This interview was conducted and written by Ruby Barcklay and first appeared on Berkeley Lab’s internal websites.
Charuleka Varadharajan, a research scientist in the Earth and Environmental Sciences Area (EESA), was interested in the nexus of environmental science and computation even as an undergraduate at the Indian Institute of Technology in Chennai. Her thesis explored the use of geographic information systems (GIS) and systems analysis for integrated coastal zone management. She continued this multidisciplinary work through her Masters and Ph.D. studies at the Massachusetts Institute of Technology, where she built software to run remote laboratory experiments, and used a combination of field work and statistical approaches to understand how freshwater lakes release methane into the atmosphere. She joined the Earth Sciences as a postdoc in 2010 to work on laboratory and synchrotron studies of groundwater quality and found a natural home for her interdisciplinary work. She shares her thoughts on this work.
What is your area of work?
I’m an environmental data scientist. Part of my research focuses on using data science approaches to understand how natural water resources such as watersheds, rivers, and groundwater bodies respond to the changing climate and to disturbances such as floods, droughts, and heat waves. I also work on building tools for data processing and integration, and facilitating data management activities that enable other scientists to make their data more openly available and reusable.
What big challenges are you hoping to solve with your work over the next 20 years?
Adapting to climate change is one of the defining problems that my generation and future generations of scientists will have to tackle. Global warming and extreme events will impact how much clean water is available for drinking, agriculture, and other purposes. Right now, we cannot accurately predict the amount of water available and its quality (e.g. water temperature, salt and oxygen content) across many different types of watersheds at seasonal to decadal timescales. This is an important problem to solve because billions of dollars are spent on water resource management annually, and decision makers need accurate models to be able to allocate water to satisfy growing demands while balancing other criteria to maintain water quality and preserve aquatic life.
I hope that in 20 years, my research will help us better predict the amount of clean water that will be available in rivers and groundwater, and determine if and how watersheds are resilient to climate stressors such as droughts and heatwaves. But the study of how climate change affects our watersheds is a very complex topic that needs a huge global community to work together. If we can get to a point where it is trivial to make scientific data open and reusable that would greatly accelerate collaborative analysis and modeling.
What steps are you taking today to accomplish this vision?
In my Department of Energy (DOE) Early Career Research Program project, we are using data science approaches to understand the impacts of disturbances such as floods and droughts on water quality in large rivers of the United States. My team is building a data-driven framework that includes machine learning models and accounts for different types of watershed characteristics such as geology, climate, and land use.
I’m also the Co-Principal Investigator (PI) for the ESS-DIVE (short for Environmental System Science Data Infrastructure for a Virtual Ecosystem) data repository, a place where scientists can archive and make their DOE-funded digital datasets public. I co-lead the data teams for the Berkeley Lab Watershed Function Science Focus Area, which studies the impact of climate-driven perturbations on a mountainous watershed in the Colorado River Basin and the NGEE-Tropics (short for Next Generation Ecosystem Experiment-Tropics) project, which focuses on how tropical forests respond to changing climate. These projects have given me an understanding of how diverse ecosystems function. They collect interdisciplinary observations, and we are building software and tools to manage data and make them more usable in models.
Who would you like to partner with at the Lab to bring this vision to life?
I have a great team of students, postdocs, and research staff who are essential to making this vision a reality. I work with lots of different people at the Lab. Within Earth Sciences I am part of large interdisciplinary teams that include a combination of field data collectors and modelers. I think it’s important to continue to use our iterative approach of co-developing models with data collection activities, and am especially interested in working with scientists who are integrating machine learning with physics-based deterministic models. There are many mentors in EESA who have, and I hope will continue to provide input into my water-related research and career development including Peter Nico, Eoin Brodie, Ken Williams, Peter Fiske, and Susan Hubbard.
I was lucky to have met Deb Agarwal with the Computational Research Division early in my career at the Lab; we have had a long-standing collaboration since 2015. I’ve been working closely with CRD researchers in her Data Sciences and Technology department— Shreyas Cholia, Valerie Hendrix, Gilberto Pasterello, Danielle Christianson and many others in the ESS-DIVE team. I’ve had an LDRD partnership with Juliane Mueller, an applied mathematician, building surrogate models for predicting groundwater levels in California watersheds. I hope to continue maintaining the long-term relationship with computational sciences, since it is critical to my work. We will also need the right scientific computing infrastructure including the high performance computing facilities from NERSC.
I also hope to work with other areas and groups at the Lab such as Biosciences, which has large projects working on making data openly available and building tools to use the data and hope to interact with them through efforts that will be housed in the BioEPIC building in development at Berkeley Lab. I am on the scientific advisory board for the National Microbiome Data Collective. And I am also part of the Lab’s data policy working group led by Joerg Heber, and believe it is important to define policies that encourage sharing data openly while providing credit to scientists who collect the data.
As a PI, I am lucky to work with phenomenal project managers and operations staff. They will continue to be a part of my future work to keep me on track with meeting milestones and helping with proposals, budgets, and navigating the complex Lab environment. HR has been a huge help with recruiting and hiring my growing group. The Early Career Enrichment Program organized by the Directorate was also helpful to me. The program encourages interdisciplinary partnerships, and through this program I’ve met people that I hope to work with in the future.
Who from the past, present, or future would you like to collaborate with? And on what?
I would love to work with the broader watershed science community to develop an understanding of the problems that will impact our water resources due to a changing climate and provide the scientific knowledge to help make optimal decisions on water management. We are studying complex, large systems, and the water research community needs to be more integrated than it is today. No single organization has the expertise or resources to collect the data or conduct modeling across all relevant spatial scales, so we need to build “networks of networks” of interdisciplinary researchers who study both natural (e.g. watershed processes) and human dimensions (e.g. dam operations, agricultural water use) to get a holistic picture.
I also believe that despite the rapid increase of environmental data available, we still do not have sufficient data to make predictions since the processes we study vary a lot in space and time. We need an order of magnitude more data, especially if we hope to make use of deep learning models. A few of us have been working on a vision of self-guiding field observatories that can help collect much more data that are optimized in real-time using models and data assimilation. Imagine if an automated drone could dynamically sample and adjust the resolution of data collected before and after a storm. To make this happen, we would need sensor and data experts from Earth Sciences partnering with networking, robotics, and computational researchers at ESNet and Computational Sciences at the Lab and elsewhere. We have a long way to go to get there, but if we manage to pull this off, it would be transformational for Earth Sciences as a whole.