Convective Storm VIL and Lightning Nowcasting Using Satellite and Weather Radar Measurements Based on Multi-Task Learning Models

Convective storms and lightning are among the most important weather phenomena that are challenging to forecast.In this study,a novel multi-task learning(MTL)encoder-decoder U-net neural network was developed to forecast convective storms and lightning with lead times for up to 90 min,using GOES-16 geostationary satellite infrared brightness temperatures(IRBTs),lightning flashes from Geostationary Lightning Mapper(GLM),and vertically integrated liquid(VIL)from Next Generation Weather Radar(NEXRAD).To cope with the heavily skewed distribution of lightning data,a spatiotemporal exponent-weighted loss function and log-transformed lightning normalization approach were developed.The effects of MTL,single-task learning(STL),and IRBTs as auxiliary input features on convection and lightning nowcasting were investigated.The results showed that normalizing the heavily skew-distributed lightning data along with a log-transformation dramatically outperforms the min-max normalization method for nowcasting an intense lightning event.The MTL model significantly outperformed the STL model for both lightning nowcasting and VIL nowcasting,particularly for intense lightning events.The MTL also helped delay the lightning forecast performance decay with the lead times.Furthermore,incorporating satellite IRBTs as auxiliary input features substantially improved lightning nowcasting,but produced little difference in VIL forecasting.Finally,the MTL model performed better for forecasting both lightning and the VIL of organized convective storms than for isolated cells.

An Observing System Simulation Experiment to Assess the Potential Impact of a Virtual Mobile Communication Tower–based Observation Network on Weather Forecasting Accuracy in China. Part 1: Weather Stations with a Typical Mobile Tower Height of 40 m

The importance of a national or regional network of meteorological stations for improving weather predictions has been recognized for many years.Ground-based automatic weather stations typically observe weather at a height of 2-10 m above ground level(AGL);however,these observations may have two major shortcomings.Large portions of data cannot be used if the station height is significantly lower than the model surface level;and such observations may contain large representativity errors as near-surface observations are often affected by the local environment,such as nearby buildings and tall trees.With the recent introduction of a significant number of mobile communication towers that are typically over40 m AGL in China,a campaign has been proposed to use such towers to build a future observing system with an observing height of 40 m.A series of observing system simulation experiments has been conducted to assess the potential utility of such a future observing system as part of a feasibility study.The experiments were conducted using the Weather Research and Forecasting model and its Rapid Update Cycle data assimilation system.The results revealed the possibility of improving weather forecasting by raising present weather stations to a height of 40 m;this would not only enable more observations to pass the terrain check,but should also reduce interpolation errors.Additionally,improvements for temperature,humidity and wind forecasting could be achieved as the accuracy of the initial conditions increases.

Unsupervised social network embedding via adaptive specific mappings

In this paper,we address the problem of unsuperised social network embedding,which aims to embed network nodes,including node attributes,into a latent low dimensional space.In recent methods,the fusion mechanism of node attributes and network structure has been proposed for the problem and achieved impressive prediction performance.However,the non-linear property of node attributes and network structure is not efficiently fused in existing methods,which is potentially helpful in learning a better network embedding.To this end,in this paper,we propose a novel model called ASM(Adaptive Specific Mapping)based on encoder-decoder framework.In encoder,we use the kernel mapping to capture the non-linear property of both node attributes and network structure.In particular,we adopt two feature mapping functions,namely an untrainable function for node attributes and a trainable function for network structure.By the mapping functions,we obtain the low dimensional feature vectors for node attributes and network structure,respectively.Then,we design an attention layer to combine the learning of both feature vectors and adaptively learn the node embedding.In encoder,we adopt the component of reconstruction for the training process of learning node attributes and network structure.We conducted a set of experiments on seven real-world social network datasets.The experimental results verify the effectiveness and efficiency of our method in comparison with state-of-the-art baselines.

An efficient method for privacy preserving location queries

Recently, the issue of privacy preserving loca- tion queries has attracted much research. However, there are few works focusing on the tradeoff between location privacy preservation and location query information collection. To tackle this kind of tradeoff, we propose the privacy persevering location query （PLQ）, an efficient privacy pre-serving location query processing framework. This frame- work can enable the location-based query without revealing user location information. The framework can also facilitate location-based service providers to collect some information about the location based query, which is useful in practice. PLQ consists of three key components, namely, the location anonymizer at the client side, the privacy query processor at the server side, and an additional trusted third party connect- ing the client and server. The location anonymizer blurs the user location into a cloaked area based on a map-hierarchy. The map-hierarchy contains accurate regions that are parti- tioned according to real landforms. The privacy query pro- cessor deals with the requested nearest-neighbor （NN） loca- tion based query. A new convex hull of polygon （CHP） algo- rithm is proposed for nearest-neighbor queries using a poly- gon cloaked area. The experimental results show that our al- gorithms can efficiently process location based queries.

Attention based simplified deep residual network for citywide crowd flows prediction

Crowd flows prediction is an important problem of urban computing whose goal is to predict the number of incoming and outgoing people of regions in the future.In practice,emergency applications often require less training time.However,there is a little work on how to obtain good prediction performance with less training time.In this paper,we propose a simplified deep residual network for our problem.By using the simplified deep residual network,we can obtain not only less training time but also competitive prediction performance compared with the existing similar method.Moreover,we adopt the spatio-temporal attention mechanism to further improve the simplified deep residual network with reasonable additional time cost.Based on the real datasets,we construct a series of experiments compared with the existing methods.The experimental results confirm the efficiency of our proposed methods.

PrF_(3)-NdF_(3)-DyF_(3)-LiF electrolyte system for preparation of Pr-Nd-Dy alloy by electrolysis

The application of Pr-Nd-Dy alloy in the field of high-performance Nd-Fe-B permanent magnet materials has great potential.The composition of the PrF_(3)-NdF_(3)-DyF_(3)-LiF(PND-LiF) electrolyte system used in the production of Pr-Nd-Dy alloys,the distribution of F,Li,RE and other elements in the electrolyte and their occurrence state were studied in this paper.The effect of temperature and lithium fluoride addition on electrolyte conductivity was revealed using the continuous conductivity cell constant(CVCC) method.The thermal analysis method was used to study the influence of lithium fluoride addition on the electrolyte’s liquidus temperature and the optimal process conditions for the production of Pr-Nd-Dy alloy were determined.The results show that the overall distribution of praseodymium neodymium fluoride and lithium fluoride is uniform in the electrolyte and dysprosium fluoride is distributed between praseodymium-neodymium fluoride and lithium fluoride.Praseodymium-neodymium oxide is embedded in praseodymium neodymium fluoride in spotty pattern.The electrolyte’s conductivity is increased as the temperature and lithium fluoride addition are going up,while the liquidus temperature is going down with increasing lithium fluoride addition.The best electrolysis process conditions for the PND-LiF system to produce praseodymium neodymium dysprosium alloy are as follows:temperature1050℃ and 15.56 wt% PrF_(3)-62.22 wt% NdF_(3)-11.11 wt% DyF_(3)-11.11 wt% LiF.

Optimal location query based on k nearest neighbours

Optimal location query in road networks is a basic operation in the location intelligence applications.Given a set of clients and servers on a road network,the purpose of optimal location query is to obtain a location for a new server,so that a certain objective function calculated based on the locations of clients and servers is optimal.Existing works assume no labels for servers and that a client only visits the nearest server.These assumptions are not realistic and it renders the existing work not useful in many cases.In this paper,we relax these assumptions and consider the k nearest neighbours(KNN)of clients.We introduce the problem of KNN-based optimal location query(KOLQ)which considers the k nearest servers of clients and labeled servers.We also introduce a variant problem called relocation KOLQ(RKOLQ)which aims at relocating an existing server to an optimal location.Two main analysis algorithms are proposed for these problems.Extensive experiments on the real road networks illustrate the efficiency of our proposed solutions.