A5073472760- Hydrology and Watershed Management Studies
- Flood Risk Assessment and Management
- Traffic Prediction and Management Techniques
- Hydrology and Sediment Transport Processes
- Hydrological Forecasting Using AI
- Hydrology and Drought Analysis
- Water Quality and Resources Studies
- Climate variability and models
- Soil and Water Nutrient Dynamics
- Data Management and Algorithms
- Human Mobility and Location-Based Analysis
- Automated Road and Building Extraction
- Cloud Computing and Resource Management
- Fish Ecology and Management Studies
- Transportation Planning and Optimization
- Cryospheric studies and observations
- Distributed systems and fault tolerance
- Advanced Data Storage Technologies
- Environmental Monitoring and Data Management
- Statistical Methods and Inference
- Data Visualization and Analytics
- Natural Language Processing Techniques
- Urban Transport and Accessibility
- Text Readability and Simplification
- Meteorological Phenomena and Simulations
NOAA Great Lakes Environmental Research Laboratory
2012-2021
National Oceanic and Atmospheric Administration
2007-2021
ORCID
2021
University of California, Berkeley
2011-2014
Many "big data" applications must act on data in real time. Running these at ever-larger scales requires parallel platforms that automatically handle faults and stragglers. Unfortunately, current distributed stream processing models provide fault recovery an expensive manner, requiring hot replication or long times, do not We propose a new model, discretized streams (D-Streams), overcomes challenges. D-Streams enable mechanism improves efficiency over traditional backup schemes, tolerates...
In this paper, we combine the most complete record of daily mobility, based on large-scale mobile phone data, with detailed Geographic Information System (GIS) uncovering previously hidden patterns in urban road usage. We find that major usage each segment can be traced to its own - surprisingly few driver sources. Based finding propose a network by defining bipartite framework, demonstrating contrast traditional approaches, which define importance solely topological measures, role depends...
We consider the problem of reconstructing vehicle trajectories from sparse sequences GPS points, for which sampling interval is between 1 s and 2 min. introduce a new class algorithms, are altogether called path inference filter (PIF), that maps data in real time, variety tradeoffs scenarios with high throughput. Numerous prior approaches map matching can be shown to special cases PIF presented this paper. present an efficient procedure automatically training on data, or without ground-truth...
We report on our experience scaling up the Mobile Millennium traffic information system using cloud computing and Spark cluster framework. uses machine learning to infer conditions for large metropolitan areas from crowdsourced data, was specifically designed support such applications. Many studies of frameworks have demonstrated scalability performance improvements simple algorithms. Our implementing a real-world learning-based application corroborates benefits, but we also encountered...
Abstract Between January 2013 and December 2014, water levels on Lake Superior Michigan‐Huron, the two largest lakes Earth by surface area, rose at highest rate ever recorded for a 2 year period beginning in ending of following year. This historic event coincided with below‐average air temperatures extensive winter ice cover across Great Lakes. It also brought an end to 15 persistently Lakes Michigan‐Huron that included several months record‐low levels. To differentiate hydrological drivers...
Hydrologic model intercomparison studies help to evaluate the agility of models simulate variables such as streamflow, evaporation, and soil moisture. This study is third in a sequence Great Lakes Runoff Intercomparison Projects. The densely populated Lake Erie watershed studied here an important international lake that has experienced recent flooding shoreline erosion alongside excessive nutrient loads have contributed eutrophication. Understanding sources pathways flows critical solve...
Abstract. This work explores the potential of distributed GEM-Hydro runoff modeling platform, developed at Environment and Climate Change Canada (ECCC) over last decade. More precisely, aim is to develop a robust implementation methodology perform reliable streamflow simulations with model large partly ungauged basins, in an efficient manner. The latest version combines SVS (Soil, Vegetation Snow) land-surface scheme WATROUTE routing scheme. has never been evaluated from hydrological point...
Key Points Some gages are better than others for extrapolation of streamflow. Increased gaged area is not necessarily extrapolation. Similarity in some key physical properties increases ARM skill.
Abstract Since the late 1990s, Laurentian Great Lakes have experienced persistent low water levels and above average over‐lake evaporation rates. During winter of 2013–2014, lakes endured most persistent, lowest temperatures highest ice cover in recent history, fostering speculation that rates might decrease rise. To address this speculation, we examined interseasonal relationships Lake Michigan's thermal regime. We find pronounced between conditions subsequent fall heat content, modest with...
Controlling and analyzing cyberphysical robotics systems is increasingly becoming a Big Data challenge. We study the case of predicting drivers' travel times in large urban area from sparse GPS traces. present framework that can accommodate wide variety traffic distributions spread all computations on cluster to achieve small latencies. Our built Discretized Streams, recently proposed approach stream processing at scale. demonstrate usefulness Streams with novel algorithm estimate vehicular...
Growing demand from the general public for centralized points of data access and analytics tools coincides with similar, well-documented needs regional international hydrology research resource management communities. To address this need within Laurentian Great Lakes region, we introduce Dashboard (GLD), a dynamic web visualization platform that brings multiple time series sets together visual analysis download. The platform's adaptable, robust, expandable Time Series Core Object Model...
Bighead carp H. nobilis and silver Hypothalmichthys molitrix (collectively bigheaded carps, BHC) are invasive planktivorous fishes that threaten to enter the Laurentian Great Lakes disrupt food webs. To assess likelihood of BHC establishment their likely effects on web Saginaw Bay, Lake Huron, we developed a multi-species individual-based bioenergetics model tracks individual bighead carp, four key fish species, seven prey biomass groups over 50 years. The daily consumption, mortality growth...
Most optimal routing problems focus on minimizing travel time or distance traveled. Oftentimes, a more useful objective is to maximize the probability of on-time arrival, which requires statistical distributions times, rather than just mean values. We propose method estimate large-scale road networks, using probe vehicle data collected from GPS. present framework that works with large input data, and scales linearly size network. Leveraging planar topology graph, computes efficiently...
We study the problem of estimating sparse precision matrices from data with missing values. show that corresponding maximum likelihood is a Difference Convex (DC) program by proving some new concavity results on Schur complements. propose algorithm to solve this based ConCave-Convex Procedure (CCCP), and we standard EM procedure weaker CCCP for problem. Numerical experiments our algorithm, called m-CCCP, converges much faster than both synthetic biology datasets.