Advances in sequential data assimilation and numerical weather forecasting: an Ensemble Transform Kalman-Bucy Filter, a study on clustering in deterministic Ensemble Square Root Filters, and a test of a new time stepping scheme in an atmospheric model

This dissertation deals with aspects of sequential data assimilation (in particular ensemble Kalman filtering) and numerical weather forecasting. In the first part, the recently formulated Ensemble Kalman-Bucy (EnKBF) filter is revisited. It is shown that the previously used numerical integration scheme fails when the magnitude of the background error covariance grows beyond that of the observational error covariance in the forecast window. Therefore, we present a suitable integration scheme that handles the stiffening of the differential equations involved and doesn’t represent further computational expense. Moreover, a transform-based alternative to the EnKBF is developed: under this scheme, the operations are performed in the ensemble space instead of in the state space. Advantages of this formulation are explained. For the first time, the EnKBF is implemented in an atmospheric model. The second part of this work deals with ensemble clustering, a phenomenon that arises when performing data assimilation using of deterministic ensemble square root filters in highly nonlinear forecast models. Namely, an M-member ensemble detaches into an outlier and a cluster of M-1 members. Previous works may suggest that this issue represents a failure of EnSRFs; this work dispels that notion. It is shown that ensemble clustering can be reverted also due to nonlinear processes, in particular the alternation between nonlinear expansion and compression of the ensemble for different regions of the attractor. Some EnSRFs that use random rotations have been developed to overcome this issue; these formulations are analyzed and their advantages and disadvantages with respect to common EnSRFs are discussed. The third and last part contains the implementation of the Robert-Asselin-Williams (RAW) filter in an atmospheric model. The RAW filter is an improvement to the widely popular Robert-Asselin filter that successfully suppresses spurious computational waves while avoiding any distortion in the mean value of the function. Using statistical significance tests both at the local and field level, it is shown that the climatology of the SPEEDY model is not modified by the changed time stepping scheme; hence, no retuning of the parameterizations is required. It is found the accuracy of the medium-term forecasts is increased by using the RAW filter.

Resolution of sharp fronts in the presence of model error in variational data assimilation

We show that the four-dimensional variational data assimilation method (4DVar) can be interpreted as a form of Tikhonov regularization, a very familiar method for solving ill-posed inverse problems. It is known from image restoration problems that L1-norm penalty regularization recovers sharp edges in the image more accurately than Tikhonov, or L2-norm, penalty regularization. We apply this idea from stationary inverse problems to 4DVar, a dynamical inverse problem, and give examples for an L1-norm penalty approach and a mixed total variation (TV) L1–L2-norm penalty approach. For problems with model error where sharp fronts are present and the background and observation error covariances are known, the mixed TV L1–L2-norm penalty performs better than either the L1-norm method or the strong constraint 4DVar (L2-norm)method. A strength of the mixed TV L1–L2-norm regularization is that in the case where a simplified form of the background error covariance matrix is used it produces a much more accurate analysis than 4DVar. The method thus has the potential in numerical weather prediction to overcome operational problems with poorly tuned background error covariance matrices.

A neurofuzzy classifier for two class problems

A neurofuzzy classifier identification algorithm is introduced for two class problems. The initial fuzzy base construction is based on fuzzy clustering utilizing a Gaussian mixture model (GMM) and the analysis of covariance (ANOVA) decomposition. The expectation maximization (EM) algorithm is applied to determine the parameters of the fuzzy membership functions. Then neurofuzzy model is identified via the supervised subspace orthogonal least square (OLS) algorithm. Finally a logistic regression model is applied to produce the class probability. The effectiveness of the proposed neurofuzzy classifier has been demonstrated using a real data set.

Online modeling with tunable RBF network

In this paper, we propose a novel online modeling algorithm for nonlinear and nonstationary systems using a radial basis function (RBF) neural network with a fixed number of hidden nodes. Each of the RBF basis functions has a tunable center vector and an adjustable diagonal covariance matrix. A multi-innovation recursive least square (MRLS) algorithm is applied to update the weights of RBF online, while the modeling performance is monitored. When the modeling residual of the RBF network becomes large in spite of the weight adaptation, a node identified as insignificant is replaced with a new node, for which the tunable center vector and diagonal covariance matrix are optimized using the quantum particle swarm optimization (QPSO) algorithm. The major contribution is to combine the MRLS weight adaptation and QPSO node structure optimization in an innovative way so that it can track well the local characteristic in the nonstationary system with a very sparse model. Simulation results show that the proposed algorithm has significantly better performance than existing approaches.

Crossbred cow adoption and milk market participation in a multivariate count data framework

Cross-bred cow adoption is an important and potent policy variable precipitating subsistence household entry into emerging milk markets. This paper focuses on the problem of designing policies that encourage and sustain milkmarket expansion among a sample of subsistence households in the Ethiopian highlands. In this context it is desirable to measure households’ ‘proximity’ to market in terms of the level of deficiency of essential inputs. This problem is compounded by four factors. One is the existence of cross-bred cow numbers (count data) as an important, endogenous decision by the household; second is the lack of a multivariate generalization of the Poisson regression model; third is the censored nature of the milk sales data (sales from non-participating households are, essentially, censored at zero); and fourth is an important simultaneity that exists between the decision to adopt a cross-bred cow, the decision about how much milk to produce, the decision about how much milk to consume and the decision to market that milk which is produced but not consumed internally by the household. Routine application of Gibbs sampling and data augmentation overcome these problems in a relatively straightforward manner. We model the count data from two sites close to Addis Ababa in a latent, categorical-variable setting with known bin boundaries. The single-equation model is then extended to a multivariate system that accommodates the covariance between crossbred-cow adoption, milk-output, and milk-sales equations. The latent-variable procedure proves tractable in extension to the multivariate setting and provides important information for policy formation in emerging-market settings

Simulated and observed fluxes of sensible and latent heat and CO2 at the WLEF-TV tower using SiB2.5

Three years of meteorological data collected at the WLEF-TV tower were used to drive a revised version of the Simple Biosphere (SiB 2.5) Model. Physiological properties and vegetation phenology were specified from satellite imagery. Simulated fluxes of heat, moisture, and carbon were compared to eddy covariance measurements taken onsite as a means of evaluating model performance on diurnal, synoptic, seasonal, and interannual time scales. The model was very successful in simulating variations of latent heat flux when compared to observations, slightly less so in the simulation of sensible heat flux. The model overestimated peak values of sensible heat flux on both monthly and diurnal scales. There was evidence that the differences between observed and simulated fluxes might be linked to wetlands near the WLEF tower, which were not present in the SiB simulation. The model overestimated the magnitude of the net ecosystem exchange of CO2 in both summer and winter. Mid-day maximum assimilation was well represented by the model, but late afternoon simulations showed excessive carbon uptake due to misrepresentation of within-canopy shading in the model. Interannual variability was not well simulated because only a single year of satellite imagery was used to parameterize the model.

Simulated variations in atmospheric CO2 over a Wisconsin forest using a coupled ecosystem-atmosphere model

Ecosystem fluxes of energy, water, and CO2 result in spatial and temporal variations in atmospheric properties. In principle, these variations can be used to quantify the fluxes through inverse modelling of atmospheric transport, and can improve the understanding of processes and falsifiability of models. We investigated the influence of ecosystem fluxes on atmospheric CO2 in the vicinity of the WLEF-TV tower in Wisconsin using an ecophysiological model (Simple Biosphere, SiB2) coupled to an atmospheric model (Regional Atmospheric Modelling System). Model parameters were specified from satellite imagery and soil texture data. In a companion paper, simulated fluxes in the immediate tower vicinity have been compared to eddy covariance fluxes measured at the tower, with meteorology specified from tower sensors. Results were encouraging with respect to the ability of the model to capture observed diurnal cycles of fluxes. Here, the effects of fluxes in the tower footprint were also investigated by coupling SiB2 to a high-resolution atmospheric simulation, so that the model physiology could affect the meteorological environment. These experiments were successful in reproducing observed fluxes and concentration gradients during the day and at night, but revealed problems during transitions at sunrise and sunset that appear to be related to the canopy radiation parameterization in SiB2.

A low-order model investigation of the analysis of gravity waves in the ensemble Kalman filter.

The behavior of the ensemble Kalman filter (EnKF) is examined in the context of a model that exhibits a nonlinear chaotic (slow) vortical mode coupled to a linear (fast) gravity wave of a given amplitude and frequency. It is shown that accurate recovery of both modes is enhanced when covariances between fast and slow normal-mode variables (which reflect the slaving relations inherent in balanced dynamics) are modeled correctly. More ensemble members are needed to recover the fast, linear gravity wave than the slow, vortical motion. Although the EnKF tends to diverge in the analysis of the gravity wave, the filter divergence is stable and does not lead to a great loss of accuracy. Consequently, provided the ensemble is large enough and observations are made that reflect both time scales, the EnKF is able to recover both time scales more accurately than optimal interpolation (OI), which uses a static error covariance matrix. For OI it is also found to be problematic to observe the state at a frequency that is a subharmonic of the gravity wave frequency, a problem that is in part overcome by the EnKF.However, error in themodeled gravity wave parameters can be detrimental to the performance of the EnKF and remove its implied advantages, suggesting that a modified algorithm or a method for accounting for model error is needed.

Four-dimensional data assimilation and balanced dynamics

The problem of spurious excitation of gravity waves in the context of four-dimensional data assimilation is investigated using a simple model of balanced dynamics. The model admits a chaotic vortical mode coupled to a comparatively fast gravity wave mode, and can be initialized such that the model evolves on a so-called slow manifold, where the fast motion is suppressed. Identical twin assimilation experiments are performed, comparing the extended and ensemble Kalman filters (EKF and EnKF, respectively). The EKF uses a tangent linear model (TLM) to estimate the evolution of forecast error statistics in time, whereas the EnKF uses the statistics of an ensemble of nonlinear model integrations. Specifically, the case is examined where the true state is balanced, but observation errors project onto all degrees of freedom, including the fast modes. It is shown that the EKF and EnKF will assimilate observations in a balanced way only if certain assumptions hold, and that, outside of ideal cases (i.e., with very frequent observations), dynamical balance can easily be lost in the assimilation. For the EKF, the repeated adjustment of the covariances by the assimilation of observations can easily unbalance the TLM, and destroy the assumptions on which balanced assimilation rests. It is shown that an important factor is the choice of initial forecast error covariance matrix. A balance-constrained EKF is described and compared to the standard EKF, and shown to offer significant improvement for observation frequencies where balance in the standard EKF is lost. The EnKF is advantageous in that balance in the error covariances relies only on a balanced forecast ensemble, and that the analysis step is an ensemble-mean operation. Numerical experiments show that the EnKF may be preferable to the EKF in terms of balance, though its validity is limited by ensemble size. It is also found that overobserving can lead to a more unbalanced forecast ensemble and thus to an unbalanced analysis.

A unified neurofuzzy model for classification

This work proposes a unified neurofuzzy modelling scheme. To begin with, the initial fuzzy base construction method is based on fuzzy clustering utilising a Gaussian mixture model (GMM) combined with the analysis of covariance (ANOVA) decomposition in order to obtain more compact univariate and bivariate membership functions over the subspaces of the input features. The mean and covariance of the Gaussian membership functions are found by the expectation maximisation (EM) algorithm with the merit of revealing the underlying density distribution of system inputs. The resultant set of membership functions forms the basis of the generalised fuzzy model (GFM) inference engine. The model structure and parameters of this neurofuzzy model are identified via the supervised subspace orthogonal least square (OLS) learning. Finally, instead of providing deterministic class label as model output by convention, a logistic regression model is applied to present the classifier’s output, in which the sigmoid type of logistic transfer function scales the outputs of the neurofuzzy model to the class probability. Experimental validation results are presented to demonstrate the effectiveness of the proposed neurofuzzy modelling scheme.

Tropical Atlantic variability modes (1979–2002). Part I: Time-evolving SST modes related to West African rainfall

This work presents a description of the 1979–2002 tropical Atlantic (TA) SST variability modes coupled to the anomalous West African (WA) rainfall during the monsoon season. The time-evolving SST patterns, with an impact on WA rainfall variability, are analyzed using a new methodology based on maximum covariance analysis. The enhanced Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) dataset, which includes measures over the ocean, gives a complete picture of the interannual WA rainfall patterns for the Sahel dry period. The leading TA SST pattern, related to the Atlantic El Niño, is coupled to anomalous precipitation over the coast of the Gulf of Guinea, which corresponds to the second WA rainfall principal component. The thermodynamics and dynamics involved in the generation, development, and damping of this mode are studied and compared with previous works. The SST mode starts at the Angola/Benguela region and is caused by alongshore wind anomalies. It then propagates westward via Rossby waves and damps because of latent heat flux anomalies and Kelvin wave eastward propagation from an off-equatorial forcing. The second SST mode includes the Mediterranean and the Atlantic Ocean, showing how the Mediterranean SST anomalies are those that are directly associated with the Sahelian rainfall. The global signature of the TA SST patterns is analyzed, adding new insights about the Pacific– Atlantic link in relation to WA rainfall during this period. Also, this global picture suggests that the Mediterranean SST anomalies are a fingerprint of large-scale forcing. This work updates the results given by other authors, whose studies are based on different datasets dating back to the 1950s, including both the wet and the dry Sahel periods.

Data assimilation with correlated observation errors: experiments with a 1-D shallow water model

Remote sensing observations often have correlated errors, but the correlations are typically ignored in data assimilation for numerical weather prediction. The assumption of zero correlations is often used with data thinning methods, resulting in a loss of information. As operational centres move towards higher-resolution forecasting, there is a requirement to retain data providing detail on appropriate scales. Thus an alternative approach to dealing with observation error correlations is needed. In this article, we consider several approaches to approximating observation error correlation matrices: diagonal approximations, eigendecomposition approximations and Markov matrices. These approximations are applied in incremental variational assimilation experiments with a 1-D shallow water model using synthetic observations. Our experiments quantify analysis accuracy in comparison with a reference or ‘truth’ trajectory, as well as with analyses using the ‘true’ observation error covariance matrix. We show that it is often better to include an approximate correlation structure in the observation error covariance matrix than to incorrectly assume error independence. Furthermore, by choosing a suitable matrix approximation, it is feasible and computationally cheap to include error correlation structure in a variational data assimilation algorithm.

The simulation of the opposing fluxes of latent heat and CO2 over various land-use types: coupling a gas exchange model to a mesoscale atmospheric model

A mesoscale meteorological model (FOOT3DK) is coupled with a gas exchange model to simulate surface fluxes of CO2 and H2O under field conditions. The gas exchange model consists of a C3 single leaf photosynthesis sub-model and an extended big leaf (sun/shade) sub-model that divides the canopy into sunlit and shaded fractions. Simulated CO2 fluxes of the stand-alone version of the gas exchange model correspond well to eddy-covariance measurements at a test site in a rural area in the west of Germany. The coupled FOOT3DK/gas exchange model is validated for the diurnal cycle at singular grid points, and delivers realistic fluxes with respect to their order of magnitude and to the general daily course. Compared to the Jarvis-based big leaf scheme, simulations of latent heat fluxes with a photosynthesis-based scheme for stomatal conductance are more realistic. As expected, flux averages are strongly influenced by the underlying land cover. While the simulated net ecosystem exchange is highly correlated with leaf area index, this correlation is much weaker for the latent heat flux. Photosynthetic CO2 uptake is associated with transpirational water loss via the stomata, and the resulting opposing surface fluxes of CO2 and H2O are reproduced with the model approach. Over vegetated surfaces it is shown that the coupling of a photosynthesis-based gas exchange model with the land-surface scheme of a mesoscale model results in more realistic simulated latent heat fluxes.

Five years of carbon dioxide fluxes measurements in a highly vegetated suburban area

Suburban areas continue to grow rapidly and are potentially an important land-use category for anthropogenic carbon-dioxide (CO2) emissions. Here eddy covariance techniques are used to obtain ecosystem-scale measurements of CO2 fluxes (FC) from a suburban area of Baltimore, Maryland, USA (2002–2006). These are among the first multi-year measurements of FC in a suburban area. The study area is characterized by low population density (1500 inhabitants km−2) and abundant vegetation (67.4% vegetation land-cover). FC is correlated with photosynthetic active radiation (PAR), soil temperature, and wind direction. Missing hourly FC is gap-filled using empirical relations between FC, PAR, and soil temperature. Diurnal patterns show net CO2 emissions to the atmosphere during winter and net CO2 uptake by the surface during summer daytime hours (summer daily total is −1.25 g C m−2 d−1). Despite the large amount of vegetation the suburban area is a net CO2 source of 361 g C m−2 y−1 on average.

Ensemble transform Kalman-Bucy filters

Two recent works have adapted the Kalman–Bucy filter into an ensemble setting. In the first formulation, the ensemble of perturbations is updated by the solution of an ordinary differential equation (ODE) in pseudo-time, while the mean is updated as in the standard Kalman filter. In the second formulation, the full ensemble is updated in the analysis step as the solution of single set of ODEs in pseudo-time. Neither requires matrix inversions except for the frequently diagonal observation error covariance. We analyse the behaviour of the ODEs involved in these formulations. We demonstrate that they stiffen for large magnitudes of the ratio of background error to observational error variance, and that using the integration scheme proposed in both formulations can lead to failure. A numerical integration scheme that is both stable and is not computationally expensive is proposed. We develop transform-based alternatives for these Bucy-type approaches so that the integrations are computed in ensemble space where the variables are weights (of dimension equal to the ensemble size) rather than model variables. Finally, the performance of our ensemble transform Kalman–Bucy implementations is evaluated using three models: the 3-variable Lorenz 1963 model, the 40-variable Lorenz 1996 model, and a medium complexity atmospheric general circulation model known as SPEEDY. The results from all three models are encouraging and warrant further exploration of these assimilation techniques.