WMO Aeronautical Meteorology Scientific Conference 2017

Transcription

1 Session 1 Science underpinning meteorological observations, forecasts, advisories and warnings 1.5 Atmospheric aerosols, volcanic ash research Modelling and data assimilation of hazardous volcanic ash plumes in the chemical-transport model MOCAGE Bojan Sic, CNRM/Météo-France bojan.sic@meteo.fr co-authors: Matthieu Plu Laaziz El Amraoui speaker: Bojan Sic Volcanic ash aerosols can be emitted in significant amounts by volcanic eruptions and once airborne they pose a serious treat to aircraft engines. Especially after the Eyjafjallajokull eruption in 2010, important efforts have been put in the improvements to the volcanic ash modelling. Our research deals with spatial and temporal improvements of volcanic ash aerosols in the chemical-transport model (CTM) MOCAGE (fr. Modèle de Chimie Atmosphérique à Grande Echelle), the operational air-quality and fast-response model of Météo-France. The chemical-transport model MOCAGE (e.g. Josse et al., 2004; Sic et al., 2015) is a global model that has embedded nested regional domains. In such case the lateral boundary conditions for the smaller domain are provided by global or larger regional domain. The nested domains are of the particular interest for the simulations of volcanic eruptions where the region around a volcanic plume source can be modelled with high resolution. MOCAGE simulates both gases and primary and secondary inorganic aerosols, each aerosol type particle size distribution is divided into six bins, and in the case of volcanic ash aerosols the bin boundaries are variable depending on the properties of each volcanic plume, and in corporate the fine-ash fraction of the total size distribution. The main uncertainty source in volcanic ash prediction is still the emission term: it is necessary to describe as much accurately as possible the mass eruption rate of the volcanic emissions and the initial vertical distribution of ash aerosols in the column. The complex processes and scarce difficult-to-make observations of the volcanic plume source have driven, in the first place, the development of the empirical parameterizations, as the attempt to define the emission term in models. Such parameterizations try to connect the height of the eruption plume (as the parameter that can be readily be observed) and the mass of the eruption aerosols ejected into the atmosphere. In the model MOCAGE, we use the Mastin et al. (2009) empirical relation which is based on the Sparks et al. (1997) work. There are continuous efforts to improve existent and to develop new parameterizations (Costa et al., 2016a). Some of downsides of such approach is that the parameterizations do not answer on the question of the aerosol vertical distribution in the eruption column, they do not or they include only simplified description of the atmospheric conditions which influence the plume, and they reflect important uncertainties of data on which they are based. An another approach to estimate the source term in the models is to understand and model the physical processes in the plume and their interaction with the atmosphere. The plume rise models get increasingly sophisticated and can provide estimations of eruption and

2 plume source parameters. To improve the estimation of the ejected mass and the particle vertical size distribution in MOCAGE we introduce the 1-D cross-section averaged plume rise model FPLUME (Folch et al., 2015), which also takes into account the effects of meteorological conditions and of important physical processes like wet aggregation, air and particle entrainment, particle sedimentation, etc. The FPLUME model is based on the turbulent buoyant plume theory, it calculates the height of an eruption plume from the eruption mass rate and the initial size distribution at the vent by solving the governing equations, and it also outputs as a result the the plume mass vertical distribution and the height dependent particle size distribution for all levels till the top of the plume. Figure 1 shows the vertical mass distribution of the plume initialized in the CTM MOCAGE obtained by the Mastin et al. (2009) parameterization and the presumed the uniform vertical distribution from one side, and the distribution obtained by the FPLUME model from the other side, for the case of 6 May 2010 eruption phase of Eyjafjall. The raw Mastin et al. (2009) relation initializes significantly more particles compared to the plume rise model. This is also due to the fact that the estimated mass eruption rate includes the sizes of particles that will be not a subject of the long-range transport, and that will be sedimented rapidly after the ejection. Therefore, the initialization of the plume mass should take this into account. In our test case, the plotted plume mass is reduced two times compared to the original estimation. The results of plume rise models in general show that the amount of rapidly removed particles from the plume is very variable and it depends on the eruption type, initial size distribution, eruption phase and external meteorological conditions and therefore, it is difficult to estimate it. On the other hand the FPLUME profile has initialized a significant mass only from the neutral buoyancy level till the plume top. Also, as the FPLUME model simulates the evolution of the size distribution, it outputs the fallout of each particle class, and it directly estimates the mass of plume that will be initialized in the model for the dispersion in each level. A physical processes that can have a significant effect on the size distribution evolution is wet aggregation of ash aerosols, and it is taken into account in FPLUME. It occurs in presence of ice and liquid water and impacts the residence time and transport of the particles inside the plume. Its influence on the relation plume height/mass eruption rate is rather weak (Macedonio et al, 2016), but its effect on the dispersed plume mass is important. In our test case, percentage of the mass eruption rate of the ash particles that will be dispersed is 19% with wet aggregation, compared to 40.5% without taking into account wet aggregation. In this case the wet aggregation is efficient as there is one of favorable aggregation conditions satisfied (a liquid water section inside the plume and presence of ice in upper parts of the plume). The occurrence of favorable conditions is impossible to predict outside of the model, and a similar effect cannot be produced with an empirical reduction of mass eruption rate in conjunction with mass eruption rate parameterizations, as usually done in the dispersion/ctm models. Figure 1. Mass vertical distribution of the initialized volcanic plume obtained by the current and FPLUME model runs

3 Figure 2 shows the difference in the size distribution of the ash particles in the plume that is initialized in the MOCAGE for the dispersion. On one side we have a prescribed empirical size distribution that tries to reflect also the distribution transformation in the plume (losing of large particles in the initial distribution due to sedimentation) and it is estimated as an unimodal distribution. On the other side we have the FPLUME averaged distribution (from the height dependent distribution) which is calculated to be a bimodal distribution by the model. These final initialized distributions depend primarily on the initial size distributions, but are also heavily modified by sedimentation and wet aggregation. The observations show that particles can have unimodal or bimodal distributions when are ejected from the vents, therefore we gave an important attention also to the estimation of the initial size distribution at the vent. Instead of prescribed distributions depending on the eruption type, we implement a semiempirical parameterization that estimates the distribution based on the plume height and magma viscosity (Costa et al., 2016b) for the cases when the observed initial size distribution is not available. Figure2. Particle size distributions of he initialized volcanic plume obtained by the current and FPLUME model runs. The comparison between the operational and the new configurations shows that differences in the source term produce important differences in extent and concentrations of the transported volcanic ash plumes. The dispersion extent is a lot larger in the case of the uniform vertical distribution of the initialized plume compared to the Suzuki shape of the plume calculated by FPLUME. But the case of the better constrained plume shape is more sensitive to the uncertainties. The uncertainties on the neutral buoyancy level can produce a

4 very different transport of the ash particles. Both approaches are readily used in models and their comparison and results confirm that uncertainties of the volcanic ash modelling are considerable. For that reason, for the further improvements, we explore the axis of combining models with available Earth observation data in means of data assimilation in the framework of the European project EUNADICS-AV. The project s goal is the monitoring and assessment system for the estimation and the forecasting of the hazardous events, like volcanic ash plumes, where the data assimilation plays an important role. We evaluate the impact of observations on the plume characteristics by assimilating lidar profiles (spaceborne and ground-based) and satellite aerosol optical depth (AOD) measurements in MOCAGE. The CTM MOCAGE has a data assimilation system based on variational methods, described in details in Sic et al. (2016). Figure 3 shows the comparison of a direct model output and the same field after data assimilation of AOD MODIS observations. Data assimilation has an important impact on the prediction of the volcanic ash plumes. The direct model run in our case has overestimated source term, and data assimilation significantly influenced the extent and intensity of the plume. The data assimilation impact is event more important when dealing with determination of threshold values in concentrations. It is important to note the fact that when we assimilate integrated quantities and when we talk about model uncertainties, it more advantageous for the assimilation that the direct model makes an overestimation than an underestimation in the terms of integrated quantity. For example, when talking about spatial extent uncertainties, an assimilated AOD observation of the volcanic plume will impact the whole vertical profile; if the volcanic plume is not present at all in the direct model profile, assimilation will instead affect aerosols that are present, often at the levels lot different than that one of the plume, and of different type. When assimilating AOD observations, we deal with two types of integrations, the vertical one to get a column, and the integration of all species and bins to get a total AOD.

5 Figure3. The aerosol optical depth over Iceland and the Atlantic ocean on 10 May 2010., simulated by the model direct run (left), simulated in MOCAGE by the MODIS assimilation model run (middle), and assimilated MODIS observations (right). We have made modelling and assimilation efforts to improve results of the volcanic plume predictions and our results confirm the significant uncertainties related to such predictions. We show that the combined approach of modelling and data assimilation is a necessary strategy in order to adequately improve the forecasts of hazards that can impact the aviation safety. References: Costa, A., Suzuki, Y.J., Cerminara, M., Devenish, B.J., Ongaro, T.E., Herzog, M., Van Eaton, A.R., Denby, L.C., Bursik, M., Vitturi, M.D.M. and Engwell, S. Results of the eruptive column model inter-comparison study. Journal of Volcanology and Geothermal Research, 326, 2-25, 2016a. Costa, A., Pioli, L. and Bonadonna, C. Assessing tephra total grain-size distribution: insights from field data analysis. Earth and Planetary Science Letters, 443, , 2016b. Folch, A., Costa, A., and Macedonio, G.: FPLUME-1.0: An integral volcanic plume model accounting for ash aggregation, Geosci. Model Dev., 9, , Josse, B., Simon, P., and Peuch, V.-H.: Radon global simulations with the multiscale chemistry and transport model MOCAGE, Tellus B, 56, , 2004 Macedonio, G., Costa, A., Folch, A. Uncertainties in volcanic plume modeling: A parametric study using FPLUME. J. Volcanol. Geoth. Res. 326, , Mastin, L.G., Guffanti, M., Servranckx, R., Webley, P., Barsotti, S., Dean, K., Durant, A., Ewert, J.W., Neri, A., Rose, W.I., Schneider, D., Siebert, L., Stunder, B., Swanson, G., Tupper, A., Volentik, A., Waythomas, C.F., X. A multidisciplinary effort to assign realistic source parameters to models of volcanic ash-cloud transport and dispersion during eruptions. J. Volcanol. Geotherm. Res. 186, 10 21, Sič, B., El Amraoui, L., Marécal, V., Josse, B., Arteta, J., Guth, J., Joly, M., and Hamer, P. D.: Modelling of primary aerosols in the chemical transport model MOCAGE: development and evaluation of aerosol physical parameterizations, Geosci. Model Dev., 8, , Sič, B., El Amraoui, L., Piacentini, A., Marécal, V., Emili, E., Cariolle, D., Prather, M., and Attié, J.-L.: Aerosol data assimilation in the chemical transport model MOCAGE during the TRAQA/ChArMEx campaign: aerosol optical depth, Atmos. Meas. Tech., 9, , Sparks, R.S.J., Bursik, M.I., Carey, S.N., Gilbert, J.S., Glaze, L.S., Sigurdsson, H., Woods, A.W., X. Volcanic Plumes. John Wiley & Sons, Chichester (574 pp.), 1997.