Geospatial Big Data Research Articles

Deep Learning-Based Weather Prediction: A Survey

Abstract Weather forecasting plays a fundamental role in the early warning of weather impacts on various aspects of human livelihood. For instance, weather forecasting provides decision making support for autonomous vehicles to reduce traffic accidents and congestions, which completely depend on the sensing and predicting of external environmental factors such as rainfall, air visibility and so on. Accurate and timely weather prediction has always been the goal of meteorological scientists. However, the conventional theory-driven numerical weather prediction (NWP) methods face many challenges, such as incomplete understanding of physical mechanisms, difficulties in obtaining useful knowledge from the deluge of observation data, and the requirement of powerful computing resources. With the successful application of data-driven deep learning method in various fields, such as computer vision, speech recognition, and time series prediction, it has been proven that deep learning method can effectively mine the temporal and spatial features from the spatio-temporal data. Meteorological data is a typical big geospatial data. Deep learning-based weather prediction (DLWP) is expected to be a strong supplement to the conventional method. At present, many researchers have tried to introduce data-driven deep learning into weather forecasting, and have achieved some preliminary results. In this paper, we survey the state-of-the-art studies of deep learning-based weather forecasting, in the aspects of the design of neural network (NN) architectures, spatial and temporal scales, as well as the datasets and benchmarks. Then we analyze the advantages and disadvantages of DLWP by comparing it with the conventional NWP, and summarize the potential future research topics of DLWP.

Applications of Google Earth Engine in fluvial geomorphology for detecting river channel change

AbstractCloud‐based computing, access to big geospatial data, and virtualization, whereby users are freed from computational hardware and data management logistics, could revolutionize remote sensing applications in fluvial geomorphology. Analysis of multitemporal, multispectral satellite imagery has provided fundamental geomorphic insight into the planimetric form and dynamics of large river systems, but information derived from these applications has largely been used to test existing concepts in fluvial geomorphology, rather than for generating new concepts or theories. Traditional approaches (i.e., desktop computing) have restricted the spatial scales and temporal resolutions of planimetric river channel change analyses. Google Earth Engine (GEE), a cloud‐based computing platform for planetary‐scale geospatial analyses, offers the opportunity to relieve these spatiotemporal restrictions. We summarize the big geospatial data flows available to fluvial geomorphologists within the GEE data catalog, focus on approaches to look beyond mapping wet channel extents and instead map the wider riverscape (i.e., water, sediment, vegetation) and its dynamics, and explore the unprecedented spatiotemporal scales over which GEE analyses can be applied. We share a demonstration workflow to extract active river channel masks from a section of the Cagayan River (Luzon, Philippines) then quantify centerline migration rates from multitemporal data. By enabling fluvial geomorphologists to take their algorithms to petabytes worth of data, GEE is transformative in enabling deterministic science at scales defined by the user and determined by the phenomena of interest. Equally as important, GEE offers a mechanism for promoting a cultural shift toward open science, through the democratization of access and sharing of reproducible code.This article is categorized under: Science of Water

Variability in individual home-work activity patterns

The way people allocate time across home and work activities determines their commuting patterns and frames much of the activities they undertake in the urban space. While inter-personal and intra-personal variability and repetitiveness in these activities have been documented, they remain largely underexplored. This study highlights the variations in and between individual home-work activity patterns by using information from metro smart card data as a proxy. To this end, the concept of individual space time usage matrix (STUM) is proposed and an analytical framework is developed in support of its use to depict how each rider allocates time in the vicinity of metro stations spatially and temporally. With this framework, we can classify space-time activity patterns that can be traced back to behavioral variability. By using Wuhan, China as a case study, variability in the number of home/work locations in personal activity patterns, and flexibility of work timeframes are investigated inter- and intra-personally. Our results show that about 25% of the population has a sophisticated home-work activity pattern that does not confirm to the ordinary 1-home 1-workplace pattern. Furthermore, even for this latter group, we find quite differentiated home and work timeframe patterns. The STUM is proved to be an effective and efficient concept to create a personal profile in analyzing the activity variability with big geo-spatial data.

Accurate Estimation of the Proportion of Mixed Land Use at the Street-Block Level by Integrating High Spatial Resolution Images and Geospatial Big Data

Mixed land use has been widely used as a planning tool to improve the functionality of cities. However, depicting mixed land use is rather difficult due to its complexities. Previous studies have decomposed urban land areas using either remote sensing images or geospatial big data. Few studies have combined these two data sources because of the lack of methodologies. This article proposed an end-to-end two-stream convolutional neural network (CNN) for combining features (CF-CNN) to estimate the proportion of mixed land use by integrating high spatial resolution (HSR) images and geospatial big data of real-time Tencent user density (RTUD) data. Two deep learning networks, one for image information extraction and other for human activity-related information extraction, are used to construct two branches of CF-CNN. The mixed land use can be described by calculating the proportions of each land use type at the street-block level. Compared with methods for using single-source data, CF-CNN obtained the highest classification accuracy. We further applied the Shannon diversity index (SHDI) to quantify the agglomerated urban mixed land use. The Spearman correlation coefficients among the SHDI, community distance, and neighborhood vibrancy were calculated to verify the effectiveness of the mixed land use composition. Our framework provided an alternative way of identifying mixed land use structures by integrating multisource data.

Fire stations siting with multiple objectives and geospatial big data

In recent years, with the rapid development of urbanization, the urban public emergency management faces increasing challenges, and the number of fire incidents has increased largely. For mitigating injury risk and reducing property loss, fire station locations need to be optimized to provide efficient fire emergency services. However, the locations of fire facilities in China are mainly determined according to administrative divisions, lacking effective data-driven applications. In addition, existing location models are designed for general purpose, and few have taken into account the unique characteristics of fire services with multiple objectives, i.e., maximum coverage, minimum overlapping coverage and balanced workloads. With the recent development of geospatial big data, lots of fire risk related data (e.g. fire incidents data and travel time data) can be obtained, which provides an unprecedented opportunity for urban fire emergency facilities planning. In this paper, we propose to establish the multi-objective maximal covering location model by accurately estimating fire rescuing demands and travel cost from geospatial big data. A case study was carried out in Nanjing, China, and the result shows that our method is effective in optimizing the locations of fire stations.

Location-aware systems or location-based services: a survey with applications to CoViD-19 contact tracking

With the CoViD-19 pandemic, location awareness technologies have seen renewed interests due to the numerous contact tracking mobile application variants developed, deployed, and discussed. For some, location-aware applications are primarily a producer of geospatial Big Data required for vital geospatial analysis and visualization of the spread of the disease in a state of emergency. For others, comprehensive tracking of citizens constitutes a dangerous violation of fundamental rights. Commercial web-based location-aware applications both collect data and—through spatial analysis and connection to services—provide value to users. This value is what motivates users to share increasingly private and comprehensive data. The willingness of users to share data in return for services has been a key concern with web-based variants of the technology since the beginning. With a focus on two privacy preserving CoViD-19 contact tracking applications, this survey walks through the key steps of developing a privacy preserving context-aware application: from types of applications and business models, through architectures and privacy strategies, to representations.

Mapping China’s regional economic activity by integrating points-of-interest and remote sensing data with random forest

Nighttime light imageries are widely used for mapping the gross domestic product (GDP) over large areas. However, nighttime light imagery is inappropriate to disaggregate agricultural GDP and inadequate to differentiate the GDP from the secondary and tertiary sectors. Points-of-interest, a kind of geospatial big data with geographic locations and textual descriptions of the category, can effectively distinguish industrial and commercial areas, and therefore have the potential to improve the precise GDP mapping from secondary and tertiary sectors. In this study, a machine learning method, random forest, was used to disaggregate the 2010 county-level census GDP data of mainland China to 1 km × 1 km grids. Six Random Forest models were constructed for different economic sectors to explore the non-linear relationships between various geographic predictors and GDP from different sectors. By fusing points-of-interest of varying categories, the spatial distribution of economic activities from the secondary and tertiary sectors was effectively distinguished. Compared to previous studies, the strategy of developing specific Random Forest models for different sectors generated a more reasonable distribution of GDP. Our results highlight the feasibility of using point-of-interest data in disaggregating non-agricultural GDP by exploiting the complementary features of the different data sources.

Seyahat Süresi Değişkenliğinin ve Güvenilirliğinin İstatistiksel Analizi

Şehir içi ulaşım ağlarında bir noktadan başka bir noktaya yapılan yolculukların seyahat sürelerinin istatistiksel olarak modellenmesi ve bunların güvenirliğinin analiz edilmesi büyük önem arz etmektedir. Gerçekleştirilen kimi çalışmalarda seyahat süresini modellemek için normal dağılım gibi teorik olarak uygun olmayan dağılımlar tercih edilmiştir. Bunun yanında birçok çalışma haftanın günleri hafta içi ve hafta sonu olmak üzere iki farklı sınıfta değerlendirmektedir. Gelişen sensör teknolojileri ile elde edilen büyük ulaşım verisi üzerinden seyahat sürelerinin istatiksel dağılımı ve güvenirliği gibi şehir hayatına etki eden faktörler analiz edilebilmektedir. Bu çalışmada New York şehrine ait açık ulaşım verisi analiz edilerek, birçok çalışmanın varsaydığı hafta içi-hafta sonu ayrımının gerçekçi olmayabileceği ve bir gün içindeki farklı zaman dilimlerinde istatistiksel sonuçların farklı çıkabileceği tespit edilmiştir. John F. Kennedy ve LaGuardia havalimanları arasında 07:00-19:00 arasında gerçekleşen taksi yolculuklarının seyahat süresi değişkenliği ve dağılımları incelendiğinde, seyahat sürelerinin Log-Lojistik (3P) dağılımına uyduğu tespit edilmiştir. Bu dağılım Salı ve Çarşamba günleri baskın dağılım değildir ve bu günlerde seyahat süresi güvenilirliğinin düşük olduğu zaman dilimlerinin bulunduğu tespit edilmiştir.

CLOUD-DESKTOP REMOTE SENSING DATA MANAGEMENT TO ENSURE TIME SERIES ANALYSIS, INTEGRATION OF QGIS AND GOOGLE EARTH ENGINE

Abstract. Our context research is conducted to investigate the possibility of common application of the remote sensing and ground-based monitoring data to detection and observation of the dynamics and change in climate and vegetation cover parameters. We applied the analysis of the annual graphs of Normalized Difference Water Index to estimate the length and time frames of growing seasons. Basing on previously gained results, we concluded that we can use the Index-based monitoring of growing season parameters as a relevant technique. We are working on automation of computations that can be applied to processing satellite imagery, computing Normalized Difference Water Index time series (in the forms of maps and annual graphs), and estimation of growing season parameters. As currently used data amounts are big (or up-to-big) geospatial data, we use the Google Earth Engine platform to process initial datasets. Our currently described experimental work incorporates investigation of the possibilities for integration of cloud computing data storage and processing with client-side data representation in universal desktop GISs. To ensure our study needs we developed a prototype of a QGIS plugin capable to run processing in GEE and represent results in QGIS.

Stand with #StandingRock: Envisioning an Epistemological Shift in Understanding Geospatial Big Data in the “Post-truth” Era

This article examines an online protest to develop a contextual interpretation of geospatial big data and to challenge the debates on post-truth. In 2016, in support of the local protests against a crude oil pipeline passing through the region, social media users around the world remotely checked in to the Standing Rock Indian Reservation in North Dakota using a technique that we call location spoofing. This collective action thus generated a massive volume of inconsistent locational data that could easily be treated as fake information. Our contextual approach, integrating anthropological methodology, interprets user-generated big data as digital traces of human activities in the broader social–technological network of the involved human and nonhuman actors. This study reveals that the online protesters’ use of location-based features and content recommendation algorithm challenged not only the political and technological authorities but also, at a more profound level, the established ways for determining what is true and who gets to decide what is true for what purpose. We argue that this case of decentralized data generation and dissemination demonstrates an ongoing reconfiguration of the previously established regime of truth that has been monopolized by scientists or politicians. Our unique contextual approach to data interpretation is a pioneering effort to update epistemological assumptions about truth in today’s data-intensive environments. We call on scholars from geography, GIScience, and other disciplines to collectively envision an epistemological shift in our continuous pursuit of knowledge and truth in the so-called post-truth era.

Introduction to Big Data Computing for Geospatial Applications

The convergence of big data and geospatial computing has brought challenges and opportunities to GIScience with regards to geospatial data management, processing, analysis, modeling, and visualization. This special issue highlights recent advancements in integrating new computing approaches, spatial methods, and data management strategies to tackle geospatial big data challenges and meanwhile demonstrates the opportunities for using big data for geospatial applications. Crucial to the advancements highlighted here is the integration of computational thinking and spatial thinking and the transformation of abstract ideas and models to concrete data structures and algorithms. This editorial first introduces the background and motivation of this special issue followed by an overview of the ten included articles. Conclusion and future research directions are provided in the last section.

Measuring Functional Urban Shrinkage with Multi-Source Geospatial Big Data: A Case Study of the Beijing-Tianjin-Hebei Megaregion

Most of the shrinking cities experience an unbalanced deurbanization across different urban areas in cities. However, traditional ways of measuring urban shrinkage are focused on tracking population loss at the city level and are unable to capture the spatially heterogeneous shrinking patterns inside a city. Consequently, the spatial mechanism and patterns of urban shrinkage inside a city remain less understood, which is unhelpful for developing accommodation strategies for shrinkage. The smart city initiatives and practices have provided a rich pool of geospatial big data resources and technologies to tackle the complexity of urban systems. Given this context, we propose a new measure for the delineation of shrinking areas within cities by introducing a new concept of functional urban shrinkage, which aims to capture the mismatch between urban built-up areas and the areas where significantly intensive human activities take place. Taking advantage of a data fusion approach to integrating multi-source geospatial big data and survey data, a general analytical framework is developed to construct functional shrinkage measures. Specifically, Landsat-8 remote sensing images were used for extracting urban built-up areas by supervised neural network classifications and Geographic Information System tools, while cellular signaling data from China Unicom Inc. was used to depict human activity areas generated by spatial clustering methods. Combining geospatial big data with urban land-use functions obtained from land surveys and Points-Of-Interests data, the framework further enables the comparison between cities from dimensions characterized by indices of spatial and urban functional characteristics and the landscape fragmentation; thus, it has the capacity to facilitate an in-depth investigation of fundamental causes and internal mechanisms of urban shrinkage. With a case study of the Beijing-Tianjin-Hebei megaregion using data from various sources collected for the year of 2018, we demonstrate the validity of this approach and its potential generalizability for other spatial contexts in facilitating timely and better-informed planning decision support.

Improved land cover map of Iran using Sentinel imagery within Google Earth Engine and a novel automatic workflow for land cover classification using migrated training samples

Accurate information about the location, extent, and type of Land Cover (LC) is essential for various applications. The only recent available country-wide LC map of Iran was generated in 2016 by the Iranian Space Agency (ISA) using Moderate Resolution Imaging Spectroradiometer (MODIS) images with a considerably low accuracy. Therefore, the production of an up-to-date and accurate Iran-wide LC map using the most recent remote sensing, machine learning, and big data processing algorithms is required. Moreover, it is important to develop an efficient method for automatic LC generation for various time periods without the need to collect additional ground truth data from this immense country. Therefore, this study was conducted to fulfill two objectives. First, an improved Iranian LC map with 13 LC classes and a spatial resolution of 10 m was produced using multi-temporal synergy of Sentinel-1 and Sentinel-2 satellite datasets applied to an object-based Random forest (RF) algorithm. For this purpose, 2,869 Sentinel-1 and 11,994 Sentinel-2 scenes acquired in 2017 were processed and classified within the Google Earth Engine (GEE) cloud computing platform allowing big geospatial data analysis. The Overall Accuracy (OA) and Kappa Coefficient (KC) of the final Iran-wide LC map for 2017 was 95.6% and 0.95, respectively, indicating the considerable potential of the proposed big data processing method. Second, an efficient automatic method was developed based on Sentinel-2 images to migrate ground truth samples from a reference year to automatically generate an LC map for any target year. The OA and KC for the LC map produced for the target year 2019 were 91.35% and 0.91, respectively, demonstrating the efficiency of the proposed method for automatic LC mapping. Based on the obtained accuracies, this method can potentially be applied to other regions of interest for LC mapping without the need for ground truth data from the target year.

Worker Collaborative group estimation in spatial crowdsourcing

Spatial crowdsourcing (SC) has gained much attention in recent years. On SC platforms, requesters can publish spatial tasks on a specific topic such as taking photos at certain locations which is reflected by some tags of the task, and workers can choose tasks according to their tags. An interesting phenomenon is that workers often form groups based on their social relationships and common interests to perform same tasks. A nature question often raised when a task is published is how to predict which (or how many) workers are attracted by the task if its tags are specified. In particular, the tags of a new task affect the willingness of the workers to choose it. On the other hand, workers are also affected by the willingness of their co-workers or friends when deciding to choose a task or not. In this paper, we study the problem of potential workers estimation for a spatial crowdsourcing task, the Worker Collaborative Group Estimation (WCGE) problem, and model whether workers will join the group of a task as a game. We present efficient algorithms to find the Nash Equilibrium of the game and estimate the potential workers for the new task. Using synthetic datasets, we experimentally study the performance of proposed solutions. Our solutions can also help understand big geospatial data for self-driving cars and more intelligent transportation applications.

Big Spatial Data Management for the Internet of Things: A Survey

The high abundance of IoT devices have caused an unprecedented accumulation of avalanches of geo-referenced IoT spatial data that if could be analyzed correctly would unleash important information. This can feed decision support systems for better decision making and strategic planning regarding important aspects of our lives that depend heavily on location-based services. Several spatial data management systems for IoT data in Cloud has recently gained momentum. However, the literature is still missing a comprehensive survey that conceptualize a convenient framework that classify those frameworks under appropriate categories. In this survey paper, we focus on the management of big geospatial data that are generated by IoT data sources. We also define a conceptual framework and match the works of the recent literature with it. We then identify future research frontiers in the field depending on the surveyed works.

Who are more exposed to PM2.5 pollution: A mobile phone data approach

BackgroundFew studies have examined exposure disparity to ambient air pollution outside North America and Europe. Moreover, very few studies have investigated exposure disparity in terms of individual-level data or at multi-temporal scales. ObjectivesThis work aims to examine the associations between individual- and neighbourhood-level economic statuses and individual exposure to PM2.5 across multi-temporal scales. MethodsThe study population included 742,220 mobile phone users on a weekday in Shenzhen, China. A geo-informed backward propagation neural network model was developed to estimate hourly PM2.5 concentrations by the use of remote sensing and geospatial big data, which were then combined with individual trajectories to estimate individual total exposure during weekdays at multi-temporal scales. Coupling the estimated PM2.5 exposure with housing price, we examined the associations between individual- and neighbourhood-level economic statuses and individual exposure using linear regression and two-level hierarchical linear models. Furthermore, we performed five sensitivity analyses to test the robustness of the two-level effects. ResultsWe found positive associations between individual- and neighbourhood-level economic statuses and individual PM2.5 exposure at a daytime, daily, weekly, monthly, seasonal or annual scale. Findings on the effects of the two-level economic statuses were generally robust in the five sensitivity analyses. In particular, despite the insignificant effects observed in three of newly selected time periods in the sensitivity analysis, individual- and neighbourhood-level economic statuses were still positively associated with individual total exposure during each of other newly selected periods (including three other seasons). ConclusionsThere are statistically positive associations of individual PM2.5 exposure with individual- and neighbourhood-level economic statuses. That is, people living in areas with higher residential property prices are more exposed to PM2.5 pollution. Findings emphasize the need for public health intervention and urban planning initiatives targeting socio-economic disparity in ambient air pollution exposure, thus alleviating health disparities across socioeconomic groups.

Coupling mobile phone data with machine learning: How misclassification errors in ambient PM2.5 exposure estimates are produced?

BackgroundMost studies relying on time-activity diary or traditional air pollution modelling approach are insufficient to suggest the impacts of ignoring individual mobility and air pollution variations on misclassification errors in exposure estimates. Moreover, very few studies have examined whether such impacts differ across socioeconomic groups. ObjectivesWe aim to examine how ignoring individual mobility and PM2.5 variations produces misclassification errors in ambient PM2.5 exposure estimates. MethodsWe developed a geo-informed backward propagation neural network model to estimate hourly PM2.5 concentrations in terms of remote sensing and geospatial big data. Combining the estimated PM2.5 concentrations and individual trajectories derived from 755,468 mobile phone users on a weekday in Shenzhen, China, we estimated four types of individual total PM2.5 exposures during weekdays at multi-temporal scales. The estimate ignoring individual mobility, PM2.5 variations or both was compared with the hypothetical error-free estimate using paired sample t-test. We then quantified the exposure misclassification error using Pearson correlation analysis. Moreover, we examined whether the misclassification error differs across different socioeconomic groups. Taking findings of ignoring individual mobility as an example, we further investigated whether such findings are robust to the different selections of time. ResultsWe found that the estimate ignoring PM2.5 variations, individual mobility or both was statistically different from the hypothetical error-free estimate. Ignoring both factors produced the largest exposure misclassification error. The misclassification error was larger in the estimate ignoring PM2.5 variations than that ignoring individual mobility. People with high economic status suffered from a larger exposure misclassification error. The findings were robust to the different selections of time. ConclusionsIgnoring individual mobility, PM2.5 variations or both leads to misclassification errors in ambient PM2.5 exposure estimates. A larger misclassification error occurs in the estimate neglecting PM2.5 variations than that ignoring individual mobility, which is seldom reported before.

Change Detection from Remote Sensing to Guide OpenStreetMap Labeling

The growing amount of openly available, meter-scale geospatial vertical aerial imagery and the need of the OpenStreetMap (OSM) project for continuous updates bring the opportunity to use the former to help with the latter, e.g., by leveraging the latest remote sensing data in combination with state-of-the-art computer vision methods to assist the OSM community in labeling work. This article reports our progress to utilize artificial neural networks (ANN) for change detection of OSM data to update the map. Furthermore, we aim at identifying geospatial regions where mappers need to focus on completing the global OSM dataset. Our approach is technically backed by the big geospatial data platform Physical Analytics Integrated Repository and Services (PAIRS). We employ supervised training of deep ANNs from vertical aerial imagery to segment scenes based on OSM map tiles to evaluate the technique quantitatively and qualitatively.

Utilization of Google Earth Engine (GEE) for land cover monitoring over Klang Valley, Malaysia

Geospatial Big Data is currently received overwhelming attention and are on highlight globally and Google Earth Engine (GEE) is currently the hot pot platform to cater big data processing for Remote Sensing and GIS. Currently few or no study regarding the usage of this platform to study land use/cover changes over years in Malaysia. The objective is to evaluate the feasibility of GEE as a free cloud-based platform by performing classification of Klang Valley area from Landsat composites of three different years (1988-2003-2018) using multiple Machine Learning Algorithms (MLA). The best classification results were then imported and further processed to quantify the changes over the years using commercial software. Although, the classification results are of high accuracy but CART shows the best accuracy with 94.71%, 97.72% and 96.57% in 1988, 2003 and 2018 in comparison with RF and SVM. Some misclassified pixels were encountered because the annual composited images were compiled without taken into considerations of crops phenological stages (paddy) which resulted to the misclassified agricultural land into urban and bare land. Hence, the selection and composition of data initially had to be structured and strategized prior to processing as they can affect the classification result and further analysis. Regardless, GEE has performed quite well and fast in term of time and processing complexity of multiple datasets with minimal human interaction and intervention. Generally, GEE has proven to be reliable in fulfilling the objectives of this study to evaluate the GEE feasibility by performing classification and quantifying the land use/cover of studied area and provide good base for further analysis using different platform.

GeoAI: Where machine learning and big data converge in GIScience

In this paper GeoAI is introduced as an emergent spatial analytical framework for data-intensive GIScience. As the new fuel of geospatial research, GeoAI leverages recent breakthroughs in machine learning and advanced computing to achieve scalable processing and intelligent analysis of geospatial big data. The three-pillar view of GeoAI, its two methodological threads (data-driven and knowledge-driven), as well as their geospatial applications are highlighted. The paper concludes with discussion of remaining challenges and future research directions of GeoAI.