Detecting gas-rich hydrothermal vents in Ngozi Crater Lake using integrated exploration tools

Gas-rich hydrothermal vents in crater lakes might pose an acute danger to people living nearby due to the risk of limnic eruptions as a result of gas accumulation in the water column. This phenomenon has been reported from several incidents, e.g., the catastrophic Lake Nyos limnic eruption. CO2 accumulation has been determined from a variety of lakes worldwide, which does not always evolve in the same way as in Lake Nyos and consequently requires a site-specific hazard assessment. This paper discusses the current state of Lake Ngozi in Tanzania and presents an efficient approach how major gas-rich hydrothermal feed zones can be identified based on a multi-disciplinary concept. The concept combines bathymetry, thermal mapping of the lake floor and gas emission studies on the water surface. The approach is fully transferable to other volcanic lakes, and results will help to identify high-risk areas and develop suitable monitoring and risk mitigation measures. Due to the absence of a chemical and thermal stratification of Lake Ngozi the risk of limnic eruptions is rather unlikely at present, but an adapted monitoring concept is strongly advised as sudden CO2 input into the lake could occur as a result of changes in the magmatic system.

input cannot be excluded. In that context Lake Ngozi is rather comparable to the geothermally heated Kelud crater lake (Indonesia) 13 or volcanic lakes in El Salvador 14 , than Lake Nyos, Lake Kivu or Lake Monoun.
Analyses of the steep crater wall gave evidence of a lake breaching event of the southern caldera rim 9 . It is assumed that the Ngozi tuff eruption (<1 ka) built up the southern caldera rim, which was then breached by the newly formed crater lake. This could still be a potential risk of flooding downstream in particular if lake water is suddenly drained as a result of activated instabilities in the crater wall. On the other hand, rapid CO 2 recharge from below could result in catastrophic CO 2 output, which might also facilitate lake breaching events. One has to remember, that Ngozi volcano is an active volcano with the last eruption <1 ka ago, and future volcanic activity in the RVP is very likely. The area-wide effects of catastrophic gas bursts have been analysed and demonstrated by numerical modeling 15 .

Methods
A 150 kHz echo sounder with a depth range from 0.7-100 m was used for the bathymetric survey. Multiple traverses have been performed using an inflatable boat with an electric motor. Depth information was recorded every 10 m at 4,681 sites. Temperature and electrical conductivity (EC) measurements have been performed with a WTW ProfiLine Multi 197i and a 100 m T-EC electrode (accuracy for temperature ±0.2 °C, for EC ±0.5% of value). 31 T-EC-logs have been performed at regularily distributed sites with data readings every 2 m. In 24 locations the sensor was lowered down to the lake floor to obtain ground temperatures. When the lake floor was touched a distinct change in EC was observed. Results from previous studies 10 have been used for comparison. CO 2 flux measurements have been performed according to the accumulation chamber technique 16 by a floating flux chamber (West Systems). The fluxmeter is equipped with a LI-COR 820 CO 2 gas analyzer (accuracy <3% of reading). 39 measurements have been conducted on the lake surface, and four reference measurements have been performed on the flank of the crater. All techniques have the great advantage of in-situ data readings, which ensure enough flexibility during field campaigns to trace and delineate the main target areas. Data processing was performed with ESRI ArcMap, OriginPro, and FluxRevision software. Data are accessible through the Research Data Repository of GFZ Data Services (dataservices.gfz-potsdam.de) or may be obtained from the author upon request.

Results and Discussion
Lake Ngozi has a maximum depth of 83 m (Fig. 1a). The western sector is characterized by a relatively flat lake floor, whereas the eastern sector appears rugged. Between both sectors four hole-like structures were identified. The mapped holes are clearly NNE-aligned and suggest the presence of a permeable fault zone, which was not known before. The holes have a maximum depth of 10 m compared to the relatively flat surrounding lake floor. Lake Ngozi is characterized by very stable temperature conditions in the water column below 10 m depth, where surface effects can be excluded (Fig. 1b). Average water temperature from 10 m depth down to the lake floor is 19.6 ± 0.1 °C. In the uppermost level of the lake, water temperature varies from 20.4 °C at the surface to 19.4 °C at 10 m depth. No significant spatial temperature variations have been observed in different depth levels of the water column. In contrast to the water column, lake floor temperatures show substantial variations from 19.6 °C to 89 °C, detected in the SE and NE of the lake. The same characteristic was observed for electrical conductivity (Fig. 1c). EC variations along the water column were minimal with an average value of 5.25 mS cm −1 , but EC values of the lake floor (presumably in sediments) varied substantially from lower to higher values (see also Kusakabe et al., 2019) 17 (min. 3.1 mS cm −1 ; max. 10.2 mS cm −1 ). The average electrical conductivity of Lake Ngozi indicates the presence of large quantities of dissolved ionic components, which can only be the result of major subaquatic hydrothermal feed zones. The highest EC values have been measured at sites characterized by highest ground temperatures (Fig. 1d), possibly indicating inflow of hydrothermal fluids through the lake floor.
With the purpose to identify minor ground temperature variations a temperature map was compiled, excluding all temperature values above 21.3 °C (Fig. 2a). The map illustrates a well-defined thermal anomaly along a central NNE-oriented corridor, which can even be identified with low temperature variations of less than d1.8 K and without any information of high ground temperatures. Ground temperatures above 21.3 °C (up to 89 °C) are illustrated as points and plot in the same area, thus confirming the pattern of the temperature map. It is expected that further high-temperature spots are located in the deep holes identified along the NNE-oriented corridor (Fig. 2b). CO 2 emissions from Lake Ngozis water surface have been determined across an area with increased lake bottom temperatures and identified hole structures (Fig. 2c,d) in order to link gas flux with permeable fractures in the subsurface 18  conclusion It is demonstrated how basic field studies can significantly improve the understanding of potential subaquatic hydrothermal feed zones in volcanic lakes. One of the striking results of the study is that minor variations in lake floor temperatures between 19.5 °C and 21.3 °C (d1.8 K) already pinpoint thermal anomalies along a narrow, NNE-oriented corridor, which suggests a strong fault control. Deep holes identified by an echo sounding survey seem to be the centres of the thermal anomalies, where gas-rich fluids enter the lake. This was confirmed by increased gas emissions from the lake surface. Increased lake floor temperatures correlate with increased electrical conductivity, suggesting intense fluid-rock interaction and possible formation of alteration minerals. The success of the presented study is based on the combination of different methodological approaches. The site-specific findings are also meaningful for a conceptual advance in the understanding of the Ngozi hydrothermal system and related geothermal exploration activities, as upflowing hydrothermal fluids have always been suspected under Ngozi volcano, but could never be proved before this study.
Lake Ngozi seems to be mainly affected by a thermally driven mixing 20 . Weak convective mixing as a result of fluid inflow from the geothermal system cannot be seen in the existing data, but also cannot be excluded due to possible variations in the magmatic-geothermal system underneath. The presented data can build the basis for the development of a site-specific monitoring concept for the main areas with increased gas flux and ground temperatures. The main hazard at Lake Ngozi is the reactivation of the magmatic system, which would have severe implications on the degassing processes. The permanent installation of a differential absorption LIDAR (DIAL) system 21 in combination with thermal sensors could be a suitable solution to detect precursory signals of volcanic unrest. Further methodological improvements can be achieved by thermal mapping of the lake floor using fibre optic measurements for integrated temperature logs, and direct sampling of hydrothermal fluids at the identified feed zones.