Have questions about these diagnostics? Please direct questions to Kyle Nardi at kmn182@psu.edu.

This website contains results from a parameter sensitivity analysis (MM5) for 17 tunable parameters in the Community Atmosphere Model version 6 (CAM6). 13 of these input parameters are from an experimental version of the parameterization of boundary layer turbulence in CAM6 (CLUBBX). The other four parameters are from the microphysics parameterization (PUMAS). The sensitivity analysis is the Morris One-at-a-Time (MOAT) method, which is a computationally-efficient procedure to screen input parameters with the goal of identifying a group of input parameters that are more influential relative to others. We run 180 CAM6 simulations with unique combinations of input parameters. From one configuration to the next, we change the value of only one input parameter and keep all others constant. This is done until all 17 input parameters have been perturbed, and then the process is repeated for nine other initial combinations of input parameters. Simulations are 3-day, 1-degree, 58-level hindcasts initialized with ERA5 atmospheric conditions, NOAA SST data, and a 12-month land spinup using the Community Land Model (CLM), which is driven by prescribed atmospheric conditions from ERA5. We repeat this process for a total of 24 random initialization dates (2 for each month) between 2010 and 2020. This results in a total of 4320 3-day CAM6 simulations.

The heatmaps below show two key sensitivity metrics: mu-star and monotonicity. Mu-star is the average magnitude of the response of the output metric (horizontal axis) to a perturbation in the input parameter (vertical axis). The mu-star heatmap shows, for a given output metric, a ranking of mu-star for each input parameter. Darker colors (higher rankings) imply a relatively large response of the output metric to perturbing the input parameter. The monotonicity is defined as the frequency with which increasing the input parameter increases the output metric. Dark reds (blues) imply that increasing the input parameter frequently increases (decreases) the output metric. We seek input parameters that produce 1) a high-magnitude response in the output metric (high mu-star) and 2) a consistent directional response (either a very low or very high monotonicity). For more information about these metrics, refer to Nardi et al. (2022).

For global plots, the output from each of the 180 unique configurations is averaged over all 24 initializations and domain-averaged over all latitudes and longitudes. Unless otherwise stated, spatial averaging employs a cosine-latitude weighting.

The tropics region covers all longitudes and latitudes from 30 degrees S to 30 degrees N.

The storm track region covers all longitudes and latitudes from 70 to 40 degrees S.

The plots below provide information about how multiple output metrics change when perturbing a particular input parameter. When increasing an input parameter and considering the directional response of two separate output metrics, there are four possible outcomes: 1) both parameters increase, 2) the first parameter increases but the second decreases, 3) the first parameter decreases but the second increases, and 4) both parameters decrease. The plots below show the probability of each individual outcome occurring for different combinations of 2 output metrics. The Gleckler-type plot synthesizes these four proababilities into one heatmap, where each quadrant represents the probability of each outcome. Another plot shows the average mu-star ranking for linear combinations of the two metrics. To calculate mu-star, we calculate the value of each output metric and scale each metric by the domain-average value of that metric over all simulations. We then average these two scaled values to get a final combined metric. Mu-star is then calculated for this combined metric, averaged over the domain of interest. The quad plots below display the probability of each directional response so that each input parameter's marker appears within the quadrant representing the most likely joint directional response. The markers are sized based on the input parameter's mu-star ranking, where larger markers imply a higher mu-star ranking.

- T2m + CAPE
- LWCF + SWCF
- Total Cloud + OLR
- Total Cloud + Stress
- SLP + UBOT
- upwp + Stress
- Stress + SWCF
- upwp + TKE
- Stress + UBOT

The plots below compare average output for configurations before and after increasing a given input parameter value. For each input parameter over 10 unique initial combinations of input parameters and 24 randomly-defined initial atmospheric/oceanic/land conditions, there are 240 instances where that input parameter is increased with all else constant. Here, we compare the average output for the 240 simulations before increasing the parameter and after increasing the parameter. In the 2D difference plots, reds (blues) imply an increase (decrease) in the output when increasing the input parameter. At the bottom of these plots, there is a comparison of the globally-averaged value of the output before and after increasing the input parameter (cosine-latitude weighting applied).

The plots below show the zonally-averaged difference in the 2D fields above. The differences in the zonal average plots are percent differences relative to the average value at each level before increasing the input parameter.

The plots below show vertical profiles of various domain-averaged quantities for the 240 configurations before and after incrasing the given input parameter. The thick lines denote the median quantity calculated at each level over the 240 configurations before/after increasing the parameter. The shading denotes the 25th-75th percentiles at each level over the 240 configurations.

The ENSO region covers latitudes from 5 degrees S to 5 degrees N and longitudes from 205 to 215 degrees E.

The South American Stratocumulus region covers latitudes from 20 to 10 degrees S and longitudes from 270 to 280 degrees E.

The Southern Ocean region covers latitudes from 60 to 50 degrees S and longitudes from 140 to 150 degrees E.

The storm track band covers latitudes from 70 to 40 degrees S and all longitudes.

The tropics band covers latitudes from 30 degrees S to 30 degrees N and all longitudes.

This region covers a box to the east of the Hawaiian Islands, from 18 to 23 degrees N and 205 to 210 degrees E.

This region covers the domain of the DYCOMS campaign off the coast of California, from 30 to 33 degrees N and 236 to 240 degrees E.

The plots below show domain-averaged vertical profiles of the terms in the CLUBB(X) prognostic momentum flux budget (upwp/vpwp) and the CLUBB vertical wind variance budget (wp2). Solid lines are the profiles averaged over all 240 simulations before increasing the given input parameter. The dashed lines are the profiles averaged over all 240 simulations after increasing the given input parameter. For more information about the individual budget terms, refer to Larson (2017).

The movies below show the evolution of domain-averaged vertical profiles from an experimental version of CM1 with CLUBB(X). CM1-CLUBBX is run with the sheared boundary layer test case and moisture removed. Horizontal grid spacing is 2 km, with vertical grid spacing of 10 m. Simulations are run for 6 hours, with movie frames every 10 minutes. The solid line denotes the profile when the input parameter is at a lower value, while the dashed line denotes the profile when the input parameter is at a higher value.

: Coefficient in near-surface term in CLUBB(X) formulation of eddy dissipation (inverse of eddy turnover timescale). Increasing this term increases eddy dissipation at the lowest model levels (where inverse of z-height coordinate is high).**C_invrs_tau_sfc**: Coefficient in shear term in CLUBB(X) formulation of eddy dissipation. Increasing this term increases eddy dissipation in the prescence of vertical wind shear.**C_invrs_tau_shear**: Coefficient in static stability term in CLUBB(X) formulation of eddy dissipation. Increasing this term increases eddy dissipation in a stably-stratified environment (e.g., a high value of the Brunt-Vaisala frequency).**C_invrs_tau_N2**: Coefficient in the formulation of dissipation that appears in the budget of scalar variances. Acts to decrease scalar variances in a stably-stratified environment.**C_invrs_tau_N2_xp2**: Coefficient in the formulation of damping that appears in the budget of vertical velocity variance (wp2). Acts to damp turbulent kinetic energy in a stably-stratified environment.**C_invrs_tau_N2_wp2**: Controls width of the Gaussian for vertical velocity. Perturbing this parameter affects the skewness of the vertical velocity PDF. Increasing this parameter increases skewness and makes clouds more cumulus and less stratocumulus.**gamma**: Controls width of the Gaussian for vertical velocity. Perturbing this parameter affects the skewness of the vertical velocity PDF. Increasing this parameter increases skewness and makes clouds more cumulus and less stratocumulus.**gamma_coefb**: Low buoyancy damping coefficient for w-skewness (wp3). Increasing this parameter decreases skewness and makes clouds more stratocumulus and less cumulus.**clubb_C11**: Newtonian damping coefficient for w-skewness (wp3). Increasing this parameter decreases skewness and makes clouds more stratocumulus and less cumulus.**clubb_C8**: Coefficient in the skewness of scalar quantities like potential temperature and moisture.**clubb_beta**: Coefficient in a pressure term that damps turbulent production of momentum flux in the presence of vertical wind shear. Increasing this parameter increases the damping of turbulent production, thus reducing the effects of turbulent production and reducing momentum flux. Also acts as a coefficient to damp the accumulation term in the budget of vertical wind variance (wp2). Increasing this term reduces the time tendency of wp2.**c_uu_shr**: Coefficient in a pressure term that damps buoyant production of vertical wind variance (wp2). Increasing this parameter acts to reduce wp2.**c_uu_buoy**: Coefficient multiplied by the friction velocity to calculate horizontal wind variances (up2 and vp2) at the surface. Increasing this parameter acts to increase the variances at the surface.**clubb_up2_sfc_coef**: Size threshold for the autoconversion of cloud ice particles to snow. Increasing this parameter will make autoconversion less frequent and produce more clouds and greater longwave cloud forcing.**micro_mg_dcs**: Multiplicative factor for calculating the ice fall velocity. Increasing this parameter increases the ice fall speed.**micro_mg_vtrmi_factor**: Multiplicative factor for calculating evaporation of precipitation. Increasing this parameter increases evaporation of precipitation.**micro_mg_pre_fact**: Enhancement factor for calculating accretion (growth of ice hydrometeors via collision with supercooled liquid water droplets). Increasing this parameter increases accretion and hydrometeor growth.**micro_mg_accre_enhan_fact**