Ensemble Sensitivity Analysis
Ensemble sensitivity analysis (ESA) possesses the unique ability to highlight forecast sensitivities to differences in the atmospheric flow through simple post-processing of ensemble output. It highlights the relationship between variables at different times as well as how changes in the earlier-time variables may propagate in time to impact the forecast. The analysis is most accurate when relationships between forecast states are linear, i.e. synoptic scale applications, but more recently it has been explored on mesoscale forecasts where linear assumptions may be constrained.
Our research group is exploring the utility of ESA applied to forecasts of convection initiation (CI) in the 12-24 hr forecast timeframe. Specifically, where bimodal forecast distributions of CI occur. Two cases of dryline initiation were investigated in April 2012 and May 2013 and ESA was found to sufficiently identify differences that led to forecast divergence of CI and identify specific features, e.g. dryline position, cold pools, upper-level forcing, that may inhibit or contribute to initiation. For surface variables, statistically-significant sensitivity was observed primarily within the 12 hours prior to initiation. The time period for statistically-significant sensitivities was expanded for variables aloft, where sensitivities to trough position and height fields were temporally consistent for the entire forecast period. The sensitivity fields were validated by subsetting CI and non-CI members at a 20-dBZ threshold and differencing them to verify the sensitive regions highlighted differences between ensemble members.
Future work regarding ESA techniques will dive into finer-resolution simulations, specifically those looking to replicate supercell and mesocyclone structure and how we can attempt to use linear assumptions in a highly-nonlinear model environment. Additional applications include using ESA to adaptively sample the atmosphere in sensitive regions that are also characterized by large uncertainty. This work was published in Monthly Weather Review (link).
Hill, A. J., C. C. Weiss, and B. C. Ancell, 2016: Ensemble sensitivity analysis for mesoscale forecasts of dryline convection initiation. Monthly Weather Review, 144, 4161-4182. doi:10.1175/MWR-D-15-0338.1
Our research group is exploring the utility of ESA applied to forecasts of convection initiation (CI) in the 12-24 hr forecast timeframe. Specifically, where bimodal forecast distributions of CI occur. Two cases of dryline initiation were investigated in April 2012 and May 2013 and ESA was found to sufficiently identify differences that led to forecast divergence of CI and identify specific features, e.g. dryline position, cold pools, upper-level forcing, that may inhibit or contribute to initiation. For surface variables, statistically-significant sensitivity was observed primarily within the 12 hours prior to initiation. The time period for statistically-significant sensitivities was expanded for variables aloft, where sensitivities to trough position and height fields were temporally consistent for the entire forecast period. The sensitivity fields were validated by subsetting CI and non-CI members at a 20-dBZ threshold and differencing them to verify the sensitive regions highlighted differences between ensemble members.
Future work regarding ESA techniques will dive into finer-resolution simulations, specifically those looking to replicate supercell and mesocyclone structure and how we can attempt to use linear assumptions in a highly-nonlinear model environment. Additional applications include using ESA to adaptively sample the atmosphere in sensitive regions that are also characterized by large uncertainty. This work was published in Monthly Weather Review (link).
Hill, A. J., C. C. Weiss, and B. C. Ancell, 2016: Ensemble sensitivity analysis for mesoscale forecasts of dryline convection initiation. Monthly Weather Review, 144, 4161-4182. doi:10.1175/MWR-D-15-0338.1
