Observation influence and imbalance in convective-scale data assimilation

Diefenbach, Theresa (2024): Observation influence and imbalance in convective-scale data assimilation. Dissertation, LMU München: Fakultät für Physik

Abstract

Numerical weather prediction (NWP) is an initial value problem where the accuracy of the initial conditions is critical to the success of the forecast. While observational data are available to define the initial state, they are insufficient compared to the requirements of the weather prediction model. To bridge this gap, data assimilation (DA) integrates observational data with a first-guess weather forecast to obtain initial conditions that represent the best estimate of the atmospheric state given all available information. This best estimate is commonly referred to as the analysis. DA is essential for NWP, however its complexity, coupled with assumptions that do not always reflect reality, may introduce noise in the form of imbalance, which has the potential to limit the effectiveness of DA. Therefore, monitoring the DA process is crucial, particularly with the advent of new and intricate observational data sources, such as satellites or ground-based remote sensing. Yet, monitoring itself presents challenges. This thesis presents two novel methods for investigating the effectiveness of a convective-scale DA system with respect to the incorporation of observational information into the analysis and the occurrence of possible imbalances due to DA. In the first part of this thesis a computationally efficient approach to quantify the analysis influence of observations is presented. We use fundamental properties of the Local Ensemble Transform Kalman Filter (LETKF), a frequently used DA method in convective-scale DA, to disentangle the analysis update into the contributions of individual observations, which are called partial analysis increments. While this method shares elements with established analysis sensitivity measures, the presented approach offers a more explicit and computationally efficient way to explore the influence of observations across various model variables. Moreover, it enables the identification of potentially detrimental observation influence and facilitates the optimisation of DA settings for enhanced analysis accuracy. The second part of this thesis deals with diagnostic techniques to assess spurious imbalances in the initial states obtained by data assimilation. This spurious imbalance often refers to a disturbance of the balance of forces that normally prevails in the atmosphere. However, the specific form of imbalance in convective-scale NWP is uncertain, and it is expected to differ from that in larger-scale models. To investigate this, three different imbalance measures each grounded in different physical rationales are implemented and applied to a convective resolving near-operational NWP system. This includes a newly developed method based on a physical balance principle that applies approximately on the convective scale, the so-called "Weak Temperature Gradient (WTG) Balance", which is being implemented and tested for the first time. The results show that the different measures seem to capture different facets of imbalance, with the WTG imbalance metric emerging as a particularly suitable candidate offering complementary results to existing methods. In the future, we hope that the two presented diagnostic tools can contribute to further progress in the field of convective-scale DA, thereby enabling improvements in forecast accuracy.

• Dokumententyp: Dissertationen (Dissertation, LMU München)
• Keywords: Numerical Weather Prediction, Meteorology, Data Assimilation, Kalman Filter, convective-scale
• Themengebiete: 500 Naturwissenschaften und Mathematik ; 500 Naturwissenschaften und Mathematik > 530 Physik
• Fakultäten: Fakultät für Physik
• Sprache der Hochschulschrift: Englisch
• Datum der mündlichen Prüfung: 22. Juli 2024
• 1. Bericht­erstatter:in: Craig, George C.