In this study, we added the post-processing module of the Ensemble Transform Sensitivity (ETS) to identify
regions sensitive to adaptive observations of different variables at different levels for typhoon forecasts. We
selected five cases between 2015 and 2020 over the Western North Pacific (WNP) using 50 ensemble forecasts
from the European Center provided by the THORPEX Interactive Grand Global Ensemble (TIGGE) portal.
Furthermore, intercomparison experiments were performed to explore the impact of adaptive observational data
assimilation with conventional observations and Himawari-8 data over the sensitive (SEN), non-sensitive
(NOSEN), and all studied regions (ALL) on typhoon forecasts using the Gridpoint Statistical Interpolation
(GSI) assimilation system of the National Centers for Environmental Prediction (NCEP) of United States. The
results show that the assimilation of less observational data in the SEN areas can achieve the forecast effect of
assimilation in ALL regions, the intensity error can be reduced by approximately 15%, and the track error is
reduced in some cases. Finally, detailed analysis of the circulation field and flux perspectives for Typhoon
Lekima, which occurred in August 2019 over the WNP, confirmed our conclusion. At 1800 UTC on August 8, the
data assimilation was performed over the SEN, NOSEN, and ALL areas. The initial circulation field, in which the
vorticity increased in the southwest of the typhoon and decreased in the southeast, benefited the westward
movement of the typhoon. The 18 h forecast indicated that the circulation in the SEN area assimilation experiment
showed a negative vorticity increase in the southwest side of the typhoon and positive potential temperature
as well as water vapor forecast differences on its north side, which contrasted with the results obtained
in the NOSEN and ALL experiments. This provided favorable weather conditions for the typhoon to change
direction and move northward. Targeted observation experiments investigating Typhoon Lekima indicated that
assimilating observations of sensitive areas greatly improves typhoon forecasting and assimilation efficiency.
However, the ALL experiment yielded the best forecast among the three cases.

Data assimilation,Adaptive observation,ETS,Sensitive area,Typhoon forecasting