.. _Section_ModelingMethodology: ******************** Modeling Methodology ******************** SAM Model ========= Simulations for LASSO-ENA use the SAM LES model version 6.10.8 with a number of custom modifications and bug fixes :cite:p:`Khairoutdinov2003`. The SAM dynamical core is built around the anelastic set of equations, making it much more efficient than WRF with its compressible dynamical core :cite:p:`Wang2019`, which was used in the earlier LASSO scenarios. SAM's reliance on doubly periodic domains is not a limitation since the nested domain capability of WRF is unnecessary for LASSO-ENA. The specfic code used for LASSO-ENA can be obtained from the ``lasso_sam_sbm`` repository at https://code.arm.gov/lasso/lasso-ena-codes/lasso_sam_sbm. We use git branches to set certain hard-wired parameters, such as turning the ice phase condensate on and off in the spectral-bin microphysics, branch names ``lasso_ena_ice`` and ``lasso_ena_noice``, respectively. The SAM model has been applied to various research applications for over twenty years, and many different versions exist. The heritage of the version we started from includes edits made for ice-phase and aerosol-cloud research at PNNL with a particular focus on the spectral-bin microphysics parameterization, and it was also used in :cite:t:`Zhang2017` for composite SGP shallow convection simulations. The primary model settings used for LASSO-ENA are listed in :numref:`TableParameterizations`. Most simulations employ spectral-bin microphysics :cite:p:`Khain2004,Fan2009`, while a smaller number use the Morrison bulk microphysics option :cite:p:`Morrison2009`. In general, the overcast cloud conditions are better simulated with the spectral-bin microphysics for our model setup. Surface fluxes are calculated online based on specified sea-surface temperature (SST) data obtained from the source of the large-scale forcing for the given simulation, which is either the ERA5 or MERRA-2 reanslysis as described below. .. table:: Important physics parameterizations and dynamics settings used for LASSO-ENA SAM simulations. :name: TableParameterizations +-----------------------+--------------------------------------+----------------------------------+ | Physics Type / | Scheme | References | | Parameter | Used | | +=======================+======================================+==================================+ | | Microphysics | | Variant of HUJI | | :cite:t:`Khain2004` | | | option 1: | | spectral bin | | :cite:t:`Fan2009` | | | spectral bin | | | +-----------------------+--------------------------------------+----------------------------------+ | | Microphysics | | Morrison | | :cite:t:`Morrison2009` | | | option 2: bulk | | | +-----------------------+--------------------------------------+----------------------------------+ | | Longwave and | | Rapid Radiation Transfer Model | | :cite:t:`Mlawer1997` | | | shortwave radiation | | for Global Climate Models | | :cite:t:`Clough2005` | | | | (RRTMG) LW and SW | | :cite:t:`Iacono2008` | +-----------------------+--------------------------------------+----------------------------------+ | | Subgrid-scale | | 1.5 order turbulent kinetic | | :cite:t:`Deardorff1980` | | | turbulence | | energy (TKE) | | +-----------------------+--------------------------------------+----------------------------------+ | | Surface layer | | Monin–Obukhov similarity theory | | :cite:t:`Bryan1996` | | | | | | | | | +-----------------------+--------------------------------------+----------------------------------+ | | Scalar advection | | Multidimensional positive definite | | :cite:t:`Smolarkiewicz1990` | | | | advection transport algorithm | | | | | (MPDATA) | | +-----------------------+--------------------------------------+----------------------------------+ Most of the simulations turn off ice-phase processes to conserve computational time. This has little to no impact on days with clouds only below the freezing level, which is very common for the closed-cell cases. The LASSO open-cell cases more often have ice-containing clouds and thus might have simulation biases due to our simplifying approach. A smaller number of cases have been simulated with both ice turned off and turned on to identify the impact of including ice processes in the spectral-bin parameterizaiton. Users requiring ice processes for cases not simulated that way for LASSO-ENA can use the input files we provide for the ice-free simulations to easily configure and run new simulations that include ice. .. figure:: images/lassoena_aerosolmode.png :name: FigAerosolSizeDist :align: center :alt: Measured aerosol size distributions :figclass: align-center Aerosol size distributions for the Aitken and accumulation modes as sampled at ENA using the SMPS and UHSAS instruments. The blue line is averaged for the period 2015–2022 when only the UHSAS was available. The red and green lines are for the period 2022–2023 and use both the SMPS and UHSAS. The green line is a subsample of the full period that only includes days with wind from the north. The gray line is the DYCOMS-II aerosol assumption, included for reference since it was used for some early LASSO-ENA simulations. The spectral-bin microphysics requires several assumptions, of which aerosol characteristics are particularly important. The period included within LASSO-ENA has surface-based aerosol measurements from the ultra-high-sensitivity aerosol spectrometer (UHSAS) that provides particle size distributions in the 60-to-1000 nm diameter range, which is useful for quantifying accumulation mode aerosol :cite:p:`Uin2024,Uin2014UHSASData`. In 2022, ARM added a scanning mobility particle sizer (SMPS) at ENA that measures smaller particles down to 10 nm diameter :cite:p:`Singh2024,Kuang2022SMPSData`, which now enables quantifying both the Aitken and accumulation modes. Using data from these instruments, an analysis has been done to identify the average aerosol number and size distributions for the LASSO-ENA cases. The details vary depending on the days sampled, as shown in :numref:`FigAerosolSizeDist`, which compares the two aerosol modes observed for the period 2022–2023 (in red) with the subset of days with winds from the north plotted separately (in green). For reference, a longer time period is averaged from 2015–2022 when only the UHSAS data is available (in blue). The modes corresponding to the green lines was used as the baseline aerosol settings. These were then supplemented with two additional simulations, one that halved the aerosol number and the other that doubled it. In total, each case has three different simulations that permit an initial indication of the cloud sensitvity to aerosol, which varies between case dates. We refer to the low, medium, and high aerosol assumptions as "aer1", "aer2", and "aer3". :numref:`TableAerosolModes` lists the aerosol mode settings corresponding to these three assumptions. Additional assumptions in the spectral-bin microphysics include the following: 1. Aerosol concentration: whenever the total aerosol plus cloud/rain (same category in spectral-bin) number concentration falls below the initial setting, new aerosol particles are added to the aerosol bins (starting from the smallest bin) so that the total number equals the initial setting. 2. Ice nucleation: whenever the total ice-particle number concentration (cloud ice + snow + graupel) falls below a constant (1.0e-3 per L) new cloud ice particles are formed. .. table:: Aerosol modes used with spectral-bin microphysics. :name: TableAerosolModes :align: center +---------------------------------+-------+-------+-------+ | Aerosol setting: | aer1 | aer2 | aer3 | +=================================+=======+=======+=======+ | Mode 1 number (cm\ :sup:`-3`\ ) | 138 | 276 | 552 | +---------------------------------+-------+-------+-------+ | Mode 1 mean radius (μm) | 0.018 | 0.018 | 0.018 | +---------------------------------+-------+-------+-------+ | Mode 1 standard deviation | 1.53 | 1.53 | 1.53 | +---------------------------------+-------+-------+-------+ | Mode 2 number (cm\ :sup:`-3`\ ) | 140.5 | 281 | 562 | +---------------------------------+-------+-------+-------+ | Mode 2 mean radius (μm) | 0.066 | 0.066 | 0.066 | +---------------------------------+-------+-------+-------+ | Mode 2 standard deviation | 1.78 | 1.78 | 1.78 | +---------------------------------+-------+-------+-------+ Domain Configuration ==================== LASSO-ENA simulations use a doubly periodic domain with 100-m horizontal grid spacing. The horizontal domain size has two options: the first is 25.6 km wide with 256 columns, which we refer to as the "small" domain; the second is 102.8 km with 1028 columns in each direction, which we refer to as the "large" domain. Vertical grid spacing is 25 m up to 6012.5 m, above which a stretched grid is used with a model top at 8087.5 m, for a total of 260 levels. The top 30% of the height employs a sponge layer to minimize reflections from the model top. This periodic modeling approach implies that each column in the simulation is statistically identical to every other colomn. We treat the lower surface as a uniform ocean with homogeneous SST imposed throughout the domain. We exclude land within the domain and treat it as a flat water surface, i.e., for Graciosa Island. The intent is to produce simulations representative of the surrounding ocean conditions. Input Data ========== Two variants are generated for almost all simulations based on using different reanalyses for the initial conditions and large-scale forcings. The two options are ERA5 :cite:p:`Hersbach2020,Copernicus2023ERA5Data` and MERRA-2 :cite:p:`Gelaro2017,GMAO2015MERRA2Data3dAsm,GMAO2015MERRA2Data2dFlx,GMAO2015MERRA2Data2dSlv`, both obtained on their model levels. Prior to downloading, the ERA5 data are resampled from their native horizontal grid to a 0.25° latitude-longitude grid, whereas MERRA-2 data are resampled to 0.625° longitude by 0.5° latitude. The gridded reanalyses are sampled for a 5° region cenetered on ENA to produce input data for SAM. Horizontal averages on height levels are done to produce large-scale forcing profiles based on instantaneous snapshots; this is done hourly for ERA5 and every three hours for MERRA-2. This is also done to generate the initial conditions for the atmosphere. SSTs are averaged hourly for both ERA5 and MERRA-2 over the same region for the lower boundary condition of SAM. These data are provided to users in the ``samin`` tar files available for each simulation.