Meng Lu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Meng Lu has a total of 93 co-authors. Here are the top co-authors by number of shared publications:
- Shunping Ji
Wuhan University
11 shared publications - Derek Karssenberg
Department of Physical Geography, Faculty of Geosciences, Utrecht University, Utrecht, The Netherlands
6 shared publications Derek Karssenberg
4 shared publicationsFoeke Boersma
Department of Geography, University of Bayreuth, Universitätsstraße 30, 95447 Bayreuth, Germany
4 shared publicationsGerard Hoek
4 shared publicationsKees de Hoogh
4 shared publications
Latest Publications
- Open Access
- 17 July 2026
- Published by MDPI AG in Remote Sensing
- Vol. 18 (14) , 2381
Assessment of Feature Selection Methods for Machine Learning-Based Chlorophyll-a Retrieval Across Optical Water Types
Accurate retrieval of Chlorophyll-a (Chla, mg m−3) from ocean color reflectance remains a challenging issue due to spectral redundancy, nonlinear optical interactions, and water type variability. This study develops and evaluates a globally representative feature selection (FS) and machine learning (ML) framework to improve Chla estimation from multispectral reflectance. Using a quality-controlled in situ global dataset aggregated to Medium Resolution Imaging Spectrometer (MERIS) bands, we evaluate seven FS methods and five ML architectures across four optical water types (OWTs). The corresponding FS-ML models for each OWT are trained and validated based on partitioned subsets of in situ data using a novel data-partitioning scheme, ‘Spatially Blocked Stratified Monte-Carlo Split’. A three-stage model evaluation and a robustness filter are utilized. Importance-driven FS methods consistently produce compact, physically interpretable predictor sets and yielded the best generalization. Robust FS–ML combinations achieve high validation performance with minimal training–validation gaps. Cross-sensor transfer tests indicate that the MERIS-trained models could be generalized to independent Moderate Resolution Imaging Spectroradiometer (MODIS) and GlobColour matchups. Distributional uncertainty diagnostics further characterize retrieval confidence spatially and temporally. Overall, the OWT-adaptive FS provides a physically grounded, statistically robust, and operationally scalable approach for global Chla retrieval from multispectral ocean color reflectances.
- 17 July 2026
- Published by MDPI AG in Remote Sensing
- Vol. 18 (14) , 2381
Open AccessAssessment of Feature Selection Methods for Machine Learning-Based Chlorophyll-a Retrieval Across Optical Water Types
Accurate retrieval of Chlorophyll-a (Chla, mg m−3) from ocean color reflectance remains a challenging issue due to spectral redundancy, nonlinear optical interactions, and water type variability. This study develops and evaluates a globally representative feature selection (FS) and machine learning (ML) framework to improve Chla estimation from multispectral reflectance. Using a quality-controlled in situ global dataset aggregated to Medium Resolution Imaging Spectrometer (MERIS) bands, we evaluate seven FS methods and five ML architectures across four optical water types (OWTs). The corresponding FS-ML models for each OWT are trained and validated based on partitioned subsets of in situ data using a novel data-partitioning scheme, ‘Spatially Blocked Stratified Monte-Carlo Split’. A three-stage model evaluation and a robustness filter are utilized. Importance-driven FS methods consistently produce compact, physically interpretable predictor sets and yielded the best generalization. Robust FS–ML combinations achieve high validation performance with minimal training–validation gaps. Cross-sensor transfer tests indicate that the MERIS-trained models could be generalized to independent Moderate Resolution Imaging Spectroradiometer (MODIS) and GlobColour matchups. Distributional uncertainty diagnostics further characterize retrieval confidence spatially and temporally. Overall, the OWT-adaptive FS provides a physically grounded, statistically robust, and operationally scalable approach for global Chla retrieval from multispectral ocean color reflectances.
- Open Access
- 15 June 2026
- Published by Springer Nature in Mathematical Geosciences
- p. 1-27
Nyquist–Shannon Guidance for Linking Sampling Adequacy and Predictive Uncertainty in Geospatial Mapping
Modeling geospatial patterns of environmental variables at increasingly fine resolutions and broader scales highlights the reliability of map products built from sparse samples as a critical issue in scientific research and practical applications. It is desirable to attribute predictive uncertainty to identifiable sources, namely, sampling density, data noise, and model limitations. To approach this, we examine whether a spatial process imposes a minimum achievable predictive variance as a function of sampling density. We first present a transition from the geometric/spatial domain to the spectral domain, linking spatial autocorrelation to bandwidth frequency through the spectrum of the Gaussian radial basis function (RBF) kernel. We then apply the Nyquist–Shannon sampling theorem to link sampling rate and predictive uncertainty. The Gaussian RBF kernel is used as a demonstration vehicle instead of a restrictive assumption. We derive the effective bandwidth as a function of the kernel length scale and subsequently generalize this to a tolerance-based prediction error floor that quantifies unresolved high-frequency variance captured by the kernel for a given sampling rate. A practical implication is that sampling density induces fundamental, model-independent limits on achievable accuracy and resolution. Furthermore, this work provides a quantitative interpretation of kernel length scale, linking predictive variance to resolvable spectral energy, and provides guidance on sampling sufficiency, resolution limits, and uncertainty interpretation. Using both simulated fields and remote sensing imagery as real-world data, one- and two-dimensional experiments with and without covariates show that the Nyquist-based limits provide practical guidance for assessing sampling adequacy and quantifying predictive uncertainty in geospatial modeling.
- 15 June 2026
- Published by Springer Nature in Mathematical Geosciences
- p. 1-27
Open AccessNyquist–Shannon Guidance for Linking Sampling Adequacy and Predictive Uncertainty in Geospatial Mapping
Modeling geospatial patterns of environmental variables at increasingly fine resolutions and broader scales highlights the reliability of map products built from sparse samples as a critical issue in scientific research and practical applications. It is desirable to attribute predictive uncertainty to identifiable sources, namely, sampling density, data noise, and model limitations. To approach this, we examine whether a spatial process imposes a minimum achievable predictive variance as a function of sampling density. We first present a transition from the geometric/spatial domain to the spectral domain, linking spatial autocorrelation to bandwidth frequency through the spectrum of the Gaussian radial basis function (RBF) kernel. We then apply the Nyquist–Shannon sampling theorem to link sampling rate and predictive uncertainty. The Gaussian RBF kernel is used as a demonstration vehicle instead of a restrictive assumption. We derive the effective bandwidth as a function of the kernel length scale and subsequently generalize this to a tolerance-based prediction error floor that quantifies unresolved high-frequency variance captured by the kernel for a given sampling rate. A practical implication is that sampling density induces fundamental, model-independent limits on achievable accuracy and resolution. Furthermore, this work provides a quantitative interpretation of kernel length scale, linking predictive variance to resolvable spectral energy, and provides guidance on sampling sufficiency, resolution limits, and uncertainty interpretation. Using both simulated fields and remote sensing imagery as real-world data, one- and two-dimensional experiments with and without covariates show that the Nyquist-based limits provide practical guidance for assessing sampling adequacy and quantifying predictive uncertainty in geospatial modeling.
- Open Access
- 17 March 2026
- Published by Elsevier in SSRN Electronic Journal
Selection of Open Building Datasets for Health Facilities and Pipe Water Infrastructure Accessibility Modelling: A Case Study of Ashaiman Municipal District, Ghana
The proliferation of open building datasets (OBDs) offers new opportunities for urban infrastructure planning, yet practitioners lack systematic methods for selecting among them. We introduce a task-specific data quality assessment framework that evaluates OBDs based on their suitability for accessibility modelling rather than absolute accuracy. Our approach combines three components: (1) a task-relevancy algorithm using locality-sensitive hashing to compare building footprint patterns against normative planning standards (WHO 1km health facility access and SDG 6.1 piped water access), (2) multi-criteria accessibility modelling incorporating pipe networks, terrain, and population factors, and (3) rank aggregation. We applied this framework to Ashaiman, Ghana—a rapidly urbanising municipality with extensive informal settlements. Our research reveals task-dependent performance variations across four OBDs. For health accessibility, Google and GBA achieve higher task-relevancy scores (0.886 and 0.868) than Microsoft (0.829) andOpenStreetMap (0.753). Water accessibility shows different patterns, with GBA achieving the highest score (0.582), followed by Google (0.551) and Microsoft (0.539), while OSM scores lowest (0.524). GBA demonstrates the highest robustness to uncertainty, maintaining top rank in 98.4% of Monte Carlo simulations. However, the sensitivity analysis identifies four critical weight thresholds at which rankings change, underscoring that dataset selection must reflect local planning priorities. Our framework shows that selecting a building dataset is not a one-size-fits-all decision. The selection depends on whether the planning objective requires an accurate representation of existing conditions (favouring OSM and GBA) or alignment with normative standards (favouring GBA and Google).
- 17 March 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessSelection of Open Building Datasets for Health Facilities and Pipe Water Infrastructure Accessibility Modelling: A Case Study of Ashaiman Municipal District, Ghana
The proliferation of open building datasets (OBDs) offers new opportunities for urban infrastructure planning, yet practitioners lack systematic methods for selecting among them. We introduce a task-specific data quality assessment framework that evaluates OBDs based on their suitability for accessibility modelling rather than absolute accuracy. Our approach combines three components: (1) a task-relevancy algorithm using locality-sensitive hashing to compare building footprint patterns against normative planning standards (WHO 1km health facility access and SDG 6.1 piped water access), (2) multi-criteria accessibility modelling incorporating pipe networks, terrain, and population factors, and (3) rank aggregation. We applied this framework to Ashaiman, Ghana—a rapidly urbanising municipality with extensive informal settlements. Our research reveals task-dependent performance variations across four OBDs. For health accessibility, Google and GBA achieve higher task-relevancy scores (0.886 and 0.868) than Microsoft (0.829) andOpenStreetMap (0.753). Water accessibility shows different patterns, with GBA achieving the highest score (0.582), followed by Google (0.551) and Microsoft (0.539), while OSM scores lowest (0.524). GBA demonstrates the highest robustness to uncertainty, maintaining top rank in 98.4% of Monte Carlo simulations. However, the sensitivity analysis identifies four critical weight thresholds at which rankings change, underscoring that dataset selection must reflect local planning priorities. Our framework shows that selecting a building dataset is not a one-size-fits-all decision. The selection depends on whether the planning objective requires an accurate representation of existing conditions (favouring OSM and GBA) or alignment with normative standards (favouring GBA and Google).
- Open Access
- 13 March 2026
- Published by Copernicus GmbH
Reliable Predictive Resolution in GeospatialModelling
High-resolution geospatial prediction and satellite image downscaling are increasingly enabled by advances in machine learning and the availability of fine-scale covariates. However, predicted maps are often delivered on arbitrary grids that are not justified by the sampling density of observations. While uncertainty can be quantified at unobserved locations, the spatial scales over which predictions are supported by the data and the modelling process are typically not characterized. Besides computational and storage costs, critical consequences including over-interpretation, modelling noise, and most importantly, the apparent predictive resolution of spatial products can be misleading for downstream applications, potentially affecting scientific conclusions. An example is the use of predicted air pollution maps in health cohort studies to assess exposure–response relationships. This raises a fundamental but under-addressed question: what is the finest spatial resolution at which predictions are meaningfully supported by the data (and model)?We investigate how to meaningfully determine the predictive resolution in regression models by linking sampling density and model parameters in the frequency domain through spectral analysis. Two challenges are 1) to identify the sampling density in the multi-dimensional feature space, where the sampling typically becomes irregular; and 2) how to relate the frequency in the feature space to the spatial resolution. Using simulated and real-world geospatial datasets, we show that some arbitrarily selected output resolutions in existing literatures could exceed the data-supported predictive resolution, and could induce unnoticed biases or change-of-support issues in downstream analyses.
- 13 March 2026
- Published by Copernicus GmbH
Open AccessReliable Predictive Resolution in GeospatialModelling
High-resolution geospatial prediction and satellite image downscaling are increasingly enabled by advances in machine learning and the availability of fine-scale covariates. However, predicted maps are often delivered on arbitrary grids that are not justified by the sampling density of observations. While uncertainty can be quantified at unobserved locations, the spatial scales over which predictions are supported by the data and the modelling process are typically not characterized. Besides computational and storage costs, critical consequences including over-interpretation, modelling noise, and most importantly, the apparent predictive resolution of spatial products can be misleading for downstream applications, potentially affecting scientific conclusions. An example is the use of predicted air pollution maps in health cohort studies to assess exposure–response relationships. This raises a fundamental but under-addressed question: what is the finest spatial resolution at which predictions are meaningfully supported by the data (and model)?We investigate how to meaningfully determine the predictive resolution in regression models by linking sampling density and model parameters in the frequency domain through spectral analysis. Two challenges are 1) to identify the sampling density in the multi-dimensional feature space, where the sampling typically becomes irregular; and 2) how to relate the frequency in the feature space to the spatial resolution. Using simulated and real-world geospatial datasets, we show that some arbitrarily selected output resolutions in existing literatures could exceed the data-supported predictive resolution, and could induce unnoticed biases or change-of-support issues in downstream analyses.
- Open Access
- 13 March 2026
- Published by Copernicus GmbH
Spatial Heterogeneity of Compound Heat and Ozone Extremes: A Multivariate Extreme Value Perspective in Southern Germany
Extreme heat and surface ozone pollution frequently co-occur during summer and pose a growing risk to human health under climate warming. This co-occurrence is expected to intensify in the future, as global climate change is projected to increase the frequency, intensity, and duration of heatwaves. At the same time, ozone remains a major air quality concern in Europe despite substantial reductions in precursor emissions, and its health impacts are well documented even at concentrations below current regulatory standards.Previous studies have shown that temperature is a key meteorological driver of high ozone episodes, particularly in summer when photochemical activity is strongest. However, recent work on compound climate extremes has demonstrated that univariate or linear approaches can substantially underestimate risk when extremes occur simultaneously, highlighting the need for multivariate extreme-value methods.Moreover, spatial heterogeneity related to urbanization, land use, topography, and local meteorological conditions is often acknowledged but rarely examined in terms of how it modifies the occurrence and strength of heat–ozone extremes at the local scale. Ozone formation is governed by complex, nonlinear interactions between temperature, emissions, boundary-layer processes, and deposition, making its response highly variable in space and time. As a result, simple correlation-based analyses may underestimate the true influence of temperature on ozone, particularly under extreme conditions and in heterogeneous environments.Previous studies in Germany and Bavaria have linked ozone and temperature observations using either nearest station matching or reanalysis products such as ERA5, often assuming limited influence of urban heat island effects on daily maximum temperature. While this approach reflects established practice in regional air-quality studies, it may introduce uncertainty in spatially heterogeneous environments, particularly for local-scale compound extremes. Matching based solely on proximity or large-scale fields may not fully capture the influence of land use and station setting. To address this methodological challenge, the present study adopts a land-use–informed, multi-scale matching perspective to evaluate the sensitivity of compound heat–ozone dependence to spatial scale.This study aims to quantify compound heat–ozone extremes in Bavaria using multivariate analysis, with a focus on (i) how urban–rural setting and local spatial context shape compound risk, and (ii) how dependence strengthens during heatwave extremes and multi-day heatwave conditions.By bridging statistical extreme-value analysis with atmospheric chemistry interpretation, this work provides a physically consistent and regionally relevant assessment of heat–ozone risks in southern Germany.
- 13 March 2026
- Published by Copernicus GmbH
Open AccessSpatial Heterogeneity of Compound Heat and Ozone Extremes: A Multivariate Extreme Value Perspective in Southern Germany
Extreme heat and surface ozone pollution frequently co-occur during summer and pose a growing risk to human health under climate warming. This co-occurrence is expected to intensify in the future, as global climate change is projected to increase the frequency, intensity, and duration of heatwaves. At the same time, ozone remains a major air quality concern in Europe despite substantial reductions in precursor emissions, and its health impacts are well documented even at concentrations below current regulatory standards.Previous studies have shown that temperature is a key meteorological driver of high ozone episodes, particularly in summer when photochemical activity is strongest. However, recent work on compound climate extremes has demonstrated that univariate or linear approaches can substantially underestimate risk when extremes occur simultaneously, highlighting the need for multivariate extreme-value methods.Moreover, spatial heterogeneity related to urbanization, land use, topography, and local meteorological conditions is often acknowledged but rarely examined in terms of how it modifies the occurrence and strength of heat–ozone extremes at the local scale. Ozone formation is governed by complex, nonlinear interactions between temperature, emissions, boundary-layer processes, and deposition, making its response highly variable in space and time. As a result, simple correlation-based analyses may underestimate the true influence of temperature on ozone, particularly under extreme conditions and in heterogeneous environments.Previous studies in Germany and Bavaria have linked ozone and temperature observations using either nearest station matching or reanalysis products such as ERA5, often assuming limited influence of urban heat island effects on daily maximum temperature. While this approach reflects established practice in regional air-quality studies, it may introduce uncertainty in spatially heterogeneous environments, particularly for local-scale compound extremes. Matching based solely on proximity or large-scale fields may not fully capture the influence of land use and station setting. To address this methodological challenge, the present study adopts a land-use–informed, multi-scale matching perspective to evaluate the sensitivity of compound heat–ozone dependence to spatial scale.This study aims to quantify compound heat–ozone extremes in Bavaria using multivariate analysis, with a focus on (i) how urban–rural setting and local spatial context shape compound risk, and (ii) how dependence strengthens during heatwave extremes and multi-day heatwave conditions.By bridging statistical extreme-value analysis with atmospheric chemistry interpretation, this work provides a physically consistent and regionally relevant assessment of heat–ozone risks in southern Germany.
- Open Access
- 13 March 2026
- Published by Copernicus GmbH
Stability of Physically Modeled versus Data-Extracted End-Members for Multispectral Decomposition of Coastal Water Reflectance
The definition of end-members plays a central role in spectral decomposition of aquatic remote sensing reflectance, especially in highly turbid coastal waters where spectral signatures are strongly mixed and sensor dependent. End-members are defined as the purest reflectance spectra of water constituents and often could not be derived directly from observational data. Such data-driven end-members are often sensitive to noise, atmospheric correction uncertainties, and the reduced spectral resolution of multispectral sensors. Here, we examine how physically modeled end-members (MEMs) and end-members extracted from observations (EEMs) compare in terms of stability across different sensor types in the Dutch Wadden Sea.Physically modeled end-members were generated using a validated bio-optical forward model constrained by realistic ranges of optically active constituents. In parallel, EEMs were extracted from in-situ hyperspectral reflectance and from Sentinel-2 MSI and Sentinel-3 OLCI data using a geometric end-member extraction approach. The stability of MEMs and EEMs was evaluated through geometric inclusion analyses, spectral similarity measures, and reflectance reconstruction following Gaussian-based spectral decomposition.The comparison shows that MEMs remain consistent across in-situ hyperspectral and satellite-derived multispectral datasets, while EEMs tend to lose representativeness when applied to multispectral observations. This degradation is mainly linked to band aggregation effects and increased sensitivity to atmospheric correction uncertainties. In contrast, MEMs preserve their spectral geometry and reconstruction capability under these conditions.By separating the role of end-member definition from subsequent retrieval steps, this study demonstrates that physically constrained end-members provide a more robust foundation for multispectral spectral decomposition in optically complex coastal waters. These findings are particularly relevant for operational satellite monitoring applications where stability and transferability are essential.KeywordsRemote Sensing, Water constituents, Spectral Separation, Bio-optical Model, Remote Sensing Reflectance
- 13 March 2026
- Published by Copernicus GmbH
Open AccessStability of Physically Modeled versus Data-Extracted End-Members for Multispectral Decomposition of Coastal Water Reflectance
The definition of end-members plays a central role in spectral decomposition of aquatic remote sensing reflectance, especially in highly turbid coastal waters where spectral signatures are strongly mixed and sensor dependent. End-members are defined as the purest reflectance spectra of water constituents and often could not be derived directly from observational data. Such data-driven end-members are often sensitive to noise, atmospheric correction uncertainties, and the reduced spectral resolution of multispectral sensors. Here, we examine how physically modeled end-members (MEMs) and end-members extracted from observations (EEMs) compare in terms of stability across different sensor types in the Dutch Wadden Sea.Physically modeled end-members were generated using a validated bio-optical forward model constrained by realistic ranges of optically active constituents. In parallel, EEMs were extracted from in-situ hyperspectral reflectance and from Sentinel-2 MSI and Sentinel-3 OLCI data using a geometric end-member extraction approach. The stability of MEMs and EEMs was evaluated through geometric inclusion analyses, spectral similarity measures, and reflectance reconstruction following Gaussian-based spectral decomposition.The comparison shows that MEMs remain consistent across in-situ hyperspectral and satellite-derived multispectral datasets, while EEMs tend to lose representativeness when applied to multispectral observations. This degradation is mainly linked to band aggregation effects and increased sensitivity to atmospheric correction uncertainties. In contrast, MEMs preserve their spectral geometry and reconstruction capability under these conditions.By separating the role of end-member definition from subsequent retrieval steps, this study demonstrates that physically constrained end-members provide a more robust foundation for multispectral spectral decomposition in optically complex coastal waters. These findings are particularly relevant for operational satellite monitoring applications where stability and transferability are essential.KeywordsRemote Sensing, Water constituents, Spectral Separation, Bio-optical Model, Remote Sensing Reflectance
- Open Access
- 2 October 2025
- Published by Copernicus GmbH in Geoscientific Model Development
- Vol. 18 (19) , 6717-6735
A close look at using national ground stations for the statistical modeling of NO2
Air pollution leads to various health and societal issues. Modeling and predicting air pollution over space have important implications in health studies, urban planning, and policy-making. Many statistical models have been developed to understand the relationships between geospatial data and air pollution sources. An important aspect often neglected is spatial heterogeneity; however, the relationships between geographically distributed variables and air pollutants commonly vary over space. This study aims to evaluate and compare various spatial and non-spatial statistical modeling (including machine learning) methods within different spatial groups. The spatial groups are defined by traffic- and population-related variables. Models are classified into local and global models. Local models use air pollution measurements from the Amsterdam area. Global models use ground station observations in Germany and in the Netherlands. We found that prediction accuracy differs substantially in different spatial groups. Predictions for places near roads with high populations show poor prediction accuracy, while prediction accuracy increases in low-population-density areas for both local and global models. The prediction accuracy is further increased in places far from roads for global models. Modeling of air pollution in different spatial groups shows that nonlinear methods can have higher prediction accuracy than linear methods. The spatial prediction patterns of global models show that nonlinear methods generally are less sensitive to extreme values compared to linear methods. Additionally, clusters of predicted air pollution differ between models within cities despite similar prediction accuracy. Also, the influence of predictors on NO2 concentrations varies across different cities. Using the local dataset of our study and explicitly accounting for spatial autocorrelation in the universal and ordinary kriging models does not improve accuracy; however, analyzing prediction performance across spatial groups provides valuable insights. Comparing local and global prediction patterns reveals that local models capture regional clusters of high air pollution, which are not detected by global models. These findings highlight the fact that solely relying on overall prediction accuracy can be insufficient and potentially misleading, underscoring the importance of considering spatial variability and model performance within different spatial groups.
- 2 October 2025
- Published by Copernicus GmbH in Geoscientific Model Development
- Vol. 18 (19) , 6717-6735
Open AccessA close look at using national ground stations for the statistical modeling of NO2
Air pollution leads to various health and societal issues. Modeling and predicting air pollution over space have important implications in health studies, urban planning, and policy-making. Many statistical models have been developed to understand the relationships between geospatial data and air pollution sources. An important aspect often neglected is spatial heterogeneity; however, the relationships between geographically distributed variables and air pollutants commonly vary over space. This study aims to evaluate and compare various spatial and non-spatial statistical modeling (including machine learning) methods within different spatial groups. The spatial groups are defined by traffic- and population-related variables. Models are classified into local and global models. Local models use air pollution measurements from the Amsterdam area. Global models use ground station observations in Germany and in the Netherlands. We found that prediction accuracy differs substantially in different spatial groups. Predictions for places near roads with high populations show poor prediction accuracy, while prediction accuracy increases in low-population-density areas for both local and global models. The prediction accuracy is further increased in places far from roads for global models. Modeling of air pollution in different spatial groups shows that nonlinear methods can have higher prediction accuracy than linear methods. The spatial prediction patterns of global models show that nonlinear methods generally are less sensitive to extreme values compared to linear methods. Additionally, clusters of predicted air pollution differ between models within cities despite similar prediction accuracy. Also, the influence of predictors on NO2 concentrations varies across different cities. Using the local dataset of our study and explicitly accounting for spatial autocorrelation in the universal and ordinary kriging models does not improve accuracy; however, analyzing prediction performance across spatial groups provides valuable insights. Comparing local and global prediction patterns reveals that local models capture regional clusters of high air pollution, which are not detected by global models. These findings highlight the fact that solely relying on overall prediction accuracy can be insufficient and potentially misleading, underscoring the importance of considering spatial variability and model performance within different spatial groups.
- Open Access
Evaluating the Quality of Open Building Datasets for Mapping Urban Inequality: A Comparative Analysis Across 5 Cities
While informal settlements lack focused development and are highly dynamic, the quality of spatial data for these places may be uncertain. This study evaluates the quality and biases of AI-generated Open Building Datasets (OBDs) generated by Google and Microsoft against OpenStreetMap (OSM) data, across diverse global cities including Accra, Nairobi, Caracas, Berlin, and Houston. The Intersection over Union (IoU), overlap analysis and a positional accuracy algorithm are used to analyse the similarity and alignment of the datasets. The paper also analyses the size distribution of the building polygon area, and completeness using predefined but regular spatial units. The results indicate significant variance in data quality, with Houston and Berlin demonstrating high alignment and completeness, reflecting their structured urban environments. There are gaps in the datasets analysed, and cities like Accra and Caracas may be under-represented. This could highlight difficulties in capturing complex or informal regions. The study also notes different building size distributions, which may be indicative of the global socio-economic divide. These findings may emphasise the need to consider the quality of global building datasets to avoid misrepresentation, which is an important element of planning and resource distribution.
Open AccessEvaluating the Quality of Open Building Datasets for Mapping Urban Inequality: A Comparative Analysis Across 5 Cities
While informal settlements lack focused development and are highly dynamic, the quality of spatial data for these places may be uncertain. This study evaluates the quality and biases of AI-generated Open Building Datasets (OBDs) generated by Google and Microsoft against OpenStreetMap (OSM) data, across diverse global cities including Accra, Nairobi, Caracas, Berlin, and Houston. The Intersection over Union (IoU), overlap analysis and a positional accuracy algorithm are used to analyse the similarity and alignment of the datasets. The paper also analyses the size distribution of the building polygon area, and completeness using predefined but regular spatial units. The results indicate significant variance in data quality, with Houston and Berlin demonstrating high alignment and completeness, reflecting their structured urban environments. There are gaps in the datasets analysed, and cities like Accra and Caracas may be under-represented. This could highlight difficulties in capturing complex or informal regions. The study also notes different building size distributions, which may be indicative of the global socio-economic divide. These findings may emphasise the need to consider the quality of global building datasets to avoid misrepresentation, which is an important element of planning and resource distribution.
- Open Access
- 24 May 2025
- Published by Copernicus GmbH in The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
- Vol. XLVIII-M-7, 29-34
Dimension Expansion-based Spatiotemporal Land Cover Change Detection: A Study Case Using Sentinel-2 Satellite Time Series
Satellite time series data enable continuous land cover change detection, classification, and monitoring across large geographical areas. Time series-based statistical methods for abrupt change detection remain widely used in understanding and monitoring environmental dynamics but face limitations, including sensitivity to noise, challenges in differentiating change classes and causes, detecting change in near real-time, and incomplete uncertainty quantification. These challenges are obvious in cultivated lands, where the seasonality and cultivated areas often alter in different years. On the other hand, change detection and classification in space-time is difficult due to the nonstationary presented in data. In this study we used dimension expansion-based approach that projects data to higher dimensionality for stationarity and understand the change of spatial stationarity over time. Our case study focuses on vegetation dynamics in a cultivated and managed terrestrial area in the Takamanda National Park in Cameroon, a protected area of significant ecological value, using Sentinel-2 satellite time series data. The results imply the possibility of new spatiotemporal approach that is robustness against noise and enables near-real-time monitoring.
- 24 May 2025
- Published by Copernicus GmbH in The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
- Vol. XLVIII-M-7, 29-34
Open AccessDimension Expansion-based Spatiotemporal Land Cover Change Detection: A Study Case Using Sentinel-2 Satellite Time Series
Satellite time series data enable continuous land cover change detection, classification, and monitoring across large geographical areas. Time series-based statistical methods for abrupt change detection remain widely used in understanding and monitoring environmental dynamics but face limitations, including sensitivity to noise, challenges in differentiating change classes and causes, detecting change in near real-time, and incomplete uncertainty quantification. These challenges are obvious in cultivated lands, where the seasonality and cultivated areas often alter in different years. On the other hand, change detection and classification in space-time is difficult due to the nonstationary presented in data. In this study we used dimension expansion-based approach that projects data to higher dimensionality for stationarity and understand the change of spatial stationarity over time. Our case study focuses on vegetation dynamics in a cultivated and managed terrestrial area in the Takamanda National Park in Cameroon, a protected area of significant ecological value, using Sentinel-2 satellite time series data. The results imply the possibility of new spatiotemporal approach that is robustness against noise and enables near-real-time monitoring.
- Open Access
Geoinformation dependencies in geographic space and beyond
The use of geospatially dependent information, which has been stipulated as a law in geography, to model geographic patterns forms the cornerstone of geostatistics, and has been inherited in many data science based techniques as well, such as statistical learning algorithms. Still, we observe hesitations in interpreting geographic dependency scientifically as a property in geography, since interpretations of such dependency are subject to model choice with different hypotheses of trends and stationarity. Rather than questioning what can be considered as trends or why it is non-stationary, in this work, we share and consolidate a view that the properties of geographic dependency, being it trending or stationary, are essentially variations can be explained further by unobserved or unknown predictors, and not intrinsic to geographic space. Particularly, geoinformation dependency properties are in fact a projection of high dimensional feature space formed by all potential predictors into the lower dimension of geographic space, where geographic coordinates are equivalent to other predictors for modelling geographic patterns. This work brings together different aspects of geographic dependency, including similarity and heterogeneity, under a coherent framework, and aligns with the understanding of modelling in high dimensional feature space with different modelling concept including the classical geostatistics, Gaussian Process Regression and popular data science based spatial modelling techniques.
Open AccessGeoinformation dependencies in geographic space and beyond
The use of geospatially dependent information, which has been stipulated as a law in geography, to model geographic patterns forms the cornerstone of geostatistics, and has been inherited in many data science based techniques as well, such as statistical learning algorithms. Still, we observe hesitations in interpreting geographic dependency scientifically as a property in geography, since interpretations of such dependency are subject to model choice with different hypotheses of trends and stationarity. Rather than questioning what can be considered as trends or why it is non-stationary, in this work, we share and consolidate a view that the properties of geographic dependency, being it trending or stationary, are essentially variations can be explained further by unobserved or unknown predictors, and not intrinsic to geographic space. Particularly, geoinformation dependency properties are in fact a projection of high dimensional feature space formed by all potential predictors into the lower dimension of geographic space, where geographic coordinates are equivalent to other predictors for modelling geographic patterns. This work brings together different aspects of geographic dependency, including similarity and heterogeneity, under a coherent framework, and aligns with the understanding of modelling in high dimensional feature space with different modelling concept including the classical geostatistics, Gaussian Process Regression and popular data science based spatial modelling techniques.