Text Classification for a Large-Scale Taxonomy Using Dynamically Mixed Local and Global Models for a Node

Abstract Accurately forecasting a tropical cyclone’s (TC) track and intensity remains one of the top priorities in weather forecasting. A dynamical downscaling approach based on the scale-selective data assimilation (SSDA) method is applied to demonstrate its effectiveness in TC track and intensity forecasting. The SSDA approach retains the merits of global models in representing large-scale environmental flows and regional models in describing small-scale characteristics. The regional model is driven from the model domain interior by assimilating large-scale flows from global models, as well as from the model lateral boundaries by the conventional sponge zone relaxation. By using Hurricane Felix (2007) as a demonstration case, it is shown that, by assimilating large-scale flows from the Global Forecast System (GFS) forecasts into the regional model, the SSDA experiments perform better than both the original GFS forecasts and the control experiments, in which the regional model is only driven by lateral boundary conditions. The overall mean track forecast error for the SSDA experiments is reduced by over 40% relative to the control experiments, and by about 30% relative to the GFS forecasts, respectively. In terms of TC intensity, benefiting from higher grid resolution that better represents regional and small-scale processes, both the control and SSDA runs outperform the GFS forecasts. The SSDA runs show approximately 14% less overall mean intensity forecast error than do the control runs. It should be noted that, for the Felix case, the advantage of SSDA becomes more evident for forecasts with a lead time longer than 48 h.

Download Full-text

Adjusting BERT’s Pooling Layer for Large-Scale Multi-Label Text Classification

Text, Speech, and Dialogue - Lecture Notes in Computer Science ◽

10.1007/978-3-030-58323-1_23 ◽

2020 ◽

pp. 214-221 ◽

Cited By ~ 1

Author(s):

Jan Lehečka ◽

Jan Švec ◽

Pavel Ircing ◽

Luboš Šmídl

Keyword(s):

Text Classification ◽

Large Scale

Download Full-text

Application of global models and satellite data for smaller-scale groundwater recharge studies

10.5194/hess-2016-410 ◽

2016 ◽

Author(s):

Rogier Westerhoff ◽

Paul White ◽

Zara Rawlinson

Keyword(s):

New Zealand ◽

Satellite Data ◽

Input Data ◽

Large Scale ◽

Regional Scale ◽

Global Scale ◽

Global Models ◽

Scale Models ◽

Boundary Issues ◽

Rainfall Recharge

Abstract. Large-scale models and satellite data are increasingly used to characterise groundwater and its recharge at the global scale. Although these models have the potential to fill in data gaps and solve trans-boundary issues, they are often neglected in smaller-scale studies, since data are often coarse or uncertain. Large-scale models and satellite data could play a more important role in smaller-scale (i.e., national or regional) studies, if they could be adjusted to fit that scale. In New Zealand, large-scale models and satellite data are not used for groundwater recharge estimation at the national scale, since regional councils (i.e., the water managers) have varying water policy and models are calibrated at the local scale. Also, some regions have many localised ground observations (but poor record coverage), whereas others are data-sparse. Therefore, estimation of recharge is inconsistent at the national scale. This paper presents an approach to apply large-scale, global, models and satellite data to estimate rainfall recharge at the national to regional scale across New Zealand. We present a model, NGRM, that is largely inspired by the global-scale WaterGAP recharge model, but is improved and adjusted using national data. The NGRM model uses MODIS-derived ET and vegetation satellite data, and the available nation-wide datasets on rainfall, elevation, soil and geology. A valuable addition to the recharge estimation is the model uncertainty estimate, based on variance, covariance and sensitivity of all input data components in the model environment. This research shows that, with minor model adjustments and use of improved input data, large-scale models and satellite data can be used to derive rainfall recharge estimates, including their uncertainty, at the smaller scale, i.e., national and regional scale of New Zealand. The estimated New Zealand recharge of the NGRM model compare well to most local and regional lysimeter data and recharge models. The NGRM is therefore assumed to be capable to fill in gaps in data-sparse areas and to create more consistency between datasets from different regions, i.e., to solve trans-boundary issues. This research also shows that smaller-scale recharge studies in New Zealand should include larger boundaries than only a (sub-)aquifer, and preferably the whole catchment. This research points out the need for improved collaboration on the international to national to regional levels to further merge large-scale (global) models to smaller (i.e., national or regional) scales. Future research topics should, collaboratively, focus on: improvement of rainfall-runoff and snowmelt methods; inclusion of river recharge; further improvement of input data (rainfall, evapotranspiration, soil and geology); and the impact of recharge uncertainty in mountainous and irrigated areas.

Download Full-text

State and Perspectives of the Concept of Large‐Scale Conditioning of Regional Climate Modelling

10.20944/preprints201807.0533.v1 ◽

2018 ◽

Author(s):

Hans von Storch ◽

Leone Cavicchia ◽

Frauke Feser ◽

Delei Li

Keyword(s):

Large Scale ◽

Dynamical Downscaling ◽

Regional Climate ◽

Ocean Dynamics ◽

Small Scale ◽

Climate Modelling ◽

Coastal Regions ◽

Global Models ◽

Atmospheric Disturbances ◽

Low Level Jets

We review the state of dynamical downscaling with scale-constrained regional and global models. The methodology, in particular spectral nudging, has become a routine and well-researched tool for hindcasting climatologies of sub-synoptic atmospheric disturbances in coastal regions. At present, the spectrum of applications is expanding to other phenomena, but also to ocean dynamics and to extended forecasting. Also new diagnostic challenges are appearing such as spatial characteristics of small-scale phenomena such as Low Level Jets.

Download Full-text

Large-scale text classification with deeper and wider convolution neural network

International Journal of Simulation and Process Modelling ◽

10.1504/ijspm.2020.106977 ◽

2020 ◽

Vol 15 (1/2) ◽

pp. 120

Author(s):

Min Huang ◽

Wei Huang

Keyword(s):

Neural Network ◽

Text Classification ◽

Large Scale ◽

Convolution Neural Network

Download Full-text

Characterizing sources of high surface ozone events in the southwestern U.S. with intensive field measurements and two global models

10.5194/acp-2019-990 ◽

2019 ◽

Author(s):

Li Zhang ◽

Meiyun Lin ◽

Andrew O. Langford ◽

Larry W. Horowitz ◽

Christoph J. Senff ◽

...

Keyword(s):

Large Scale ◽

Las Vegas ◽

Surface Ozone ◽

Field Measurements ◽

Ambient Air ◽

Daily Maximum ◽

Global Models ◽

High Background ◽

Stratospheric Intrusions ◽

Surface O3

Abstract. The detection and attribution of high background ozone (O3) events in the southwestern U.S. is challenging but relevant to the effective implementation of the lowered National Ambient Air Quality Standard (NAAQS; 70 ppbv). Here we leverage intensive field measurements from the Fires, Asian, and Stratospheric TransportLas Vegas Ozone Study (FAST-LVOS) in MayJune 2017, alongside high-resolution simulations with two global models (GFDL-AM4 and GEOS-Chem), to pinpoint the sources of O3 during high-O3 events. We show stratospheric influence on four out of the ten events with daily maximum 8-hour average (MDA8) surface O3 above 65 ppbv in the greater Las Vegas region. While O3 produced from regional anthropogenic emissions dominates pollution in the Las Vegas Valley, stratospheric intrusions can mix with regional pollution to push surface O3 above 70 ppbv. GFDL-AM4 captures the key characteristics of deep stratospheric intrusions consistent with ozonesondes, lidar profiles, and co-located measurements of O3, CO, and water vapor at Angel Peak, whereas GEOS-Chem has difficulty simulating the observed features and underestimates observed O3 by ~ 20 ppbv at the surface. The two models also differ substantially during a wildfire event, with GEOS-Chem estimating ~ 15 ppbv greater O3, in better agreement with lidar observations. At the surface, the two models bracket the observed MDA8 O3 values during the wildfire event. Both models capture the large-scale transport of Asian pollution, but neither resolves some fine-scale pollution plumes, as evidenced from aerosol backscatter, aircraft, and satellite measurements. U.S. background O3 estimates from the two models differ by 5 ppbv on average and up to 15 ppbv episodically. Our multi-model approach tied closely to observational analysis yields process insights, suggesting that elevated background O3 may pose challenges to achieving a potentially lower NAAQS level (e.g., 65 ppbv) in the southwestern U.S.

Download Full-text

CharTeC-Net: An Efficient and Lightweight Character-Based Convolutional Network for Text Classification

Journal of Electrical and Computer Engineering ◽

10.1155/2020/9701427 ◽

2020 ◽

Vol 2020 ◽

pp. 1-7 ◽

Cited By ~ 2

Author(s):

Aboubakar Nasser Samatin Njikam ◽

Huan Zhao

Keyword(s):

Text Classification ◽

Building Block ◽

Large Scale ◽

State Of The Art ◽

Building Blocks ◽

Training Data ◽

Superior Performance ◽

Classification Problems ◽

Computationally Efficient ◽

Convolutional Network

This paper introduces an extremely lightweight (with just over around two hundred thousand parameters) and computationally efficient CNN architecture, named CharTeC-Net (Character-based Text Classification Network), for character-based text classification problems. This new architecture is composed of four building blocks for feature extraction. Each of these building blocks, except the last one, uses 1 × 1 pointwise convolutional layers to add more nonlinearity to the network and to increase the dimensions within each building block. In addition, shortcut connections are used in each building block to facilitate the flow of gradients over the network, but more importantly to ensure that the original signal present in the training data is shared across each building block. Experiments on eight standard large-scale text classification and sentiment analysis datasets demonstrate CharTeC-Net’s superior performance over baseline methods and yields competitive accuracy compared with state-of-the-art methods, although CharTeC-Net has only between 181,427 and 225,323 parameters and weighs less than 1 megabyte.

Download Full-text

Quantifying pollution inflow and outflow over East Asia in spring with regional and global models

Atmospheric Chemistry and Physics ◽

10.5194/acp-10-4221-2010 ◽

2010 ◽

Vol 10 (9) ◽

pp. 4221-4239 ◽

Cited By ~ 69

Author(s):

M. Lin ◽

T. Holloway ◽

G. R. Carmichael ◽

A. M. Fiore

Keyword(s):

Boundary Layer ◽

Air Pollution ◽

High Resolution ◽

Boundary Conditions ◽

East Asia ◽

Large Scale ◽

Regional Model ◽

Global Model ◽

Global Models ◽

High Resolution Models

Abstract. Understanding the exchange processes between the atmospheric boundary layer and the free troposphere is crucial for estimating hemispheric transport of air pollution. Most studies of hemispheric air pollution transport have taken a large-scale perspective using global chemical transport models with fairly coarse spatial and temporal resolutions. In support of United Nations Task Force on Hemispheric Transport of Air Pollution (TF HTAP; www.htap.org), this study employs two high-resolution atmospheric chemistry models (WRF-Chem and CMAQ; 36×36 km) driven with chemical boundary conditions from a global model (MOZART; 1.9×1.9°) to examine the role of fine-scale transport and chemistry processes in controlling pollution export and import over the Asian continent in spring (March 2001). Our analysis indicates the importance of rapid venting through deep convection that develops along the leading edge of frontal system convergence bands, which are not adequately resolved in either of two global models compared with TRACE-P aircraft observations during a frontal event. Both regional model simulations and observations show that frontal outflows of CO, O3 and PAN can extend to the upper troposphere (6–9 km). Pollution plumes in the global MOZART model are typically diluted and insufficiently lofted to higher altitudes where they can undergo more efficient transport in stronger winds. We use sensitivity simulations that perturb chemical boundary conditions in the CMAQ regional model to estimate that the O3 production over East Asia (EA) driven by PAN decomposition contributes 20% of the spatial averaged total O3 response to European (EU) emission perturbations in March, and occasionally contributes approximately 50% of the total O3 response in subsiding plumes at mountain observatories (at approximately 2 km altitude). The response to decomposing PAN of EU origin is strongly affected by the O3 formation chemical regimes, which vary with the model chemical mechanism and NOx/VOC emissions. Our high-resolution models demonstrate a large spatial variability (by up to a factor of 6) in the response of local O3 to 20% reductions in EU anthropogenic O3 precursor emissions. The response in the highly populated Asian megacities is 40–50% lower in our high-resolution models than the global model, suggesting that the source-receptor relationships inferred from the global coarse-resolution models likely overestimate health impacts associated with intercontinental O3 transport. Our results highlight the important roles of rapid convective transport, orographic forcing, urban photochemistry and heterogeneous boundary layer processes in controlling intercontinental transport; these processes may not be well resolved in the large-scale models.

Download Full-text