Guido Cervone

Multi-sensor studies of the Sumatra earthquake and tsunami of 26 December 2004

Multi sensor satellites are now capable of monitoring the globe during day and night and provide information about the land, ocean and atmosphere. Soon after the Sumatra tsunami and earthquake of 26 December 2004, multi‐sensors data have been analysed to study the changes in ocean, land, meteorological and atmospheric parameters. A pronounced changes in the ocean, atmospheric and meteorological parameters are observed while comparing data prior and after the Sumatra main event of 26 December 2004. These changes strongly suggest a strong coupling between land, ocean and atmosphere associated with the Sumatra event.

Using Twitter for tasking remote-sensing data collection and damage assessment: 2013 Boulder flood case study

ABSTRACT A new methodology is introduced that leverages data harvested from social media for tasking the collection of remote-sensing imagery during disasters or emergencies. The images are then fused with multiple sources of contributed data for the damage assessment of transportation infrastructure. The capability is valuable in situations where environmental hazards such as hurricanes or severe weather affect very large areas. During these types of disasters it is paramount to ‘cue’ the collection of remote-sensing images to assess the impact of fast-moving and potentially life-threatening events. The methodology consists of two steps. First, real-time data from Twitter are monitored to prioritize the collection of remote-sensing images for evolving disasters. Commercial satellites are then tasked to collect high-resolution images of these areas. Second, a damage assessment of transportation infrastructure is carried out by fusing the tasked images with contributed data harvested from social media such as Flickr and Twitter, and any additional available data. To demonstrate its feasibility, the proposed methodology is applied and tested on the 2013 Colorado floods with a special emphasis in Boulder County and the cities of Boulder and Longmont.

The Development of the AQ20 Learning System and Initial Experiments

This research has been supported in part by the National Science Foundation under grant IIS-9906858 and by a UMBC Grant under LUCITE Task #32.

EFFECT OF DUST STORM ON OCEAN COLOR AND SNOW PARAMETERS

ConclusionThe present study clearly shows the influence of dust storms on chlorophyll bloom in the offshore region of the Arabian Sea, with a time lag of few days, during the pre-monsoon season. Various satellite derived parameters over the Arabian Sea, Himalayan and Tibet snow covered regions show large changes due to the influence of dust storms. The MODIS snow albedo gives unreliable values under the influence of dust storms due to increase in the aerosol loading over these regions and snow albedo product must be used in combination with snow pixel counts during the dust storm season. A detailed study is required for the quantitative evaluation of dust storms on the chlorophyll blooms in the Arabian Sea region and on the snow parameters in the Himalayan region.

Comment on “Satellite altimetry and the intensification of Hurricane Katrina”

In a recent Eos article, Scharroo et al. [2005] reported that the dynamic sea topography anomalies along the track of Hurricane Katrina were the most prominent factors causing the intensification of Katrina as it passed over these anomalous regions in the Gulf of Mexico. They show that the sea surface temperature (SST) in the entire Gulf of Mexico was uniformly ∼30°C and was not associated with the rapid intensification of Katrina. We partly agree with their findings based on the results of dynamic topography associated with Katrinas intensification; however, we do not concur with their idea that SST was not linked with the rapid intensification of Katrina. Here, we show the significant impact of high SST anomaly in the Gulf on Katrinas rapid intensification and the role of anomalous SST in governing the air-sea interactions during its intensification.

Experimental validations of the learnable evolution model

A recently developed approach to evolutionary computation, called Learnable Evolution Model or LEM, employs machine learning to guide processes of generating new populations. The central new idea of LEM is that it generates new individuals not by mutation and/or recombination, but by processes of hypothesis generation and instantiation. The hypotheses are generated by a machine learning system from examples of high and low performance individuals. When applied to problems of function optimization and parameter estimation for nonlinear filters, LEM significantly outperformed the standard evolutionary computation algorithms used in experiments, sometimes achieving two or more orders of magnitude of evolutionary speed-up (in terms of the number of births). An application of LEM to the problem of optimizing heat exchangers has produced designs equal to or exceeding the best human designs. Further research needs to explore trade-offs and determine best areas for LEM application.

Monte Carlo source detection of atmospheric emissions and error functions analysis

A Monte Carlo algorithm is iteratively run to identify candidate sources for atmospheric releases. The values of the ground measurements of concentration are synthetically generated by a benchmark simulation of a Gaussian dispersion model. At each iteration, a Gaussian reflected plume model is applied to compute the dispersion from a candidate source, and the resulting concentrations are compared with the measurements at fixed points on the ground. Iterative algorithms for detection of atmospheric release sources are based on the optimization of an error function between numerical simulations and observations. However, the definition of error between observations and simulations by an atmospheric dispersion model is not univocal. In this paper, the comparisons are made using various error functions. The characteristics of different error functions between model predictions and sensor measurements are investigated, with a statistical analysis of the results. Sensitivity to domain size and addition of random noise to the measurements are also investigated.

DisasterMapper: A CyberGIS framework for disaster management using social media data

Traditional GIS tools and systems are powerful for analyzing geographic information for various applications but they are not designed for processing dynamic streams of data. This paper presents a CyberGIS framework that can automatically synthesize multi-sourced data, such as social media and socioeconomic data, to track disaster events, to produce maps, and to perform spatial and statistical analysis for disaster management. Within our framework, Apache Hive, Hadoop, and Mahout are used as scalable distributed storage, computing environment and machine learning library to store, process and mine massive social media data. The proposed framework is capable of supporting big data analytics of multiple sources. A prototype is implemented and tested using the 2011 Hurricane Sandy as a case study.

Using Non-authoritative Sources During Emergencies in Urban Areas

During emergencies in urban areas, it is paramount to assess damage to people, property, and environment in order to coordinate relief operations and evacuations. Remote sensing has become the de facto standard for observing the Earth and its environment through the use of air-, space-, and ground-based sensors. These sensors collect massive amounts of dynamic and geographically distributed spatiotemporal data daily and are often used for disaster assessment, relief, and mitigation. However, despite the quantity of big data available, gaps are often present due to the specific limitations of the instruments or their carrier platforms. This chapter presents a novel approach to filling these gaps by using non-authoritative data including social media, news, tweets, and mobile phone data. Specifically, two applications are presented for transportation infrastructure assessment and emergency evacuation.

Speeding Up Evolution through Learning: LEM

This paper reports briefly on the development of a new approach to evolutionary computation, called the Learnable Evolution Model or LEM. In contrast to conventional Darwinian-type evolutionary algorithms that employ mutation and/or recombination, LEM employs machine learning to generate new populations. At each step of evolution, LEM determines hypotheses explaining why certain individuals in the population are superior to others in performing the designated class of tasks. These hypotheses are then instantiated to create a next generation. In the testing studies described here, we compared a program implementing LEM with selected evolutionary computation algorithms on a range optimization problems and a filter design problem. In these studies, LEM significantly outperformed the evolutionary computation algorithms, sometimes speeding up the evolution by two or more orders of magnitude in the number of evolutionary steps (births). LEM was also applied to a real-world problem of designing optimized heat exchangers. The resulting designs matched or — outperformed the best human designs.