Geographically weighted regression (GWR) offers a local approach to modeling spatial data, considering geographical location and spatial relationships between observations.A salient feature of GWR is the emphasis on geographical proximity, in accordance with Tobler's First Law of Geography, which assumes that closer entities have a greater influence on the target location.Traditional GWR models have been augmented to consider various forms of physical distances aimed at enhancing model performance, and they often disregarded the potential influence of other data attributes, a shortcoming that extends to most GWR extensions.In this study, we introduce a novel weight matrix construction, which integrates data attribute similarity alongside the conventional geographically weighted matrix.The two weights are integrated in a manner that results in improved model performance.The proposed model, called Similarity and Geographically Weighted Regression or SGWR, was applied to five distinct datasets: housing prices, crime rates, and three health outcomes including mental health, depression, and HIV.Results show that SGWR significantly improved model performance based on several statistical measures, outperforming the global regression model and the traditional GWR.

Street view imagery such as Google Street View is widely used in people’s daily lives. Many studies have been conducted to detect and map objects such as traffic signs and sidewalks for urban built-up environment analysis. While mapping objects in the horizontal dimension is common in those studies, automatic vertical measuring in large areas is underexploited. Vertical information from street view imagery can benefit a variety of studies. One notable application is estimating the lowest floor elevation, which is critical for building flood vulnerability assessment and insurance premium calculation. In this article, we explored the vertical measurement in street view imagery using the principle of tacheometric surveying. In the case study of lowest floor elevation estimation using Google Street View images, we trained a neural network (YOLO-v5) for door detection and used the fixed height of doors to measure doors’ elevation. The results suggest that the average error of estimated elevation is 0.218 m. The depthmaps of Google Street View were utilized to traverse the elevation from the roadway surface to target objects. The proposed pipeline provides a novel approach for automatic elevation estimation from street view imagery and is expected to benefit future terrain-related studies for large areas.

Nighttime satellite imagery has been applied in a wide range of fields. However, our limited understanding of how observed light intensity is formed and whether it can be simulated greatly hinders its further application. This study explores the potential of conditional Generative Adversarial Networks (cGAN) in translating multispectral imagery to nighttime imagery. A popular cGAN framework, pix2pix, was adopted and modified to facilitate this translation using gridded training image pairs derived from Landsat 8 and Visible Infrared Imaging Radiometer Suite (VIIRS). The results of this study prove the possibility of multispectral-to-nighttime translation and further indicate that, with the additional social media data, the generated nighttime imagery can be very similar to the ground-truth imagery. This study fills the gap in understanding the composition of satellite observed nighttime light and provides new paradigms to solve the emerging problems in nighttime remote sensing fields, including nighttime series construction, light desaturation, and multi-sensor calibration.

Big climate data offers great opportunities for scientific discovery but demands efficient and effective analytics to investigate unknown and complex patterns. Most existing online processing and analytics systems for climate studies only support fixed user interface with predefined functions. These systems are often not scalable to handle massive climate data that could easily accumulate terabytes daily. To address the major limitations of existing online systems for climate studies, this paper presents a scalable online visual analytic system, known as SOVAS, to balance both usability and flexibility. SOVAS, enabled by a set of key techniques, supports large-scale climate data analytics and knowledge discovery in a scalable and sharable environment. This research not only contributes to the community an efficient tool for analyzing big climate data but also contributes to the literature by providing valuable technical references for tackling spatiotemporal big data challenges.

In a cloud data center, servers are always over-provisioned in an active state to meet the peak demand of requests, wasting a large amount of energy as a result. One of the options to reduce the power consumption of data centers is to reduce the number of idle servers, or to switch idle servers into low-power sleep states. However, the servers cannot process the requests immediately when transiting to an active state. There are delays and extra power consumption during the transition. In this paper, we consider using state-of-the-art servers with multi-sleep modes. The sleep modes with smaller transition delays usually consume more power when sleeping. Given the arrival of incoming requests, our goal is to minimize the energy consumption of a cloud data center by the scheduling of servers with multi-sleep modes. We formulate this problem as an integer linear programming (ILP) problem during the whole period of time with millions of decision variables. To solve this problem, we divide it into sub-problems with smaller periods while ensuring the feasibility and transition continuity for each sub-problem through a Backtrack-and-Update technique. We also consider using DVFS to adjust the frequency of active servers, so that the requests can be processed with the least power. Our simulations are based on traces from real world. Experiments show that our method can significantly reduce the power consumption for a cloud data center.

Disasters require quick response times, thought-out preparations, overall community, and government support to ensure the prevention of loss of life and reduce possible damages. Hurricane Irma can be recognized as a more popular recent disaster in terms of social media attention and made landfall in the US with significant time to prepare, making it a good model for an evaluation of disaster response. The objective of this research is to establish a pattern regarding sentiment trends over the progression of the storm totality using sentiment analysis and produce a viable set of topic models for its and Latent Dirichlet Allocation (LDA) topic modeling. The results from this study demonstrate that sentiment analysis can measure changes in users's emotions during natural disasters and that simple topic models can be formed from the twitter data. Information like this can be used by authorities to limit the damage and effectively recover from the disaster as well as adjust future response efforts accordingly. This research can be further improved by incorporating sentiment analysis methods for short texts, classifying emoticons and non-textual components such as videos or images, and optimizing data collection and preparation methods.

Flood inundation probability is critical for situation awareness, flood mitigation, emergency response, and postevent damage assessment. Current flood inundation mapping approaches can be categorized into real-time (RT) and near-RT (NRT) processes based on the timing of data acquisition. However, the intrinsic limitations of each category largely hamper their applications for flood mapping. Taking the 2015 South Carolina flood in downtown Columbia as a case study, this paper proposes a flood inundation reconstruction model by enhancing the NRT normalized difference water index (NDWI) derived from remote sensing imagery with the RT data including stream gauge readings and social media (tweets). Splitting into three modules: water height module, global enhancement module, and local enhancement module, the proposed model first incorporates the gauge readings and the NDWI image to reconstruct a macroscale flood probability layer, which is then locally enhanced using the verified flood-related tweets. The final output of the model matches well with the U.S. Geological Survey inundation map and its surveyed high-water marks. Results suggest that by enhancing NRT imagery with RT data sources, the proposed flood inundation probability reconstruction model renders a more robust, spatially enhanced flood probability index for emergency responders to quickly identify areas in need of urgent attention.

Climate observations and model simulations are producing vast amounts of array-based spatiotemporal data. Efficient processing of these data is essential for assessing global challenges such as climate change, natural disasters, and diseases. This is challenging not only because of the large data volume, but also because of the intrinsic high-dimensional nature of geoscience data. To tackle this challenge, we propose a spatiotemporal indexing approach to efficiently manage and process big climate data with MapReduce in a highly scalable environment. Using this approach, big climate data are directly stored in a Hadoop Distributed File System in its original, native file format. A spatiotemporal index is built to bridge the logical array-based data model and the physical data layout, which enables fast data retrieval when performing spatiotemporal queries. Based on the index, a data-partitioning algorithm is applied to enable MapReduce to achieve high data locality, as well as balancing the workload. The proposed indexing approach is evaluated using the National Aeronautics and Space Administration (NASA) Modern-Era Retrospective Analysis for Research and Applications (MERRA) climate reanalysis dataset. The experimental results show that the index can significantly accelerate querying and processing (~10× speedup compared to the baseline test using the same computing cluster), while keeping the index-to-data ratio small (0.0328%). The applicability of the indexing approach is demonstrated by a climate anomaly detection deployed on a NASA Hadoop cluster. This approach is also able to support efficient processing of general array-based spatiotemporal data in various geoscience domains without special configuration on a Hadoop cluster.

The high power consumption of cloud data centers has aroused great concern on environmental implications such as global warming. Therefore, cloud service providers like Facebook and Green House Data have built their own wind or solar farms to power the data centers to reduce both energy cost and carbon emissions. In this paper, we propose an optimization- based framework to make planing for green cloud data centers, where the objective functions range from minimizing energy cost to minimizing carbon emissions. Our planning is based on the optimized scheduling of the users' requests to each data center, while considering time-varying and location-varying electricity prices and the supply of sustainable energy under different weather conditions. Our plan includes: 1) How many servers each data center should have. 2) How many wind turbines and solar panels should be used to power each data center. 3) What capacity of energy storage device (ESD) should be equipped for each data center. We formulate each problem as a constraint optimization problem, and solve it using Cplex. Extensive experiments have been done on traces from real world. As far as we know, our work is the first to explore the issue of planning for green cloud data centers through optimized scheduling methods.

Driver response to variable message signs' recommendation is analyzed. Each driver has two strategies: accepting the recommendation or refusing the recommendation. A decision making model of strategy selection of drivers is proposed. The process by which the proportion of strategies changed in the population over time is studied using evolutionary game theory. The stable states and stable strategy are analyzed. The results show that the proportion of drivers accepting the recommendation will converge to 100% over time if and only if the driver's payoff of accepting the recommendation is larger than the payoff of refusing recommendation. For VMS strategies, it is necessary that driver's payoff of accepting the recommendation should be strictly larger than the payoff of refusing recommendation and the difference of the two payoffs should be as large as possible. Meanwhile, VMS should make a positive impression on them when drivers use the VMS recommendation at the first time. These provide decision support for VMS strategy making.