) and b), respectively. Steam fluxes estimated fromMultiGAS traverses show that, using CO as a marker, the min/mean/max values are: 0.35/0.52/0.86,0.15/0.30/0.47 and 0.29/0.44/0 .
67 kg / s for CS, G56 and TAS, respectively. Heat flux estimates basedon these data give 0.93/1.36/2.28, 0.40/0.79/1.25 and 0.78/1.12/1 .
79 MW for CS, G56 and TAS,respectively. Considering the relative standard error of 40%, we find that the MultiGAS fluxes haveremained stable since regular estimates began in mid 2018, subsequent to the M4.1 earthquake.The temporal variations in fumarole steam flux calculated by the Pitot-tube for the three ventsat CS are also shown in Fig. 4a. These data indicate that the fluxes can show a large degree ofvariation in short time periods which is especially true during periods of accelerated unrest such asfrom March–May 2018 (Moretti et al., 2020a). We find steam fluxes to have min/mean/max valuesof 0.01/0.12/0 .
31 kg / s at CSC, 0.22/0.70/1 .
50 kg / s at CSN and 0.82/2.71/3 .
85 kg / s at CSS (Fig. 4a),which equate to heat fluxes of 0.03/0.29/0 .
75 MW for CSC vent, 0.53/1.69/3 .
64 MW for CSN vent, and1.99/6.56/9 .
31 MW for CSS vent (Fig. 4b). Fluxes were also measured at NAPN and we found meanmass and heat fluxes of 0 .
03 kg / s and 0 .
07 MW with very little variation over time, including duringthe 2018 unrest. The contribution of NAPN to the total heat and mass budget is thus negligible.We note that, due to an improved method for estimating vent area based on head-on thermal imagescompared to visual estimation during measurements (see Methods), the vent heat fluxes presented hereare quantitatively lower than reported in Moretti et al. (2020a), although the qualitative temporalvariation is the same. The Pitot-tube data show that vent fluxes at CS were strongly affected by anddecreased during the 2018 unrest phase, but have since settled to around 4 kg / s and 10 MW for massand heat flux, respectively. Fig. 4b) shows, although the coefficients for heat capacity and latent heatvary with temperature, the same trends as per Fig. 4a), indicating that the heat flux depended muchmore strongly on the variations in mass flux than temperature changes during this period.At CS, we have overlap in the Pitot-tube and MultiGAS instrument data that allows us to comparethe data collected by these instruments from closely spaced outings. For example, in terms of steamflux the Pitot-tube data from 15 June 2020 show that CSC+CSN+CSS emitted around 4 . ± . / s.The flux estimated from MultiGAS measurements at CS on 22 May 2020 were 0 . ± . / s. TheseMultiGAS estimates are almost an order of magnitude times lower than those from the Pitot-tube,and Fig. 4 indicates that this is systematically the case. Whilst we have attempted to correct forthe quantity of condensed vapour that is undetectable by the MultiGAS, additional errors in thiscalculation are likely to be primarily responsible for the difference between these two values, althoughthey agree to within an order of magnitude and appear to show qualitatively the same temporalvariations. 8 hermal springs heat and mass fluxes In Fig. 5, we present the mass flow rate and temperature measured over the period between 2000 and2020 (a subset of the entire data set, see Villemant et al., 2005, Fig. 5a and b), as well as the heat fluxcalculated from this via Eq. (10) (Fig. 5c). Whilst the flow rate and temperature measurements havecontinued until the present day, there are gaps in the mass and heat flux data during 2014–2016 dueto instrument failures. The GA, Tarade spring (TA), Bains Jaunes (BJ) and Pas du Roy (PR) springsare amongst the most accessible and this is reflected in both the abundance and persistence of themeasurements in the OVSG database. They are also the most representative of acid-sulphate thermalsprings linked to La Soufri`ere’s hydrothermal activity. This record does not reflect the absolute totalmass/thermal output of the thermal springs as i) other sites are known but are far less accessible orimpractical to measure and ii) some sites may not yet have been discovered. However, particularlyas GA and TA have the largest known flow rates, it is likely that these calculations are nonethelessrepresentative of the total budget for the thermal springs. We fitted linear trends to the data for TA,GA and PR, and extrapolated where necessary to project the values to the current date.Overall, we see that both mass flow rate and water temperature have slowly and steadily increasedover time in an approximately linear fashion. For example, the flow rate at TA increased from around1 . / s in 2010 to 2 . / s at present whilst its temperature rose from 309 to 318 K . Only the TAand PR sites have data that cover the whole data range and manual measurements stopped at GAin 2014. Historically, GA dominates the heat budget for the thermal springs, and has almost doublethe output of TA. Summing over these three sites, we find that the total heat flux from the thermalsprings is around 0 .
57 MW (including the extrapolated trend for GA).
Discussion
Comparison of steam and heat flux estimation methods
Fumarole flux
Our measurements (Fig. 6) show that the plume mass and heat fluxes have not undergone extensiveevolution since August 2005. With this in mind, we must consider that the fumarole plume heatand mass flux estimates of Gaudin et al. (2016) to be excessively high. In their discussion, the orderof magnitude discrepancy with the estimations from MultiGAS traverses (Allard et al., 2014) wasmostly attributed to the MultiGAS studies not accounting for condensed water vapour. However, wenote several key assumptions in Gaudin et al. (2016) that may have led to systematic errors in theirestimations: • Plume thickness grows linearly with distance from the vent , x , and not x / . This latter scalingis for the height of the plume axis from the ground (Slawson and Csanady, 1967), so that theirmass flux integral overestimated the plume area. • The plume section was assumed to be axisymmetric though this is generally not true: wind-blownplumes from smoke stacks, cooling towers and in laboratory experiments have been shown to bemore broad (horizontally) than thick (vertically, Contini and Robins, 2001). Thus, transmissivitycalculated looking horizontally through a plume would have been lower than was actually thecase, leading to an overestimation of plume temperature. Consequently the plume density andthus the mass flow rate should be lower than the given estimations. • The vapour carrying capacity of the plume was assumed to be equal to that of the atmosphere.However, as plume temperatures are higher than atmospheric and more water vapour can there-fore be carried without condensing, this relationship does not hold.Overall, this suggests that a more realistic plume flux for 2010 would be more in line with theMultiGAS (taking into account condensed vapour) and Pitot-tube measurements, that is a steam fluxof 5 . / s for CS. Thus taking the Pitot-tube measurement at CS as the ground truth, we found ascaling factor for the MultiGAS measurements. Using this to scale the MultiGAS flux for TAS gives9 . / s. Likewise, we find heat fluxes of 13 . . O/CO ratio determined from Giggenbach bottle sampling (OVSG-IPGP2017-2020; Moretti et al., 2020b) and multiply this by the CO flux estimated from the MultiGAS data.As this ratio is measured at the vent, it is not subject to a loss of matter due to condensation contraryto measurements within the plume. The resulting fluxes at CS resemble much more closely the Pitot-tube-derived fluxes (see “Reworked CS MG data” in Fig. 4). This correlation starkly indicates thedifficulties in accounting for condensed volcanogenic vapour in the MultiGAS steam-flux estimations.Nevertheless, in a monitoring context, either or both methods could be applied in various volcanoesworldwide to estimate their mass and heat fluxes. Ground flux
Although we have used the same model for h c as Gaudin et al. (2016), we obtain slightly differentvalues simply due to differing weather conditions (compare h c = 41 . ± . . ± . / (m K)for the summit and flanks, respectively, with h c = 30 . ± . / (m K) as derived from data in Table2 of Gaudin et al., 2013). Thus, similarity between the results of our study and those of Gaudinet al. (2013) at the same site would be suggestive of a decrease in temperature at that site. A goodcomparison can be made at the FTY sites. We note in particular that the mean total heat fluxdensity for FTY0 + FTY1 sites combined, 228 ±
14 W / m , is in strong agreement with the heat fluxdensity calculated from temperature gradient measurements of 265 ±
45 W / m (Gaudin et al., 2013),which suggests that, on average, temperatures have not decreased (the ambient temperature duringthe 2010 survey and ours was approximately 17 ° C in both cases). It is somewhat unclear preciselywhere Gaudin et al. (2016) defined the boundary of FTY and, indeed, their Fig. 1 suggests that thismight extend into what we define as Ravine Claire, so judging the evolution of the extent of this site isdifficult. However, taking uncertainties into account, the present-day total flux for FTY0+FTY1+RCof 1 . ± .
23 MW is not too dissimilar to the 1 . ± . . ×
13 cm (78 cm ) fumarole in a 169 cm pixel was lower than the actual temperature by 40 ° C.Taking this into account, we must consider that the fluxes that we calculate are minimum estimates,emphasising the importance of the ground heat flux for La Soufri`ere.
Total heat budget
As noted by Gaudin et al. (2016), some heat loss may be undetectable by the methods describedin this work, due to either vegetation cover (e.g. on the flanks) or temperature changes that arebelow the instrument resolution (cf. summit in the region of CS). As per their work, we note thatthe “background” heat conducted through the system, as deduced from borehole measurements andextrapolated to the scale of the dome adds only an additional 0 .
013 MW. Furthermore, we note thatsome heat will be transported by gases other than steam, notably CO in the plume and, in particular,CO soil degassing which is a widely-used proxy for heat flux (cf. Chiodini et al., 1998; Bloomberget al., 2014; Harvey et al., 2015). A detailed study is beyond the scope of this present work but wemay make progress under the following assumptions:1. Passive CO degassing occurs in the same areas and to the same extent as the ground thermalanomalies. 10. The ground heat flux, Q soil , calculated above equals the underground convection of steam, ˙ m H O c p,H O ( T − T amb ).3. The CO /H2O ratio in areas of soil degassing is the same as in the fumaroles (Chiodini et al., 2001).Under assumption (2), ˙ m H O = 135 . / s based on a typical anomaly temperature of 80 ° C andambient temperature of 15 ° C. Based on a H O/CO ratio of 43.5 (OVSG-IPGP, 2020), this gives˙ m CO = 3 . / s and thus Q CO2 = ˙ m CO2 c p , CO2 ( T − T amb ) = 0 .
19 MW.Clearly, although comparable to the contributions of certain thermal springs, heat transport byCO does not add significantly to the total budget. Nevertheless, due to the accelerating spread ofthe altered zones and the fact that the area over which CO degassing occurs may be far greater thanthat involved in degassing of water vapour (cf. Chiodini et al., 2005), the OVSG has begun to carryout joint surveys of soil temperature profiling and CO flux (via the accumulation chamber technique,Chiodini et al., 1998) in order to further constrain ground heat losses and we will return to this issuein a forthcoming paper.Summing the fumarole (28 . ± . . ± . .
56 MW 1.5%), we obtain a total heat output of 36 . ± . Temporal evolution
La Soufri`ere has undergone a significant evolution of its activity during 2010–2020 as described andanalysed in detail by Moretti et al. (2020a) and OVSG-IPGP (2014-2020). This can be summarisedas follows:1. The appearance of new fumarolic vents and the reactivation of pre-existing fumaroles with localhigh-velocity degassing.2. Vegetation die-off near-to and far-from active vents (see Supplementary Material Fig. 1).3. The enlargement of a major extensive area of heated ground on the summit areas that progressestowards the North from the Fracture Napol´eon (see below and Supplementary Material Fig. 2).4. More frequent and stronger seismic events including felt events (M4.1, April 2018).5. An acceleration in the opening rate of several summit fractures.6. The appearance of new mineralised water springs at the base of the volcano as a result of the rapidcooling of hydrothermal fluids.Undoubtedly the greatest phenomenological change at the summit of La Soufri`ere is indeed theappearance and spread of the ground thermal anomaly in the ZFNN region. For example, the heatedarea at the summit has gone from an estimated 610 m in 2010 (Gaudin et al., 2016) to 14 070 m for thepresent study. Indeed, whereas Gaudin et al. (2016) identified thermal anomalies along the Napol´eonfracture and in the craters containing the CS fumaroles, they calculated that the associated heat losseswere 0 .
01 MW by radiation and 0 . . ± .
07 and 4 . ± .
49 MW have increased by an order of magnitude between 2010 and2020, which is in large part due to this increase in heated area and also, to a lesser degree, because ofincreased temperatures. The total heat flux density presently estimated at 403 ±
26 W / m is greaterthan the 2010 estimate of 326 . ± . / m , and thus suggests an increase in thermal intensity atthe summit, though these values are within the bounds of measurement uncertainty.Apart from the pulse of unrest around March–April 2018, the fumarolic fluxes have not changedconsiderably since early 2018 (Fig. 4), and the thermal spring fluxes have increased only slightly (Fig. 5)over the past 20 years. To gain a greater perspective of the overall temporal evolution of plume fluxesover a similar period, we plot in Fig. 6 our data along with the steam and heat fluxes taken from Allardet al. (2014), Gaudin et al. (2016) and Tamburello et al. (2019). This figure shows that our current dataare, given the natural variability of these fluxes, consistent with the previous MultiGAS measurementsof Allard et al. (2014) and Tamburello et al. (2019). They are far more consistent with the Pitot-tubemeasurement for the CSN + CSC vents cited in Gaudin et al. (2016) than the fluxes that they derived11rom analysis of thermal images (compare their value of 5 . ± . / s to the contemporary sum forCSN + CSC of around 1 . ± . / s, Fig. 4a).Three major swarms of VT earthquakes occurred from 1 February to 28 April 2018, with the thirdswarm initiated by the off-volcanic axis M4.1 earthquake which struck at 00:32 UTC on 28 April andwas widely felt throughout Guadeloupe. In particular, as reported in Moretti et al. (2020a), a short-lived increase in plume flux occurred concurrently with temperature increases before the earthquakes,but both observables had returned to background levels before the M4.1 event on 27 April 2018. Hence,our Pitot-tube flux results illustrate the importance of these flux estimations for close monitoring ofvolcanic activity.Our results combined with published data indicate that plume flux has decreased overall sincethese records began. Thus, given a lack of increased VT seismicity or other signs of sudden evolutionin 2010, and in the light of the errors discussed above, it seems that the values reported in Gaudinet al. (2016) are anomalous. In order to provide a better comparison with the present study, weattempt to re-estimate the 2010 plume fluxes given the available data from this period (Allard et al.,2014; Gaudin et al., 2016). We suppose that, despite possible overestimation, the ratio of CS/TASfluxes was correctly established in 2010 and that the relative standard error will remain unchanged.Thus scaling with the 2005 Pitot-tube data, for 2010 we find steam fluxes of 5 . ± . / s for CS and6 . ± . / s for TAS, and heat fluxes of 13 . ± . . ± . . ± . . ± . . ± . . ± . ±
24 W / m is slightlyhigher than the 337 . ± . / m estimated in 2010. Using the reworked values from Gaudin et al.(2016), fumarole heat flux has decreased at CS and TAS, decreasing from a total of 28 . ± . . ± . . Comparison with other hydrothermal systems
La Soufri`ere’s total heat budget is on par with other major hydrothermal system volcanoes. Forexample, the heat output at Vulcano (Italy) was estimated at 10–12 MW from combining ground basedradiometer and ASTER measurements (Mannini et al., 2019). At Whakaari, heat output estimatedusing crater floor soil CO degassing as a proxy was found to be 20 . ± . ±
24 W / m (Table 1) which, if we consider the total heat budget overthe total heated area of 26 280 m , the mean flux density of the currently active part of the La Soufri`erecomplex climbs to 1366 ±
82 W / m (Tables 1 and 2). Based on the data compiled in Harvey et al.(2015) from heat flux density calculated from CO flux, this ranks La Soufri`ere amongst the world’smost powerful heat producing volcanoes, well above Whakaari (205 W/m2), Vulcano (140 W / m ,Mannini et al., 2019), Campi Flegrei (118 W / m ) and Nisyros (19 W / m ), and roughly on par withIschia (764 W / m ). Similar to La Soufri`ere, Vulcano, Whakaari and Ischia are also dome volcanoesand the larger heat flux densities here may indicate optimal steam transport to the surface alonghigh-permeable pathways associated with dome emplacement: Ischia, in particular, has a fumarolicH O/CO ratio similar to that at la Soufri`ere, (H O/CO =147 in 2001, Chiodini et al., 2005). This,due to the very extensive hydrothermal system at La Soufri`ere, indicates the dominance of heat andmass transport by water vapour generated through the interaction of hot magmatic fluids and thewater table. Taken together, especially with respect to the recent evolution at the summit, thesefindings indicate the importance of ground heating and thermal anomalies as a precursor for unrestof volcanic sites such as La Soufri`ere which may be far more relevant than at caldera-type sites (e.g.Campi Flegrei or Nisyros) where CO degassing is far more pervasive and heat loss through the groundis dominant. Conclusions
La Soufri`ere is an andesitic stratovolcano in the lesser Antilles arc with an extensive hydrothermalsystem that has undergone six phreatic/hydrothermal eruptions since 1635 C.E. Here, we have con-centrated on using thermal measurements to highlight the changes to the system over the past twodecades which cover most of the current unrest since its onset in 1992. Direct measurements were madeof the temperature and mass flux at the key fumarolic emission sites and at numerous thermal springslinked to the hydrothermal activity of La Soufri`ere. The ground temperature at sites showing extensivethermal anomalies was determined from airborne thermal imagery. From these and ancillary measure-ments for ambient conditions, we have deduced heat and mass fluxes as well as heat flux densities.We have compared and discussed our measurements in light of historic data available in the literature.Based on a reinterpretation of previously published data, we deduce that fumarolic output has propor-tionally decreased from 95% of the total heat budget in 2010 to 78% currently, whereas ground heatinghas increased from 4% in 2010 to 21% currently. The present-day convective and radiative heat fluxesin the summit area of 4.94 and 0 .
74 MW, respectively, have increased by an order of magnitude in thepast decade which is largely due to an increase in heated area and also, to a lesser degree, because ofincreased temperatures. The total heat flux density presently estimated at 403 ±
24 W / m is greaterthan the 2010 estimate of 326 ±
69 W / m , and thus suggests an increase in thermal intensity at thesummit, though these values are within the bounds of measurement uncertainty. These changes areexplained partly by a spreading in fumarolic sites over the dome during the past decade but also,13nd more importantly, that ground thermal anomalies on the summit have propagated significantlyin recent years. Fractures on the dome along with steady horizontal radial displacements of 3–10mm/year have been observed over the same period. The thermal spring activity has changed little in20 years although several of the thermal springs closest to the dome (GA, TA, BJ, PR) have shownsince 2000 a steady linear increase of their temperature and heat flux rate.We find that, in terms of heat flux density (heat loss per unit area), La Soufri`ere is amongst the mostintense emitters of heat for volcanoes worldwide, and that its ranking has dramatically increased inrecent years. With recent unrest events in mind, plus petrological evidence and geochemical analysis ofmagmatic fluids, we must consider that conditions with the potential to lead to phreatic/hydrothermalevents currently exist at La Soufri`ere. Hence, La Soufri`ere remains the subject of continued andenhanced surveillance and research strategies to better understand the origin of unrest and track itsdynamic evolution. Acknowledgements
The authors thank: the Assistant Editor, M. R. James and two anonymous authors for their con-structive comments and suggestions; the OVSG-IPGP team for logistical support and help with datacollection, Pierre Agrinier, and especially Gilbert Hammouya and Olivier Crispi for data collectionbefore 2013; Pascal Allemand and IGN for DEMs and orthophotos; the Pr´efecture de Guadeloupe andthe pilots of the Dragon 971 helicopter base in Guadeloupe (S´ecurit´e Civile, Minist`ere de l’Int´erieur) forproviding helicopter support; the Parc National de Guadeloupe for assistance and authorisation of re-search and monitoring on La Soufri`ere; IPGP, INSU-CNRS through the Service National d’Observationen Volcanologie (SNOV), and the Minist`ere pour la Transition ´Ecologique et Solidaire (MTES) for fi-nancial support. This work has been supported by the ANR DOMOSCAN, ANR DIAPHANE, theAO-IPGP 2018 project “Depth to surface propagation of fluid-related anomalies at La Soufri`ere deGuadeloupe volcano (FWI): timing and implications for volcanic unrest” (coord.: R. Moretti), andthe European Union’s through EUROVOLC (project No 731070). This study contributes to the IdExUniversit´e de Paris ANR-18-IDEX-0001, is IPGP contribution number 4164 and is LabEx ClerVolccontribution number 426. MJH acknowledges funding via the INSU-CNRS project “Assessing the roleof hydrothermal alteration on volcanic hazards”.
Author Contributions
DEJ collected and analysed the field data, prepared the figures and wrote the manuscript. SM, RM,DG, JCK and VR also collected and analysed field data. All authors contributed in the writing anddiscussion of the manuscript, and consented to its submission.
References
Allard, P., Aiuppa, A., Beauducel, F., Gaudin, D., di Napoli, R., Calabrese, S., Parello, F., Crispi,O., Hammouya, G., Tamburello, G., 2014. Steam and gas emission rate from la Soufri`ere volcano,guadeloupe (lesser antilles): implications for the magmatic supply during degassing unrest. Chem.Geol. 384, 76–93.Aubert, M., 10 1999. Practical evaluation of steady heat discharge from dormant active volcanoes:case study of Vulcarolo fissure (Mount Etna, Italy). J. Volcanol. Geoth. Res. 92 (3–4), 413–429.Aubert, M., Auby, R., Bourlet, F., Bourlet, Y., 1984. Contribution `a la surveillance de l’activit´e del’Etna `a partir de l’´etude des zones fumerolliennes. Bull. Volcanol. 47 (4), 1039–1050.Berk, A., Hawes, F., 2017. Validation of MODTRAN ® Table 1: comparison of radiative and convective fluxes, and heat flux density for the ground-heatedsites around La Soufri`ere from 2019 aerial imagery. In the right-hand column, mean values are givenfor the flux densities, and total values for all other quantities.
Summit Matylis (lower) Ravine Claire FTY0 FTY1 Total/Mean ∗ Rad. flux density, q rad /[W/m ] 52 76 50 71 62 54.59Conv. flux density, q conv /[W/m ] 350 455 310 432 383 351.16Total heat flux density/[W/m ] 403 531 360 504 446 405.75Radiative flux, Q rad /[MW] 0.74 0.12 0.40 0.10 0.07 1.43Convective flux, Q conv /[MW] 4.94 0.74 2.49 0.63 0.42 9.23Total ( Q rad + Q conv )/[MW] 5.67 0.87 2.89 0.74 0.49 10.66Heated area/[m ] 14070 1630 8010 1460 1100 26270 Table 2:
Comparison of the mass and heat flux estimates in 2010 (Gaudin et al., 2016) and for 2020(present study). Values in parentheses are reworked fluxes which, for 2010, are based on the likelyevolution of fumarole fluxes (see Fig. 6) and the thermal springs data (Fig. 5) and for the present-dayfumarole estimates are the MultiGAS traverse results scaled to the Pitot-tube measurements. Dashesindicate that no data was acquired at that site. *Encompasses RC and Matylis.
Flux Year Fumaroles Ground thermal anomaly Hot springs TotalCS G56 TAS Total Summit FTY Flanks ∗ Total GA PR TA TotalMass[kg/s] 2010 19 . ± . . ± . . ± . . ± . . ± .
1) (6 . ± .
2) (11 . ± .
3) (2.51) (0.44) (1.61) (4.56) (16 . ± . . ± . . ± .
8) (2 . ± .
1) (9 . ± .
5) - - - - 3.35 0.44 1.61 5.40 (15 . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± .
6) (15 . ± .
4) (28 . ± .
0) (0.25) (0.02) (0.12) (0.39) (29 . ± . . ± . . ± .
2) (7 . ± .
0) (28 . ± .
8) 5 . ± . . ± . . ± . . ± . . ± . i g h f l u x f u m a r o l e s T h e r m a l s p r i n g s r o a d s w a t e r w a y s S o u f r i e r e - Z F NN F l a n k a l t e r a t i o n z o n e s G e o l o g i c a l f e a t u r e s C r a t e r s a ) b ) Figure 1: i g h f l u x f u m a r o l e s r o a d s / f oo t p a t h s Z F NN F l a n k a l t e r a t i o n z o n e s F r a c t u r e s S u r f a c e t e m p e r a t u r e / [ ° C ] >= . . . . . . . . . . . a ) b ) Figure 2: a) Georeferenced thermal images of La Soufri`ere and surroundings taken from the helicopteron 22 November 2019 between 05:40 and 06:05 local time b) zoom showing the summit thermalanomalies. The base map is the 2017 IGN aerial orthophoto (BDOrtho). The thermal images areshown in greyscale where white and black denote hot and cold, respectively.21 igure 3:
Conceptual model of the heat flux measured by a thermal camera, L b viewing and displayedas the “brightness” temperature, T b . The incoming heat fluxes (left) from solar radiation, L sol , andfrom the atmosphere, L atm , are reflected from the surface in proportion to the surface albedo, α , and1 − (cid:15) , respectively, where (cid:15) is the emissivity of the surface. For a long-wave infrared sensor such asthe thermal camera used here, α ≈ − (cid:15) . The emitted radiation of the surface, L (which depends onsurface temperature, T , though the Planck function, P ), is added to these reflected fluxes which arriveat the camera having been transmitted through a mixture of atmospheric and volcanogenic gases attemperature, T g , and having transmissivity, τ . 22 S t e a m flu x / [ k g/ s ] CS MG CO2 markerTAS MG CO2 markerG56 MG CO2 markerReworked CS MG data Pitot tube, CSCPitot tube, CSNPitot tube, CSSPitot tube, NAPN a) Date T o t a l h e a t flu x / [ M W ] b) Figure 4:
Time series of the summit fumarole fluxes since the last quarter of 2017 as estimated frompitot tube and multigas data. Steam fluxes are shown in a) and heat fluxes in b). Vertical grey barsindicate the record of VT earthquakes with magnitude > M2 . flux multiplied by the H O/CO ratio determined from Giggenbach bottleanalyses. 23 M a ss fl o w r a t e / [ k g/ s ] BCM/EVBJB/BJ CEGA HRPR RM3TA VH-1 a) T w / [ K ] mean air temp. b) Date . . . . . H e a t flu x / [ M W ] c) Figure 5:
Mass flow rate, water temperature and heat flux for the thermal springs monitored by theOVSG for the period from 2000-2020. See text for site codes. The colour code for each site is identicalbetween plots and dashed lines show linear trends for the GA, TA and PR sites.24 S t e a m flu x / [ k g/ s ] True fluxes?