This project aims at developing an AI system to provide a reliable assessment of the structural damage caused by wildfires in the first burn period. Our approach uses multimodal data, including multispectral aerial images, historical post-fire damage assessment data, and building footprints, to create an association between damage data and structure footprints. We use these associations to generate features and use machine learning methods to assess the level of damage to structures. The resulting AI-driven system can be used to provide wildfire-induced structural damage assessments in near-real-time using only aerial images for future fires. We provide damage assessment results on several megafires in California to demonstrate the applicability of our approach to real wildfire scenarios.

Timely prediction of debris flow probabilities in areas impacted by wildfires is crucial to mitigate public exposure to this hazard during post-fire rainstorms. This paper presents a machine learning approach to amend an existing dataset of post-fire debris flow events with additional features reflecting existing vegetation type and geology, and train traditional and deep learning methods on a randomly selected subset of the data. The developed methods achieve AUC (area under the receiver operational characteristic curve) values of 0.93 (random forest) and 0.92 (neural network) on the test set, representing a significant improvement over a logistic regression model currently used (AUC 0.79). The paper also overviews a distributed, Kubernetesbased big data processing pipeline to efficiently retrieve features in areas impacted by new fires, and deploy the methods for real-time prediction of debris flow hazards.

Knowledge networks/graphs provide a powerful approach for data discovery, integration, and reuse. The NSF's new Convergence Accelerator program, which focuses on transitioning research to practice and translational research, announced Track A on the Open Knowledge Network (OKN). The program calls for multidisciplinary and multi-sector teams to work together to build a cooperative and shared open knowledge network infrastructure to drive innovation across science, engineering, and humanities. This workshop aims to invite researchers, practitioners, and the general public to brainstorm the ideas related to OKN, collaboratively build KGs for different domains or applications, develop AI algorithms to provide intelligent services based on OKN, and discuss the social and economic implications related to OKN.

A key takeaway from the COVID-19 crisis is the need for scalable methods and systems for ingestion of big data related to the disease, such as models of the virus, health surveys, and social data, and the ability to integrate and analyze the ingested data rapidly. One specific example is the use of the Internet of Things and wearables (i.e., the Oura ring) to collect large-scale individualized data (e.g., temperature and heart rate) continuously and to create personalized baselines for detection of disease symptoms. Individualized data, when collected, has great potential to be linked with other datasets making it possible to combine individual and societal scale models for further understanding the disease. However, the volume and variability of such data require novel big data approaches to be developed as infrastructure for scalable use. This paper presents the data pipeline and big data infrastructure for the TemPredict project, which, to the best of our knowledge, is the largest public effort to gather continuous physiological data for time-series analysis. This effort unifies data ingestion with the development of a novel end-to-end cyberinfrastructure to enable the curation, cleaning, alignment, sketching, and passing of the data, in a secure manner, by the researchers making use of the ingested data for their COVID-19 detection algorithm development efforts. We present the challenges, the closed-loop data pipelines, and the secure infrastructure to support the development of time-sensitive algorithms for alerting individuals based on physiological predictors illness, enabling early intervention.

The Neuroscience domain stands out from the field of sciences for its dependence on the study and characterization of complex, intertwining structures. Understanding the complexity of the brain has led to widespread advances in the structure of large-scale computing resources and the design of artificially intelligent analysis systems. However, the scale of problems and data generated continues to grow and outpace the standards and practices of neuroscience. In this paper, we present an automated neuroscience reconstruction framework, called NeuroKube, for large-scale processing and labeling of neuroimage volumes. Automated labels are generated through a machine-learning (ML) workflow, with data-intensive steps feeding through multiple GPU stages and distributed data locations leveraging autoscalable cloud-native deployments on a multi-institution Kubernetes system. Leading-edge hardware and storage empower multiple stages of machine-learning, GPU-accelerated solutions. This demonstrates an abstract approach to allocating the resources and algorithms needed to elucidate the highly complex structures of the brain. We summarize an integrated gateway architecture, and a scalable workflow-driven segmentation and reconstruction environment that brings together image big data with state-of-the-art, extensible machine learning methods.

High-resolution satellite imagery is a rich source of data applicable to a variety of domains, ranging from demo-graphics and land use to agriculture and hazard assessment. We have developed an end-to-end analysis pipeline that uses deep learning and unsupervised learning to process high-resolution satellite imagery and have applied it to various applications in previous work. As high-resolution satellite imagery is large-volume data, scalability is important to be able to analyze data from large geographical areas. To add scalability to our process, we converted our original pipeline, implemented using the Caffe deep learning library and the Python machine learning library Scikit-Learn, to other platforms that make use of distributed computation. Specifically, to add scalability, we use Keras for deep learning, and evaluate two different distributed platforms, Spark and Dask, for unsupervised learning. We report on results in scaling up our satellite analysis pipeline.

In the history of Computer Science, the act of `delegation' has been the greatest multiplier of society's problem-solving ability. A scientist working on detecting anomalies in a phenomenon, does not need to re-invent matrix multiplication techniques to solve her problem. Scientific workflows provide ultimate `delegation' mechanism - where a domain scientist can completely forget the specifics of `how' her program will execute on a large cluster in an efficient and cost-effective manner and can instead focus on the mathematical formulation and theoretical robustness of her solution. We present here an approach that directly aims to make the execution of Scientific Workflows more reliable, robust and efficient. We aim that the work presented in this paper will propel the larger effort, from the scientific workflow community, of making scientific workflow execution as simple, efficient and robust as a JOIN operation in a modern database. Specifically, we apply Deep Learning techniques to develop a mechanism that forecasts the final state (success or failure) of a dynamic job in a large-scale particle physics experiment, with minimal data gathering, and as early as possible in job's life cycle. The key advantage of having a predictive mechanism to identify and anticipate failure-prone jobs is the potential for designing intelligent Fault Tolerance mechanisms to handle anomalous events. We achieve a 14% improvement in computational resources utilization, and an overall classification accuracy of 85% on real tasks executed in a High Energy Physics Computing workflow. To the best of our knowledge, this is the most exhaustive and first of its kind study of neural network architectures in context of a real-dataset profiled from a large-scale scientific workflow.

Land cover classification analysis from satellite imagery is important for monitoring change in ecosystems and urban growth over time. However, the land cover classifications that are widely available in the United States are generated at a low spatial and temporal resolution, so that the spatial distribution between vegetation and urban areas in the wildland urban interface is difficult to measure. High spatial and temporal resolution analysis is essential for understanding and managing changing environments in these regions. This paper describes an end to end satellite data ingestion and analysis pipeline using deep learning on high resolution satellite imagery for generating pixel-based land cover classification.

Southern California's dry climate and fire-prone vegetation make the area vulnerable to extreme wildfire conditions. These conditions are exacerbated by Santa Ana weather patterns, which are characterized by very low humidity and gusty winds blowing in from the deserts. We present an approach using unsupervised learning to model and detect Santa Ana conditions based on sensor measurements from weather stations. Our approach uses cluster analysis to capture weather patterns specific to the region surrounding each weather station. A method is provided to automatically determine the Santa Ana cluster for each cluster model using dynamic, data-driven criteria. The resulting cluster models are applied to real-time sensor measurements to provide location-specific and time-specific detection of Santa Ana conditions. The Spark distributed platform is leveraged to scale the system to large datasets from multiple weather stations, and the Kepler workflow system is used to provide a GUI-based, easy-to-use interface to the underlying system. Results of testing our approach on an existing network of weather stations are presented. Our scalability experiment shows that the approach can process up to one million live sensor measurements in less than one minute on one machine. The proposed system can be used to aid in wildfire management and prevention by focusing firefighting efforts on regions with increased wildfire risks.

Advances in satellite imagery presents unprecedented opportunities for understanding natural and social phenomena at global and regional scales. Although the field of satellite remote sensing has evaluated imperative questions to human and environmental sustainability, scaling those techniques to very high spatial resolutions at regional scales remains a challenge. Satellite imagery is now more accessible with greater spatial, spectral and temporal resolution creating a data bottleneck in identifying the content of images. Because satellite images are unlabeled, unsupervised methods allow us to organize images into coherent groups or clusters. However, the performance of unsupervised methods, like all other machine learning methods, depends on features. Recent studies using features from pre-trained networks have shown promise for learning in new datasets. This suggests that features from pre-trained networks can be used for learning in temporally and spatially dynamic data sources such as satellite imagery. It is not clear, however, which features from which layer and network architecture should be used for learning new tasks. In this paper, we present an approach to evaluate the transferability of features from pre-trained Deep Convolutional Neural Networks for satellite imagery. We explore and evaluate different features and feature combinations extracted from various deep network architectures, and systematically evaluate over 2,000 network-layer combinations. In addition, we test the transferability of our engineered features and learned features from an unlabeled dataset to a different labeled dataset. Our feature engineering and learning are done on the unlabeled Draper Satellite Chronology dataset, and we test on the labeled UC Merced Land dataset to achieve near state-of-the-art classification results. These results suggest that even without any or minimal training, these networks can generalize well to other datasets. This method could be useful in the task of clustering unlabeled images and other unsupervised machine learning tasks.

Santa Ana conditions are hot, dry, windy weather conditions that can greatly increase the dangers of wildfires in southern California. We present a machine learning approach to detect Santa Ana conditions based on sensor measurements from weather stations. Cluster analysis is performed on historical weather data to build models to identify Santa Ana patterns. A separate model is built using data from each weather station to capture the patterns specific to the microclimate of each region. Real-time sensor data from a weather station can then be processed to determine if the region surrounding that station is experiencing Santa Ana conditions. Results can be used as a warning system to focus firefighting efforts on regions with increased wildfire risks. Through the use of the Kepler workflow system and distributed computing with Spark, data from several weather stations can be processed in parallel using a scalable clustering algorithm, allowing our approach to scale to large datasets from multiple weather stations.

Today's dynamic computing deployment for commercial and scientific applications is propelling us to an era where minor inefficiencies can snowball into significant performance and operational bottlenecks. Data center operations is increasingly relying on sensors based control systems for key decision insights. The increased sampling frequencies, cheaper storage costs and prolific deployment of sensors is producing massive volumes of operational data. However, there is a lag between rapid development of analytical techniques and its widespread practical deployment. We present empirical evidence of the potential carried by analytical techniques for operations management in computing and data centers. Using Machine Learning modeling techniques on data from a real instrumented cluster, we demonstrate that predictive modeling on operational sensor data can directly reduce systems operations monitoring costs and improve system reliability.

Ability to track provenance is a key feature of scientific workflows to support data lineage and reproducibility. The challenges that are introduced by the volume, variety and velocity of Big Data, also pose related challenges for provenance and quality of Big Data, defined as veracity. The increasing size and variety of distributed Big Data provenance information bring new technical challenges and opportunities throughout the provenance lifecycle including recording, querying, sharing and utilization. This paper discusses the challenges and opportunities of Big Data provenance related to the veracity of the datasets themselves and the provenance of the analytical processes that analyze these datasets. It also explains our current efforts towards tracking and utilizing Big Data provenance using workflows as a programming model to analyze Big Data.

Most scientists conduct analyses and run models in several different software and hardware environments, mentally coordinating the export and import of data from one environment to another. The Kepler scientific workflow system provides domain scientists with an easy-to-use yet powerful system for capturing scientific workflows (SWFs). SWFs are a formalization of the ad-hoc process that a scientist may go through to get from raw data to publishable results. Kepler attempts to streamline the workflow creation and execution process so that scientists can design, execute, monitor, re-run, and communicate analytical procedures repeatedly with minimal effort. Kepler is unique in that it seamlessly combines high-level workflow design with execution and runtime interaction, access to local and remote data, and local and remote service invocation. SWFs are superficially similar to business process workflows but have several challenges not present in the business workflow scenario. For example, they often operate on large, complex and heterogeneous data, can be computationally intensive and produce complex derived data products that may be archived for use in reparameterized runs or other workflows. Moreover, unlike business workflows, SWFs are often dataflow-oriented as witnessed by a number of recent academic systems (e.g., DiscoveryNet, Taverna and Triana) and commercial systems (Scitegic/Pipeline-Pilot, Inforsense). In a sense, SWFs are often closer to signal-processing and data streaming applications than they are to control-oriented business workflow applications.

We illustrate how a domain scientist can perform a complex scientific task by interleaving data access, querying, and manipulation, as well as analytical steps and computations in complex, problem specific ways. We show how our system is used by a geneticist for solving the problem of discovering the so-called "co-regulated" genes by interlinking data and computation from several Web sites, local computations, as well as local and remote databases. The main distinctive features of our system (compared, e.g., to the ZOO environment (Ioannidis et al., 1996)) include: (i) executable workflows run as Web services; (ii) abstract workflows employ concept names and semantic types that are higher-level (and thus more "scientist friendly") than executable workflows; and (iii) our system supports automatic translation of the latter into the former.

The authors present an approach for compiling "scientist-friendly" abstract workflow specifications into real-world executable workflows of Web service invocations, using a set of abstract-as-view definitions from a repository of abstract tasks. There have been a number of proposals and systems for scientific workflow management. However, our approach features unique aspects, in particular: the separation of abstract and concrete executable workflows; and the use of database mediation techniques to automatically translate AWFs into EWFs.