[A-AS06] Advances in Tropical Cyclone Research: Past, Present, and Future

[AAS06-P02] Experiment on Assimilating In-Situ Observation Data from a Wave Glider Just Below a Typhoon

*Yusuke Umemiya¹, Kosuke Ito^2,3, Fumiaki Moriyama², Naoko Kosaka¹, Tsuneko Kura¹, Masaki Hisada¹, Kazuhisa Tsuboki^2,4, Masaki Satoh^2,5, Shuichi Mori^2,6, Hitoshi Tamura^2,7, Fudeyasu Hironori² (1.NTT Space Environment and Energy Laboratories, 2.Yokohama National University, Typhoon Science and Technology Research Center, 3.Disaster Prevention Research Institute, Kyoto University, 4.Institute for Space-Earth Environmental Research, Nagoya University, 5.Atmosphere and Ocean Research Institute, The University of Tokyo., 6.Japan Agency for Marine-Earth Science and Technology., 7.National Institute of Maritime, Port and Aviation Technology)

Keywords:Typhoon forecasting, Data assimilation, Typhoon observation, Wave Glider, Weather Research and Forecasting Model (WRF), Three-dimensional variational data assimilation system

In recent years, typhoons have intensified, causing significant damages to social infrastructure. An early response, leveraged by improved accuracy in typhoon forecasts, is key to prevent and mitigate these damages. This demands augmentation of actual measurement data and effective utilization of the observation data [1]. In this regard, we have been exploring methods to assimilate our typhoon observation data into numerical weather prediction models, aiming to establish an optimal method for enhancing typhoon forecasting. In this study, we conducted assimilation experiments in some settings to evaluate the impact of background errors.
In this experiment, we assessed the impact of various background errors on the assimilation results using our sea level pressure (SLP) and sea level temperature (SLT) observations. The assimilation was timed for 12 UTC on Aug. 31, 2022, when the Wave Glider was nearest to the typhoon's center. This timing corresponds with the time resolution of NCEP GFS/FNL, which provides data every six hours for initial and boundary conditions. At the time, the Wave Glider was approximately 23 km from the typhoon's center. Our observations were 951.226 hPa for SLP and 302.257 K for SST. Both were averaged every five minutes using a water-resistant Baro Troll (by In-situ Inc.) mounted 50 cm above the Wave Glider's mast [2].
The assimilation experiment utilizes WRFDA for 3DVAR, included in the Weather Research & Forecasting Model (WRF, ver. 4.5.1), and two methods implemented in WRF (CV3, CV5) are used to prepare the background errors. CV3 refers to a global-scale background error, applicable universally across any region; however, it cannot accommodate region-specific or seasonal variations [3]. Contrarily, CV5 generates a background error specific to the experiment's region and period, thus expected to more precisely represent the characteristics of the targeted phenomenon.
Initial values are derived from the NCEP GFS/FNL. We set the initial values in the WRF for 06 UTC on Aug. 31, 2022, and six hours later, at 12 UTC, we assimilate the observational data using each type of background error and further generate a 24-hour forecast. As for the background errors, the ones prepared in WRF are used for CV3, whereas the NMC method is applied to multiple initial values for 12-hour (T+12) and 24-hour (T+24) forecasts in line with the WRF user guide for CV5 [4]. The initial values are created every 12 hours.
As a result, the analysis moved closer to the observational values due to the assimilation in both experiments at the assimilation time. In the case of CV5, the adjustment made by assimilation was smaller compared to CV3, showing a value closer to the control run. Conversely, with CV3, the values significantly deviated from the control run even after the assimilation time, indicating that the effect of corrections was too dispersed, and that the intensity was overestimated during the forecast period. Considering the structure of the typhoon, it appears reasonable to adopt CV5 as the background error, although further verification might be required. In future work, we plan to enhance our 3D variational method to perform 1-hour typhoon assimilations by incorporating observational data and applying it to the results of CV5.

[1] K. Ito, et al., “Forecasting a Large Number of Tropical Cyclone Intensities around Japan Using a High-Resolution Atmosphere–Ocean Coupled Model,” WAF, 30(3), pp.793-808, 2015.
[2] N. Kosaka, et al., “Synchronous observations of atmosphere and ocean directly under typhoons using autonomous surface vehicles,” SOLA, 19, 116−125, 2023. doi:10.2151/sola.2023-016
[3] Y. Umemiya, et al., “Impacts of Assimilating Sea Surface Observation directly under Super Typhoon Hinnamnor (2022) in the Northwest Pacific,” IWTRC2023.
[4] National Center for Atmospheric Research, “Chapter 6: WRF Data Assimilation (WRFDA)” https://www2.mmm.ucar.edu/wrf/users/docs/user_guide_v4/v4.4/users_guide_chap6.html, accessed on February 7, 2023.