An analysis of variability and predictability of organised deep convection and its divergent upper tropospheric outflow
Doctoral thesis published by Johannes Gutenberg University in Mainz (Germany)
Abstract:
The consequences of convective organisation, aggregation and convective momentum transport for upper tropospheric divergent outflows from deep convection are explored. Furthermore, the variability and predictability of these outflows is thereby connected to other aspects of dynamics and predictability of the convective systems. Different approaches to the simulation of convection are investigated, in which the conditional dependence of divergent outflow, on net latent heating rate, differs as a consequence of different methods to represent convective systems.
The theoretical understanding of the convective outflows is addressed first, by investigating a comprehensive set of idealised Large Eddy Simulations. The experiments, with four prototypes of convective systems, reveal that convective organisation and net latent heat release (convertible to precipitation rate) shape the patterns in magnitude of the divergent outflows. Dimensionality of convective outflows (2D convection versus 3D convection, or a mixed/intermediate regime) bounds an envelope of divergent outflow variability. This outcome is mostly consistent with convective outflows, if represented in older linear gravity wave models.
Investigating these convective outflows in the NWP model ICON for an event on 10th-11th of June 2019 over Central Europe, the divergent outflows in a parameterised and an explicit representations of deep convective systems are intercompared. Near-linear response of deep convective outflows to net latent heating is found in parameterised convection, while coherent patterns in variability are found in convection-permitting simulations, at 1 km horizontal grid spacing. Convective organisation and aggregation induce a non-linear increase in the magnitude deep convective outflows, with increasing net latent heating. This non-linearity is demonstrated by the confidence interval of the best fit, between power transformed net latent heating and detected magnitude of outflows. Other statistical patterns also support the representation of that pattern in the studied case. However, mixed and weaker than expected signals are found, in an attempt to detect the representation of dimensionality of the convection and its consequences for the divergent outflows. To detect the representation, an ellipse fitting algorithm that describes the elongation of the intense (convective) precipitation systems is used. These signals are understandable and suggest the need of further investigation. Convective momentum transport is suggested to slightly increase the magnitude of divergent outflows, in the studied case.
In a subset of the Large Eddy Simulations, in which a so-called squall line is triggered, error or difference growth is investigated in relation to dynamics and precipitation variability, amongst others. During the two hour simulations, the first stage of convective initiation is associated with crucial gravity wave activity, which induces de-correlation between ensemble members. After an initial trigger of convection (about 20 minutes into the simulations), a second phase of convective initiation (at 30 minutes) determines much of the structure in the ensemble spread, for the next hour or so. Directly after that second phase of convective initiation, spread in cold pool acceleration is found, while cold pool propagation velocity is maintained afterwards (t=45 to t=100 minutes). Coherent flow anomalies, initiated directly after the second phase of convective initiation, are also maintained on the time scale of an hour. They dissipate after about 80 to 100 minutes simulation time. When flow is evaluated in a frame relative to cold pool edge, it is shown that error or difference growth in terms of zonal wind, within the ensemble, is substantially smaller than in the Eulerian perspective. Furthermore, feedbacks acting within the squall line are not dominating this difference growth: much of the difference is directly explained by differences in cold pool propagation. Much of the ensemble spread still maintained in the cold pool-relative framework, such as in precipitation and downdrafts, is also strongly related to the decisive second phase of convective triggering.
Looking at convective variability from a (Bayesian) perspective, conditional on precipitation rate, the often subtle threshold behaviour in convective initiation is bypassed. However, the approach demonstrates that a conditional view can shed important light on convective variability and how it is represented in NWP. Here, it shows contrasts in between idealised Large Eddy Simulations, convection-permitting NWP and deep convection parameterising NWP, where implicit assumptions on divergent convective outflows are identified. Strong coupling between dynamics, predictability and precipitation is accentuated. In representativity studies of other aspects in an NWP (e.g. microphysics, turbulence, radiation) and predictability studies, the applied conditional approach may be fruitful.
The theoretical understanding of the convective outflows is addressed first, by investigating a comprehensive set of idealised Large Eddy Simulations. The experiments, with four prototypes of convective systems, reveal that convective organisation and net latent heat release (convertible to precipitation rate) shape the patterns in magnitude of the divergent outflows. Dimensionality of convective outflows (2D convection versus 3D convection, or a mixed/intermediate regime) bounds an envelope of divergent outflow variability. This outcome is mostly consistent with convective outflows, if represented in older linear gravity wave models.
Investigating these convective outflows in the NWP model ICON for an event on 10th-11th of June 2019 over Central Europe, the divergent outflows in a parameterised and an explicit representations of deep convective systems are intercompared. Near-linear response of deep convective outflows to net latent heating is found in parameterised convection, while coherent patterns in variability are found in convection-permitting simulations, at 1 km horizontal grid spacing. Convective organisation and aggregation induce a non-linear increase in the magnitude deep convective outflows, with increasing net latent heating. This non-linearity is demonstrated by the confidence interval of the best fit, between power transformed net latent heating and detected magnitude of outflows. Other statistical patterns also support the representation of that pattern in the studied case. However, mixed and weaker than expected signals are found, in an attempt to detect the representation of dimensionality of the convection and its consequences for the divergent outflows. To detect the representation, an ellipse fitting algorithm that describes the elongation of the intense (convective) precipitation systems is used. These signals are understandable and suggest the need of further investigation. Convective momentum transport is suggested to slightly increase the magnitude of divergent outflows, in the studied case.
In a subset of the Large Eddy Simulations, in which a so-called squall line is triggered, error or difference growth is investigated in relation to dynamics and precipitation variability, amongst others. During the two hour simulations, the first stage of convective initiation is associated with crucial gravity wave activity, which induces de-correlation between ensemble members. After an initial trigger of convection (about 20 minutes into the simulations), a second phase of convective initiation (at 30 minutes) determines much of the structure in the ensemble spread, for the next hour or so. Directly after that second phase of convective initiation, spread in cold pool acceleration is found, while cold pool propagation velocity is maintained afterwards (t=45 to t=100 minutes). Coherent flow anomalies, initiated directly after the second phase of convective initiation, are also maintained on the time scale of an hour. They dissipate after about 80 to 100 minutes simulation time. When flow is evaluated in a frame relative to cold pool edge, it is shown that error or difference growth in terms of zonal wind, within the ensemble, is substantially smaller than in the Eulerian perspective. Furthermore, feedbacks acting within the squall line are not dominating this difference growth: much of the difference is directly explained by differences in cold pool propagation. Much of the ensemble spread still maintained in the cold pool-relative framework, such as in precipitation and downdrafts, is also strongly related to the decisive second phase of convective triggering.
Looking at convective variability from a (Bayesian) perspective, conditional on precipitation rate, the often subtle threshold behaviour in convective initiation is bypassed. However, the approach demonstrates that a conditional view can shed important light on convective variability and how it is represented in NWP. Here, it shows contrasts in between idealised Large Eddy Simulations, convection-permitting NWP and deep convection parameterising NWP, where implicit assumptions on divergent convective outflows are identified. Strong coupling between dynamics, predictability and precipitation is accentuated. In representativity studies of other aspects in an NWP (e.g. microphysics, turbulence, radiation) and predictability studies, the applied conditional approach may be fruitful.
Canonical Valuations and the Birational Section Conjecture
Abstract:
We develop a notion of a `canonical $\mathcal{C}$-henselian valuation' for a class $\mathcal{C}$ of field extensions, generalizing the construction of the canonical henselian valuation of a field. We use this to show that the $p$-adic valuation on a finite extension $F$ of $\mathbb{Q}_p$ can be recovered entirely (or up to some indeterminacy of the residue field) from various small quotients of $G_F$, the absolute Galois group of $F$. In particular, it can be recovered fully from the maximal solvable quotient. We use this to prove several versions of the birational section conjecture for varieties over $p$-adic fields.Characterizing uncertainty in deep convection triggering using explainable machine learning
Journal of the Atmospheric Sciences American Meteorological Society
Abstract:
Realistically representing deep atmospheric convection is important for accurate numerical weather and climate simulations. However, parameterizing where and when deep convection occurs (“triggering”) is a well-known source of model uncertainty. Most triggers parameterize convection deterministically, without considering the uncertainty in the convective state as a stochastic process. In this study, we develop a machine learning model, a random forest, that predicts the probability of deep convection, and then apply clustering of SHAP values, an explainable machine learning method, to characterize the uncertainty of convective events. The model uses observed large-scale atmospheric variables from the Atmospheric Radiation Measurement constrained variational analysis dataset over the Southern Great Plains, US. The analysis of feature importance shows which mechanisms driving convection are most important, with large-scale vertical velocity providing the highest predictive power for more certain, or easier to predict, convective events, followed by the dynamic generation rate of dilute convective available potential energy. Predictions of uncertain convective events instead rely more on other features such as precipitable water or low-level temperature. The model outperforms conventional convective triggers. This suggests that probabilistic machine learning models can be used as stochastic parameterizations to improve the occurrence of convection in weather and climate models in the future.Climate SPHINX: evaluating the impact of resolution and stochastic physics parameterisations in climate simulations
Geoscientific Model Development European Geosciences Union