Geosci. Model Dev., 14, 409–436, 2021 https://doi.org/10.5194/gmd-14-409-2021 © Author(s) 2021. This work is distributed under the Creative Commons Attribution 4.0 License. FALL3D-8.0: a computational model for atmospheric transport and deposition of particles, aerosols and radionuclides – Part 2: Model validation Andrew T. Prata1, Leonardo Mingari1, Arnau Folch1, Giovanni Macedonio2, and Antonio Costa3 1Barcelona Supercomputing Center (BSC), Barcelona, Spain 2Istituto Nazionale di Geofisica e Vulcanologia, Osservatorio Vesuviano, Naples, Italy 3Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Bologna, Bologna, Italy Correspondence: Andrew T. Prata ([email protected]) Received: 27 May 2020 – Discussion started: 17 June 2020 Revised: 4 November 2020 – Accepted: 19 November 2020 – Published: 25 January 2021 Abstract. This paper presents model validation results for and FMS scores greater than 0.40 indicate acceptable agree- the latest version release of the FALL3D atmospheric trans- ment with satellite retrievals of volcanic ash and SO2. In ad- port model. The code has been redesigned from scratch to dition, we show very good agreement, across several orders incorporate different categories of species and to overcome of magnitude, between the model and observations for the legacy issues that precluded its preparation towards extreme- 2013 Mt. Etna and 1986 Chernobyl case studies. Our results, scale computing. The model validation is based on the new along with the validation datasets provided in the publicly FALL3D-8.0 test suite, which comprises a set of four real available test suite, form the basis for future improvements case studies that encapsulate the major features of the model; to FALL3D (version 8 or later) and also allow for model in- namely, the simulation of long-range fine volcanic ash dis- tercomparison studies. persal, volcanic SO2 dispersal, tephra fallout deposits and the dispersal and deposition of radionuclides. The first two test suite cases (i.e. the June 2011 Puyehue-Cordón Caulle ash cloud and the June 2019 Raikoke SO2 cloud) are val- 1 Introduction idated against geostationary satellite retrievals and demon- strate the new FALL3D data insertion scheme. The metrics FALL3D-8.0 is the latest major version release of FALL3D used to validate the volcanic ash and SO2 simulations are the (Costa et al., 2006; Folch et al., 2009), an open-source code structure, amplitude and location (SAL) metric and the figure with a 15-yearC track record and a growing number of of merit in space (FMS). The other two test suite cases (i.e. users in the volcanological and atmospheric science com- the February 2013 Mt. Etna ash cloud and associated tephra munities. A companion paper (Folch et al., 2020) details fallout deposit, and the dispersal of radionuclides resulting the physics and the novel numerical implementation of the from the 1986 Chernobyl nuclear accident) are validated with code, which has been redesigned and rewritten from scratch scattered ground-based observations of deposit load and lo- in the framework of the EU Centre of Excellence for Ex- cal particle grain size distributions and with measurements ascale in Solid Earth (ChEESE). From the point of view from the Radioactivity Environmental Monitoring database. of model physics, a relevant improvement in the new ver- For validation of tephra deposit loads and radionuclides, we sion (v8.x) has been the generalisation of the code to deal use two variants of the normalised root-mean-square error with atmospheric species other than tephra including other metric. We find that FALL3D-8.0 simulations initialised with types of particles (e.g. mineral dust), gases and radionuclides data insertion consistently improve agreement with satellite (see Table 3 in Folch et al., 2020, for details). These differ- retrievals at all lead times up to 48 h for both volcanic ash ent categories and subcategories of species can be simulated and SO2 simulations. In general, SAL scores lower than 1.5 using independent sets of bins that allow for dedicated pa- Published by Copernicus Publications on behalf of the European Geosciences Union. 410 A. T. Prata et al.: FALL3D-8.0 – Part 2 rameterisations for physics, emissions (source terms) and in- ically passed when substantial model updates in the master teractions among bins (e.g. aggregation, chemical reactions, branch of the code are committed to the repository and, es- radioactive decay). In FALL3D, “tephra” species are subdi- sentially, contain idealised 1-D or 2-D cases with known an- vided into four subcategories, depending on particle diame- alytical solutions for verification purposes (see, e.g. Figs. 3, ter, d (Folch et al., 2020): (i) fine ash (d ≤ 64µm), (ii) coarse 4 and 5 in Folch et al., 2020). In contrast, the test suite in- ash (64 µm < d ≤ 2 mm), (iii) lapilli (2 mm < d ≤ 64 mm) cludes larger-size, real-case simulations aimed at model vali- and (iv) bomb (d > 64 mm). In terms of model performance, dation. Model users can download the public test suite repos- the new model version contains a much more accurate and itory files to run the model and to check whether it has been less diffusive solver, as well as a better memory management properly installed and configured on their local machines. and parallelisation strategy that notably outperforms the scal- This paper presents the four cases from the FALL3D-8.0 test ability and the computing times of the precedent code ver- suite listed in Table1; namely, Puyehue-2011 (simulation of sion (v7.x) (Folch et al., 2020). This paper complements the the June 2011 Puyehue-Cordón Caulle ash cloud), Raikoke- Folch et al.(2020) companion paper by presenting a detailed 2019 (simulation of the June 2019 Raikoke SO2 cloud), set of validation examples, all included in the new test suite Etna-2013 (simulation of the 23 February 2013 Mt. Etna of the code. This paper also contains some novel aspects re- ash cloud and related tephra fallout deposit) and Chernobyl- garding geostationary satellite detection and retrieval of vol- 1986 (simulation of the dispersal and deposition of radionu- canic ash (AppendixA) and SO 2 (AppendixB), as well as the clides resulting from the April 1986 Chernobyl nuclear acci- FALL3D-8.0 data insertion methodology. To furnish the ini- dent). Note that the names of the validation cases are shown tial model condition required for data insertion, the satellite in bold throughout this paper. The FALL3D-8.0 test suite retrievals are collocated with lidar measurements of cloud- repository contains independent folders (one per validation top height and thickness from the CALIPSO (Cloud-Aerosol case). All case folders have the same subfolder structure: Lidar and Infrared Pathfinder Satellite Observation) platform – InputFiles. This subfolder contains all the necessary in- (Winker et al., 2009). The data insertion scheme is a prelim- put files to run the case. The only exception is mete- inary step towards model data assimilation using ensembles, orological data because these typically involve substan- a novel functionality currently under development. tially large files (tens of gigabytes) that make the storage The paper is organised as follows. Section2 provides an and the transfer to/from the GitLab public repository un- overview of the model test suite, detailing the file structure practical/unfeasible. and contents for each of the four case studies considered for validating and testing FALL3D-8.0. Section3 provides a de- – Utils/Meteo. This subfolder contains all the necessary scription of each of the events making up the test suite, which files to obtain meteorological data depending on the me- include simulations of the June 2011 Puyehue-Cordón Caulle teorological driver (for possible options, see Table 12 ash cloud, the June 2019 Raikoke SO2 cloud, the February in Folch et al., 2020). For global datasets such as the 2013 Mt. Etna ash cloud and associated tephra fallout de- ERA5 dataset (Copernicus Climate Change Service, posit, and the dispersal of radionuclides resulting from the 2017) used in the Puyehue-2011 case, Python and shell 1986 Chernobyl nuclear accident. The datasets used for val- scripts are provided so that the user can download and idation and the FALL3D model configurations used in each merge meteorological data consistently with the SetDbs case are also contained in Sect.3. Section4 describes the val- pre-process task (see Sect. 5 in Folch et al., 2020). For idation metrics, which include the structure, amplitude and mesoscale datasets (e.g. the WRF-ARW dataset used in location (SAL) metric (Wernli et al., 2008) and the figure of the Etna-2013 case), the subfolder includes the meteo- merit in space (FMS; Galmarini et al., 2010; Wilkins et al., rological model “namelists” and scripts to download the 2016) to quantitatively compare model results with satellite global data driving the corresponding mesoscale model retrievals, along with the root-mean-square error (RMSE) for simulation. validation of the ground deposit simulations of tephra and ra- – Utils/Validation. This subfolder contains all the nec- dionuclides. Section5 presents a detailed discussion of the essary files to validate the FALL3D-8.0 model execu- validation results for the four test suite cases. Finally, Sect.6 tion results, including a file with the expected validation presents the conclusions of the paper and outlines the next metric results. steps in terms of model development and applications. Tables2 and3 list the files in the test suite for the Puyehue-2011 (Sect. 3.1) and the Etna-2013 (Sect. 3.3) 2 The FALL3D-8.0 test suite cases, respectively. The Raikoke-2019 (Sect. 3.2) and Chernobyl-1986 (Sect. 3.4) filenames look very similar to FALL3D-8.0 includes both a benchmark and a test suite. The the Puyehue-2011 and Etna-2013 filenames and are not ex- benchmark suite consists of a series of non-public small-case plicitly shown here.