16:00 〜 16:15
[MGI29-09] Controllability of Extreme Events with the Lorenz-96 Model
★Invited Papers
The successful development of numerical weather prediction (NWP) helps better preparedness for extreme weather events. Weather modifications have also been explored, for example, when enhancing rainfalls by cloud seeding [1]. However, it is generally believed that the tremendous energy involved in extreme events prevents any attempt of human interventions to avoid or to control their occurrences. In this study, we investigate the controllability of chaotic dynamical systems by using small perturbations to generate powerful effects and prevent extreme events. The high sensitivity to initial conditions would ultimately lead to modifications of extreme weather events with infinitesimal perturbations.
In a preliminary experiment with the Lorenz-63 model, it has been shown that the Lorenz attractor stays in a regime, without shifting to the other, if small perturbations are suitably applied to the nature run [2]. In the present study, we extend this approach to the Lorenz-96 40-variable model [3], as a simple prototype of weather system on some latitude circle, and design a new control simulation experiment (CSE). In this framework, our goal is to avoid extreme values in a 100-year run, with these extreme values defined by the 200 biggest values of the nature run over 100 years.
Given any initial time tm, we run a T days forecast with the analysis ensemble given by the LETKF. A control is then applied to the nature run if at least one ensemble member shows an extreme value in the forecast. This control consists in applying a perturbation signal every 0.01 time steps (0.05 = 6 hours) before the next observation, see Fig. 1. These perturbations are obtained by rescaling the differences between two ensemble members. This forecast-control-analysis process is repeated each 6-hour over the 100 years run. A comparison of histograms between the states values with and without control is provided in Fig. 2. We also study the efficiency of the control with various choices of the amplitude of the perturbation, the forecast length, the localized perturbation and the partial observations. It is expected that this control method can be applied to more complicated weather systems and to other chaotic dynamical systems not limited to NWP.
[1] Flossmann, A. I., et al., 2019: Bull. Amer. Meteorol. Soc., 100, 1465-1480.
[2] Miyoshi, T., Sun, Q., 2021: Nonlin. Proc. Geophys. Discuss. [in press].
[3] Lorenz, E., 1996: Seminar on Predictability, Vol. I, ECMWF.
In a preliminary experiment with the Lorenz-63 model, it has been shown that the Lorenz attractor stays in a regime, without shifting to the other, if small perturbations are suitably applied to the nature run [2]. In the present study, we extend this approach to the Lorenz-96 40-variable model [3], as a simple prototype of weather system on some latitude circle, and design a new control simulation experiment (CSE). In this framework, our goal is to avoid extreme values in a 100-year run, with these extreme values defined by the 200 biggest values of the nature run over 100 years.
Given any initial time tm, we run a T days forecast with the analysis ensemble given by the LETKF. A control is then applied to the nature run if at least one ensemble member shows an extreme value in the forecast. This control consists in applying a perturbation signal every 0.01 time steps (0.05 = 6 hours) before the next observation, see Fig. 1. These perturbations are obtained by rescaling the differences between two ensemble members. This forecast-control-analysis process is repeated each 6-hour over the 100 years run. A comparison of histograms between the states values with and without control is provided in Fig. 2. We also study the efficiency of the control with various choices of the amplitude of the perturbation, the forecast length, the localized perturbation and the partial observations. It is expected that this control method can be applied to more complicated weather systems and to other chaotic dynamical systems not limited to NWP.
[1] Flossmann, A. I., et al., 2019: Bull. Amer. Meteorol. Soc., 100, 1465-1480.
[2] Miyoshi, T., Sun, Q., 2021: Nonlin. Proc. Geophys. Discuss. [in press].
[3] Lorenz, E., 1996: Seminar on Predictability, Vol. I, ECMWF.