Advice on publishing earth observation research
A Q&A from a top reviewer in the field
A few weeks ago, I came across Dr. Andrew Feldman’s profile on Twitter. Dr. Feldman completed a Ph.D. in Civil and Environmental Engineering from MIT, and he is currently a Postdoctoral Research Fellow at NASA Goddard Space Flight Center.
He has recently been named one of the top reviewers for Remote Sensing of Environment. I thought it would be great to ask him some questions about his work as a reviewer and his thoughts on the future of earth observation research. This article contains a list of questions I presented to him, and his associated responses.
I learned a lot from Dr. Feldman’s take on publishing, and will be applying some of these principles as I publish my first couple papers. I hope you also find value in his insights.
Can you tell us a little bit about your background and how you became involved in reviewing papers for an academic journal focused on earth observation?
“Currently, I am in my second year as a NASA Postdoctoral Program fellow at NASA Goddard Spaceflight Center in Greenbelt, Maryland, USA. I am a hydrologist and biospheric scientist that mainly uses satellite remote sensing to understand how the global water and carbon cycles respond to climate variability. Specifically, I am studying how vegetation across the globe responds to variability from daily rainfall frequency changes and to more extreme variability from droughts. Much of my work has focused on drylands that have water-limited vegetation.
I became involved in reviewing shortly after I submitted my first paper in my third year of graduate school for review in Remote Sensing of Environment (RSE). The RSE associate editor for my manuscript is an expert in my specific field. Coincidentally, I met this same editor at a remote sensing conference briefly the same year. When I returned home from the conference, I had a review request in my inbox from this editor for a submission at RSE. I now review papers regularly for RSE. This same thing happened the next year for another journal with an interaction with an editor. I am now receiving many review requests.
I suspect you start getting to review papers when you start meeting associate editors via submitting a paper for review or by getting to know them at conferences and meetings. Associate editors are often mid-career researchers who do the hard work of finding reviewers to evaluate submissions for a journal and are always looking for early career researchers to review for them.”
How do you evaluate the quality and rigor of papers submitted for review, and what criteria do you use to determine whether a paper is suitable for publication?
“I mostly focus on the research questions and methods. I start with what the research questions or objectives are. I then determine whether the methods are rigorous enough to address the research questions/objectives and support the main conclusions. I am usually less critical of the introduction and discussion sections because they mostly detail what previous work has done or how their work connects to other work. If the claims do not fit with the methods tested for, then I am usually critical of the work. Much of this is based on “feel” and knowing uncertainties with specific methods from my own research. I typically don’t worry too much how “impactful” the results are. I leave that for the editor to decide if the submission is worth sending to review. However, the work does need to be original and have a clear purpose. If the work is original and purposeful, that is enough for me to keep reading.
It is important for the authors to make the title, abstract, and introduction sound interesting. Reviewers are not always assessing how interesting these parts are, just whether the work is “correct” and is a new contribution to the field. In my first few papers, I often made the mistake of being too specific or niche with my title and abstract, which I think chased many suitable readers away. Reviewer’s aren’t necessarily there to check that, only to make sure that the work is sound.
I’ll note that reviewing is very subjective with no written standards or training on how to do it. How I review may be very different than how others do it – and often is! This is why papers get two to four and sometimes more reviewers.”
What are some of the most interesting or important papers you have reviewed in this field, and why do you find them compelling?
“It is difficult to single out specific papers or research questions. I’ve been fortunate to review many high quality submissions, but I haven’t reviewed a paper that I found far more important than others. I’ve felt most papers I have reviewed were important in some way: they either make a small step to advance a method or detail a new finding.
That said, papers that excite me most are those that bring a truly novel and creative idea. In cases of completely novel thinking, I tend to be slightly more forgiving since other researchers often follow up and test these novel ideas in future papers. Most of the papers I review are effectively slight variations from previous work, which will often check all the boxes for making it through the review. However, they may only have a short-lived impact or small impact on their field. It is very difficult as a researcher to truly find something new – it may take years. So don’t put pressure on yourself to get there. However, I think most researchers in our community appreciate when someone takes a risk and does something new.”
What advice would you give to researchers and authors who are looking to publish papers in this field, and what common mistakes or oversights do you see in papers that are submitted for review?
“I suggest being as rigorous with your methods as possible. I like to try alternative approaches to my methods to see if I get the same result. Can you convince yourself that your results and conclusions are right? When working with satellite observations, we use data science and statistics approaches, which have many confounding factors. While you can never fully get rid of confounders, have you tried multiple means to diagnose whether a confounder is altering your results? For example, this might include retrying your methods by replacing one soil moisture dataset with a completely independent soil moisture dataset to see if your overarching results still hold. These tests often lead to new insights and a far better paper.
I often get frustrated that Earth observation papers were formulated too quickly. The Earth data was already available. One only needs to download it and not spend hours manually collecting it in the field or laboratory. The authors then write a quick analysis. However, as I get more experienced, I can see the common flaws more easily, and these less developed papers don’t always do the extra work to show robustness of their results.
My advice: (1) have clear research questions. These are essential to tell your readers where you are going. Without them, a reader can make assumptions about what you are promising and start off in the wrong direction. (2) To get a feel for how confident in your results you should be, take your time doing the analysis and spend more time than you think you should in trying a new or alternative analysis. Journal articles take a long time to do the analysis for and write. It is often obvious that the authors did not take the extra time to check the results of their paper with new or alternative analyses. These papers tend to struggle in review.”
In your opinion, what are some of the biggest challenges facing earth observation research today, and how are researchers and scientists working to address them?
“With increasing temporal and spatial resolution, we now have large amounts of data. In graduate school, I started working with SMAP soil moisture data at either a 9km or 36km grid scale. It was sometimes slow to run an analysis pixel by pixel, but it was manageable on a personal laptop. With a supercomputer, I can do nearly any analysis I can think of with SMAP data via MATLAB or python with parallel computing within a reasonable amount of time.
However, I recently started working with ECOSTRESS land surface temperature data at less than 100 m resolution. A few swaths of observations can quickly add up to 1TB, which is not feasible on a personal computer if I want to do an analysis over a large spatial area. It forces us to turn to cloud computing tools. I am aware of NASA researchers developing ways to link their Earth observation databases with cloud computing tools through private industry like Amazon Web Services. Such tools are convenient, but can be costly. This is not yet a solved problem. I think this will be a challenge for remote sensing scientists for years to come.”
Can you discuss some recent advancements or breakthroughs in earth observation that you find particularly exciting, and what potential implications do they have for our understanding of the planet and its systems?
“A big breakthrough that comes to mind are the development of smaller, cheaper satellite instruments like “CubeSats.” Traditionally, space agencies like NASA have relied on billion dollar “flagship” missions that have instruments onboard their own spacecraft. While they typically result in highly valuable measurements, these missions take well over a decade to plan and implement and cost large fractions of a space agency’s budget. Scientists and engineers have found creative ways to develop satellite instruments at a fraction of the cost, for example, with constellations of CubeSats or smaller instruments that will attach onboard of the international space station (ISS). These allow instruments like ECOSTRESS on the ISS that measure land surface temperature at a 70 m resolution or small satellite constellations like CYGNSS which can track ocean winds and hurricanes at kilometer resolutions. These missions cost far less than one billion dollars and took less than a decade to plan. However, these types of instruments often do not have the accuracy of the flagship missions and/or may be more constrained in space and time. Satellite missions both big and small have value, but the rise of cheaper instruments has given the science community new dimensions of data availability.”
How do you balance the need for scientific accuracy and rigor with the need to communicate complex ideas and findings in a way that is accessible to a broader audience?
“This is a tough balance! This one depends on what audience your message is for. Some of my research includes technical, methodological advances that are useful to other scientists. I am usually not breaking my language into layman’s terms when writing these detailed findings for a technical journal. However, some of my other research cuts across multiple subfields of Earth science and I often need to be very accessible with my wording. Ultimately, I never sacrifice accuracy in these scientific publications and will always err on the side of being complex and correct rather than simple, but incorrect. When speaking at a conference, I may make a slightly incorrect analogy to make a concept make sense to broader range of people. However, this is rare and I will always state that the analogy is not 100% correct when that is the case.”
Can you discuss any trends or changes you have observed in the field of earth observation over the course of your career, and how do you see the field evolving in the coming years?
“A noticeable one is an expansion of Earth observation into the private sector. I don’t know the statistics, but I have heard there are now more private than public (government-owned) spacecraft measuring the Earth’s surface. There also appear to be increasing job opportunities in industry for remote sensing scientists, though I don’t know much about the differences of working for these companies instead of a university or government agency. There are so many “high profile” processes to track on the Earth’s surface, like oil and farming industries, that it is easy to see that one can monetize that. I am also happy that these industry positions appear to be on the rise, especially with noticeable increasing demands and requirements on academic researchers to get tenure and obtain funding. I think it has created competition with academia at least in the United States and may have been a driver of the recent push to increase US researcher salaries, especially on the earlier career side in PhD and postdoctoral positions.
I don’t know where these industries are headed. However, I do know that Earth observation is here to stay for the foreseeable future in some form, especially for science being done in the academic sector. NASA and many other agencies across the world continue to fund Earth observing missions given the high value that their measurements bring to science.”
What impact do you believe earth observation research can have on society and the world at large, and how important is it for the general public to be informed about this research?
“One application I focus on: with signs of climate change already increasing drought and heatwave intensity, Earth observation is going to be of increasing value in monitoring the time evolution of climate extremes. We are likely to see some bad, unprecedented extremes by the end of the century and scientists continue to predict increases in human health risks and losses in agricultural yields that come with these more intense climate extremes. An example of vulnerable areas that need continuous monitoring are drylands, or land surfaces that receive some of the lowest rainfall on the Earth’s surface such as much of Sub-Saharan Africa, the Western US, Northwest China, and Australia. A third of Earth’s population lives in these dry places and they often rely on rain-fed agriculture. Satellite observations of precipitation, soil moisture, temperature, and vegetation stress will help scientists track extreme climate event impact on humans and agriculture in near real time in these locations and many others. This is one of so many applications. Crop yield prediction, deforestation mapping, landslide detection, improved weather prediction, and so many others have so much value to guide us in our endeavors to improve (and save!) human and ecological life.
The general public should be more aware of the sheer amount of information Earth observation gives us. Whether they know it or not, the public already uses Earth observation regularly. For example, Google Maps is an Earth observation tool that relies on satellite-based and automobile-based sensors. Farmers are starting to use crop health monitoring applications developed by startup companies. With Google Earth Engine and other applications making Earth observation more user-friendly, anyone with a personal computer can interact directly with the same data that scientists use for Earth science research. Industries and startups that continue to develop and improve these applications will inherently make the general public more aware of these satellites’ capabilities and why they need to be funded. In this technological age, whether or not the public thinks it is important, Earth observation will certainly be increasingly used in their daily lives.”
Finally, what advice would you give to someone who is interested in pursuing a career in earth observation research or related fields, and what skills and qualities do you believe are most important for success in this area?
“Perhaps with any job, my advice is to only pursue the career if you truly enjoy it. The careers I am familiar with in Earth observation tend to require graduate degrees. In this case, you spend years making smaller paychecks and potentially working harder than you might in industry. I love Earth science and I love my research career! However, if I didn’t, I would have been able to use my engineering undergraduate degree in a range of consulting jobs and be compensated similarly for my effort. I say this without knowing much about the rise of careers in industry in Earth observation and whether graduate degrees are always needed. For example, I have seen many geographical information system (GIS) jobs across industry and government sectors that do not require a graduate degree.
For success in Earth observation research, my advice is to stick to fundamental science as much as possible when doing research. It is important to always keep asking “why” and breaking the problem down into smaller units to better understand driving processes. With the rise of larger datasets and longer processing time, we can easily get lost “plugging” and “chugging” into models or “black-box” tools like machine learning. We then forget to think about what is going on in between. These tools are certainly helpful, but we should ultimately use whatever we can to understand Earth’s fundamental processes which will require combinations of statistical analysis of observations and running process models. For example, never move on from a research project without thinking about whether your results fit with known theory. It can be easy to let the data do all the work and forget to use our brain.
A skill that will never stop being useful is probability and statistics. With bigger and bigger data sets and many data sources, it is more important than ever to be able to quantify the uncertainty level on the dataset and method you are using. During my PhD, in addition to probability coursework, I read most of an Econometrics textbook that focuses on time series analysis. It is has been one of the most useful skillsets I have. I use it to tease apart environmental processes driving plant function and estimate the error bars on my results.”
I hope you enjoyed Dr. Feldman’s insights as much as I did. If you have any questions for him, you can contact him here or find him on Twitter at @a_feldman24
Thanks for reading Earthbound!