B6 - New data assimilation approaches to better predict tropical convection
Other researcher: Yvonne Ruckstuhl (PostDoc)
Forecasting precipitation in the tropics remains a large challenge for numerical weather prediction (NWP) today. The practical predictability realized with operational systems appears to be much further away from theoretical intrinsic predictability than at higher latitudes. Amongst the many reasons for this are the large analysis errors at synoptic to planetary scales in the tropics, which at least partly stem from the introduction of unphysical features through data assimilation (DA) systems optimized for the extratropics. These features may then further amplify through the coupling of tropical waves with convection, which occurs through wave-induced modification of environmental conditions for convection on one hand and wave triggering by convective latent and radiative heating on the other hand, as investigated in Project C2. During Phase 1, fundamental research was conducted on designing DA algorithms that can preserve pre-defined physical properties. In particular, preservation of mass, positivity, energy, and enstrophy were examined for atmospheric state estimation. Constraining or not constraining the analysis ensemble with the physical properties, had a profound effect on ensemble forecasts as well as representation of uncertainty in idealized experiments.
Phase 2 will build on these results and expand them to experiments with the full-blown NWP Icosahedral Non- hydrostatic (ICON) model. In order to specifically concentrate on the atmospheric dynamics most relevant for low latitudes, simulations will be run at high resolution in a tropical channel configuration with zonally symmetric sea-surface temperatures (SSTs) and no continents. Particular attention will be paid to the coupling of convection (and thus rainfall) with tropical waves. This coupling is suspected to provide a potential source of predictability on synoptic to planetary scales, but current state-of-the-art NWP models struggle to realistically represent the involved multi-scale interactions. We will investigate several potential reasons for this, using our idealized tropical channel experiments: (a) resolution: do we need grid spacings fine enough to explicitly represent deep moist convection? Are there other effects of resolution on the convection-wave coupling? (b) Clouds: how important is a detailed and realistic treatment of uncertain microphysical and cloud-radiative processes for the interaction with the waves, e.g., through modifications to vertical profiles of heating and divergence? (c) Data assimilation: how can we best constrain model forecasts in the tropics given the complex behavior of wave-convection couplings? How can we avoid unrealistic projections onto the wave structures and spurious convection? In order to answer these questions, we combine approaches from DA, cloud physics, and atmospheric dynamics. Ensemble DA experiments will be based on virtual observations taken from a high-resolution nature run using the ICON model and different physical constraints (e.g., energy conservation) will be applied. The characteristics of wave features (e.g. propagation, intensity, vertical structure) and their influence on convection will be extracted using filtering tools based on research in Project C2. Finally the role of clouds will be investigated through targeted sensitivity experiments built on insight into the most important microphysical controls (e.g., ice nucleation) from Project B1. In the long run, this kind of interdisciplinary approach can be extended to more realistic settings and thus has potential to improve practical predictability at low latitudes.