| dc.description.abstract | This study aims to explore the presence of negative bias in lower-level data in GNSS-RO refractivity data caused by ducting. Since ducting occurrences cannot be directly detected from observational data, this study relies on model data for assessment. Comparisons show that when ducting occur in the model, the significant negative bias exists in corresponding observational data below the ducting layer.
An examination of thermal profiles during ducting occurrences in the model identifies three primary factors contributing to ducting: mixing inversion in boundary layer, cold pools from rainfall systems causing severe near-surface cooling, and subsidence inversions induced by other systems like anticyclone, with sharp vertical temperature and humidity gradients acting as primary drivers.
To evaluate the impact of these negatively biased GNSS-RO data on the model, a heavy rainfall event on May 22, 2020, is selected as the focal case. The study divides experiments into four groups: 1. CTRL, without assimilating GNSS-RO data; 2. REF_ALL, assimilating all GNSS-RO data; 3. REF_LSW, using LSW>35% as a QC strategy; and 4. REF_NODUCT, removing data below the ducting layer. After three and a half days of data assimilation cycles, differences in analysis fields, forecast fields, and forecast rainfall are compared among the experiments.
Analysis reveals that assimilating GNSS-RO data significantly adjusts the dynamical and thermal environments in the region from the South China Sea to South China, making them more conducive to increased frontal intensity. Additionally, the utilization of QC strategies further enhances water vapor content. However, the performance of the dynamic field varies based on the QC strategy employed: removing more data in REF_LSW leads to weakened low-level convergence and decreased frontal strength, while retaining more data in REF_NODUCT yields the strongest low-level convergence and frontal strength, showing the importance of mitigating negative bias and preserving lower-level data.
As for forecast outcomes, REF_NODUCT exhibits the highest rainfall, REF_ALL is comparable to REF_LSW, and CTRL produces the weakest results. However, skill scores suggest that REF_ALL performs the best, followed by CTRL, while the performances of REF_NODUCT and REF_LSW are less satisfactory. This could be attributed to the eastward shift of the southwest vortex in South China caused by the removal of lower-level data, resulting in a faster movement of the front compared to its original speed.
Incremental analysis experiments indicate that assimilating GNSS-RO low-level data leads to a significant reduction in local water vapor and can cause its movement due to the model advection. However, the impact on water vapor transport depends on its interaction with the dynamic field.
This study′s reliance on model data for ducting determination may lead to erroneous data removal due to model errors. Furthermore, removing data could lead to the eastward shift of the southwest vortex, causing rainfall displacement. Future efforts should focus on integrating techniques for directly determining ducting from GNSS-RO observational data and reconstructing data below the ducting layer to address these issues and maximize the utility of GNSS-RO lower-level data. | en_US |