Here we present for the first time a proof of concept for an emulation-based method that uses a large-eddy simulations (LESs) to present sub-grid cloud processes in a general circulation model (GCM). We focus on two key variables affecting the properties of shallow marine clouds: updraft velocity and precipitation formation. The LES is able to describe these processes with high resolution accounting for the realistic variability in cloud properties. We show that the selected emulation method is able to represent the LES outcome with relatively good accuracy and that the updraft velocity and precipitation emulators can be coupled with the GCM practically without increasing the computational costs. We also show that the emulators influence the climate simulated by the GCM but do not consistently improve or worsen the agreement with observations on cloud-related properties, although especially the updraft velocity at cloud base is better captured. A more quantitative evaluation of the emulator impacts against observations would, however, have required model re-tuning, which is a significant task and thus could not be included in this proof-of-concept study. All in all, the approach introduced here is a promising candidate for representing detailed cloud- and aerosol-related sub-grid processes in GCMs. Further development work together with increasing computing capacity can be expected to improve the accuracy and the applicability of the approach in climate simulations.

The number of cloud droplets formed at the cloud base depends on both the properties of aerosol particles and the updraft velocity of an air parcel at the cloud base. As the spatial scale of updrafts is too small to be resolved in global atmospheric models, the updraft velocity is commonly parameterised based on the available turbulent kinetic energy. Here we present alternative methods through parameterising updraft velocity based on high-resolution large-eddy simulation (LES) runs in the case of marine stratocumulus clouds. First we use our simulations to assess the accuracy of a simple linear parameterisation where the updraft velocity depends only on cloud top radiative cooling. In addition, we present two different machine learning methods (Gaussian process emulation and random forest) that account for different boundary layer conditions and cloud properties. We conclude that both machine learning parameterisations reproduce the LES-based updraft velocities at about the same accuracy, while the simple approach employing radiative cooling only produces on average lower coefficient of determination and higher root mean square error values. Finally, we apply these machine learning methods to find the key parameters affecting cloud base updraft velocities.

Significant discrepancies remain in estimates of climate impacts of anthropogenic aerosols between different general circulation models (GCMs). Here, we demonstrate that eliminating differences in model aerosol or radiative forcing fields results in close agreement in simulated globally averaged temperature and precipitation responses in the studied GCMs. However, it does not erase the differences in regional responses. We carry out experiments of equilibrium climate response to modern-day anthropogenic aerosols using an identical representation of anthropogenic aerosol optical properties and the first indirect effect of aerosols, MACv2-SP (a simple plume implementation of the second version of the Max Planck Institute Aerosol CLimatology), in two independent climate models (NorESM, Norwegian Earth System Model, and ECHAM6). We find consistent global average temperature responses of −0.48 ( ±0.02 ) and −0.50 ( ±0.03 ) K and precipitation responses of −1.69 ( ±0.04 ) % and −1.79 ( ±0.05 ) % in NorESM1 and ECHAM6, respectively, compared to modern-day equilibrium climate without anthropogenic aerosols. However, significant differences remain between the two GCMs' regional temperature responses around the Arctic circle and the Equator and precipitation responses in the tropics. The scatter in the simulated globally averaged responses is small in magnitude when compared against literature data from modern GCMs using model intrinsic aerosols but same aerosol emissions − (0.5–1.1) K and − (1.5–3.1) % for temperature and precipitation, respectively). The Pearson correlation of regional temperature (precipitation) response in these literature model experiments with intrinsic aerosols is 0.79 (0.34). The corresponding correlation coefficient for NorESM1 and ECHAM6 runs with identical aerosols is 0.78 (0.41). The lack of improvement in correlation coefficients between models with identical aerosols and models with intrinsic aerosols implies that the spatial distribution of regional climate responses is not improved via homogenizing the aerosol descriptions in the models. Rather, differences in the atmospheric dynamic and snow/sea ice cover responses dominate the differences in regional climate responses. Hence, even if we would have perfect aerosol descriptions inside the global climate models, uncertainty arising from the differences in circulation responses between the models would likely still result in a significant uncertainty in regional climate responses.

Using reports of forest losses caused directly by large scale windstorms (or primary damage, PD) from the European forest institute database (comprising 276 PD reports from 1951–2010), total growing stock (TGS) statistics of European forests and the daily North Atlantic Oscillation (NAO) index, we identify a statistically significant change in storm intensity in Western, Central and Northern Europe (17 countries). Using the validated set of storms, we found that the year 1990 represents a change-point at which the average intensity of the most destructive storms indicated by PD/TGS > 0.08% increased by more than a factor of three. A likelihood ratio test provides strong evidence that the change-point represents a real shift in the statistical behaviour of the time series. All but one of the seven catastrophic storms (PD/TGS > 0.2%) occurred since 1990. Additionally, we detected a related decrease in September–November PD/TGS and an increase in December–February PD/TGS. Our analyses point to the possibility that the impact of climate change on the North Atlantic storms hitting Europe has started during the last two and half decades.

Low-bandwidth transmission of synthetic digital content to the end user device in the form of a scene of 3-D meshes requires efficient compression of the mesh geometry. For applications in which the meshes are observed from a single viewpoint, this work explores the use of the image rendering-based distortion measures in rate allocation to their surface regions for view-dependent mesh geometry compression. It is experimentally demonstrated that the image rendering-based distortion measures yield far superior performance (the quality of the rendered image of the reconstructed scene from a viewpoint at a given rate) in optimal rate allocation than other previously proposed distortion measures. A fast rate allocation method is also proposed for use with the image rendering-based measures for real-time or interactive applications. Not only does this method have significantly lower complexity than the optimal rate allocation method due to the rendering of the images of the reconstructed meshes at only judiciously selected rate–distortion operating points, but also its coding performance is just as competitive. Further complexity reduction in rate allocation, through rendering of only the coded regions of the meshes, is also investigated.

In this article, we consider a Bayesian inverse problem associated to elliptic partial differential equations in two and three dimensions. This class of inverse problems is important in applications such as hydrology, but the complexity of the link function between unknown field and measurements can make it difficult to draw inference from the associated posterior. We prove that for this inverse problem a basic sequential Monte Carlo (SMC) method has a Monte Carlo rate of convergence with constants which are independent of the dimension of the discretization of the problem; indeed convergence of the SMC method is established in a function space setting. We also develop an enhancement of the SMC methods for inverse problems which were introduced in Kantas et al. (SIAM/ASA J Uncertain Quantif 2:464–489, 2014); the enhancement is designed to deal with the additional complexity of this elliptic inverse problem. The efficacy of the methodology and its desirable theoretical properties, are demonstrated for numerical examples in both two and three dimensions.

In this paper, the problem of rate allocation to the surface regions of 3-D meshes for view dependent compression of their geometries is addressed. The view image MSE based distortion measure is first experimentally shown to yield higher decoded view image Peak-Signal-To-Noise-Ratio (PSNR) than the distortion measures previously proposed in the literature when used in the framework for optimal rate allocation to regions. In place of the optimal rate allocation method, a logarithmic search rate allocation method is also proposed for use with the view image MSE based measure that requires far fewer view image renderings and thereby offers lower complexity in return for a small drop in coding performance.

Modeling and prediction of student success is a critical task in education. In this paper, we employ machine learning methods to predict course grade performance of Computer Engineering students. As features, in addition to the conventional course grades we use fine grained student performance measurements corresponding to different goals (ABET outcomes) of a course. We observe that, compared to using only previous course grades, addition of outcome grades can significantly improve the prediction results. Using the trained model enables interpretation of how different courses affect performance on a specific course in the future. We think that even more detailed and systematically produced course outcome measurements can be beneficial in modeling students university performance.

This paper presents the results of main part-of-speech tagging of Turkish sentences using Conditional Ran- dom Fields (CRFs). Although CRFs are applied to many different languages for part-of-speech (POS) tagging, Turkish poses interesting challenges to be modeled with them. The challenges include issues related to the statistical model of the problem as well as issues related to computational complexity and scaling. In this paper, we propose a novel model for main-POS tagging in Turkish. Furthermore, we pro- pose some approaches to reduce the computational complexity and allow better scaling characteristics or improve the performance without increased complexity. These approaches are discussed with respect to their advantages and disadvantages. We show that the best approach is competitive with the current state of the art in accuracy and also in training and test durations. The good results obtained imply a good first step towards full morphological disambiguation.

The concept of Web 2.0 or "semantic web" has been getting more and more popular during the last half decade. The potential of very subtle yet important emergent semantics hidden in such environments calls for equally elegant and powerful methods to "mine" them. However, much of the previous work on model based recommender systems for folksonomies considered user to resource and resource to tag similarity separately, ignoring the dependency of users' interest to both the tags and the corresponding resources. In this paper, we propose a probabilistic personalized recommendation model, Latent Interest Model, that accounts for users, tags and resources jointly. The proposed method's performance is evaluated on real data sets obtained from a popular online bookmarking site using different performance measures for tag and resource recommendation tasks. Our experimental results show that our model captures personal preferences for tag usage and resource selection. Performance evaluation of Latent Interest Model indicates that the proposed personalized method yields significant improvement of recommendation accuracy.

Folksonomies constitute an important type of Web 2.0 services, where users collectively annotate (or "tag") resources to create custom categories. Semantic relation of these categories hint at the possibility of another categorization at a higher level. Discovering these more general categories, called "topics", is an important task. One problem is to discover these semantically coherent topics and the accompanying small sets of tags that cover these topics in order to facilitate more detailed item search. Another important problem is to find words/phrases that describe these topics, i.e. labels or "meta-tag"s. These labeled topics can immensely increase the item search efficiency of users in a folksonomy service. However, this possibility has not been sufficiently exploited to date. In this paper, a probabilistic model is used to identify topics in a folksonomy, which are then associated with relevant, descriptive meta-tags. In addition, a small set of diverse and relevant tags are found which cover the semantics of the topic well. The resulting topics form a personalized categorization of folksonomy data due to the personalized nature of the model employed. The results show that the proposed method is successful at discovering important topics and the corresponding identifying meta-tags.

We propose a method to solve the inter region/inter object rate allocation problem for 3D scenes using the rendered 2D images of the scene. Even though the problem has previously been addressed by approximations applied in 3D space, a benchmark method for assessing the performance of rate allocation methods has not yet been developed. The proposed method fullfills this need.