Multicomponent Spatial-Temporal Graph Attention Convolution Networks for Traffic Prediction with Spatially Sparse Data

Predicting traffic data on traffic networks is essential to transportation management. It is a challenging task due to the complicated spatial-temporal dependency. The latest studies mainly focus on capturing temporal and spatial dependencies with spatially dense traffic data. However, when traffic data become spatially sparse, existing methods cannot capture sufficient spatial correlation information and thus fail to learn the temporal periodicity sufficiently. To address these issues, we propose a novel deep learning framework, Multi-component Spatial-Temporal Graph Attention Convolutional Networks (MSTGACN), for traffic prediction, and we successfully apply it to predicting traffic flow and speed with spatially sparse data. MSTGACN mainly consists of three independent components to model three types of periodic information. Each component in MSTGACN combines dilated causal convolution, graph convolution layer, and the weight-shared graph attention layer. Experimental results on three real-world traffic datasets, METR-LA, PeMS-BAY, and PeMSD7-sparse, demonstrate the superior performance of our method in the case of spatially sparse data.

Exchange rate pass-through in Russian regions: Searching for spatial dependencies

Introduction. The article examines the spatial heterogeneity and factors of the exchange rate pass-through effect in consumer prices in Russian regions. Two hypotheses are tested. The first hypothesis is that there are differences in the magnitude of the passthrough between the Russian regions, the second is the significant influence of spatial relationships between regions on the magnitude of the pass-through effect. Theoretical analysis. The factors of the interregional differences in the pass-through effect are analyzed: the share of imports in the consumption structure, the share of value added produced in the domestic market in the final price of goods, transaction costs, the level of competition and the market structure. Empirical analysis. First pass-through estimates were obtained by means of vector autoregression model. Then the spatial dependence of the exchange rate pass-through was investigated on the basis of the global Moran and Geary indices, LISA, SAR and SEM models. Results. The results indicate the heterogeneity of the pass-through effect in Russian regions, which confirms the first of the hypotheses put forward. Confirmation of the second hypothesis was found only for food products in the short term, which is due to the nature of commodity flows between Russian regions. It is concluded that it is necessary to study the spatial relationships of the pass-through effect based on disaggregated prices.

Gambling Outlets Location in Urban Areas: A Case Study of Madrid

Cities are certainly a key factor in the location of gambling facilities. This paper aims to map the location of gambling outlets in urban areas and to examine potential links between neighborhoods socioeconomic and demographic characteristics and gambling supply, taking into account spatial dependencies of neighboring areas. This correlation is of interest because neighborhood characteristics may attract sellers, and because the presence of gambling sellers may cause changes in neighborhood demographics. Using detailed official data from the city of Madrid for the year 2017, three spatial econometric approaches are considered: spatial autoregressive (SAR) model, spatial error model (SEM) and spatial lag of X (explicative variables) model (SLX). Empirical analysis finds a strong correlation between neighborhoods characteristics and co-location of gambling outlets, highlighting a specific geographic patterning of distribution within more disadvantaged urban areas. This may have interesting implications for gambling stakeholders and for local governments when it comes to the introduction and/or increase of gambling availability.

Restoration of missing structural health monitoring data using spatiotemporal graph attention networks

For structural health monitoring systems with many low-cost sensors, missing data caused by sensor faults, power supply interruptions and data transmission errors are almost inevitable, significantly affecting structural diagnosis and evaluation. Considering the inherent spatial and temporal correlations in the sensor network, this study proposes a spatiotemporal graph attention network for restoration of missing data. The proposed model was stacked with a graph convolutional layer and several spatiotemporal blocks composed of spatial and temporal layers. The monitoring data of normal sensors were first mapped to all sensors through the graph convolutional layer, and attention mechanisms were used in the spatiotemporal blocks to model the spatial dependencies of sensors and the temporal dependencies of time steps, respectively. The extracted spatiotemporal features were assembled through a fully connected layer to reconstruct the missing signals. In this study, both homogeneous and heterogeneous monitoring items were used to calculate the spatial attention coefficients. The data restoration accuracy with and without the multi-source data fusion was discussed. Application on a long-span cable-stayed bridge to restore missing cable forces demonstrates that spatiotemporal attention modelling can achieve satisfactory restoring accuracy without any prior analysis.

A 10-year monitoring of soil properties dynamics and soil fertility evaluation in Chinese hickory plantation regions of southeastern China

Monitoring the temporal and spatial variation of soil properties is helpful to understand the evolution of soil properties and adjust the management method in time. Soil fertility evaluation is an urgent need to understand soil fertility level and prevent soil degradation. Here, we conducted an intensive field investigation in Chinese hickory (Carya cathayensis Sarg.) plantation to clarify the spatial and temporal variation of soil properties and its influencing factors, and to evaluate the change of soil fertility. The results showed that the soil pH and soil organic carbon (SOC) significantly increased from 2008 to 2018, while available nitrogen (AN) significantly decreased from 2008 to 2018. The semi-variance revealed that except available phosphorus (AP), the spatial dependencies of soil properties increased from 2008 to 2018. An increasing south-north gradient was found for soil AN, AP, available potassium (AK) and SOC and a decreasing south-north gradient was found for soil pH. The average soil fertility in the whole area was increased from 2008 to 2018. Our findings demonstrated that the changes of the management measures were the reason for the change of soil properties from 2008 to 2018. Therefore, rational fertilization strategies and sod cultivation are recommended to maintain the long-term development of the producing forest.

Exploring the Potential of the Cost-Efficient TAHMO Observation Data for Hydro-Meteorological Applications in Sub-Saharan Africa

The Trans-African Hydro-Meteorological Observatory (TAHMO) is a promising initiative aiming to install 20,000 stations in sub-Saharan Africa counteracting the decreasing trend of available measuring stations. To achieve this goal, it is particularly important that the installed weather stations are cost-efficient, appropriate for African conditions, and reliably measure the most important variables for hydro-meteorological applications. Since there exist no performance studies of TAHMO stations while operating in Africa, it is necessary to investigate their performance under different climate conditions. This study provides a first analysis of the performance of 10 selected TAHMO stations across Burkina Faso (BF). More specifically, the analysis consists of missing value statistics, plausibility tests of temperature (minimum, maximum) and precipitation, spatial dependencies (correlograms) by comparison with daily observations from synoptical stations of the BF meteorological service as well as cross-comparison between the TAHMO stations. Based on the results of this study for BF for the period from May 2017 to December 2020, it is concluded that TAHMO potentially offers a reliable and cost-efficient solution for applications in hydro-meteorology. The usage of wind speed measurements cannot be recommended without reservation, at least not without bias correcting of the data. The limited measurement period of TAHMO still prevents its usability in climate (impact) research. It is also stressed that TAHMO cannot replace existing observation networks operated by the local meteorological services, but it can be a complement and has great potential for detailed spatial analyses. Since restricted to BF in this analysis, more evaluation studies of TAHMO are needed considering different environmental and climate conditions across SSA.

Deep Active Context Estimation for Automated COVID-19 Diagnosis

Many studies on automated COVID-19 diagnosis have advanced rapidly with the increasing availability of large-scale CT annotated datasets. Inevitably, there are still a large number of unlabeled CT slices in the existing data sources since it requires considerable consuming labor efforts. Notably, cinical experience indicates that the neighboring CT slices may present similar symptoms and signs. Inspired by such wisdom, we propose DACE, a novel CNN-based deep active context estimation framework, which leverages the unlabeled neighbors to progressively learn more robust feature representations and generate a well-performed classifier for COVID-19 diagnosis. Specifically, the backbone of the proposed DACE framework is constructed by a well-designed Long-Short Hierarchical Attention Network (LSHAN), which effectively incorporates two complementary attention mechanisms, i.e., short-range channel interactions (SCI) module and long-range spatial dependencies (LSD) module, to learn the most discriminative features from CT slices. To make full use of such available data, we design an efficient context estimation criterion to carefully assign the additional labels to these neighbors. Benefiting from two complementary types of informative annotations from -nearest neighbors, i.e., the majority of high-confidence samples with pseudo labels and the minority of low-confidence samples with hand-annotated labels, the proposed LSHAN can be fine-tuned and optimized in an incremental learning manner. Extensive experiments on the Clean-CC-CCII dataset demonstrate the superior performance of our method compared with the state-of-the-art baselines.

Reconfiguring Periphery: Localizing Spatial Dependencies of Capitalism in West Asia and North Africa

This article discusses the relationship between processes of peripherialization, agency, and knowledge production in West Asia and North Africa, and, in so doing, introduces the META special issue on "Periphery".

Are Regional Differences in Psychological Characteristics and Their Correlates Robust? Applying Spatial-Analysis Techniques to Examine Regional Variation in Personality

There is growing evidence that psychological characteristics are spatially clustered across geographic regions and that regionally aggregated psychological characteristics are related to important outcomes. However, much of the evidence comes from research that relied on methods that are theoretically ill-suited for working with spatial data. The validity and generalizability of this work are thus unclear. Here we address two main challenges of working with spatial data (i.e., modifiable areal unit problem and spatial dependencies) and evaluate data-analysis techniques designed to tackle those challenges. To illustrate these issues, we investigate the robustness of regional Big Five personality differences and their correlates within the United States (Study 1; N = 3,387,303) and Germany (Study 2; N = 110,029). First, we display regional personality differences using a spatial smoothing approach. Second, we account for the modifiable areal unit problem by examining the correlates of regional personality scores across multiple spatial levels. Third, we account for spatial dependencies using spatial regression models. Our results suggest that regional psychological differences are robust and can reliably be studied across countries and spatial levels. The results also show that ignoring the methodological challenges of spatial data can have serious consequences for research concerned with regional psychological differences.