1 James D. Doyle and Clark Amerault Naval Research Laboratory, Monterey, CA James D. Doyle and Clark Amerault Naval Research Laboratory, Monterey, CA Sensitivity and Predictability of Atmospheric Rivers and Cyclones: Preliminary CalWater Results UW CIMSS TPW Feb 14, 2014

2 Adjoint allows for the mathematically rigorous calculation of forecast sensitivity of a response function to changes in initial state COAMPS Adjoint Model Sensitivity of response function (J) at time t n to the state at time t 0 COAMPS ® Moist Adjoint Model Dynamics: nonhydrostatic, nested Physics: PBL, surface flux, microphysics Response Function, J: - kinetic energy in a box (1 km deep) Resolution:  x=45 km, 15 km (36 & 24h) Calwater Demonstration Cases 8-10 Feb 2014: nonhydrostatic, nested Feb 2014: nonhydrostatic, nested 45 km 15 km KE Response Function (location can vary) M T is the adjoint or transpose of the tangent linear model

3 Nested Adjoint Sensitivity 36-h Sensitivity (Analysis) 12Z 8 February 2014 (Final Time 00Z 10 February) Geopotential Heights and Water Vapor 850 hPa 36-h forecast sensitivity calculations for 12Z 8 Feb (valid at 00Z 10 Feb). Moisture sensitivity is a maximum along the atmospheric river and just to the north of the AR. Moisture sensitivity is 2X larger than the temperature and 3X larger than the wind component sensitivity (assume analysis errors are ~1 K, 1 m/s, 1 g/kg) 36-h Water Vapor Sensitivity 850 hPa (q v shown in gray)

4 Nested Adjoint Sensitivity 24-h Sensitivity (Analysis) 00Z 9 February 2014 (Final Time 00Z 10 February) Geopotential Heights and Water Vapor 850 hPa 24-h forecast sensitivity calculations for 00Z 9 Feb. (valid at 00Z 10 Feb.) Moisture sensitivity is more confined along the AR and closer to CA. 24-h moisture sensitivity is weaker than the 36-h sensitivity. 24-h Water Vapor Sensitivity 850 hPa (q v shown in gray)

5 Nested Adjoint Sensitivity 36-h Sensitivity (Analysis) 12Z 8 February 2014 (Final Time 00Z 10 February) PV Sensitivity (Adjoint Optimal Pert.) Geopotential Heights (850 hPa) PV adjoint optimal perturbation (proxy for PV sensitivity) shows two maxima near shortwave trough at 850-hPa Sea surface temperature sensitivity (at initial time) is a maximum along AR and just to the north coincident with the shortwave trough. 36-h SST Sensitivity SST (color)

6 Nested Adjoint Sensitivity 36-h Sensitivity (Analysis) 00Z 13 February 2014 (Final Time 12Z 14 February) Geopotential Heights and Water Vapor 850 hPa 36-h forecast sensitivity calculations for 00Z 13 Feb. (valid at 00Z 14 Feb.) Moisture sensitivity is strongest along AR axis; located > 2000 km upstream Moisture sensitivity is 2-3X larger than for 8 Feb. case. 24-h Water Vapor Sensitivity 850 hPa (q v shown in gray)

7 PV Sensitivity (Adjoint Optimal Pert.) Geopotential Heights (850 hPa) PV adjoint optimal perturbation maxima near 850-hPa shortwave trough Sea surface temperature sensitivity is a maximum north of the AR. PV and SST sensitivity is larger for the 13 Feb. case than for 8 Feb. 36-h SST Sensitivity SST (color) Nested Adjoint Sensitivity 36-h Sensitivity (Analysis) 00Z 13 February 2014 (Final Time 12Z 14 February)

8 Nested Adjoint Sensitivity 36-h Evolved Optimal Perturbations (Grid 2) 700-mb Heights and Optimal U Perturbation Evolved u-wind component and qv optimal perturbations grow rapidly; after 36 h, u-wind component grows from 1 to 20 m s -1. Analysis errors that project on to the initial sensitivity regions will rapidly contaminate the forecast, in particular water vapor flux and convergence.

9 U-Wind Optimal Perturbation 850 hPa Water Vapor Optimal Perturbation 850 hPa Nested Adjoint Sensitivity 36-h Evolved Optimal Perturbations (Grid 2) 12Z 14 February m/s ● Evolved u-wind component and qv optimal perturbations grow rapidly; after 36 h, u-wind component grows from 1 to 20 m s -1. Analysis errors that project on to the initial sensitivity regions will rapidly contaminate the forecast, in particular water vapor flux and convergence.

10 Sensitivity to moisture (often within ARs) is most important for many of the cyclones we have examined, despite substantial differences in structure & evolution of the storms. Initial condition sensitivity is a maximum: within AR (14 Feb) and in AR & to north (8 Feb) near dynamically active regions. Mesoscale predictability of the landfalling storms and precipitation is limited by very rapid growth in some cases. -Growth in 13 Feb case more rapid than 8 Feb event. Additional observations in sensitive regions could be beneficial for NWP model initial conditions and forecasts CalWater2 plans: -Possibly run adjoint in real time to support field observing strategies. -Sensitivity calculations for SSTs could be useful for AXBT observing strategies in real time (Sue Chen will be involved with this). Summary and Plans