Quantifying the Impact of Assimilating Ground-based GNSS in Operational NWP Models

Zenith Total Delay (ZTD) from networks of ground based GNSS receivers has been assimilated in both the Met Office global and high-resolution regional numerical weather predication (NWP) models for over a decade. It is useful to be able to quantify the impact of assimilating these data and compare this with the impact of assimilating other observation types. This helps inform observing network evolution.

Forecast Sensitivity to Observation Impact (FSOI) analysis is an established method for monitoring the collective impact of an observing network on the quality of an NWP forecast. FSOI uses an adjoint-based method to compute the observation sensitivity for each assimilated observation in a global model forecast simultaneously. The sensitivity value represents the change in forecast error for a unit observation innovation; this information is combined with innovation data from the model forecast to provide an estimate for the resulting change in error for the total 24-hour energy norm. The computational efficiency of the FSOI process makes it a good alternative to traditional Observation System Experiments (OSE) when considering the benefits of an observing network.

Unfortunately, the underpinning assumptions in the FSOI methodology mean that the method does not easily translate to higher resolution regional NWP models. For these models, observation impacts can still be determined through OSEs through which observations of specific type or types are denied to the model and the results compared with a Control model run with all types of observations included.

FSOI analysis has previously been used to compare and contrast the benefits associated with different observing method types. This study uses the original FSOI methodology but refines the output to collate impact data at the observing-site level (in relation to a specified observing network).  This enables geographic variations in the impact associated with individual sites within an observing network to be visualised and the tool can be used to assist with decisions on network design and performance.

Results are presented that illustrate the impact of GNSS ZTD data on global NWP model forecasts on a seasonal and annual basis. A number of metrics are used to assess the benefits associated with a GNSS site, including the total impact over a fixed period of time as well as the mean impact per observation on both a global and country scale. The results highlight the significant impact that individual sites in areas of low data density (such as the southern hemisphere) have on model forecasts.

The impact of GNSS ZTD data on the 1.5 km resolution UKV model are also presented. The contrasting impacts in summer and winter, and variation of forecast impact as lead-time increases are explored.