1

The Fourth Annual Meeting of NSF Expeditions in Computing Award # 1029711

June 30- July 2, 2014

Workshop Venue: National Center for Atmospheric Research

Mesa Labs Boulder, CO

www.climatechange.cs.umn.edu

Fourth Workshop on

Understanding Climate Change from Data

2

3

Understanding Climate Change from Data June 30 – July 2, 2014

Table of Contents

Table of Contents pg 3 Final Program Schedule pg 5 Abstracts, June 30, 2014 Presenters pg 9 Abstracts, July 1, 2014 Presenters pg 13 Abstracts, July 2, 2014 Presenters pg 19 Panel Discussion: Data Science and Climate Science: Narrowing Gaps pg 20 Invited Participant Bios pg 21 Expeditions in Computing Team pg 31 Poster Session pg 41 Attendee Contact Information pg 57

4

5

Agenda: Fourth Workshop on Understanding Climate Change from Data

Monday, June 30, 2014 pg

11:00 Registration opens, lunch begins- catered at Mesa Labs 12:20 Welcoming Remarks

Session 1 Chair: Doug Nychka

9

12:30 James Hurrell, National Center for Atmospheric Research KEYNOTE: Climate Predictions and Projections in the Coming Decades: Uncertainty due to Natural Variability

1:05

Vipin Kumar, University of Minnesota Introduction to the NSF Expeditions in Computing on Understanding Climate Change: A Data Driven Approach

1:40 Auroop Ganguly, Northeastern University Informing Climate Adaptation with Big Data and Bigger Models

1:55 Break

Session 2 Chair: William Hendrix

10

2:30 Noah Diffenbaugh, Stanford University Quantifying the influence of global warming on the likelihood of unprecedented extreme climate events

2:55 Arindam Banerjee, University of Minnesota Estimating High-Dimensional Dependencies: Applications to Multi-task Learning for Combining Climate Model Outputs

3:10 Timothy DelSole, George Mason University Using Climate Models to Constrain Learning Algorithms

3:35 Break

Session 3 Chair: Raju Vatsavai

11

3:55 James Faghmous, University of Minnesota Monitoring Mesoscale Ocean Eddies From Space: A Theory-Guided Data Mining Perspective

4:10 Stefan Liess, University of Minnesota Different modes of variability over the Tasman Sea

4:25 Nagiza Samatova, North Carolina State University Modulatory Networks for Climate Extremes

4:40

Networking

5:00

Poster Session & Dinner, catered, at Mesa Labs

6

Tuesday, July 1, 2014 13 8:50 Registration opens

Session 4 Chair: Auroop Ganguly

13

9:00 Warren Washington, National Center for Atmospheric Research KEYNOTE: Future Development of Climate and Earth System Models and Their Data Needs

9:35 Dimitris Giannakis, New York University Extraction and predictability of Madden-Julian oscillation signals in infrared brightness temperature data

10:00 Clara Deser, National Center for Atmospheric Research Unforced versus forced climate trends over North America

10:25 Break

Session 5 Chair: Abdollah Homaifar

14

11:00 Ghassem R Asrar, University of Maryland The Role of Data in Integrated Human-Earth Systems Modeling and Assessment

11:25 Ramakrishna Nemani, NASA Ames Research Center NASA Earth Exchange (NEX): Collaborative computing for global change science

11:50 Forrest M. Hoffman, Oak Ridge National Laboratory and University of California - Irvine Representativeness-based Sampling Network Design for NGEE and Identifying Phenoregions for the Conterminous U.S.

12:15 Lunch Break, NCAR Cafeteria

Session 6 Chair: Sucharita Gopal 15

1:45 Nikunj C. Oza, NASA Ames Research Center Data Mining for Earth Science at NASA

2:10 Abdollah Homaifar, North Carolina Agricultural & Technical University Multiple linear trend analysis for non-stationary climatic time series

2:25 Alison Baker, National Center for Atmospheric Research Evaluating the Impact of Data Compression on Climate Simulation Data

2:50 Break

Session 7 Chair: Scott Sellars

16

3:20 Bala Rajaratnam, Stanford University A Methodology for Robust Multiproxy Paleoclimate Reconstructions

3:45 Wei Ding, University of Massachusetts Boston A Data-Driven Machine Learning Framework for Long-Lead Flood Forecasting

4:10 Ansu Chatterjee, University of Minnesota A study of mixed-source variability and dependence in precipitation data over India

4:25 Richard (Ricky) Rood, University of Michigan Ann Arbor Climate Informatics: Human Experts and the End-to-End System

4:50 Reception, light refreshments

7

Wednesday, July 2, 2014 19

8:50 Registration opens

Session 8 Chair: James Faghmous

19

9:00 Kevin Trenberth, National Center for Atmospheric Research KEYNOTE: Climate change: It’s about the data isn’t it?

9:35 Imme Ebert-Uphoff, Colorado State University Weakening of atmospheric information flow in a warming climate - preliminary results

10:00 Break

10:30

Panel Discussion: Data Science and Climate Science: Narrowing Gaps Moderator: Doug Nychka Panelists: Laurence Buja, National Center for Atmospheric Research Alicia Karspeck, National Center for Atmospheric Research Richard Loft, National Center for Atmospheric Research Linda Mearns, National Center for Atmospheric Research Srini Parthasarthy, Ohio State University Stephen Sain, National Center for Atmospheric Research Shashi Shekhar, University of Minnesota Claudia Tebaldi, National Center for Atmospheric Research

20

12:40 Closing remarks

8

9

Presentations, Session 1, June 30 James Hurrell – National Center for Atmospheric Research Keynote speaker, Title: Climate Predictions and Projections in the Coming Decades: Uncertainty due to Natural Variability

A grand challenge problem is the prediction and projection of the consequences of natural and anthropogenic climate variability and change at regional and global scales. This includes understanding the impacts of climate change on the water cycle, water availability, weather extremes, and the health and functioning of marine and terrestrial ecosystems, the potential for abrupt changes in climate, and understanding the limits and options society has to respond to climate change. In this talk I will discuss several sources of uncertainty in predictions and projections of the future, highlighting that unpredictable, internally generated climate fluctuations make a substantial contribution to climate trends projected for the next 50 years. Results are based on large ensembles of climate change integrations with the Community Earth System Model (CESM). I also will show that the large-scale atmospheric circulation is responsible for much of the diversity in climate change projections across the individual ensemble members.

Vipin Kumar – University of Minnesota Title: NSF Expeditions in Computing on Understanding Climate Change: A Data Driven Approach

Climate change is the defining environmental challenge facing our planet, yet there is considerable uncertainty regarding the social and environmental impact due to the limited capabilities of existing physics-based models of the Earth system. Consequently, important questions relating to food security, water resources, biodiversity, and other socio-economic issues over relevant temporal and spatial scales remain unresolved. A new and transformative approach is required to understand the potential impact of climate change. Data driven approaches that have been highly successful in other scientific disciplines hold significant potential for application in environmental sciences. This Expeditions project addresses key challenges in the science of climate change by developing methods that take advantage of the wealth of climate and ecosystem data available from satellite and ground-based sensors, the observational record for atmospheric, oceanic, and terrestrial processes, and physics-based climate model simulations. These innovative approaches help provide an improved understanding of the complex nature of the Earth system and the mechanisms contributing to the adverse consequences of climate change, such as increased frequency and intensity of hurricanes, precipitation regime shifts, and the propensity for extreme weather events that result in environmental disasters. Methodologies developed as part of this project will be used to gain actionable insights and to inform policymakers. This presentation provides an overview of the challenges being addressed in this multi-disciplinary, multi-institutional project and includes highlights of some of the results obtained over the past year.

Auroop Ganguly – Northeastern University Title: Informing Climate Adaptation with Big Data and Bigger Models

Predictive insights on climate variability, extremes and uncertainty, especially at local to regional scales, may provide useful information for adaptation. However, while physics-based models may lose their ability to generalize with greater parameterization, especially when the parameters cannot be directly or indirectly estimated from data, insights obtained by mining massive volumes of observed and model-simulated data may have limited interpretability, particularly when explanations are attempted based

10

on simple physical intuition. The latter may even yield spurious results when complex dependence patterns are ignored. Another major set of challenges stem from the fact that the climate system is nonlinear and dynamical, even chaotic, but contaminated with random noise, which may include low-frequency, and even 1/f, components. While arguments and frameworks have been advanced to change the null hypothesis in adaptation decisions from stationary climate to warming environments, fundamental and achievable limits to predictability may need to be considered. This presentation discusses new findings and old perspectives, as well as prevailing beliefs in the scientific and adaptation communities, and suggests the need to consider both non-traditional tools and non-intuitive insights while being mindful of their potential pitfalls. Presentations, Session 2, June 30 Noah Diffenbaugh – Stanford University Title: Quantifying the influence of global warming on the likelihood of unprecedented extreme climate events

The social and biological impacts of extreme climate events create tremendous - and usually immediate - interest in the possible role of global warming, particularly for events that are unprecedented in the observed record. In addition, observed trends in many climate quantities combined with projections of robust future changes in the occurrence of extreme events suggest that global warming could alter the probability of unprecedented events. However, the rarity of such events and the difficulty in accurately simulating the frequency of occurrence of the most extreme events combine to make objective, hypothesis-based conclusions a daunting scientific challenge. I will present recent work that combines analysis of physical climate dynamics with objective statistical approaches to quantify the likelihood that the probability of occurrence of different unprecedented extreme events is different in the current climate compared with a climate in which there were no human forcings.

Arindam Banerjee – University of Minnesota Title: Estimating High-Dimensional Dependencies: Applications to Multi-task Learning for Combining Climate Model Outputs Estimating dependencies between several variables from a small number of high-dimensional samples is a key problem in statistical machine learning. The ability to find such dependencies can advance our scientific understanding of complex phenomenon in several domains. The brute force approach, based on computing conditional mutual information with small number of samples in high-dimensions, is both statistically unreliable and computationally infeasible. This talk will present advances in estimating such dependencies for multi-variate Gaussian and Gaussian copula distributions even in presence of missing values. The estimators are statistically effective, i.e., can correctly estimate the dependencies with a small number of samples, and computationally efficient, e.g., seamlessly scales to millions of variables with trillions of potential dependencies out of which a few are real. We show applications of the approach to multi-task learning in the context of combining climate model outputs, with promising preliminary results.

Timothy DelSole – George Mason University Title: Using Climate Models to Constrain Learning Algorithms

11

There is great interest is identifying persistent components of the climate system that could provide a basis for prediction on seasonal and climate time scales. Identifying such components based on observations is problematic because of the short sample size. In this talk, I discuss methods for incorporating information from imperfect climate simulations into this identification problem. A guiding principle is that persistent components on seasonal or longer time scales tend to originate from the ocean and tend to be large-scale. This principle motivates a regularized regression methodology for identifying persistent components in observations. The solution of the regularized regression problem is facilitated by representing data in terms of the eigenfunctions of the Laplacian operator over the domain of interest. Presentations, Session 3, June 30 James Faghmous – University of Minnesota Title: Monitoring Mesoscale Ocean Eddies From Space: A Theory-Guided Data Mining Perspective

Mesocale ocean eddies are ubiquitous dynamic phenomena that propagate across the global oceans. Eddies play a critical role in the transport of heat, momentum, and nutrients and are instrumental in balancing our planet's energy. We present the results from our OpenEddy project, a theory-guided data mining effort to catalog daily ocean eddy activity on a global scale using satellite altimeter data. We conclude by presenting preliminary results of the impact of ocean eddies on tropical cyclones on global scale. For more information or to download our data and source code please see: www.ucc.umn.edu/eddies

Stefan Liess – University of Minnesota Title: Different modes of variability over the Tasman Sea

A new approach based on Shared Reciprocal Nearest Neighbors is used to detect atmospheric teleconnections without being bound by orthogonality (such as Empirical Orthogonal Functions). One selected pattern is a teleconnection between the Tasman Sea and the Southern Ocean, which is related to the El Niño/Southern Oscillation (ENSO), the Indian Ocean Dipole (IOD), and the Southern Annular Mode (SAM). The teleconnection is significantly correlated with SAM during austral summer, fall and winter, with IOD during spring, and with ENSO in summer. Increasing pressure over the Tasman Sea not only leads to higher (lower) surface temperature over eastern Australia (southwestern Pacifc) in all seasons, but is also related to reduced surface temperature over Wilkes Land and Adélie Land in Antarctica during austral fall and winter. Precipitation responses are generally negative over New Zealand. Over eastern Australia, precipitation anomalies are negative during austral spring and summer and positive during fall and winter for one standard deviation thresholds of the selected teleconnection. When doubling the threshold, the size of the anomalous high-pressure center increases and precipitation anomalies are generally negative over Australia and New Zealand. Eliassen-Palm fluxes quantify the seasonal dependence of SAM, ENSO and IOD influences.

Nagiza Samatova – North Carolina State University Title: Modulatory Networks for Climate Extremes

Decades of hypothesis-driven and/or first-principles research have been applied towards the discovery and explanation of the mechanisms that drive climate phenomena, such as those modulating African rainfall or Atlantic hurricane variability. Although connections between various climate factors have

12

been theorized, not all of the key relationships are fully understood. To complement such efforts, data-driven approaches show a promise in identifying candidate players, or “hot spots,” suggesting modulatory relationships influencing a climate phenomenon of interest, improving the predictive power of empirical forecast models, and ultimately facilitating our mechanistic understanding. A suite of such methodologies will be illustrated in the context of data-driven, climatologically relevant inference of modulatory networks for the African rainfall and Atlantic hurricane variability.

13

Presentations, Session 4, July 1 Warren Washington – National Center for Atmospheric Research Keynote Title: Future Development of Climate and Earth System Models and Their Data Needs

The development of climate and Earth system models has been regarded primarily as the making of scientific tools to study the complex nature of the Earth’s climate. These models have a long history starting with very simple physical models based on fundamental physics in the 1960s. Over time they have become much more complex with atmospheric, ocean, sea ice, land/vegetation, biogeochemical, glacial and ecological components. The policy use aspects of these models did not start in the 1960s and 1970s as decision making tools but were used to answer fundamental scientific questions such as what happens when the atmospheric carbon dioxide concentration increases. They gave insights into the various interactions and were extensively compared with observations. It was realized that models of the earlier time periods could only give first order answers to many of the fundamental policy questions. As societal concerns about climate change rose, the policy questions of anthropogenic climate change became better defined; they were mostly concerned with the climate impacts of increasing greenhouse gases, aerosols, and land cover change. In the late 1980s, the United Nations set up the Intergovernmental Panel on Climate Change to perform assessments of the published literature. Thus, the development of climate and Earth system models became intimately linked to the need to not only improve our scientific understand but also answering fundamental policy questions. In order to meet this challenge, the models became more complex and realistic so that they could address these policy oriented science questions such as rising sea level. The presentation will discuss the past and future development of global climate and Earth system models for science and policy purposes along with the associated data issues. Also to be discussed is their interactions with economic integrated assessment models, regional and specialized models such as river transport or ecological components. Computational and data challenges will also be part of the discussion along with a few results on future climate change projections.

Dimitris Giannakis – New York University Title: Extraction and predictability of Madden-Julian oscillation signals in infrared brightness temperature data

We discuss spatiotemporal modes of variability of tropical convection recovered from infrared brightness temperature (Tb) data using nonlinear Laplacian spectral analysis (NLSA); a manifold-based data analysis technique. Applied to Tb data from the CLAUS multi-satellite archive, NLSA yields a hierarchy of modes spanning interannual to diurnal timescales. Here, of particular interest are intraseasonal traveling modes including the Madden-Julian oscillation (MJO) and the northward-propagating Boreal summer intraseasonal oscillation. The recovered modes are used in conjunction with information-theoretic methods to assess the influence of ENSO and other large-scale patterns on MJO predictability.

Clara Deser – National Center for Atmospheric Research Title: Unforced versus forced climate trends over North America

This talk will highlight the relative importance of internally-generated vs. externally-forced climate trends during the past 35 years and the next 50 years at local and regional scales over North America in two global coupled model ensembles. Both ensembles contain a large number of integrations (17 and

14

40), each of which is subject to the same anthropogenic radiative forcing but starts from a slightly different atmospheric state. The large ensemble size allows for a robust estimate of the anthropogenically-forced component of the trends in each model, as well as unambiguous determination of the contribution of internal variability to the trends in each individual run. The results show that unpredictable, intrinsic variability of the climate system strongly influences the pattern and magnitude of simulated surface air temperature and precipitation trends in any single realization, both for the past and the future. Implications of the results for model validation, inter-model comparisons, and interpretation of observed climate trends are discussed. Presentations, Session 5, July 1 Ghassem R. Asrar – Joint Global Change Research Institute, University of Maryland Title: The Role of Data in Integrated Human-Earth Systems Modeling and Assessment

The scientific understanding of changes in human-Earth systems, and interactions among them, has been a research priority at the national and international level for the past few decades. We have made significant scientific progress in understanding the Earth and Human systems individually; however progress on understanding of their interactions and representation of such interactions in fully integrated Human-Earth system models has been relatively limited. Despite limited progress, the knowledge gained from research efforts has benefited greatly the science-based environmental assessments by the Intergovernmental Panel on Climate Change, Millennium Development, Water and Energy, etc. The Joint Global Change Research Institute (JGCRI) of the Pacific Northwest National Laboratory (PNNL) hosted by the University of Maryland in College Park, was established to focus on socioeconomic and policy aspects of global change research– topics such as energy, environment and economy; energy, water and food security; energy technologies such as carbon-based, renewable and nuclear systems; and carbon management and pricing options including carbon capture and storage technologies. A combination of data, interactive Human-Earth system models, and integrated assessment and analyses are used in these studies. Recent studies include the competition for land and water for food and energy production, the role of natural and managed terrestrial ecosystems in cycling carbon, energy technology choices, policies and socioeconomic implications. Such holistic approach to studying complex scientific problems at the interface of Earth system domains with consideration of technology and policy choices and their socioeconomic appeal is key to devising acceptable and affordable solutions to global, regional and national environmental problems.

Nikunj C. Oza – NASA AMES Research Center Title: Data Mining for Earth Science at NASA

In this talk, I will describe two of the NASA Data Sciences Group's threads of work involving data mining applied to Earth science. The first thread of work involves the use of virtual sensors, which involves the prediction of one sensor measurement given other measurements for situations when the target measurement is missing or excessively noisy. We applied this method to estimate a spectrum for an older instrument where it was not available. The second thread of work is the development of BlockGP, which takes advantage of multiple modalities in data to make Gaussian Process Regression scalable. BlockGP's scalability is demonstrated on MODIS data over California.

Forrest M. Hoffman - Oak Ridge National Laboratory and University of California - Irvine

15

Title: Representativeness-based Sampling Network Design for NGEE and Identifying Phenoregions for the Conterminous U.S.

The Arctic and the tropics contain vast stores of carbon in soils and vegetation, respectively, and are hosts to potentially vulnerable ecosystems that are poorly represented in today's Earth system models (ESMs). Improving process representation for these critical ecosystems is required to reduce uncertainties in ESM predictions of future climate. The U.S. Dept. of Energy has initiated combined observational and modeling activities, called Next Generation Ecosystem Experiments (NGEE), to improve our understanding of fundamental mechanisms controlling the cycling of carbon in both Arctic and tropical environments. Resource and logistical constraints limit the frequency and extent of environmental observations, particularly in harsh Arctic and tropical climates, necessitating the development of a systematic sampling strategy to maximize coverage and objectively represent environmental variability at desired scales. We developed a quantitative methodology for stratifying sampling domains, informing site selection, and determining the representativeness of measurement sites and networks. We applied multivariate spatio-temporal clustering (MSTC) to gridded data for the State of Alaska and to globally gridded data for producing quantitative ecoregions and for understanding the representativeness of candidate sampling sites and distributed sampling networks. For the Arctic study, we used 37 bioclimatic characteristics from downscaled general circulation model results and data for the State of Alaska at 4 sq km resolution to define multiple sets of bioclimatic ecoregions for decadal time periods representing the present (2000-2009) and future (2090-2099). We subsequently applied this methodology to 17 climatic, topographic, and edaphic variables for the globe, also at a 4 sq km resolution, to produce quantitative ecoregions and delineate sampling domains for tropical forests. These analyses provide model-inspired insights into optimal sampling strategies, offer frameworks for upscaling measurements, and provide downscaling approaches for integration of models and measurements. We applied similar techniques to the full record of normalized difference vegetation index (NDVI) observations from the MODIS satellite-based instruments at 230 m resolution for the conterminous U.S. to produced phenoregions, and developed a method for assigning labels to improve the utility of our unsupervised classification. Presentations, Session 6, July 1 Ramakrishna Nemani – NASA AMES Research Center Title: NASA Earth Exchange (NEX): Collaborative computing for global change science

There is increasing pressure on the science community not only to understand how recent and projected changes in climate affect Earth's environment and the natural resources on which society depends but also to design solutions to mitigate or cope with the likely impacts. Responding to this multidimensional challenge requires new tools and research frameworks that assist scientists in collaborating to rapidly investigate complex interdisciplinary science questions of critical societal importance. One such collaborative research framework, within the NASA Earth science program, is the NASA Earth Exchange (NEX). NEX is a collaborative platform that combines state-of-the-art supercomputing, Earth system modeling, remote sensing data from NASA and other agencies, and a scientific social network to provide an environment in which users can explore and analyze large Earth science data sets, run modeling and analysis codes, collaborate on new or existing projects, and share results within and/or among communities. NEX has been supporting the National Climate Assessment (NCA) by providing a platform for climate modeling, climate downscaling and climate impacts assessment through sound science that is transparent and reproducible.

16

Abdollah Homaifar – North Carolina Agricultural & Technical University Title: Multiple linear trend analysis for non-stationary climatic time series

The latest IPCC report has confirmed that the Earth is warming due to human activities. While the global mean temperature is an excellent measure of this warming on a global scale, analysis of temperature trend changes on local scales is important as well. In order to analyze the trend of this non-stationary climatic time series, it is useful to model it with multiple linear trends and locate the time points of significant changes. However, it is very important to ensure that the detected change points are only due to natural phenomena and occurred at all stations and are not caused by changes in measurement devices. Segmentation of multi-dimensional non-stationary time series is challenging, because the problem is mathematically ill-posed. However, with Bounded-variation (BV) segmentation, we can convert the problem into a convex optimization. This method does not rely on Gaussian/Markovian models to analyze the data from all the stations in parallel. It finds change points that are due to natural phenomena only. We found the optimal number of change points by the well-known information theory criteria. We have applied the BV-segmentation to the temperature time series of North Carolina (NC) and United States and the results represent region-specific climate variability despite higher frequency harmonics in the climatic time series.

Alison Baker – National Center for Atmospheric Research Title: Evaluating the Impact of Data Compression on Climate Simulation Data

Abstract: High-resolution climate simulations require tremendous computing resources and can generate massive datasets. At present, preserving the data from these simulations consumes vast storage resources at institutions such as the National Center for Atmospheric Research. The historical data generation trends are economically unsustainable, and storage resources are already beginning to limit science objectives. To mitigate this problem, we investigate the use of data compression techniques on climate simulation data from the Community Earth System Model. Ultimately, to convince climate scientists to compress their simulation data, we must be able to demonstrate that the reconstructed data reveals the same mean climate as the original data. As a first step towards that goal, we present an approach for verifying the climate data and use it to evaluate several compression algorithms. Presentations, Session 7, July 1 Bala Rajaratnam – Stanford University Title: A Methodology for Robust Multiproxy Paleoclimate Reconstructions

Great strides have been made in the field of reconstructing past temperatures based on models relating temperature to temperature-sensitive paleoclimate proxies. One of the goals of such reconstructions is to assess if current climate is anomalous in a millennial context. These regression based approaches model the conditional mean of the temperature distribution as a function of paleoclimate proxies (or vice versa). Some of the recent focus in the area has considered methods which help reduce the uncertainty inherent in such statistical paleoclimate reconstructions, with the ultimate goal of improving the confidence that can be attached to such endeavors. In this presentation we introduce novel statistical methodology for achieving this goal using quantile regression with autoregressive residual structure. Our approach requires both estimation of model parameters and the development of a rigorous framework for specifying uncertainty estimates of quantities of interest. We show that by using the above methodology we can demonstrably produce a more robust reconstruction than is possible by

17

using conditional-mean-fitting methods. Our reconstruction shares some of the common features of past reconstructions, but we also gain useful insights. More importantly, we are able to demonstrate a significantly smaller uncertainty than that from previous regression methods. (Joint work with L. Janson)

Wei Ding – University of Massachusetts Boston Title: A Data-Driven Machine Learning Framework for Long-Lead Flood Forecasting

Regional flooding is often produced by long sequences of slowly moving, low-pressure or frontal storm systems including decaying hurricanes or tropical storms occurring over periods of several days to several weeks. Accurate precipitation forecasts by numerical weather prediction models are limited to a few days lead-time because the non-linearity in the governing equations of the atmosphere creates a sensitive dependence on initial conditions. We develop an integrated machine learning framework using complex, high-dimensional, imbalanced atmospheric data to identify precursors of precipitation regimes that lead to major floods up to two weeks in advance. The project targets at: 1) constructing local spatio-temporal features using cuboid-window to capture the dynamic behavior of weather patterns, 2) applying streaming feature selection to online identify atmospheric precursors of precipitation sequences to deal with enormous feature space and non-faithful data distribution, 3) learning a biased discriminative subspace to favor rare flooding events, and 4) incrementally building an online classifier of large-scale atmospheric and hydrological data using least squares-based optimization.

Ansu Chatterjee – University of Minnesota Title: A study of mixed-source variability and dependence in precipitation data over India Understanding the patterns in the available data on precipitation in India, and in particular that due to Indian monsoons, is a challenging problem to statisticians and climatologists alike. The standard Physics-based approach for precipitation modeling, using the Clausius-Clapeyron relation for example, needs to be supplemented by using statistical approaches to deal with multiple sources of uncertainty and dependence. We present a mixed-effect model to partially address address this issue. This is joint work with Lindsey Dietz, Ujjal Mukherjee, Megan Heyman and Abhishek Nandy (UMN).

18

Richard (Ricky) Rood – University of Michigan Ann Arbor Climate Informatics: Human Experts and the End-to-End System The scientific community has expended significant effort to provide access to climate observations and climate-model projections. Still, however, a recent estimate states that scientists and engineers spend more than 60% of their time just preparing the data for model input and data-model intercomparison. When the need for climate data extends outside of the close community of scientists, the usability barriers are often so high as to prohibit the use of the data. This seminar explores the usability of climate data in planning for climate change adaptation and resource management. I describe the gap between what scientists perceive as useful and what planners and decision makers consider usable. I outline barriers and the steps in translating climate data from data providers to data users. Improved usability requires collaboration and interaction between scientists and decision makers that takes into consideration not only the decision environment of potential users but also different ways to manage uncertainty in decision making.

19

Presentations, Session 8, July 2 Kevin Trenberth – National Center for Atmospheric Research KEYNOTE Title: Climate change: It’s about the data isn’t it?

Abstract: The first rule of management is “You can’t manage what you can’t measure” and so data are essential. We have a lot of data and facts about how the climate is changing, and “you are entitled to your own opinion but not your own facts” according the Patrick Daniel Moynihan. A brief outline will be given of the facts: how the global climate is changing with a focus on temperatures and more detail on the recent hiatus in the rise in global mean temperature. We explore the seasonality of the changes, and possible explanations of whether global warming has gone away or not. Changes in the oceans prove to be a key but are not as well known as we would like (not enough good data!) What we do about this is up to all of us, and denial of climate change facts by some politicians ought to be called out.

Imme Ebert-Uphoff – Colorado State University Title: Weakening of atmospheric information flow in a warming climate - preliminary results

Abstract: Causal discovery theory provides methods for identifying potential causal relationships from data. In this talk we look at one application of causal discovery to climate science, namely the construction of graphs of information flow around the globe, and show what those graphs can reveal about likely changes in a warming climate. The key idea is to interpret large-scale atmospheric dynamical processes as information flow around the globe and to identify their pathways from observed data using methods from causal discovery, thus revealing interaction pathways around the globe. We introduce the basic method and present graphs obtained using the daily geopotential height data from the Community Climate System Model Version 4.0 (CCSM4.0)’s 20th century climate simulation and 21st century climate projection. One of the results obtained is that in the CCSM4.0 model under enhanced greenhouse gases forcing, prominent midlatitude information pathways in the midtroposphere weaken and shift poleward, while major tropical information pathways start diminishing. All results are preliminary, since they are based on a single climate model, a specific future emission scenario and a single isobaric level. We discuss our recent progress in extending the methods to generating graphs of 3-dimensional information flow (representing several isobaric layers) and conclude with topics for future research.

20

July 2, 2014 Panel Discussion: Data Science and Climate Science: Narrowing Gaps Moderator: Douglas Nychka, National Center for Atmospheric Research Topic: Understanding and narrowing gaps between Data Science and Climate Science for actionable insights towards understanding impacts and adaptation Abstract: The role of Data Science (e.g., Statistics, Machine Learning, Data Mining, Visualization) in Climate Science is well acknowledged. However, recent literature is reporting challenges: “Failure to account for dependence between models, variables, locations, and seasons is shown to yield misleading results” [1]; “… no scientist thinks you can solve this problem by crunching data alone, no matter how powerful the statistical analysis …” [2]. As the volume and variety of climate data increases [3] and so expands opportunities for the use of data science, many crucial questions arise: What are strengths and limitations of current data science in context of climate data and climate science questions, e.g., impacts? How can these limitations be addressed via either next-generation Data Science methods, new approaches to modeling, or novel observations of the Climate System? Which current Data Science methods leverage concepts and theories of Climate Science? How may next-generation Data Science methods become more aware of concepts and theories of Climate Science to facilitate interpretation of mined patterns by scientists? Are there some kinds of data analysis that will break as climate simulations and remotely sensed data exceed particular resolutions or data volumes? References: [1] Statistical significance of climate sensitivity predictors obtained by data mining, P. M. Caldwell et al., Geophys. Res. Lett., 41:1803-1808, 2014. [2] Eight (No, Nine!) Problems With Big Data, Gary Marcus et al., New York Times, April 6th, 2014. www.nytimes.com/2014/04/07/opinion/eight-no-nine-problems-with-big-data.html?_r=0 [3] Climate data challenges in the 21st century, J. Overpeck et al., Science, 331:700–702, 2011, doi:10.1126/science.1197869. Panelists: Lawrence Buja, National Center for Atmospheric Research Alicia Karspeck, National Center for Atmospheric Research Richard Loft, National Center for Atmospheric Research Linda Mearns, National Center for Atmospheric Research Srini Parthasarthy, Ohio State University Stephen Sain, National Center for Atmospheric Research Shashi Shekhar, University of Minnesota Claudia Tebaldi, National Center for Atmospheric Research

21

Invited Participant Bios (alphabetical order):

Ghassem R. Asrar – Joint Global Change Research Institute, University of Maryland Ghassem.Asrar@pnnl.gov Web: http://www.globalchange.umd.edu Ghassem Asrar is the Director of Joint Global Change Research Institute of the Pacific Northwest National Laboratory located at the University of Maryland, College Park, Maryland, USA. Prior to this position, he served as the Director of World Climate Research Program (WCRP) in Geneva, Switzerland from 2008-2013, after 20 years of service with U.S. Department of Agriculture and National Aeronautics and Space

Administration. He was the chief scientist for the Earth Observing System in the Office of Earth Science at NASA Headquarters prior to being named as the Associate Administrator for Earth Science in 1998. In his position of chief scientist, he led an international team developing the scientific priorities and measurements to be obtained from a series of advanced Earth-orbiting satellites that provided fundamental new insights into the connections between Earth’s land, oceans, atmosphere, cryopshere and life. He established the NASA Earth System Science graduate fellowship and the New Investigators Post-doctoral Programs to support training of the next generation of Earth scientists and engineers. He is the recipient of U.S. Presidential Distinguished Executive Award (2002), an elected Fellow of American Meteorological Society (2001), and IEEE (2000). He has received numerous awards and honors, including the NASA Exceptional Performance Award in 1997, the AIAA Goddard Memorial Lecture Medal in 1998, NASA Exceptional Service Medal, 1999, NASA Distinguished Leadership Medal, 2000, the U.S. Presidential Executive Service Award, 2002, the Space System Award from the American Institute of Aeronautics and Astronautics, 2006, and Distinguished and Outstanding Alumni Awards from the Michigan State University, 2008.

Allison Baker – National Center for Atmospheric Research abaker@ucar.edu Allison H. Baker is a project scientist in the Application Scalability and Performance group in the Technology Development Division at NCAR. Dr. Baker works primarily with the Community Earth System Model (CESM). Her research interests include high-performance computing, software for scientific computing, performance analysis, iterative linear solvers, and, more recently, data compression and verification techniques. Before joining NCAR in 2012,

she worked at the Center for Applied Scientific Computing (CASC) at Livermore National Laboratory, where she was primarily involved with the hypre software project, a library of high-performance preconditioners, with a focus on parallel algebraic multigrid methods and exascale computing. She earned her Ph.D. in Applied Mathematics from the University of Colorado at Boulder in 2003, and her bachelor's degree in Mechanical Engineering from Rice University.

Timothy DelSole – Smithsonian Institution tdelsole@gmu.edu Timothy DelSole conducts research on climate variability and predictability using stochastic turbulence models and multivariate statistics. After completing his doctorate, Dr. DelSole was a Global Change Distinguished Postdoctoral Fellow for two years and a National Research Council Associate for two years at NASA Goddard. Currently, Dr. DelSole is a full professor in the Department of Atmospheric, Oceanic, and Earth Sciences at George Mason University, and a senior research scientist at the Center for Ocean-Land-Atmosphere Studies. Dr.

22

DelSole currently serves as editor of Journal of Climate. Clara Deser – National Center for Atmospheric Research cdeser@ucar.edu

Clara Deser is the head of the Climate Analysis Section within the Climate and Global Dynamics Division at NCAR. Her research interests include diagnostic analysis of observed climate variability in the coupled atmosphere-ocean-ice system, as well as future climate change. Clara is a co-chair of the CESM Climate Variability and Change Working Group.

Dr. Deser received a B.S. in Earth and Planetary Sciences at MIT and earned her PhD in Atmospheric Sciences from the University of Washington. She is also a Fellow of the American Meteorological Society and has been recognized with several professional awards, including the Meisinger Award from the American Meteorological Society and the Editor's Award from the Journal of Climate.

Noah Diffenbaugh – Stanford University diffenbaugh@stanford.edu

Dr. Noah Diffenbaugh is an Associate Professor and Senior Fellow at Stanford University. He studies the climate system, including the processes by which climate change could impact agriculture, water resources, and human health. Dr. Diffenbaugh is currently an Editor of the peer-review journal Geophysical Research Letters, and a Lead

Author for Working Group II of the Intergovernmental Panel on Climate Change (IPCC). Dr. Diffenbaugh is a recipient of the James R. Holton Award from the American Geophysical Union, a CAREER award from the National Science Foundation, and a Terman Fellowship from Stanford University. He has also been recognized as a Kavli Fellow by the U.S. National Academy of Sciences, and as a Google Science Communication Fellow.

Wei Ding – University of Massachusetts Boston ding@cs.umb.edu Web: www.cs.umb.edu/~ding

Wei Ding received her Ph.D. degree in Computer Science from the University of Houston in 2008. She has been an Assistant Professor of Computer Science in the University of Massachusetts Boston since 2008. Her research interests include data mining, machine learning, artificial intelligence, computational semantics, and with applications to environmental sciences, astronomy, geosciences, and health science. She has published more than 85 referred research papers, 1 book, and has 2 patents. She is an Associate Editor of Knowledge and Information Systems (KAIS) and an

editorial board member of the Journal of Information System Education (JISE), the Journal of Big Data, and the Social Network Analysis and Mining Journal. She is the recipient of a Best Paper Award at the 2011 IEEE International Conference on Tools with Artificial Intelligence (ICTAI), a Best Paper Award at the 2010 IEEE International Conference on Cognitive Informatics (ICCI), a Best Poster Presentation award at the 2008 ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems (SIGSPAITAL GIS), and a Best PhD Work Award between 2007 and 2010 from the University of Houston. Her research projects are currently sponsored by NASA and DOE. She is an IEEE senior member and an ACM member.

23

Imme Ebert-Uphoff – Colorado State University iebert@engr.colostate.edu Web: http://www.engr.colostate.edu/~iebert/

Imme Ebert-Uphoff is a research faculty member in the Department of Electrical and Computer Engineering at Colorado State University. Her current research focuses on causal discovery, i.e. learning potential causal relationships from data using machine learning algorithms. Most of her current research applies causal discovery to climate science, generating new hypotheses about causal relationships between different climate variables, as well as generating "graphs of information flow" around the globe. These graphs may help climate scientists

to better understand certain dynamical processes of our planet's climate. Before joining Colorado State she was an assistant professor, then associate professor (with tenure) in the Mechanical Engineering department at Georgia Tech. She received a M.S. degree in mathematics from the University of Karlsruhe, Germany, followed by M.S and Ph.D. degrees in Mechanical Engineering from the Johns Hopkins University (Baltimore, MD). Rewards she received include an NSF CAREER award (2000) and the CETL BP/Amoco Junior Faculty Teaching Excellence Award at Georgia Tech (2000), the SME Outstanding Young Manufacturing Engineer Award (2001) and the IEEE Conference on Robotics and Automation Best Video Award (2004). Recently she co-chaired the Third International Workshop on Climate Informatics in 2013 and is on the steering committee of the fourth workshop in 2014.

Ivy Frenger – Princeton University ifrenger@princeton.edu Ivy Frenger is a Postdoctoral Research Fellow at the Program in Atmospheric and Oceanic Sciences, Princeton University, working with Jorge Sarmiento. Most of her work concerns ocean mesoscale eddies and their effects, with the Southern Ocean as focus area. Projects of hers include ocean eddy tracking and statistical analysis,

Southern Ocean water masses, and the impacts of a changing climate therein. To this end, she uses satellite data, in-situ observations and high resolution model simulations carried out by Princeton University in collaboration with the Geophysical Fluid Dynamics Laboratory, NOAA. Ivy earned her Ph.D. in Environmental Physics at ETH in Zurich, Switzerland, in 2013. She received her Master's degree in Meteorology from the University of Hamburg, Germany, in 2009.

Dimitris Giannakis – New York University dimitris@cims.nyu.edu Web: www.cims.nyu.edu/~dimitris

Dimitris Giannakis is an Assistant Professor of Mathematics at New York University’s Courant Institute of Mathematical Sciences. He received his Bachelors and Masters degrees in Physics from Cambridge University (2001) and a PhD in Physics from the University of Chicago (2009). Prior to joining the faculty of NYU he was a postdoctoral researcher at Courant’s Center for Atmosphere Ocean Science. Dr. Giannakis’ research interests are in applied mathematics for climate atmosphere ocean science.

He is currently working on methods for spatiotemporal decomposition of large-scale data from climate models and observations. This work has applications in the detection and forecasting of climate patterns on seasonal to interannual timescales.

24

Suchi Gopal – Boston University suchi@bu.edu Web: http://blogs.bu.edu/suchi/ Suchi Gopal received her Ph.D. degree in Geography and Spatial Sciences from the University of California, Santa Barbara. She is a full Professor in the department of Earth and Environment, Boston University as well as a Research Professor at several centers including the Center for Remote Sensing, Center for Environment and Energy Systems, Cognitive and Neural Systems (CNS) Tech lab, BU’s Marine

Program (BUMP) and the Pardee Center of the Study of the Longer-Range Future. Her research interests include remote sensing and GIS technologies, spatial accuracy, multi-scale models, neural networks, and data visualization applied in a variety of fields including marine spatial planning, public health, and global change. She has published numerous papers. Her most recent publications includes marine spatial planning in Belize to protect coral reefs, characterization of urban landcover using fuzzy sets, spatial modeling of health facility availability and accessibility by pregnant women in Kalomo, Zambia, and examining the impact of development and climate change in Cambodia’s Tonle Sap lake. Her work is currently funded by the National Science Foundation and MacArthur Foundation. She is an editorial board member of Geographical Analysis.

Forrest M. Hoffman - Oak Ridge National Laboratory and University of California – Irvine forrest@climatemodeling.org Web: http://www.climatemodeling.org/∼forrest Forrest M. Hoffman is a Senior Computational Climate Scientist in the Computer Science & Mathematics Divsion and the Earth System Modeling Theme Lead for the Climate Change Science Institute (CCSI) at Oak Ridge National Laboratory (ORNL). Forrest is also pursuing a Ph.D. in the Department of Earth System

Science at the University of California, Irvine. With a focus on terrestrial biogeochemical cycles, Forrest performs development and evaluation of global Earth system models (ESMs). He has a special interest in applying data mining and machine learning techniques to observational data and model results for improving our understanding and the representation of feedbacks between biogeochemistry and the climate system in ESMs.

Jim Hurrell – National Center for Atmospheric Research jhurrell@ucar.edu Web: http://www.cgd.ucar.edu/staff/jhurrell/

James (Jim) W. Hurrell is the Director of the National Center for Atmospheric Research (NCAR). He is also a Senior Scientist in the Climate and Global Dynamics Division (CGD) of the NCAR Earth System Laboratory (NESL). Jim received a Ph. D. (1990) in Atmospheric Science from Purdue University. He is the former Chief Scientist of Community Climate Projects in CGD, which includes the Community Earth System Model (CESM), and a former Director of both CGD and NESL. Jim also spent a year as a visiting scientist at the

Hadley Centre for Climate Prediction and Research in the U.K. Jim's research has centered on empirical and modeling studies and diagnostic analyses to better understand climate, climate variability and climate change. He has authored or co-authored around 100 peer-reviewed journal articles and book chapters, as well as dozens of other planning documents, workshop papers, and editorials. Jim has also edited several books, and he has been acknowledged as a Highly Cited Researcher by Thomson-ISI (2004). Jim has given more than 150 professional invited and

25

keynote talks, as well as many other contributed presentations at national and international conferences. Jim has convened around two-dozen national and international workshops, and he has served several national and international science-planning efforts. Jim has been extensively involved in the World Climate Research Programme (WCRP) on Climate Variability and Predictability (CLIVAR), including roles as co-chair of the Scientific Steering Group (SSG) of both U.S. and International CLIVAR, Chair of the Scientific Organizing Committee for the WCRP Open Science Conference (2011), and membership on several other CLIVAR panels. He has also served the International Geosphere-Biosphere Programme (IGBP) as a member of the Global Ocean Ecosystem Dynamics (GLOBEC) SSG and the CLIVAR-PAGES (Past Global Changes) working group. Jim has been involved in assessment activities of the Intergovernmental Panel on Climate Change (IPCC) and the U.S. Climate Change Science Program (CCSP). He has served on several National Research Council (NRC) panels, and he has provided briefings and testimonies to both the U.S. Senate and the House of Representatives on climate change science.

Ramakrishna Nemani – NASA Ames Research Center Rama.nemani@ nasa.gov

Rama Nemani is a research scientist with the Advanced Supercomputing division at Ames Research Center. His research deals with integration of satellite data into simulation models for understanding and predicting the state and function of terrestrial ecosystems. He published on a variety of topics including remote sensing, global ecology, ecological forecasting and climatology. He has received several awards from NASA including the exceptional scientific achievement medal

in 2008 and the outstanding leadership medal in 2012.

Douglas Nychka – National Center for Atmospheric Research Web: https://www2.image.ucar.edu/staff/nychka-doug

Douglas Nychka is a statistical scientist with an interest in the problems posed by geophysical data sets. His Ph. D. (1983) is from the University of Wisconsin and he subsequently spent 14 years as a faculty member at North Carolina State University. His research background in fitting curves and surfaces lead to an interest in the analysis of spatial and environmental data. Pursuing this area of application, he assumed leadership of the Geophysical Statistics Project at the

National Center for Atmospheric Research (NCAR) in 1997, an NSF program to build collaborative research and training between statistics and the geosciences. In 2004, he became Director of the Institute of Mathematics Applied to Geosciences, an interdisciplinary component at NCAR with a focus on transferring innovative mathematical models and tools to the geosciences. His current interests are in quantifying the uncertainty of numerical experiments that simulate the Earth's present and possible future climate and spatial statistics applied to large data sets. He has received the Jerry Sacks Award for Multidisciplinary Research (2004), the Achievement Award given by the Committee for the 12th International Meetings on Statistical Climatology (2013) and is a Fellow of the American Statistical Association.

26

Nikunj C. Oza – NASA Ames Research Center nikunj.c.oza@nasa.gov Web: https://c3.nasa.gov/dashlink/members/27/

Nikunj Oza is the leader of the Data Sciences Group at NASA Ames Research Center. He additionally leads the Discovery of Precursors to Safety Incidents (DPSI) team which applies data mining to aviation safety. He is an Ames Center Program Manager for the Advanced Information Systems Technology (AIST) program, which funds research and development efforts in Information

Technology with application to Earth science. Dr. Oza’s 40+ research papers represent his research interests which include data mining, machine learning, fault detection, and their applications to Aeronautics and Earth Science. He received the Arch T. Colwell Award for co-authoring one of the five most innovative technical papers selected from 3300+ SAE technical papers in 2005, and the 2010 NASA Aeronautics Research Mission Directorate Associate Administrator team award for Technology and Innovation, . Dr. Oza is on the editorial board of the peer-reviewed journal Information Fusion (Elsevier). He received his B.S. in Mathematics with Computer Science from MIT in 1994, and M.S. (in 1998) and Ph.D. (in 2001) in Computer Science from the University of California at Berkeley.

Srinivasan Parthasarathy – Ohio State University Web: http://www.cse.ohio-state.edu/~srini/ Dr. Srinivasan Parthasarathy (PhD, University of Rochester), is currently a Professor of Computer Science and Engineering at the Ohio State University (OSU). His research interests are broadly in the areas of Data Mining, Databases, Bioinformatics and Parallel and Distributed Computing. He is a recipient of an Ameritech Faculty Fellowship in 2001, a US National Science Foundation CAREER

award in 2003, a US Department of Energy Early Career Award in 2004, multiple IBM Faculty Awards in 2007 and 2010, and a Google Research Award in 2009. His papers have received six best paper awards or similar honors from among ten nominations in leading conferences in the field, including ones at SIAM international conference on data mining (SDM), IEEE international conference on data mining (ICDM), Intelligent Systems for Molecular Biology (ISMB), the Very Large Databases Conference (VLDB) and at the ACM Knowledge Discovery and Data Mining (SIGKDD). He has served on the program, organizational and steering committees of leading conferences in the fields of data mining, databases, and high performance computing. He currently serves on the editorial boards of several journals including the Data Mining and Knowledge Discovery Journal (DMKDJ), the Distributed and Parallel Databases Journal (DAPDJ), the Journal of Parallel and Distributed Computing (JPDC), and the ACM Transactions on Knowledge Discovery and Data Mining (ACM-TKDD). Prof. Parthasarathy is also a co-director of the newly minted undergraduate program on Data Analytics (https://data-analytics.osu.edu/) at Ohio State University -- among the first of its kind nationwide.

27

Bala Rajaratnam – Stanford University brajarat@stanford.edu Web: https://woods.stanford.edu/about/woods-faculty/bala-rajaratnam

Bala Rajaratnam is currently a faculty member at Stanford in the Department of Statistics and the Department of Environmental Earth System Science. His work in statistics, machine learning and environmental sciences span the areas of graphical models, network models, high dimensional regression, Bayesian

inference for high dimensional data, covariance matrix estimation and spatial and environmental statistics.

Pradeep Ravikumar – University of Texas at Austin pradeepr@cs.utexas.edu Web: http://www.cs.utexas.edu/~pradeepr/ Pradeep Ravikumar received his B.Tech. in Computer Science and Engineering from the Indian Institute of Technology, Bombay, and his PhD in Machine Learning from the School of Computer Science at Carnegie Mellon University. He was then a postdoctoral scholar at the Department of Statistics at the University of California, Berkeley. He is now an Assistant Professor in the Department of Computer Science, at the University of Texas at Austin. He is also affiliated with

the Department of Statistics and Data Sciences, and the Institute for Computational Engineering and Sciences at UT Austin. His thesis has received honorable mentions in the ACM SIGKDD Dissertation award and the CMU School of Computer Science Distinguished Dissertation award. He is a Sloan Fellow, a Siebel Scholar, a recipient of the NSF CAREER Award; and was Program Chair for AISTATS 2013.

Richard (Ricky) Rood – University of Michigan Ann Arbor rbrood@umich.edu Web: http://aoss.engin.umich.edu/people/rbrood Richard Rood is a Professor in the Department of Atmospheric, Oceanic and Space Sciences and in the School of Natural Resources and the Environment at the University of Michigan. He teaches a cross-discipline graduate course on climate change, which addresses critical analysis and complex problem solving.

Prior to moving to Michigan in 2005, Rood was a researcher and manager of both scientific and computational organizations at NASA. He writes the climate change blog for Wunderground.com.

Kevin E. Trenberth – National Center for Atmospheric Research trenbert@ucar.edu Web: http://www.cgd.ucar.edu/staff/trenbert/ Dr. Kevin E. Trenberth is a Distinguished Senior Scientist in the Climate Analysis Section at the National Center for Atmospheric Research. From New Zealand, he obtained his Sc. D. in meteorology from Massachusetts Institute of Technology. He was a lead author of the 1995, 2001 and 2007 Scientific Assessment of Climate Change reports from the Intergovernmental Panel on Climate Change (IPCC), and

shared the 2007 Nobel Peace Prize which went to the IPCC. He served from 1999 to 2006 on the Joint Scientific Committee of the World Climate Research Programme (WCRP), and he chaired the WCRP

28

Observation and Assimilation Panel from 2004 to 2010 and now chairs the Global Energy and Water Exchanges (GEWEX) scientific steering group. He has also served on many national committees. He is a fellow of the American Meteorological Society (AMS), the American Association for Advancement of Science, the American Geophysical Union, and an honorary fellow of the Royal Society of New Zealand. In 2000 he received the Jule G. Charney award from the AMS; in 2003 he was given the NCAR Distinguished Achievement Award; and in 2013 he was awarded the Prince Sultan Bin Abdulaziz International Prize for Water. He edited a 788 page book Climate System Modeling, published in 1992 and has published 500 scientific articles or papers, including 55 books or book chapters, and over 225 refereed journal articles. He has given many invited scientific talks as well as appearing in a number of television, radio programs and newspaper articles. He is listed among the top 20 authors in highest citations in all of geophysics.

Raju Vatsavai – Oak Ridge National Laboratory rvatsavai@acm.org Web: http://web.ornl.gov/~r7v Currently, Raju is the lead data scientist for the computational sciences and engineering division of the Oak Ridge National Laboratory (ORNL). Raju works at the intersection of big data management, data analytics, and high performance computing with applications in national security, geospatial intelligence, natural resources, climate change, location-based services, and human terrain mapping.

Raju has published more than 70 peer reviewed research articles, served as PC member of several international conferences including ACM KDD and ACM SIGSPATIAL GIS, co-organized several workshops including SSTDM with ICDM, BigSpatial with ACM SIGSPATIAL GIS, BDAC with SC, and served on several NSF panels. He has more than 20 years of research and development experience in large-scale spatiotemporal data management and geographic knowledge discovery. He holds MS and PhD degrees in computer science from the University of Minnesota. Raju will be joining the Department of Computer Science, NC State University in August 2014 as Associate Professor of Geospatial Analytics in the Chancellor’s Faculty Excellence Program.

Warren Washington – National Center for Atmospheric Research Web: http://www.cgd.ucar.edu/ccr/warren/ WARREN M. WASHINGTON is a senior scientist and former head of the Climate Change Research Section and director of the Climate and Global Dynamics Division at the National Center for Atmospheric Research (NCAR) in Boulder, Colorado. His expertise is in atmospheric and climate Research. He has engaged in research for early 50 years, and he has given advice, testimony, and lectures on global climate

change. Dr. Washington has been a member the President’s National Advisory Committee on Oceans and Atmosphere and has had presidential appointments under the Carter, Reagan, Clinton, and Bush administrations. More recently, he served on the National Science Board as a member and as its chair. He has more than 150 publications and co-authored with Claire Parkinson a book that is considered a standard reference on climate modeling, An Introduction to Three-Dimensional Climate Modeling, and an autobiography, Odyssey in Climate Modeling, Global Warming, and Advising Five Presidents. Dr. Washington has many awards, including being a member of the National Academy of Engineering, the American Meteorological Society (former president), the American Philosophical Society, and the American Academy of Arts and Sciences. Members of his group at NCAR shared in the 2007 Nobel Peace Prize as significant contributors to the Inter-governmental Panel of Climate Change (IPCC) Assessment. Dr. Washington has honorary degrees from Oregon State University, Bates College and University of

29

Massachusetts, Amherst. He is also principal investigator for the University for Atmospheric Research and U. S. Department of Energy cooperative agreement that carried out climate research. In 2010, he was awarded the National Medal of Science by President Obama, the nation’s highest science award. Dr. Washington earned a B.S. in physics and a M.S. in meteorology from Oregon State University and a Ph.D. in meteorology from Pennsylvania State University. He has served on a number of National Research Committees of the National Academies, and is currently serving as chair of the Committee to Advise the U.S. Global Change Research Program.

30

31

Expeditions in Computing: Understanding Climate Change, A Data Driven Approach NSF Awards: 1029711, 1029166, 1029731, 1028746 Project Homepage: climatechange.cs.umn.edu

Expeditions in Computing Team: The project team, led by the University of Minnesota, includes faculty and researchers from Minnesota's College of Science and Engineering, College of Food, Agricultural and Natural Resource Sciences, College of Liberal Arts, and the Institute on the Environment, as well as researchers from North Carolina A & T State University, North Carolina State University, Northwestern University, and Northeastern University.

Vipin Kumar – University of Minnesota PI of Expeditions in Computing Project kumar@cs.umn.edu, www.cs.umn.edu/~kumar Vipin Kumar is currently William Norris Professor and Head of the Computer Science and Engineering Department at the University of Minnesota. Kumar received the B.E. degree in Electronics & Communication Engineering from Indian Institute of Technology Roorkee (formerly, University of Roorkee), India, in 1977, the M.E. degree

in Electronics Engineering from Philips International Institute, Eindhoven, Netherlands, in 1979, and the Ph.D. degree in Computer Science from University of Maryland, College Park, in 1982. Kumar's current research interests include data mining, high-performance computing, and their applications in Climate/Ecosystems and Biomedical domains. Kumar is the Lead PI of a 5-year, $10 Million project, "Understanding Climate Change - A Data Driven Approach", funded by the NSF's Expeditions in Computing program that is aimed at pushing the boundaries of computer science research. He also served as the Director of Army High Performance Computing Research Center (AHPCRC) from 1998 to 2005. His research has resulted in the development of the concept of isoefficiency metric for evaluating the scalability of parallel algorithms, as well as highly efficient parallel algorithms and software for sparse matrix factorization (PSPASES) and graph partitioning (METIS, ParMetis, hMetis). He has authored over 300 research articles, and has coedited or coauthored 11 books including widely used text books ``Introduction to Parallel Computing'' and ``Introduction to Data Mining'', both published by Addison Wesley. Kumar has served as chair/co-chair for many international conferences and workshops in the area of data mining and parallel computing, including IEEE International Conference on Data Mining (2002) and International Parallel and Distributed Processing Symposium (2001). Kumar co-founded SIAM International Conference on Data Mining and served as a founding co-editor-in-chief of Journal of Statistical Analysis and Data Mining (an official journal of the American Statistical Association). Currently, Kumar serves on the steering committees of the SIAM International Conference on Data Mining and the IEEE International Conference on Data Mining, and is series editor for the Data Mining and Knowledge Discovery Book Series published by CRC Press/Chapman Hall. Kumar is a Fellow of the ACM, IEEE and AAAS. He received the Distinguished Alumnus Award from the Indian Institute of Technology (IIT) Roorkee (2013), the Distinguished Alumnus Award from the Computer Science Department, University of Maryland College Park (2009), and IEEE Computer Society's Technical Achievement Award (2005). Kumar's foundational research in data mining and its applications to scientific data was honored by the ACM SIGKDD 2012 Innovation Award, which is the highest award for technical excellence in the field of Knowledge Discovery and Data Mining (KDD).

32

Ankit Agrawal– Northwestern University ankitag@eecs.northwestern.edu Ankit Agrawal is a Research Associate Professor in the Dept. of Electrical Engineering and Computer Science at Northwestern University. He received his PhD in Computer Science from Iowa State University, USA in 2009, and was awarded the Research Excellence Award and Peer Research Award for outstanding research accomplishments. He received B.Tech in Computer Science and Engineering from

the Indian Institute of Technology, Roorkee, India in 2006, where he was the graduating topper of his batch and was awarded Institute Silver medals for obtaining the highest GPA and the best B.Tech project. His research interests include high performance data mining and its applications in bioinformatics, climate science, social media, materials science, etc., and has published more than 80 papers in various peer-reviewed journals and conferences. His research is supported by National Science Foundation, Department of Energy, Air Force Office of Sponsored Research, and National Institute of Standards and Technology.

Arindam Banerjee – University of Minnesota banerjee@cs.umn.edu Arindam Banerjee is an Associate Professor at the Department of Computer and Engineering and a Resident Fellow at the Institute on the Environment at the University of Minnesota, Twin Cities. His research interests are in machine learning, data mining, convex analysis and optimization, and their applications in complex real-

world problems, including climate science, ecology, aviation, and the web. He has won several awards, including the IBM Faculty Award (2013), the Yahoo Faculty Research Engagement Program Award (2013), the NSF CAREER award (2010), the McKnight Land-Grant Professorship at the University of Minnesota, Twin Cities (2009–2011), the IBM PhD fellowship (2003-05), and five Best Paper awards in top-tier conferences.

Ruben Buaba – North Carolina Agricultural & Technical State University rbuaba@ncat.edu Ruben Buaba is an Adjunct Assistant Professor and Postdoctoral Researcher in the Department of Electrical and Computer Engineering at North Carolina Agricultural and Technical State University. He received his Ph.D. in Electrical Engineering from North Carolina Agricultural and Technical State University in 2012. His research interests include Controls, Data Mining, Image Processing, Machine Learning, Search Algorithms and High Performance and Scalable

Computing. He obtained his B.S. in Electrical and Electronic Engineering from Kwame Nkrumah University of Science and Technology, Ghana in 2007. In the summer of 2009, he was an Intern at NOAA/NGDC, Boulder, Colorado, where he investigated a Similarity Search Algorithm for satellite image retrieval application. He has recently developed software called SIBRA (Satellite Image-Based Retrieval Application), which provides an interactive image search platform for image search analyses. His

33

research work is partially supported by the NSF Expedition in Computing. He has published more than 9 papers in peer-reviewed journals and conferences.

Snigdhansu Chatterjee – University of Minnesota chatterjee@stat.umn.edu Ansu Chatterjee is Associate Professor in the School of Statistics, University of Minnesota. After graduating from the Indian Statistical Institute, he worked at the University of Manchester in England and at University of Nebraska-Lincoln before joining University of Minnesota, where he is currently tenured. He has published in the

Annals of Statistics, Annals of Applied Statistics, Annals of the Institute of Statistical Mathematics, Bioinformatics, and other journals. His research interests include climate statistics, small area statistics, Bayesian statistics, change detection methods, resampling techniques and other nonparametric methodology.

Zhengzhang Chen – Northwestern University zhengzhang.chen@northwestern.edu Zhengzhang Chen is currently a Research Staff Member at NEC Laboratories America, and an Adjunct Research Assistant Professor in the Electrical Engineering and Computer Science department at Northwestern University. He earned his PhD in Computer Science from North Carolina State University in 2012, and his research

interests include data mining, bioinformatics, graph algorithms, social computing, and machine learning.

Alok Choudhary – Northwestern University choudhar@eecs.northwestern.edu Alok Choudhary is a John G. Searle Professor of Electrical Engineering and Computer Science at Northwestern University. He is the founding director of the Center for Ultra-scale Computing and Information Security (CUCIS). Prof. Choudhary was a co-founder and VP of Technology of Accelchip Inc. in 2000. Accelchip, Inc., was eventually acquired by Xilinx. He received the National Science

Foundation's Young Investigator Award in 1993. He also received an IEEE Engineering Foundation award, an IBM Faculty Development award and an Intel Research Council award. He is a fellow of IEEE, ACM and AAAS. His research interests are in high-performance computing, data intensive computing, scalable data mining, computer architecture, high-performance I/O systems and software and their applications in many domains including information processing (e.g., data mining, CRM, BI) and scientific computing (e.g., scientific discoveries). Alok Choudhary has published more than 350 papers in various journals and conferences and has graduated 27 PhD students. Techniques developed by his group can be found on every modern processor and scalable software developed by his group can be found on most supercomputers. Alok received his Ph.D. degree in Electrical and Computer Engineering from the University of Illinois, Urbana-Champaign, in 1989.

James Faghmous – University of Minnesota jfagh@cs.umn.edu

34

James H. Faghmous obtained his Ph.D. in computer science from the University of Minnesota -Twin Cities. As part of the Expeditions team, James developed theory-guided data mining algorithms to analyze large climate datasets with applications to tropical cyclone forecasting and mesoscale ocean eddy monitoring. James' Ph.D. thesis was awarded the 2014 Best Dissertation Award in Physical Sciences and Engineering at the University of Minnesota. James was amongst the earliest data scientists to publish on "theory-guided data science" which earned him Best Student Paper Award at the 2011 NASA Conference on Intelligent Data Understanding.

James' research has been generously funded by an NIH Neuro-Physical-Computational Graduate Fellowship, an NSF Graduate Research Fellowship, an NSF Nordic Research Opportunity Fellowship, and a University of Minnesota Doctoral Dissertation Fellowship. James graduated in 2006 with a B.Sc. in computer science from the City of College of New York where he was a Rhodes and a Gates Scholar nominee.

Jonathan Foley – University of Minnesota jfoley@umn.edu Jonathan Foley is the director of the Institute on the Environment (IonE) at the University of the Minnesota, where he is a professor and McKnight Presidential Chair in the Department of Ecology, Evolution and Behavior. He also leads the IonE’s Global Landscapes Initiative. Foley’s work focuses on the sustainability of

our civilization and the global environment. He and his students have contributed to our understanding of global food security, global patterns of land use, the behavior of the planet’s climate, ecosystems and water cycle, and the sustainability of the biosphere. This work has led him to be a regular advisor to large corporations, NGOs and governments around the world.

Poulomi Ganguli – Northeastern University p.ganguli@neu.edu Poulomi Ganguli is a post-doctoral research fellow in the Civil and Environmental Engineering department at Northeastern University. She earned PhD in Civil Engineering from Indian Institute of Technology Bombay in 2013. Her research interests include hydrological extremes, hydro-climatology and assessment of climate change and variability in

surface and sub-surface hydrology.

Auroop Ganguly – Northeastern University a.ganguly@neu.edu Auroop Ganguly directs the Sustainability and Data Sciences Laboratory (SDS Lab) at Northeastern University in Boston, MA, where he joined in Fall 2011, as an associate professor of civil and environmental engineering. He was at the Oak Ridge National Laboratory in their computational sciences and engineering division for exactly 7 years from 2004-2011, most recently as a senior scientist. He led or co-led projects funded by DARPA, DOD, DHS, DOE, ONR, and ORNL, and

received three significant event awards and two outstanding mentor awards at ORNL. He received an

35

outstanding faculty award from the University of Tennessee in Knoxville where he held a joint appointment with ORNL. His research on climate extremes and uncertainty has been published in Proceedings of the National Academy of Sciences, Nature Climate Change, Geophysical Research Letters, Journal of Geophysical Research, and on hydrology in Water Resources Research, Advances in Water Resources, Journal of Hydrometeorology, and Nonlinear Processes in Geophysics. His work on computational methods and complex systems has been published in Physical Review E, IEEE Transactions on Intelligent Transportation Systems, and Intelligent Data Analysis, as well as in peer-reviewed conferences in computer science such as SIAM Data Mining, besides ACM KDD and IEEE ICDM workshops. He was the primary founding organizer of a 2006-2012 workshop series on sensor data mining at the ACM KDD, the first of which resulted in an edited book by CRC Press called "Knowledge Discovery from Sensor Data". He holds associate editor positions in the journal Water Resources Research published by the American Geophysical Union and Journal of Computing in Civil Engineering published by the American Society of Civil Engineers. In addition, he is an elected member of the Artificial Intelligence Committee of the American Meteorological Society. He was on a visiting faculty position at the University of South Florida in Tampa, FL, for about 10 months, and currently holds a visiting faculty position at the Indian Institute of Technology Bombay in Mumbai, India, within their climate change interdisciplinary program. He was employed at Oracle Corporation for about 5 years, first as a time series software developer for a year in their core database kernel, and then as the product manager of their demand forecasting and planning product, which he managed from inception to market acceptance. For a year, he was the product manager for analytics and strategy at a best-of-breed semi-startup company on demand-driven supply chain, the company later got acquired by Oracle Corporation. Ganguly has a PhD from MIT in hydrology from their civil and environmental engineering department, and research experience at their Sloan school of management in supply chain, information sciences, and data mining.

William Hendrix – Northwestern University whendrix@northwestern.edu William Hendrix is a Postdoctoral Research Fellow in the Electrical Engineering and Computer Science department at Northwestern University. He earned his PhD in Computer Science from North Carolina State University in 2010, and his research interests include graph algorithms, high performance computing, and data mining.

Abdollah Homaifar – North Carolina Agricultural & Technical University homaifar@ncat.edu Abdollah Homaifar received his B.S. and M.S. degrees from the State University of New York at Stony Brook in 1979 and 1980, respectively, and his Ph.D. degree from the University of Alabama in 1987, all in electrical engineering. He is currently the Duke Energy Eminent professor in the Department of Electrical and Computer Engineering at North Carolina A&T State University (NCA&TSU). He is also the director of the Autonomous Control and Information Technology center at NCA&TSU, and Thrust Area

Leader for Data Fusion, data mining and Distributed Architecture, NOAA ISET Center, at NCA&TSU. His research interests include machine learning, climate data processing, optimization, optimal control, flexible robotics, signal processing, soft computing and modeling. He is the author and co-author of over 200 articles in journals and conference proceedings, one book, and three chapters of books. He has participated in six short courses, serves as an associate editor of the Journal of Intelligent Automation and Soft

36

Computing, and is a reviewer for IEEE Transactions on Fuzzy Systems, Man Machines & Cybernetics, and Neural Networks. He is a member of the IEEE Control Society, Sigma Xi, Tau Beta Pi, and Eta Kapa Nu.

Joseph F. Knight – University of Minnesota jknight@umn.edu Joseph Knight is an Assistant Professor of Remote Sensing in the Department of Forest Resources at the University of Minnesota, Twin Cities. Dr. Knight studies how changing land use affects both natural resources and humans. He uses geospatial science methods such as remote sensing, image processing, and

geographic information systems (GIS) in applications such as: identifying and characterizing natural and anthropogenic landscape change to assess impacts on natural resources, wetlands mapping and characterization, describing landscape-human interactions that lead to exposure to infectious diseases, and thematic accuracy assessment methods development. Dr. Knight teaches three courses at the University of Minnesota: Remote Sensing of Natural Resources and Environment, Field Remote Sensing and Resource Survey, and Issues in the Environment. He holds a Ph.D. from North Carolina State University and previously worked as a Biologist with the United States Environmental Protection Agency. He is an author of numerous publications, including peer-reviewed journal articles, book chapters, and technical reports. Dr. Knight is a recipient of the 2007 U.S. Environmental Protection Agency Science and Technology Achievement Award.

Wei-keng Liao– Northwestern University wkliao@ece.northwestern.edu Wei-keng Liao is a Research Professor in the Electrical Engineering and Computer Science Department. He received a Ph.D. in computer and information science from Syracuse University in 1999. Prof. Liao's research interests are in the area of high-performance computing, parallel I/O, parallel file systems, data mining, and data management for large-scale scientific applications.

Stefan Liess – University of Minnesota liess@umn.edu Stefan Liess is a Research Associate in Atmospheric Sciences at the University of Minnesota. He analyzes climate dynamics and climate change with observations, general circulation models, and high-resolution regional models. His specific research interests are intraseasonal variability and predictability on the order of a

few weeks to multiple months, interactions of climate and vegetation including regional projections of future climate and vegetation pattern, and teleconnections in global and tropical climate. Previously, he worked as the responsible scientist and administrator for the SPARC (Stratospheric Processes and Their Role in Climate) Data Center and studied the impacts of tropopause characteristics on tropical convection. Stefan’s research has been funded by the National Science Foundation, the National Aeronautics and Space Administration, the National Oceanic and Atmospheric Administration, and the Max-Planck Institute for Meteorology in Germany. He received his PhD in Atmospheric Sciences from

37

Max-Planck Institute for Meteorology in 2002. In 1997 he earned a MS in Meteorology from the University of Hamburg in Germany.

Rachindra Mawalagedara – Northeastern University Rachindra Mawalagedara is a Postdoctoral Research Associate in the Sustainability and Data Sciences Lab at Northeastern University. She earned her Ph.D. in Earth and Atmospheric Sciences from University of Nebraska-Lincoln in 2013. Her work is primarily focused on regional climate modeling and climate change with particular

emphasis on studying the climatic impacts due to changes in land-use.

Nagiza Samatova – North Carolina State University samatova@csc.ncsu.edu Dr. Nagiza F. Samatova is a Professor in Computer Science Department of North Carolina State University and a Senior Research Scientist in Computer Science and Mathematics Division of Oak Ridge National Laboratory. She received the B.S. degree in applied mathematics from Tashkent State University, Uzbekistan, in 1991

and her Ph.D. degree in mathematics from the Computing Center of Russian Academy of Sciences (CCAS), Moscow, in 1993. She also obtained an M.S. degree in Computer Science in 1998 from the University of Tennessee, Knoxville, USA. Dr. Samatova specializes in Graph Theory and Algorithms, High Performance Data Analytics, Climate Informatics, Bioinformatics, Systems Biology, Data Management, Scientific and High Performance Computing, and Machine Learning. She is the author of over 200 publications in peer-reviewed journals and conference proceedings. She co-edited the book, “Practical Graph Mining with R.” She is the recipient of various awards and honors, including the R&D 100 Award for ADIOS, R&D 100 Award for VIPAR, IEEE Distinguished Contribution to Public Service, Euro-Par Distinguished Paper Award, HPDC Best Paper Awards, NASA CIDU Best Paper Award. See her Dossier at http://www.csc.ncsu.edu/directories/faculty_info.php?id=2362.

Fredrick Semazzi – North Carolina State University fred_semazzi@ncsu.edu Dr. Semazzi has served in several senior positions of scientific organizations in the US, Europe, and Africa. He was a lecturer in the department of meteorology at the University of Nairobi, Kenya; Research Associate Scientist at NASA Goddard Space Flight Center, Greenbelt Maryland; US National Science Foundation (NSF) Climate Dynamics Program Associate Program Director, Washington DC; Founding Director

of Climate Information & Prediction Services Program of the World Meteorological Organization at the United Nations, Geneva Switzerland; Senior Scientist, World Climate Research Program (WCRP) CLIVAR International Project Office, Southampton, England; Director of the Climate-PSM degree program & Professor at the North Carolina State University, Raleigh NC, USA. Dr. Semazzi has served in capacities of review editor & author for the Intergovernmental Panel on Climate Change (IPCC) climate change assessment. In March 2009 he received a certificate of special recognition from Dr. Rajendra K. Pachauri, Chairman of IPCC, for distinguished contribution resulting in the award of the Nobel Peace Prize for 2007 to the IPCC. This recognition was extended, ‘only to those who have contributed substantially to the

38

work of the IPCC over the years since the inception of the organization’. Dr. Semazzi is a member of the Joint Scientific Committee (JSC) for the World Climate Research Programme (WCRP; 2009-present). WCRP is sponsored by the World Meteorological Organization, the International Council for Science (ICSU) and the Intergovernmental Oceanographic Commission (IOC) of UNESCO. Dr. Semazzi has directed over 20 Masters and PhD degree theses. He has authored over sixty scientific publications & served as principal investigator and co-investigator on many grants, with total funding of more than $25 million.

Shashi Shekhar – University of Minnesota shekhar@cs.umn.edu http://www.cs.umn.edu/~shekhar Shashi Shekhar is a Mcknight Distinguished University Professor at the University of Minnesota (Computer Science faculty). For contributions to geographic information systems (GIS), spatial databases, and spatial data mining, he received the IEEE-CS Technical Achievement Award and was elected an IEEE Fellow as well as an AAAS Fellow. He was also named a key difference-maker for the field of GIS by the most

popular GIS textbook. He has a distinguished academic record that includes 280+ refereed papers, a popular textbook on Spatial Databases (Prentice Hall, 2003) and an authoritative Encyclopedia of GIS (Springer, 2008). Shashi is serving as a member of the Computing Community Consortium Council (2012-15), a co-Editor-in-Chief of Geo-Informatica : An International Journal on Advances in Computer Sciences for GIS (Springer), a series editor for the Springer-Briefs on GIS, and as a member of the National Research Council (NRC) committee on Geo-targeted Disaster Alerts and Warning (2013). Earlier, he served on multiple NRC committees including Future Workforce for Geospatial Intelligence (2011), Mapping Sciences (2004-2009) and Priorities for GEOINT Research (2004-2005). He also served as a general or program co-chair for the Intl. Conference on Geographic Information Science (2012), the Intl. Symposium on Spatial and Temporal Databases (2011) and ACM Intl. Conf. on Geographic Information Systems (1996). He also served on the Board of Directors of University Consortium on GIS (2003-4), as well as the editorial boards of IEEE Transactions on Knowledge and Data Eng. and IEEE-CS Computer Sc. & Eng. Practice Board. In early 1990s, Shashi's research developed core technologies behind in-vehicle navigation devices as well as web-based routing services, which revolutionized outdoor navigation in urban environment in the last decade. His recent research results played a critical role in evacuation route planning for homeland security and received multiple recognitions including the CTS Partnership Award for significant impact on transportation. He pioneered the research area of spatial data mining via pattern families (e.g. collocation, mixed-drove co-occurrence, cascade), keynote speeches, survey papers and workshop organization. Shashi received a Ph.D. degree in Computer Science from the University of California (Berkeley, CA).

Peter Snyder – University of Minnesota pksnyder@umn.edu Peter Snyder is an atmospheric scientist studying an array of research problems related to atmospheric physics, land-atmosphere interactions, hydrometeorology, climate change, and the biosphere. His research areas span the Arctic, the tropics, and North America. Particular research problems include the role of the Great

Plains Low Level Jet on moisture transport and precipitation events in the upper Midwest, the role of climate change on the frequency of extreme events, the influence of Arctic warming on the boreal forest and feedback mechanisms, monitoring and mitigation of urban heat islands around the world, and the

39

climate response to boreal afforestation for carbon sequestration. He uses observations as well as global climate models, regional weather models, and land surface models to investigate these problems.

Michael Steinbach – University of Minnesota steinbac@cs.umn.edu Michael Steinbach earned his B.S. degree in Mathematics, a M.S. degree in Statistics, and M.S. and Ph.D. degrees in Computer Science from the University of Minnesota. He is currently a research associate in the Department of Computer Science and Engineering at the University of Minnesota, Twin Cities. Previously, he held a variety

of software engineering, analysis, and design positions in industry at Silicon Biology, Racotek, and NCR. His research interests are in the area of data mining, bioinformatics, and statistics. He has authored over 30 research articles, and is a co-author of the data mining textbook, Introduction to Data Mining, published by Addison-Wesley. He is a member of the IEEE Computer Society and the ACM.

40

41

Poster Session

Monday, June 30, 2014 Listed in Alphabetical order of Last Name of Presenter

42

Presenter : Norbert Agana, North Carolina Agricultural & Technical University Title: Analysis of extremes of precipitation events Contributors: Agana, Gorji-Sefidmazgi, Homaifar (NCAT) Considerable attention has been devoted to the statistical modelling of climatic extreme events. Most notably is the modeling of stationary extreme events of which the models assume stationarity of the observed process. Extremes are rare events and lack of long-term datasets is always a very challenging issue in modeling extremes. To accurately assess the risk of extreme events such as hurricanes, drought, earthquakes and floods, there is the need to develop accurate extreme value statistical models that will take into account factors that might have influence on the occurrence of these events. This is because extreme events often display non-stationarity, varying continuously with a number of covariates or in most cases with time. Hence, incorporating these covariates in the models will lead to more reliable estimates of the extreme events distributions. In this work, we derive a statistical model to incorporate the effects of covariates. Specifically, we analyze as an example the extreme precipitation events in North Carolina, where the potential impacts of El-Nino Southern Oscillation (ENSO) and North Atlantic Oscillation (NAO) phenomena on extreme precipitation events are incorporated into the model. For this research, both the location parameter and the scale parameter of the Generalized Extreme Value (GEV) distribution will be modelled as polynomial functions of the covariates. Maximum Likelihood Estimation method is employed to estimate the parameters of the GEV distribution. Likelihood ratio test will be performed to assess the impact of incorporating these covariates into the model.

Presenter : Saurabh Agrawal, University of Minnesota Title: Finding Climate Teleconnections and Evaluating Climate Models Contributors: Kawale (Adobe), Agrawal, Ormsby, Liess, Steinbach, Steinhaeuser, Chatterjee, Snyder, V Kumar (UMN), Ganguly (NEU), Samatova, Semazzi (NCSU) Pressure dipoles are important long distance climate phenomena (teleconnection) characterized by pressure anomalies of opposite polarity appearing at two different locations at the same time. Such dipoles have proven important for understanding and explaining the variability in climate in many regions of the world, e.g., the El Nino climate phenomenon is known to be responsible for precipitation and temperature anomalies at large parts of the globe. Besides, they are typically used as standard benchmarks by climate scientists for evaluating climate models. We present a novel graph based approach called shared reciprocal nearest neighbors that considers only reciprocal positive and negative edges in the shared nearest neighbor graph to find dipoles. The proposed approach can generate a single snapshot picture of all the dipole interconnections on the globe in a given dataset and thus makes it possible to study changes in dipole interactions and movements. Also, the data driven climate indices generated from our algorithm perform better than CPC indices in terms of capturing impact on global temperature and precipitation. We introduce our ongoing work on evaluating climate models using impact maps of dipoles. Impact

43

maps capture the impact relationships of a dipole all over the globe. The proposed evaluating scheme gives a similarity score to the impact map obtained from a climate model based on its similarity to the one obtained from the observations. The basic similarity measures like Root Mean Square Error suggest consistent improvement from CMIP3 to CMIP5 for impact of ENSO on variables like sea surface temperature and precipitation. The future work focuses on advanced similarity measures like Earth Mover's Distance and its variations which can handle inherent spatial variability present in the data more appropriately.

Presenter : Michael Angus, North Carolina State University Title: The Role of the Greater Horn of Africa in Modulating Atlantic Hurricane Variability Contributors: Angus, Bello, Harlalka, Smith, Semazzi, Samaotva (NCSU), Waniha (NCSU, Tanzania Meteorological Agency), Kumar (UMN) Atlantic hurricane variability has been primarily attributed to the influence of climatic factors operating over the Atlantic and Pacific Oceans. Here, we attempt to widen that scope by identifying global relative humidity “hotspots”, regions highly correlated with the Atlantic hurricane season. We then assess these hotspots in terms of the variance of Atlantic hurricane variability that they explain. We find that a hotspot over the east Africa region, referred to as the Greater Horn of Africa Climate Index (GHACI), has higher unique information content about hurricane activity than both the other hotspots we have discovered, and traditional climate indices. This new index shows a strong negative correlation with Atlantic hurricane seasonal count. The robustness of this correlation is examined by comparing years with high hurricane count (≥8 hurricanes per year) with years with low hurricane count (≤ 4 hurricanes per year). Fundamental differences between the two sets of years are observed in the regional atmospheric circulation over the Greater Horn of Africa, identified in reanalysis data. During June, in years of high (low) hurricane activity, the coastal sea surface temperatures are generally cooler (warmer), the wind direction is northward (southward) along the coast and an associated area of higher (lower) pressure forms. This gives us confidence that the relationship between the GHACI and Atlantic hurricane activity is climatologically driven, and not a chance statistical finding. The results of this study have important implications for the potential improvement of empirical Atlantic hurricane forecast models, and the interpretation of climate change projections of Atlantic hurricane activity, which currently do not include critical conditions over the GHA region.

Presenter : Gonzalo Bello, North Carolina State University Title: Response-Associated Clustering of Spatiotemporal Data: An Application to Climate Index Discovery for Seasonal Rainfall in the Greater Horn of Africa Contributors: Bello, Angus, Pedemane, Harlalka, Semazzi, Samatova (NCSU), Waniha (NCSU, Tanzania Meteorological Agency), Kumar (UMN) Identifying clusters with physical or functional meaning associated with a response of interest in a complex system, such as rainfall variability or hurricane activity in the case of the climate system, is an important task in many scientific domains. In this work, we propose a novel methodology for the discovery of response-associated clusters in multivariate spatiotemporal data. Our proposed methodology aims to find clusters with predictive power for a given response by incorporating information of the response during the clustering process. We apply our methodology to the task of climate index discovery, a key application of spatiotemporal data mining in the climate science domain. Climate indices provide a generalized overview of persistent, large-scale patterns in the global climate

44

system, and are commonly used to analyze the influence of these patterns on regional weather. Specifically, we aim to discover climate indices with predictive power for seasonal rainfall variability in the Greater Horn of Africa. Experiments were performed using October-December seasonal rainfall at synoptic stations located in the Northeastern Highlands of Tanzania as the response of interest. Response-associated clusters persistently identified include regions known to influence seasonal rainfall variability in the Greater Horn of Africa, such as the tropical Pacific Ocean (ENSO) and the Indian Ocean (IOD). This suggests that our methodology is able to capture the underlying physical processes that drive seasonal rainfall in this region. We use these response-associated clusters as predictors in classification and regression models for the prediction of the October-December rainfall season. Our results show improvements on the accuracy of the prediction with respect to state-of-the-art methods for data-driven climate index discovery and official forecasts issued by Tanzania Meteorological Agency.

Presenter : Shrutilipi Bhattacharjee, Indian Institute of Technology Kharagpur Title: Prediction of Meteorological Parameters based on Semantic Analysis of Land-Atmospheric Interaction Model Contributors: Bhattacharjee, Ghosh (IIT-Kharagpur) Meteorological parameters, related to earth surface, are considered to be important factors for modelling climate dynamics. Examples of these parameters include land surface temperature, vegetation index, built-up index, etc. According to NASA Earth Observatory, the land surface temperature is an influential factor to determine climate patterns. Our research work focuses on the prediction and forecasting of these meteorological parameters/attributes. We have proposed a new Kriging based spatial interpolation technique, namely Semantic Kriging, considering the semantic relationships among geospatial objects for better prediction. The semantic relationships between different LULC (land use/land cover) objects play significant role in the prediction of meteorological parameters. Though Kriging is reported as the most popular predictor, it fails to use this semantic knowledge of LULC for modelling land atmospheric interaction. In Semantic Kriging, the autocorrelation calculation considers both the locations of the sample points and their representative land-covers. The ontology hierarchy of spatial land-cover captures the semantic knowledge and the analysis of this spatial ontology along with the spatial correlation, facilitates better prediction of meteorological attributes. Our research focuses on the following aspects:

• Identification of semantic knowledge of the terrain, their quantification using ontology • Extending the traditional two dimensional spatial autocorrelation into three dimensional spatio-semantic autocorrelation, with this semantic knowledge • Enhancing the prediction accuracy with enhanced semantic land-atmospheric interaction model

This framework is now being extended for the prediction of urban heat islands. Further, identification of different influential meteorological variables and modelling their spatiotemporal behaviors with semantic knowledge of the terrain, are being investigated.

Presenter: Xi Chen, University of Minnesota Title: Unsupervised Framework to Monitor Lake Dynamics Contributors: Chen, Khandelwal, Shi, Anderson, Blank, Boriah and Kumar (UMN)

45

Approximately two-thirds of the Earth's surface is covered by water in the form of streams, rivers, lakes, wetlands and oceans. Surface water is not only critical to the ecosystems as a key component of the global hydrologic cycle but also is extensively used in our daily life (e.g., electricity generation, drinking water, agriculture, transportation, and industrial purposes). Traditional monitoring/event reporting systems for surface water primarily rely on human observations or in situ sensors. Due to the massive area that water covers, a comprehensive effort that maps changes in global surface water is lacking. This limits our understanding of the hydrologic cycle, hinders water resource management and also compounds risks. Remote sensing instruments collect large quantities of data every year and provide an opportunity to monitor global surface water automatically. In this poster, we propose an unsupervised algorithm for monitoring the surface area of lakes that overcomes many challenges including: (1) Pixels containing the shoreline are usually a mixture of water and other surface classes (such as vegetation or soil). (2) Hydrologic cycles are often irregular and do not follow a regular annual cycle. (3) There is considerable data heterogeneity on the global scale. We also develop an innovative methodology for quantitative evaluation of algorithm performance by using a combination of independent validation data and physics-guided labelling, and finally show that the proposed method is better for monitoring the evolution of water surface than the state-of-art method in three different lakes.

Presenter : Poulomi Ganguli, Northeastern University Title: Severity-duration-frequency curves suggest decreasing severity and relatively stationary spatial variability of U.S. meteorological droughts Contributors: Ganguli, Ganguly (NEU) Drought characterization remains challenging, especially because attributes such as severity, duration, spatial coverage, and frequency exhibit complex interdependence. Meteorological droughts relate to supply–driven water stress relative to baseline conditions, and may be harbingers of longer–lasting agricultural or hydrological droughts. The Standardized Precipitation Index (SPI) has been used to characterize meteorological droughts, while copulas developed in probability theory have been used to track the dependence among drought attributes. However, spatiotemporal patterns of meteorological droughts over the conterminous United States (CONUS), and the dependence among drought attributes, require further study. This paper exploits joint dependence with copulas between SPI-based drought severity and duration across a range of frequencies to develop regional severity-duration-frequency (SDF) curves with observed precipitation data. These SDF curves generate statistically–grounded insights for water resources planning and the design of water–storage and distribution infrastructures over climatologically-homogeneous regions. Analyses of changing SDF relations over two 30-year time–windows (1950–1979 versus 1980–2009) in the CONUS point to decreasing trends in drought severity and show no evidence of statistically significant trends in spatiotemporal variability across all return periods. While extreme droughts have shown a tendency to grow more severe in drought prone regions of the CONUS, the pattern is not homogeneous. The findings are specific to meteorological droughts and suggest the need for continuously monitoring extremes of such droughts especially in drought-prone regions.

Presenter: André Ricardo Gonçalves, University of Minnesota Title: Global climate models combination using Multitask Sparse Structure Learning Contributors: Das, Chatterjee, Sivakumar, Banerjee (UMN)

46

Global Climate Models (GCMs) are mathematical models designed to represent physical climate processes and are run as computer simulations to analyze and predict climate variables such as temperature, pressure, and precipitation. However, these models have a high variance in their projection of future climate variables at local scales, which produce uncertainty in analysis based on these predictions. Combining multiple GCMs outputs can significantly reduce the variability in projections. Thus, climate scientists have been interested in ways to combine properly GCM outputs, reducing forecast variability without loss of physical significance. We propose a Multitask Sparse Structure Learning (MSSL) algorithm for GCMs outputs combination. MSSL performs coefficients learning for all locations simultaneously and by doing that exploit information from related locations to provide more accurate predictions on individual locations. Further, it encourages similar locations, including neighbors, to have similar model coefficients, which for some climate variables is more physically plausible. The main difference between MSSL and the previous multitask learning methods is that MSSL does not assume any dependency structure among locations, but learns it from the data instead. Experiments on multi-model combination are performed in land surface temperature prediction in South America, which presents a variety of climate, ranging from hot and wet Amazon rainforest to very cold regions in southernmost parts, passing through arid regions with one of the driest deserts in the world. The obtained results show that using MSSL for GCMs multi-model combination not only provides more accurate prediction than baseline approaches, including the average as well as the best GCM, but also captures informative correlated behaviors within locations.

Presenter : Doel González, North Carolina State University Title: On the data-driven inference of modulatory networks in climate science: an application to West African rainfall Contributors: González, Angus, Tetteh, Bello, Padmanabhan, Pendse, Srinivas, Yu, Semazzi, Samatova (NCSU), Kumar (UMN) Decades of hypothesis-driven and/or first-principles research have been applied towards the discovery and explanation of the mechanisms that drive climate phenomena, such as western African Sahel summer rainfall variability. Although connections between various climate factors have been theorized, not all of the key relationships are fully understood. We propose a data-driven approach to identify candidate players in this climate system, which can help explain underlying mechanisms and/or even suggest new relationships, to facilitate building a more comprehensive and predictive model of the modulatory relationships influencing a climate phenomenon of interest. We applied coupled heterogeneous association rule mining (CHARM), Lasso multivariate regression, and Dynamic Bayesian networks to find relationships within a complex system, and explored means with which to obtain a consensus result from the application of such varied methodologies. Using this fusion of approaches, we identified relationships among climate factors that modulate Sahel rainfall. These relationships fall into two categories; well-known associations from prior climate knowledge, such as the relationship with the El Niño Southern Oscillation (ENSO) and putative links, such as the North Atlantic Oscillation (NAO), that invite further research.

Presenter : Mohammad Gorji-Sefidmazgi, North Carolina Agricultural & Technical University Title: Change detection in temperature trend of US during 1900-2012

47

Contributors: Gorji-Sefidmazgi, Homaifar (NCAT), Liess (UMN) In order to analyze low frequency variability of climate, analyzing the trend in climate time series is useful. It is well known that the simple linear trend is not suitable for modeling the complex climate time series. When piecewise linear trends are used as an alternative, the problem of trend detection is converted to finding change points in the linear trend. In this work, the method of bounded-variation segmentation is applied for change detection in temperature trends over the US from 1184 stations during 1900-2012. The method finds multiple linear trends and locates times of significant changes. Based on the reasonable assumption that the number of change points is bounded, the original non-convex optimization problem of change detection can be converted to convex linear programming and ordinary least squares which can be solved iteratively. The Genetic Algorithm is used for improving the performance of optimization. Bayesian Information Criterion is used to find the optimum number of change points. This method analyzes multidimensional time series, all of the stations’ data are processed in parallel and thus the results show only changes due to natural reasons, not those caused by sensor relocation. An important attribute of the proposed method is that it is independent of any restrictive Gaussian or Markovian assumptions. The result shows a spatio-temporal pattern of climate change over US and also indicate that there is a statistically significant change in 1958 for minimum, maximum and average temperature of US.

Presenter: William Hendrix, Northwestern University Title: Accurate Forecasting of Extreme Events in Spatio-temporal Systems Contributors: Chen, Xie, Cheng, Zhang, Agrawal, Liao, Choudhary (NWU), Samatova, (NCSU) The accurate forecasting of extreme events, such as Atlantic hurricanes, remains a difficult yet critical problem for sustainability. While physics-based models have their own merits, they have difficulty in producing reliable predictions for certain variables of interest, notably precipitation. In this poster, we describe techniques we have developed for forecasting these extreme events in under-determined spatio-temporal contexts and for detecting and correcting forecast errors. By combining these approaches, we can achieve 73-83% accuracy in forecasting extreme events such as tropical cyclones in the Northern Hemisphere, hurricanes in the North Atlantic, and North African rainfall.

Presenter: Zhe Jiang, University of Minnesota Title: Learning focal-test-based spatial decision tree. Contributors: Jiang, Shekhar, Zhou, Knight, Corcoran (UMN) Given learning samples from a raster dataset, the spatial decision tree learning problem aims to learn a decision tree classifier that minimizes classification errors as well as salt-and-pepper noise. The problem has important societal applications such as land cover classification for understanding climate change, natural resource management and disaster management. However, the problem is challenging due to spatial autocorrelation, i.e., class labels of nearby samples are correlated instead of independently identically distributed. Related work relies on local tests (i.e., testing feature information of a location) and cannot adequately model the spatial autocorrelation effect, resulting in salt-and-pepper noise. In contrast, we recently proposed a focal-test-based spatial decision tree (FTSDT), in which the tree traversal direction of a sample is based on both local and focal (neighborhood) information. Preliminary results on real world datasets showed that FTSDT reduces classification errors and salt-and-pepper noise. We also addressed the computational challenges. We first identify the computational bottleneck:

48

focal function computation. We then design a refined algorithm that reuses focal values across candidate thresholds. Theoretical analysis shows that the refined algorithm is correct and more scalable. Experiment results also confirm that our refined algorithm significantly reduces computational time.

Presenter: Anuj Karpatne, Univeristy of Minnesota Title: Global lake mapping in the presence of heterogeneous land and water types Contributors: Khandelwal, Chen, Mithal, Kumar (UMN) Mapping the extent of lakes at a global scale is critical for enhancing our understanding about the global hydrologic cycle, as well as for assessing the impact of climate change and human activities on the global water ecosystem. Classification approaches using remote sensing data can be utilized to achieve this task, where each pixel needs to be classified as either water or land. However, developing classification approaches for global lake mapping is challenging in the presence of a wide variety of land and water types, leading to a rich heterogeneity in the two classes. Furthermore, obtaining labeled information about land and water classes is expensive and is often carried out infrequently or at localized regions of study. This results in limited availability of labels for supervised learning, which are often under-representative of the distribution of global data. Remote sensing datasets are also subject to noise and outliers, such as cloud and aerosol occlusions, which can significantly impact classification performance. We present a classification framework that is able to capture the heterogeneity in the training data using multi-modal modeling of the two classes. By learning an ensemble of classifiers, each representative of a labeled mode in the training data, we are able to provide high precision labels on a majority of land and water pixels. We are also able to leverage the data heterogeneity for identifying uncertain test instances, which either appear at the boundaries of land or water, or belong to an unknown class, such as clouds and their shadows.

Presenter: Ankush Khandelwal, University of Minnesota Title: K-means for non i.i.d. data Contributors: Khandelwal & Kumar (UMN) Traditional K-means algorithm aims at reducing overall sum of squared error (SSE) in feature space to achieve reliable clustering but it neglects relationship between different data instances i.e. it assumes the data to be i.i.d. for various applications where data is non i.i.d. Hence cluster assignment of a data instance should depend not only its feature values but also on cluster assignment of related data instances. Recently, a variation of K-means, K-means - - (K-means Minus Minus) has been proposed which simultaneously reduces SSE and detects outliers. Here, we propose a new approach that builds on the concept of K-Means - - and aims to create clusters which have both lower SSE and high conformity to constraints due to non i.i.d. nature of data. Specifically, the proposed approach forces K-means to converge to a local optimum which have high level of conformity with the relationship between data instances. We test the performance of the proposed approach on remote sensing satellite data which shows non i.i.d. nature due to spatial autocorrelation. The desired clusters should not only lead to low SSE in feature space but should also be spatially smooth in geographical space. Through various experiments we show that the proposed approach has been successful in obtained high quality spatially smooth clusters.

49

Presenter: Devashish Kumar, Northeastern University Title: Uncertainty characterization and delineation of nonstationarity in intensity-duration-frequency curves of precipitation relevant for infrastructural design Contributors: Kumar & Ganguly (NEU) Understanding the degree of stationarity in intensity-duration-frequency (IDF) curves of precipitation extremes relates to the design and maintenance of hydraulic infrastructures and hence to flood resilience. Statistical analysis of observations and climate models forced with projected emissions scenarios point to more frequent and intense precipitation extremes at global and century scales. A statistical delineation of changes in IDF curves at scales relevant for infrastructures has proved elusive, owing to significant spatial variability, data and model quality, and estimation uncertainties. The contribution of natural climate variability, in addition to thresholds and intermittences, has not received much attention in the literature. However, this relatively irreducible component of the uncertainty may dominate for decadal planning horizons, especially at the spatial scales relevant for hydraulic infrastructures. A characterization of the irreducible uncertainties, which in turn yields an envelope of plausible scenarios, may need to be a critical pathway for resilient engineering. Statistical and information theoretic methods are developed or adapted from the literature to characterize predictability and nonstationarity of IDF curves. Risk management methods are examined to develop best practices for translating the characterization of irreducible uncertainty to guidelines for infrastructural decisions and resource allocations.

Presenter: Chaunté Lacewell, North Carolina Agricultural & Technical University Title: SCOT: Selective Clustering based Oversampling Technique Contributors: Lacewell & Homaifar (NCAT) Data imbalance is an essential source of low performance in learning of rare events because most classifiers assume to have balanced data. There is a high demand in both academia and industry to accurately identify rare events, which are usually more important than events that occur more frequently. In most real-world applications, the observed data is highly imbalanced and classifiers are biased to the larger class. For example, identification of cloud clusters which will develop into a tropical cyclone would be considered an imbalanced data set because the number of non-developing cloud clusters is expected to outnumber the number of developing cloud clusters. In this climatology application, classifying a developing cloud cluster accurately is of great importance. In this case, misclassifying a developing cloud cluster is more costly than misclassifying a non-developing cloud cluster. Therefore, we proposed the Selective Clustering based Oversampling Technique (SCOT) which addresses data imbalance in a selective manner and can be used with a standard classifier. This method identifies hard-to-learn minority samples that are located near the decision boundary and uses clustering of the hard-to-learn minority samples to identify the most relevant minority samples to assist in generating synthetic minority samples. SCOT ensures that generated synthetic samples are inside the minority class region and are not in regions of the majority class which prevents overlapping between classes and misclassification of samples. SCOT is evaluated on multiple real world time series and

50

multivariate data sets and the simulation results verify that the proposed technique outperforms the commonly used techniques.

Presenter: Mengqian Lu, Columbia University Title: Predicting 30 Days Extreme Precipitation using a Global SST-SLP Correlation Network Contributors: Lu, Lall (Columbia), Kawale (Adobe), Liess, Kumar (UMN) Correlation networks identified from financial, genomic, ecological, epidemiological, social and climate data are being used to provide useful topological insights into the structure of high dimensional data. Strong convection over the oceans and the atmospheric moisture transport and flow convergence indicated by atmospheric pressure fields may determine where and when extreme precipitation occurs. Here, the spatiotemporal relationship between climate and extreme global precipitation is explored using a graph-based approach that uses the concept of reciprocity to generate cluster pairs of locations with similar spatiotemporal patterns at any time lag. A global time-lagged relationship between pentad sea surface temperatures (SST) anomalies and pentad sea level pressure (SLP) anomalies is investigated to understand the linkages and influence of the slowly changing oceanic boundary conditions on the development of the global atmospheric circulation. We explore the use of this correlation network to predict extreme precipitation globally over the next 30 days, using a Principal Component logistic regression on the strong global dipoles found between SST and SLP. Unprecedented success of the predictive skill under cross validation for 30 days precipitation higher than the 90th percentile is indicated for selected global regions for each wet season considered.

Presenter: Varun Mithal, University of Minnesota Title: Event detection in Spatio-temporal data: Application in land surface monitoring Contributors: Khandelwal, Nayak, Boriah, Kumar (UMN) The physical surface of the earth consists of material such as trees, grass, crops, water bodies or impervious land. Based on its constituent physical material, land is often classified into discrete categories known as land cover. A variety of natural and anthropogenic processes such as forest fires, insect infestation, urban growth, agricultural expansion, shrinking water bodies and deforestation have been changing the land cover. However, due to an increase in the human population and the standard of living, changes in land cover are now occurring at unprecedented rate and have become a threat to the sustainability of existing ecosystems. There is an urgent need to develop an ability to monitor these changes so that policy makers can take the necessary decisions to preserve the biodiversity of the planet and mitigate climate change and its impact. In this poster we will present our recent developments in spatio-temporal data mining approaches to address challenges in land surface monitoring from remote sensing data. These novel approaches have enabled accurate and timely monitoring of forest fires globally.

Presenter: Stephen Ranshous (NCSU) Title: Community Detection with Knowledge Priors in Climate Networks

51

Contributors: Harenberg, Seay, Ranshous, Bello, Boyuka, Harlalka, Lakshminarasimhan, O’Brien, Padmanabhan, Patowary, Schendel, Chirkova, Samatova(NCSU), Hendrix, Choudhary(NWU), Kumar (UMN). Community detection is an important problem in graph mining and knowledge discovery. However, the large scale of many real-world graphs, including climate networks, and the large output space generated by many traditional methods for community detection, often make the application of these methods on real-world graphs impractical. For this reason, we introduce the notion of knowledge-biased community detection, a query-driven reformulation of the community detection problem that aims to reduce the output space, computation time, and peak memory usage. Mining for communities with query nodes as knowledge priors allows for filtering out irrelevant information and enriching end-users knowledge associated with the problem of interest. Additionally, to further address the computational challenges associated with community detection in large-scale real-world graphs, we explored out-of-core and indexing techniques for improved memory efficiency, and proposed a partial computation strategy for improved query response time. In this work, we apply knowledge-biased community detection to discover associations between global climate patterns and climate phenomena in a climate network of sea surface temperature (SST). Specifically, we aim to discover associations with Atlantic hurricane activity by using the vertices located in the Atlantic Ocean's Main Development Region (MDR) as the query set. Knowledge-biased communities discovered contain nodes that belong to climate indices known to be associated with Atlantic hurricane variability, such as ENSO in the tropical Pacific Ocean and AMO, AMM, and NAO in the North Atlantic Ocean.

Presenter: Scott Sellars, Univeristy of California, Irvine Title: Machine Learning in Hydrometeorology and Hydroclimate at the University of California, Irvine – The Center for Hydrometeorology and Remote Sensing (CHRS) Contributors: Sellars, Tao, Karbalee (UCI) Learning from earth science data continues to make great strides using computational science algorithms. The Center for Hydrometeorology and Remote Sensing (CHRS) – University of California, Irvine (UCI), building on over a decade of research for estimating precipitation from satellite data Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) (Hsu et al., 1998; Sorooshian et al., 2000), has sought to push the boundaries of hyperspectral and hyper dimensional learning for estimating precipitation and understanding the variability in hydrometeorology and hydroclimate. This poster will report on the research directions supported by the National Science Foundation (NSF) Cyber-Enabled Sustainability Science and Engineering (CyberSEES) program and NASA Earth and Space Science Fellowship (NESSF) program. The first research direction, supported by CyberSEES program, is focused on harnessing state of the art advancements in Machine Learning for precipitation estimation. One approach uses “stacked auto-encoders (SAE)”, a typical deep neural network algorithm, support vector machines as well as typical ensemble methods, applied to data of PERSIANN-Cloud Classification System (PERSIANN-CCS) to improve its accuracy of the precipitation estimation. SAE algorithms are used to extract features from massive precipitation image patches and construct an effective bias removal model. The other approach uses a “self-organizing feature map (SOM)” to classify and map Geostationary Earth Orbiting (GEO) satellite data to rainfall rates. The Low Earth Orbiting (LEO) satellite data are used to improve GEO-based estimation by applying a probability matching method (PMM).

52

Secondly, the poster will illustrate results from our research supported by the NESSF, the investigation of “constrained” learning models applied to the PERSIANN-CONNected precipitation object dataset ((PERSIANN-CONNECT, Sellars et al., 2013) for feature selection and prediction of monthly hydrometeorological conditions (e.g. wet, normal or dry months) in the Western United States. These conditions are defined as deviations from the climatological mean precipitation conditions for each month. Constrained linear models are often used to reduce the complexity of a model by “shrinking” the number of parameters in accordance with a user specified regularization term in the optimization scheme. Our results suggest that predictability of monthly hydroclimate conditions using this approach can provide comparable results to using climatology. The poster will give an overview of the challenges (linear functional form) and advantages (use of the key characteristics and features) of the application of this approach in the context of forecasting hydrometeorology mean monthly conditions.

Presenter: Deepti Singh, Stanford University Title: Extreme rainfall events in the South Asian Summer Monsoon season Contributors: Tsiang, Rajaratnam, Diffenbaugh (Stanford) The South Asian Summer Monsoon directly affects the lives of over 1/6th of the world’s population. There is substantial variability within the monsoon season, including fluctuations between periods of heavy rainfall (“wet spells”) and low rainfall (“dry spells”). These fluctuations can cause extreme wet and dry regional conditions that adversely impact agricultural yields, water resources, infrastructure, and human systems. Through a comprehensive statistical analysis of precipitation observations (1951-2011), we show that statistically significant decreases in peak-season precipitation over the core-monsoon region have co-occurred with statistically significant increases in daily-scale precipitation variability. Further, we find statistically significant increases in the frequency of dry spells and intensity of wet spells, and statistically significant decreases in the intensity of dry spells. These changes in extreme wet and dry spell characteristics are supported by increases in convective available potential energy (CAPE) and low-level moisture convergence, along with changes to the large-scale circulation aloft in the atmosphere. The observed changes in wet and dry extremes during the monsoon season are relevant for managing climate-related risks, with particular relevance for water resources, agriculture, disaster preparedness, and infrastructure planning. Presenter: Vidyashankar Sivakumar, University of Minnesota Title: An Exploratory Quantile Based Analysis of Indian Summer Monsoon Rainfall (ISMR) Contributors: Sivakumar, Chatterjee, Goncalves, Banerjee (UMN) Indian monsoon has a direct impact on the economies of south Asian countries and the lives of a significant fraction of the total world population. Prediction of monsoon rainfall in the Indian subcontinent has been one of the toughest problems in the climate science domain. The regional precipitation in India can have large year-to-year fluctuations, the interannual variability of total India rainfall is about 10% of the mean rainfall. In our work here, we carry out a data driven exploratory analysis of Indian monsoon precipitation. We use the framework of sparse quantile regression for the analysis. We consider the precipitation in a particular region as the predictand. To analyze the effect of different climate indices on the different quantiles of precipitation data we fit quantile regression curves with the climate indices as covariates.

53

Previous work has shown that the precipitation in a particular region is dependent only on a few of the covariates and hence we introduce sparsity inducing regularizers. We also analyze spatial relationships using the concept of association analysis. In particular we study the co-occurence patterns of high and low precipitation events in the different regions. In particular our work differs from other previous work in that we consider the quantiles of the precipitation data intead of the mean. Traditional analysis has used linear least squares regression which assumes that the conditional distribution is Gaussian. Also the estimator gives the mean of the predictand given the predictors. In quantile regression we fit curves for each quantile thus giving an idea about the true conditional distribution and leads to a richer analysis.

Presenter: Daniel Swain, Stanford University Title: The Extraordinary California Drought of 2013-2014: Character, Context, and the Role of Climate Change Contributors: Swain, Tsiang, Haugen, Singh, Charland, Rajaratnam, Diffenbaugh (Stanford) Nearly the entire state of California experienced extremely dry conditions during calendar year 2013. State-wide 12-month accumulated precipitation was less than 34% of average, leading to a wide range of adverse impacts upon both human and natural systems. The California drought occurred in tandem with a highly persistent region of positive geopotential height (GPH) anomalies over the northeastern Pacific Ocean(associated with the so-called “Ridiculously Resilient Ridge”), which displaced the jet stream well to the north of its climatological mean position and prevented winter storms from reaching California during the canonical wet season. Remarkably, this anomalous high-amplitude ridging recurred during at least a portion of two consecutive winter seasons in California, accelerating the onset of exceptionally severe drought conditions. We find that the extreme GPH values during 2013 are observationally unprecedented over a vast region covering most of the North Pacific Ocean and adjacent land areas. We also report that California experienced both its driest calendar year and driest consecutive 12-month period in at least 119 years. Finally, we assess the role of climate change in affecting the likelihood of occurrence of the large-scale conditions linked to the California drought using data from an ensemble of climate models run as part of the CMIP5 experiment. While a 2013-magnitude event appears to be rare in both preindustrial and present-day climates, we find that the GPH values associated with the 2013 event are at least three times more likely to occur in the present climate.

Presenter: Danielle Touma, Stanford University Title: A Multi-Model and Multi-index Evaluation of Drought Characteristics in the 21st Century Contributors: Touma, (Stanford & ORNL), Ashfaq, Kao (ORNL), Nayak (IIHR, UofIA), Diffenbaugh (Stanford) Drought is one of the most costly and least understood disasters that impact the agricultural, health, social and political status of a region. Moreover, an increase in global greenhouse gas forcing is expected to change the characteristics and impacts of drought in the 21st century. We use four drought indices, the Standardized Precipitation Index (SPI), the Standardized Runoff Index (SRI), the Standardized Precipitation-Evapotranspiration Index (SPEI) and the Supply-Demand Drought Index (SDDI), and 15 Global Climate Models (GCMs) data from the fifth phase of the Coupled Model Intercomparison Project

54

(CMIP5) to assess the frequency, duration, number of events and time of emergence of drought in two future sub-periods (2010-2054 and 2055-2099) under the Representative Concentration Pathway (RCP) 8.5 as compared to the baseline period (1961-2005). We find a general increase in the frequency, occurrences and duration of drought over the globe in the 21st century, with larger changes occurring in the later future sub-period and when using the SPEI and SDDI. However, there are large uncertainties among the GCMs in the magnitude of these increases and, in some regions, the sign of the change of the drought characteristic. The Amazon, Northern and western Africa and the Mediterranean, which are already water-stressed, show the largest changes in drought characteristics using all indices and the earliest time of emergence in the frequency of drought. These findings strengthen the ongoing dialogue about adaptation to future drought events and can help steer both local and inter-regional water management policies.

Presenter: Erin Towler, National Center for Atmospheric Research Title: Improving future predictions of drought: A new hybrid statistical-dynamical downscaling technique Contributors: Towler, PaiMazumder, Holland (NCAR) There is great interest in how drought may change in the future, especially at the local scale where adaptation decisions are made. As such, the goal of this research is to improve local drought prediction, as measured by the Palmer Drought Severity Index, through use of a new hybrid statistical-dynamical technique that utilizes the respective capabilities of Global Climate Models (GCMs) and Regional Climate Models (RCMs). The hybrid technique uses two predictors: (i) local temperature and (ii) a large-scale index that serves as a proxy for local precipitation. For the former, we use dynamically downscaled temperature from multiple RCMs driven by the same GCM. For the latter, we investigate observed large-scale patterns that are associated with local precipitation; from this we derive a large-scale index that we obtain from the GCM. The approach circumnavigates the uncertainty in regional precipitation projections to better inform local decision-making efforts. The technique is demonstrated for predicting late summer drought in South-central Oklahoma, but the method is general and could be applied to other regions. In short, we demonstrate that a hybrid technique built around tailored indices from large- scale patterns has promise for regional scale downscaling. This technique could potentially be generalized by using data mining, or other data driven approaches, to identify the large-scale predictors. Presenter: Dawei Wang, University of Massachusetts Boston Title: Local Learning on High Dimension, Imbalanced, and Noisy Data: A Framework for Long-Lead Extreme Precipitation Clusters Forecasting Contributors: Wang, Ding, Mu (UMB), Small, Islam (Tufts) Extreme Flood is usually a consequence of a sequence of precipitation events occurring over from several days to several weeks. Certain atmospheric regimes (e.g., blocking) can lead to sequence of precipitation events. However, the task of long-term (5-15 days) forecasting of precipitation clusters will suffer from overwhelming number of relevant features and high imbalanced, multimodal, and noisy sample sets. Existing atmospheric models which rely on nonlinear deterministic systems cannot deal with such a huge feature space to provide accurate long-range predictability of weather. In this work, we develop a data mining framework for long-lead extreme precipitation cluster forecasting through the identification of atmospheric regime precursors. We synthesize a representative feature set that describes the atmosphere motion, and then we design a novel Bi-Class Streaming Feature Selection (BCSFS) algorithm for feature selection on imbalanced data. After the dimension reduction step, we

55

apply the Local Discriminative Distance Metrics Ensemble Learning (LDDM) algorithm, which learns distance metrics according to different training samples and predicts a test sample by classifiers ensemble, to learn from the multimodal and noisy sample set and make prediction through classification. An extensive empirical study is conducted on historical precipitation and associated flood data collected in the State of Iowa.

Presenter: Marcia Zilli, University of California Santa Barbara Title: Frequency Analysis of Extreme Events based on Precipitation Station Data over Southeastern Brazil Contributors: Zilli, Carvalho (UCSB) The southeastern coast (SE) of Brazil is among the most densely populated areas of eastern South America with population largely concentrated in urban centers. Due to complex terrain and chaotic urbanization, this region is subject to a variety of natural disasters that frequently occur during the austral summer (September to March), when the South American Monsoon System (SAMS) and the South Atlantic Convergence zone (SACZ) are in their most active period. Previous studies showed increasing trends in daily precipitation and consequently in extreme events in this area; some observational studies have already demonstrated a positive trend in these events in particular locations. Nevertheless, these analyses either focus on one single station or investigate relatively short periods. This study further investigates interannual to multiannual variations and changes in the frequency and intensity of daily extreme precipitation events using long time series (70 years of data with less than 5% missing) from a set of rain stations located between 20°S and 25°S, along Brazilian coast. The period of analysis spans roughly from 1935 to 2012, with some small variations from station to station. The analysis of the frequency of extreme events is based on extreme precipitation indices and respective trends tested using a non-parametric Mann-Kendall test. At least 25% of the stations exhibited a positive and significant trend in extreme daily rainfall rates. The geographic distribution of the observed trends is further investigated using a varied of spatial statistical analysis.

56

57

Attendee Contact Information

Listed in Alphabetical order by Last Name

58

Norbert Agana North Carolina Agricultural & Technical State University acitcenter.ncat.edu/Agana.html Ankit Agrawal Northwestern University ankitag@eecs.northwestern.edu Saurabh Agrawal University of Minnesota Michael Angus North Carolina State University mpangus@ncsu.edu Ghassem R. Asrar National Aeronautics and Space Administration Ghassem.Asrar@pnnl.gov Alison Baker National Center for Atmospheric Research abaker@ucar.edu Arindam Banerjee University of Minnesota banerjee@cs.umn.edu Shrutilipi Bhattacharjee Indian Institute of Technology Kharagpur www.dak.iitkgp.ernet.in/phd/profile.php?roll=11IT91P01 Gonzalo Bello North Carolina State University gabellol@ncsu.edu Shyam Boriah FirstFuel Software sboriah@cs.umn.edu Ruben Buaba

North Carolina Agricultural & Technical State University rbuaba@ncat.edu Lawrence Buja National Center for Atmospheric Research southern@ucar.edu Ansu Chatterjee University of Minnesota chatterjee@stat.umn.edu Xi Chen University of Minnesota chen@cs.umn.edu Timothy DelSole George Mason University tdelsole@gmu.edu Clara Deser National Center for Atmospheric Research cdeser@ucar.edu Noah Diffenbaugh Stanford University diffenbaugh@stanford.edu Wei Ding UMass Boston ding@cs.umb.edu www.cs.umb.edu/~ding Imme Ebert-Uphoff Colorado State University iebert@engr.colostate.edu www.engr.colostate.edu/~iebert/ James Faghmous University of Minnesota jfagh@cs.umn.edu

57

www.cs.umn.edu/~jfagh Ivy Frenger Princeton University ifrenger@princeton.edu www.princeton.edu/aos/people/research_staff/frenger/index.xml Poulomi Ganguli Northeastern University p.ganguli@neu.edu Auroop Ganguly Northeastern University a.ganguly@neu.edu Dimitris Giannakis New York University dimitris@cims.nyu.edu www.cims.nyu.edu/~dimitris André Gonçalves University of Minnesota andre@cs.umn.edu www-users.cs.umn.edu/~andre/ Doel González North Carolina State University dlgonzal@ncsu.edu Sucharita Gopal Boston University suchi@bu.edu www.bu.edu/earth/people/faculty/sucharita-gopal/ Mohammad Gorji-Sefidmazgi North Carolina Agricultural & Technical State University mgorjise@aggies.ncat.edu acitcenter.ncat.edu/gorji.html Matz Haugen Stanford University mahaugen@stanford.edu William Hendrix Northwestern University whendrix@northwestern.edu

Forrest Hoffman Oak Ridge National Laboratory forrest@climatemodeling.org www.climatemodeling.org/∼forrest Abdollah Homaifar North Carolina Agricultural & Technical State University homaifar@ncat.edu Jim Hurrel National Center for Atmospheric Research jhurrell@ucar.edu www.cgd.ucar.edu/staff/jhurrell/ Zhe Jiang University of Minnesota zhe@cs.umn.edu www-users.cs.umn.edu/~zhe/ Anuj Karpatne University of Minnesota karpa009@umn.edu Alicia Karspeck National Center for Atmospheric Research aliciak@ucar.edu Woodrow Keifenheim University of Minnesota keif0006@umn.edu Ankush Khandelwal University of Minnesota ankush@cs.umn.edu Devashish Kumar Northeastern University devashishkr@gmail.com www.northeastern.edu/sds/kumar.html Vipin Kumar University of Minnesota kumar@cs.umn.edu www.cs.umn.edu/~kumar Chaunté Lacewell North Carolina Agricultural & Technical State University

58

cwlacewe@ncat.edu acitcenter.ncat.edu/Lacewell.html Todd Leen National Science Foundation tleen@nsf.gov Wei-King Liao Northwestern University wkliao@ece.northwestern.edu Stefan Liess University of Minnesota liess@umn.edu Richard Loft National Center for Atmospheric Research loft@ucar.edu Mengqian Lu Columbia University Linda Mearns National Center for Atmospheric Research lindam@ucar.edu www.isse.ucar.edu/staff/mearns/index.php Varun Mithal University of Minnesota mitha011@umn.edu Guruprasad Nayak University of Minnesota nayak013@umn.edu Ramakrishna Nemani NASA- Ames Research Center rama.nemani@nasa.gov Anita Nickolich National Science Foundation anikolic@nsf.gov Douglas Nychka National Center for Atmospheric Research nychka@ucar.edu www2.image.ucar.edu/staff/nychka-doug Nikunj Oza

National Aeronautics and Space Administration nikunj.c.oza@nasa.gov Srini Parthasarathy Ohio State University parthasarathy.2@osu.edu Bala Rajaratnam Stanford University brajarat@stanford.edu woods.stanford.edu/about/woods-faculty/bala-rajaratnam Stephen Ranshous North Carolina State University smransho@ncsu.edu Pradeep Ravikumar University of Texas- Austin pradeepr@csutexas.edu www.cs.utexas.edu/~pradeepr/ Richard Rood National Center for Atmospheric Research rbrood@ucar.edu Stephen Sain National Center for Atmospheric Research ssain@ucar.edu www.image.ucar.edu/~ssain Nagiza Samatova North Carolina State University samatova@csc.ncsu.edu Scott Sellars University of California Irvine scott.sellars@uci.edu Shashi Shekhar University of Minnesota shekhar@cs.umn.edu Deepti Singh Stanford University singhd@stanford.edu Vidyashankar Sivakumar University of Minnesota sivak017@umn.edu

59

Michael Steinbach University of Minnesota steinbac@cs.umn.edu Daniel Swain Stanford University dlswain@stanford.edu Xun Tang University of Minnesota tangx456@umn.edu Claudia Tebaldi National Center for Atmospheric Research tebaldi@ucar.edu Danielle Touma Stanford University detouma@stanford.edu Erin Towler National Center for Atmospheric Research towler@ucar.edu staff.ucar.edu/users/towler Kevin Trenberth National Center for Atmospheric Research trenbert@ucar.edu www.cgd.ucar.edu/staff/trenbert/ Raju Vatsavai Oak Ridge National Laboratory vatsavairr@ornl.gov www.ornl.gov/~r7v Dawei Wang University of Massachusetts Boston Warren Washington National Center for Atmospheric Research www.cgd.ucar.edu/ccr/warren/ Marcia Zilli University of California, Santa Barbara