CNOP-based sensitive areas identification for tropical cyclone adaptive observations with PCAGA method

Abstract

In this paper, conditional nonlinear optimal perturbation (CNOP) was investigated to identify sensitive areas for tropical cyclone adaptive observations with principal component analysis based genetic algorithm (PCAGA) method and two tropical cyclones, Fitow (2013) and Matmo (2014), were studied with a 120 km resolution using the fifth-generation Mesoscale Model (MM5). To verify the effectiveness of PCAGA method, CNOPs were also calculated by an adjoint-based method as a benchmark for comparison on patterns, energies, and vertical distributions of temperatures. Comparing with the benchmark, the CNOPs obtained from PCAGA had similar patterns for Fitow and a little different for Matmo; the vertically integrated energies were located closer to the verification areas and the initial tropical cyclones. Experimental results also presented that the CNOPs of PCAGA had a more positive impact on the forecast improvement, which gained from the reductions of the CNOPs in the whole domain containing sensitive areas. Furthermore, the PCAGA program was executed 40 times for each case and all the averages of benefits were larger than the benchmark. This also proved the validity and stability of the PCAGA method. All results showed that the PCAGA method could approximately solve CNOP of complicated models without computing adjoint models, and obtain more benefits of reducing the CNOPs in the whole domain.