A –Streams of Thought– contribution by Sina Khatami.
Asst/Prof. Grey Nearing is a hydrologist at the Department of Geological Sciences at The University of Alabama (UA). Prior to joining UA, he has worked as Project Scientist at the National Center for Atmospheric Researchand, and Research Scientist at the NASA Hydrologic Sciences Lab. I’ve enjoyed an ongoing dialectical debate with Grey, intense yet delightful, on the philosophy of science particularly hydrological uncertainty. It’s been a pleasure to interview Grey.
Can you tell us a little about your background and education?
I studied Math in undergraduate because I felt that this would
keep my options open in terms of future career paths. I went into the
Environmental Sciences mostly because this is where I found a graduate
assistantship (through the US Department of Agriculture). I chose my PhD
adviser because I enjoyed reading his papers.
Was becoming a scientist your career plan when you were a student? Tell us about the journey how you got here?
I started college to be a software engineer but decided that I
didn’t want to spend all my time writing code (I currently spend about a
third to half of my time programming). Several of my friends from high
school and college that did become software engineers ended up being
fairly bored with their careers after a few years. Then again, some of
the young software engineers that I work with (e.g., at consulting and
tech companies) really love their work and have skills that allow them
to accomplish more in the field than any trained Hydrologist that I
know.
Your research has a particular focus on Information Theory
(IT) and its application in hydrological sciences. How did you become
interested in IT, and do you think it’s a worthwhile research focus to
pursue in hydrological sciences?
I’m not sure that information theory is a worthwhile research topic. I got into it because I was unsatisfied with the philosophical
discussion about uncertainty in Hydrology journals. It didn’t make
sense to me to try to quantify what we don’t know (i.e., to quantify
uncertainty), and information theory seemed to provide a way to reframe
questions about inference under partial information. There are certain
boundedness theorems in information theory that don’t exist in
probability theory that I think might help provide coherency to
questions about imperfect inference; if we learn how to ask scientific
questions in the context of information theory rather than probability
theory. I’m not certain about this, but I am certain that I don’t like
the existing literature on uncertainty quantification.
You’ve moved from academia (University of Arizona) to the
public sector (NCAR and NASA) and then moved back to academia again
(University of Alabama). How did you find these transitions?
I’m not sure it’s that big of a jump from academia to
government research, but there are differences. Projects at government
labs tend to be high-profile, and you work on them with large(r) teams.
If you want to be part of a big project and don’t mind having to meet
deadlines related to code production, then this is a good option. My
work as a project scientist at NCAR and NASA was definitely less
stressful than academia. But I had less creative freedom.
And what skills do you think you need to develop now for an academic career that you didn’t need to previously?
Time management, perpetual self-motivation, and the ability to
handle rejection are the skills that matter most in academia. It’s hard
to work so much, and have your grant proposals rejected so often,
without losing the sense of excitement that keeps the work interesting.
Academia is hyper-competitive, which was exciting at first, but starts
to be a little tedious after a few years.
What are your current research projects?
I’m working on the intersection of machine learning and
process-modeling. How can we take the best parts of both and combine
them?
You’ve been recognized as an AGU’s Outstanding Reviewers of 2017 by the Water Resources Research Editorial.
What do you think led to this great recognition, and what is your
advice on paper review to young and early career hydrologists?
Some of the advice I sometimes get from senior scientists about
paper and proposal reviews is that this should not be a priority for
young scientists. But if you don’t do your fair share then you are a
burden on the community. My rule is to review at least 3x the number of
papers that I submit, otherwise I’m asking others for more review effort
than I am willing to give. I spend a lot of time on (most) of my
reviews, and I think I am often the third reviewer because I recommend
to reject a lot.
What do you account as your major breakthrough so far in your research career as an early career scientist?
I don’t have a major breakthrough. There aren’t many major
breakthroughs in Hydrology. I’ve not seen one in the field in several
years. My opinion is that Hydrology isn’t a field defined by
breakthroughs – it is a field of incremental progress toward relatively
well-defined objectives.
What do you account as your major challenges ahead of you in your research career as an early career academic?
Success boils down to funding. That’s really all there is to
it. If I’m successful in getting proposals funded, I will have the
resources (people) to pursue bigger projects. The other path to success
in science is fundamentally new theory development, but I don’t see this
as likely in a field as saturated as Hydrology.
How do you describe your research style? Is novelty your main criterion for research?
I am a contrarian, so not only is novelty the
main criteria for me in choosing a research direction, but I often
choose to work on projects just because they are contrary to what others
are working on.
How do you approach creativity in your research? In other
words, what people and activities helped you develop the mindset for
“outside the box” thinking.
When I’m being creative, I just think a lot. I think about
projects when I’m at the gym, walking to class, eating dinner, going to
bed, etc. If I think about something long enough, understand it well
enough, and keep myself active, I’ll often have an interesting thought.
In the ~10 years that I’ve been in the field I’ve gone in and out of
phases where I do this vs. where I just work on routine stuff that needs
to get done.
Do you think that creativity and success are correlated?
Probably not. The big names in Hydrology got there because they
led a big project with societal relevance. My personal assessment is
that creativity and success in Hydrology are anti-correlated, if
anything. Creativity is rewarded in science when there are new things to
discover, but in an incremental, applied field like Hydrology, the
qualities that lead to success are the ability to clearly articulate and
implement the next incremental step.
What are your main hobbies besides work?
I don’t do anything other than work. Honestly, I don’t think
it’s realistic for most people to have a work-life balance as a
non-tenured academic. Maybe an exceptional person can pull this off.
How are you balancing your work and life? Any regrets or advice for early career and aspiring hydrologists?
I don’t have any regrets. My job is what I’m interested in and
what I love doing. My only advice is that you really have to love doing
this work if you go into academia. There are plenty of jobs in industry
that pay more and require less time commitment and stress.
What are the grand gaps/questions of hydrological sciences
that in your opinion the community, particularly the early career
hydrologists, should tackle?
Right now, I think the big challenge that Hydrology might
participate in is about understanding the interface between physical
science and data-driven discovery. In its current form, this is about
how we can merge physics with machine learning, but it’s
fundamentally the same problem as understanding the roles of
hypotheses/theory vs. statistics in traditional scientific inference. At
a philosophical level, merging physics with machine learning is the
same problem as uncertainty quantification, which is just the problem of
merging probability with physics.
I also still see some potential for finding scale-relevant hydrological laws. I know that this is generally considered to be a failed project, but I think there may have been some opportunities missed in previous efforts to constrain interactions between multi-timescale models using maximum entropy and maximum entropy production. This has been tried for decades (going back to the 1950s), but I’m not sure that it was done well. An example of this is the study by Wang and Bras (2011) on maximum entropy production models of evapotranspiration, which are fantastic.
I really only see two fundamental challenges in the field – scale-relevant theories and uncertainty quantification, in whatever form the latter might take (I would argue that it’s really a problem of inference under partial information). Both of these projects have been around for a long time with some practical advances but no real theoretical advances. If someone could make a real dent against one of these problems, it would be a big deal. Of course, there are a lot of practical challenges in Hydrology – predicting effects of climate change on local hydrological systems, predictive modeling anthropogenic influences on watersheds, closing the water cycle with remote sensing, partitioning evaporation and transpiration over large scales, etc. – but these aren’t what I would call Grand Challenges. The Grand Challenges in the field have not changed since Beven outlined the two I mentioned above in his speech at the Water for the Future conference published in 1987.
What are the biggest challenges and opportunities for
hydrologists in the next 10 years? In the next 50 years? Especially the
ones that interest you.
To be honest, I think our discipline is at an existential tipping point. We are firmly in the era of Big Data,
and machine learning is better, almost across the board, at simulating
hydrological systems than our best process-based models that have
resulted from the previous decades of incremental development. Soon, I
expect that most water managers will buy their water-related information
products from companies like IBM and Google; the latter is apparently
working on a global streamflow modeling system.
I think the major question facing our community right now is to understand where, when, and under what conditions our traditional physical science adds value in the emerging Age of Artificial Intelligence. Maybe calling this an existential crisis is a little dramatic, but I do think that understanding this cohesion (science + machine learning) will be the challenge that ends up having the largest overall impact on our field over the next 50 years.
About the author
Sina Khatami, a PhD Candidate in hydrology and uncertainty at
the University of Melbourne, is currently the Secretary of Young
Hydrologic Society (YHS) and an Editor of YHS Blogs. He is also a
committee member of AGU’s Hydrology Section Hydrological Uncertainty
Technical Committee since 2018, and Student Subcommittee (H3S) since
2017. Correspondence to sina.khatami@unimelb.edu.au.
References
Beven, K. J. (1987) Towards a new paradigm in hydrology. In Water for the future: hydrology in perspective, International Association of Hydrological Sciences, Rome Symposium (eds JC Rodda, NC Matalas), IAHS Publ. No. 164, pp. 393–403. London, UK: IAHS.
Wang, J., and Bras, R. L. ( 2011), A model of evapotranspiration based on the theory of maximum entropy production, Water Resour. Res., 47, W03521, doi:10.1029/2010WR009392.