Personnel public expenditure and fiscal heterogeneity in Brazil: addressing spatial dependence in the presence of spatial nonstationarity

Spatial interpolation uses the geospatial features with known values and spatial relationships to predict unknown values. Spatial weights matrix is the most conventional way to represent spatial relationships when interpolating, where Euclidean distance is employed as the similarity measurement to tackle the spatial dependence problem. But spatial relationships of geospatial features should build on spatial nonstationarity besides spatial dependence. Therefore Euclidean distance is not suitable to represent spatial relationships over the whole geospatial space and local spatial analysis methods, which address the spatial nonstationarity should be employed to construct spatial weights matrix. This paper proposes accumulated similarity surface and brings in curve evolution and fast marching method to calculate the similarity of geographical features. Thus spatial weights matrix is constructed using accumulated similarity surfaces according to both spatial dependence and spatial nonstationarity. Experiments are conducted using ASTER DEM as experimental data. The interpolation results show that spatial weights matrix based on accumulated similarity surfaces performs better than Euclidean-distance-based spatial weights matrix.

Spatial dependencies and spatial drift in public transport seasonal ticket revenue data

The purpose of this research is to examine a multilevel approach to property hedonic models. Because individual properties are nested within neighborhoods within a city, a multilevel method is used to develop hedonic models. The author analyses several statistical and substantive reasons for explicitly modeling individual properties as belonging to neighborhoods. The reasons include spatial dependencies, non-stationarity and spatial heterogeneity. The conclusion is that multilevel models are not the only means of addressing issues of spatial dependencies and non-stationarity. The advantage of using a multilevel approach for property hedonic models is that they can analyze variables from different scales simultaneously, while also accounting for issues of spatial dependencies and non-stationarity.

Deep neural networks are being increasingly used for short-term traffic flow prediction, which can be generally categorized as convolutional (CNNs) or graph neural networks (GNNs). CNNs are preferable for region-wise traffic prediction by taking advantage of localized spatial correlations, whilst GNNs achieves better performance for graph-structured traffic data. When applied to region-wise traffic prediction, CNNs typically partition an underlying territory into grid-like spatial units, and employ standard convolutions to learn spatial dependence among the units. However, standard convolutions with fixed geometric structures cannot fully model the nonstationary characteristics of local traffic flows. To overcome the deficiency, we introduce deformable convolution that augments the spatial sampling locations with additional offsets, to enhance the modeling capability of spatial nonstationarity. On this basis, we design a deep deformable convolutional residual network, namely DeFlow-Net, that can effectively model global spatial dependence, local spatial nonstationarity, and temporal periodicity of traffic flows. Furthermore, to better fit with convolutions, we suggest to first aggregate traffic flows according to pre-conceived regions or self-organized regions based on traffic flows, then dispose to sequentially organized raster images for network input. Extensive experiments on real-world traffic flows demonstrate that DeFlow-Net outperforms GNNs and existing CNNs using standard convolutions, and spatial partition by pre-conceived regions or self-organized regions further enhances the performance. We also demonstrate the advantage of DeFlow-Net in maintaining spatial autocorrelation, and reveal the impacts of partition shapes and scales on deep traffic flow prediction.

Purpose – This paper aims to consider spatial effects in the analysis of the relationship of revenue and service quality. When firms’ customers are located in spatially dispersed areas, it can be difficult to manage service quality on a geographically small scale because the relative importance of service quality might vary spatially. Moreover, standard approaches discussed so far in the marketing science literature usually neglect spatial effects, such as spatial dependencies (e.g. spatial autocorrelation) and spatial drift (spatial non-stationarity). Design/methodology/approach – The authors propose a comprehensive but intelligible approach based on spatial econometric methods that cover spatial dependencies and spatial drift simultaneously. In particular, they incorporate the spatial expansion method (spatial drift) into spatial econometric models (e.g. spatial lag model). Findings – Using real company data on seasonal ticket revenue (dependent variable) and service quality (independent variables) of a regional public transport service provider, the authors find that the elasticity for the length of the public transport network is between 0.2 and 0.5, whereas the elasticity for the headway is between −0.2 and 0.6, for example. The authors control for several socio-economic, socio-demographic and land-use variables. Practical implications – Based on the empirical findings, the authors show that addressing spatial effects of service data can improve management’s ability to implement programs aimed at enhancing seasonal ticket revenue. Therefore, they derive a spatial revenue response function that enables managers to identify small-scale areas that are most efficient in terms of increasing revenue by service improvement. Originality/value – The paper addresses the need to account for spatial effects in revenue response functions of public transport companies.

Exploring the spatial relationships between various geological features and mineralization is not only conducive to understanding the genesis of ore deposits but can also help to guide mineral exploration by providing predictive mineral maps. However, most current methods assume spatially constant determinants of mineralization and therefore have limited applicability to detecting possible spatially non-stationary relationships between the geological features and the mineralization. In this paper, the spatial variation between the distribution of mineralization and its determining factors is described for a case study in the Dingjiashan Pb–Zn deposit, China. A local regression modeling technique, geological weighted regression (GWR), was leveraged to study the spatial non-stationarity in the 3D geological space. First, ordinary least-squares (OLS) regression was applied, the redundancy and significance of the controlling factors were tested, and the spatial dependency in Zn and Pb ore grade measurements was confirmed. Second, GWR models with different kernel functions in 3D space were applied, and their results were compared to the OLS model. The results show a superior performance of GWR compared with OLS and a significant spatial non-stationarity in the determinants of ore grade. Third, a non-stationarity test was performed. The stationarity index and the Monte Carlo stationarity test demonstrate the non-stationarity of all the variables throughout the area. Finally, the influences of the degree of non-stationary of all controlling factors on mineralization are discussed. The existence of significant non-stationarity of mineral ore determinants in 3D space opens up an exciting avenue for research into the prediction of underground ore bodies.

Spatial data exists practically everywhere, including the oil and gas industry. Several factors drive the distribution of the location of oil and gas wells: performance of existing wells, available acreage, need for operators to maintain a certain amount of production and to stay competitive. Some of the important parameters to consider in the design of a completion job for an unconventional oil and gas well are the length of lateral (and by extension perforated interval), number of stages, total pounds of proppants, total volume of fluid pumped, injection pressure and injection rate. In big data analytics and building of a regression model to capture the effects of these parameters on oil production, the practice has been to analyze wells in similar formations or similar basins, even when these wells are miles apart. Due to the presence of spatial autocorrelation and non-stationarity in such data, the recommended practice should be to take these spatial dependencies into account by using geographically weighted regression (GWR). In this paper, we present an application of GWR in location-based regression modeling to capture the effect of these completion parameters on the first six months of oil production in 5700 wells in the Bakken and Three Forks formation in North Dakota. GWR builds different models for every location, leading to a spatial distribution of variable coefficients. This model is well suited to capture both local and global variations in our dependent variable. We also compare the results obtained with that of three other models: multiple regression model, artificial neural network model and universal kriging. Just like the use of kriging, GWR model resulted in a much-improved prediction of oil production as captured by the goodness-of-fit diagnostics (R squared, AIC, and RMSPE), compared to the other two non-location-based models. We recommend the use of the GWR model in the prediction of oil or gas production when spatial non-stationarity exists.

Comparisons of spatial and non-spatial models for predicting soil carbon content based on visible and near-infrared spectral technology

Understanding where, when, what type and why crashes are occurring can help determine the most appropriate initiatives to reduce road trauma. Spatial statistical analysis techniques are better suited to analysing crashes than traditional statistical techniques as they allow for spatial dependency and non-stationarity. For example, crashes tend to cluster at specific locations (spatial dependency) and vary from one location to another (non-stationarity). Several spatial statistical methods were used to examine crash clustering in metropolitan Melbourne, including Global Moran’s I statistic, Kernel Density Estimation and Getis-Ord Gi* statistic. The Global Moran’s I statistic identified statistically significant clustering on a global level. The Kernel Density Estimation method showed clustering but could not identify the statistical significance. The Getis-Ord Gi* method identified local crash clustering and found that 15.7 per cent of casualty crash locations in metropolitan Melbourne were statistically significant hot spots at the 95 per cent confidence level. The degree, location and extent of clustering was found to vary for different crash categories, with fatal crashes exhibiting the lowest level of clustering and bicycle crashes exhibiting the highest level of clustering. Temporal variations in clustering were also observed. Overlaying the results with land use and road classification data found that hot spot clusters were in areas with a higher proportion of commercial land use and with a higher proportion of arterial and sub-arterial roads. Further work should investigate network based hot spot analysis and explore the relationship between crash clusters and influencing factors using spatial techniques such as Geographically Weighted Regression.

景观生态视角下草地生态质量的空间差异及其影响因素——以内蒙古四子王旗为例

Accurate short-term weather forecasting plays a vital role in disaster response, agriculture, and energy management, where timely and reliable predictions are essential for decision-making. Graph neural networks (GNNs), known for their ability to model complex spatial structures and relational data, have achieved remarkable success in meteorological forecasting by effectively capturing spatial dependencies among distributed weather stations. However, most existing GNN-based approaches rely on pairwise station connections, limiting their capacity to represent higher-order spatial interactions. Moreover, their dependence on supervised learning makes them vulnerable to spatial heterogeneity and temporal non-stationarity. This paper introduces a novel spatial–temporal pretraining framework, Hypergraph-enhanced Meteorological Pretraining (HyMePre), which combines hypergraph neural networks with self-supervised learning to model high-order spatial dependencies and improve generalization across diverse climate regimes. HyMePre employs a two-stage masking strategy, applying spatial and temporal masking separately, to learn disentangled representations from unlabeled meteorological time series. During forecasting, dynamic hypergraphs group stations based on meteorological similarity, explicitly capturing high-order dependencies. Extensive experiments on large-scale reanalysis datasets show that HyMePre outperforms conventional GNN models in predicting temperature, humidity, and wind speed. The integration of pretraining and hypergraph modeling enhances robustness to noisy data and improves generalization to unseen climate patterns, offering a scalable and effective solution for operational weather forecasting.

A key function of local governments is to provide a wide array of public services. The supply of these services has been found to create spatial spillovers among neighbouring municipalities. Although it is generally agreed that spillovers are present in models that explain government expenditures, their type—whether endogenous, exogenous or residual—and sign—whether positive or negative—remain ambiguous. In most cases, a subjective process is used to select the type of spatial regression model used in analysis, with mixed results. Per capita expenditures of ten subprogrammes (Security, Housing, Welfare, Environment, Social services, Employment promotion, Health, Education, Culture and Sport) are analysed for all Spanish municipalities with more than 5000 inhabitants in the 2010–2012 period. A Spatial Seemingly Unrelated Regression methodology in a panel framework is used to incorporate correlation between different subprogrammes and spatial dependence. Our results show that the three types of spatial effects are present. Furthermore, substantive dependence is positive in most cases, while negative residual dependence is observed in some.

Measuring spatial effects in the presence of institutional constraints: The case of Italian Local Health Authority expenditure

ABSTRACTIn the public economics literature expenditure needs, allocative efficiency and spatial dependence of local governments costs have been widely analysed separately implying bias estimations of the expenditure needs at local level. An original procedure that simultaneously takes into account the standard level of services, the allocative efficiency and the spatial proximity among Municipalities, has been proposed. The estimation strategy has been applied on a very detailed database of more than 4,000 Italian Municipalities for the year 2013.

Spatial nonstationarity in the stochastic frontier model: An application to the Italian wine industry