I defended my PhD thesis* in January, and decided it would be nice to put it online in HTML format. I have posted it here. The PDF is available via France’s open science portal HAL.

Heberger, Matthew. 2024. Improved observation of the global water cycle with satellite remote sensing and neural network modeling. PhD thesis. Sorbonne University, Paris, France. https://theses.hal.science/tel-04517802v1

My research had to do with using remote sensing data to describe the water cycle at the global scale, and explored methods to improve these data. There have been quite a few studies on this subject in the last several years, but we did a few things differently, namely using a much larger database of observations than previous studies, and using neural network models with some unique features. A few members of my committee said found my writing “pedagogic” and plan to share certain parts with their students. I hope that it is of some use to students or others interested in hydrology and remote sensing.

The PDF is formatted nicely and better for print. But the web version is nicer to read on a wider range of devices. It can also be resized or reformatted for easier reading. (When I’m reading long online documents, I like to use Firefox’s reader view.)

I used Latex and Overleaf to prepare the manuscript. I was hoping that it would be straightforward to convert to HTML. Unfortunately, it was not! I used the software pandoc to convert the document, but it required a lot of manual cleanup. Please let me know if you see any weird formatting or typos!

* In the US, it is customary to talk about a Master’s thesis, and a PhD dissertation. Under the British educational system, it’s the opposite. And in France, to obtain a doctoral degree, one writes une thèse.

I put together this list of journals with some info on their fees during my PhD research, and I’m sharing it here in case others may find it helpful. This information is likely to change fast, so make sure to verify.

Link to the Google Sheet.

During my PhD research, I became a convert to Latex for scientific writing, specifically using the website Overleaf.com. However, one feature that is missing is a good spelling and grammar checker. I like to use Microsoft Word for copyediting, as it has good built-in tools, and there are also plenty of addins available, like grammarly and ProWritingAid.

But first you need to get a clean export from Latex to Word, which is not straightforward.

Here is a method that works using Google Drive. It does not do a good job converting figures, tables, and equations, so I suppress those before continuing. It also helps to turn off hyphenation.

1. Add the following block to the preamble of the Latex document.

\usepackage[document]{ragged2e}
\usepackage[none]{hyphenat} % Turn off hyphenation
\usepackage{environ}
\RenewEnviron{figure}{} % Removes figures
\RenewEnviron{table}{} % Removes tables
\usepackage[nolists,nomarkers]{endfloat}
\renewcommand{\processdelayedfloats}{}
\pagestyle{empty} % Removes page numbers

2. Create a PDF document.

3. Upload the PDF to Google Drive.

4. In Drive, click the file to view it. At the top, click “Open with Google Docs.”

5. Choose File > Download > Microsoft Word (.docx)

Now you can download the Word file and copy edit at your leisure.

If you are interested in the Earth Sciences, you should consider attending the annual American Geophysical Union annual meeting. It is massive, with over 50,000 attendees, and thousands of scientific presentations, posters, lectures, and films. This can also make it feel overwhelming. Here are some highlights of the December 2023 conference in San Francisco from my point of view.

A former professor of mine, Rich Vogel from Tufts University, gave the Langbein Lecture. Being invited to give a named lectures is a big deal, equivalent to a lifetime achievement award. Even though he retired several years ago, he still publishes tons of papers and hasn’t lost any of his pep. It was a lot of fun.

I have really mixed feelings about poster sessions. On the one hand, it can be fun to get a quick view of a huge amount of research, and to meet researchers in person and talk to them about their work. On the other hand, it can be hard to learn much given the format. The Better Poster movement seems to have made very little impact. A good 95% of the posers are inscrutable walls of text.

I was honored to present my PhD research at the Annual Meeting of the American Geophysical Union in San Francisco, California, on December 15, 2023. My research is in the field of remote sensing and large scale hydrology. Here’s a copy of my slides if you’re interested:

Heberger_AGU_2023-12-15.pdf (2.5 MB)

For those interested in a deep dive, here is a link to the draft of my thesis (to be finalized after my defense in January). Here is a link to the final version of my PhD thesis:

• pdf
• html

Suppose you are interested in recreating my calculations or doing something related. The input datasets I used can be freely obtained via the sources listed in thesis Section 2, Datasets. The Appendix contains a much longer list of all the datasets I reviewed or considered using. I think it’s a good snapshot of remote sensing datasets describing the hydrologic cycle that were available circa 2023. The compiled data and Matlab scripts needed to perform the analysis can be downloaded from: https://doi.org/10.5281/zenodo.8101659.

I was delighted by the look of this watershed. Can you figure out where it is?

Just a few kilometers west of Bolivia’s capital, Sucre, lies the Maragua Crater, the site of a long-ago asteroid impact.

All of the top search results are about hiking out to visit it. There are 2,000 year old cave paintings, fossil dinosaur footprints, and incredible multicolor rock formations. Wow, new place to add to the bucket list!

Photo by Cody Hinchliff, flickr.

I recently became aware that the Global Watersheds app occasionally goes down when it’s under heavy load. At its peak, the app is delivering a watershed every 1.2 seconds. You all are loving the site to death! 🙂

I’ve upgraded my hosting plan to 2GB of RAM. Let’s see how much that helps. It costs a few hundred dollars per year, which is a lot for a hobby project by a full-time student. If you’ve enjoyed using the app, or it’s been valuable in your work, would you please support it by sending a few dollars? Thank you! 🙏 💕

The Global Watersheds web app has a nice feature that it seems only a few people have discovered. You can create a permalink to an interactive map of your watershed (or flowpath). Then you can embed the map on a web page, send it to a friend, or post it on your favorite social media site. Just look for these buttons at the bottom left after you have created a new watershed:

Since the app was launched in October 2022, users have created 770 shared watersheds. Here is a map and a table showing all of them. In that same time, users have made over 237,000 watersheds and over 36,000 flow paths.

Happy exploring!

Matlab is great for producing high-quality graphs and plots for all kinds of science and engineering problems. Plots can exported to many formats, and inserted directly into reports or journal articles. 

Matlab is also great when you want to create multiple plots together on a single figure. The standard way to do this is with the subplot() command. Indeed, you can create attractive figures with a grid of plots. The following code creates a 3 x 3 set of 9 plots:

figure('units','normalized','outerposition',[0 0 1 1]);
for i = 1:9
   subplot(3, 3, i)
   plot(rand(1, 20));
end

A common complaint about the subplot function is that it leaves too much whitespace in between the plots. This is especially noticeable when you are making plots without tick labels on the x-axis or y-axis: 

tiledlayout()

In 2019, Matlab introduced the tiledlayout() function as an alternative to subplot(), which gives you more control over the amount of space between the individual plots. Here’s an example of it’s use:

figure('units', 'normalized', 'outerposition', [0 0 1 1]);
t = tiledlayout(3, 3, "TileSpacing", "tight");
for i = 1:9
   nexttile
   plot(rand(1, 20));
end

The ‘TileSpacing’ parameter can take the values ‘tight’, ‘none’, ‘compact’, or ‘loose’. This gif shows the difference: 

tight_subplot()

The tiledlayout function works well, most of the time. If you need even more control over the layout of your subplots, head over to the Matlab File Exchange, and download the function tight_subplot(). There are many similar contributions, but this one appears to be the most popular, with 47K downloads as of early 2023. It works perfectly and is simple to use once you get the hang of it.

With the tight_subplot() function, you can specify exactly how much space you want between plots. 

You can separately adjust the vertical spacing and horizontal spacing between subplots. 

To do this, adjust the parameters gap_h and gap_w for the gap distance between plots in the horizontal and vertical directions. 

[ha, pos] = tight_subplot(rows, cols, [gap_h, gap_w]);

The function returns two variables. The first is an array of handles of the axes objects. You have to activate the axis in order to make a plot. For example:

[ha, pos] = tight_subplot(3, 3, [0.05, 0.05]);
for i = 1:9
    axes(ha(i));     % Don't forget to 'activate' the axis. 
    plot(rand(10, 1); 
end

The values for gap_h and gap_v are “normalized units” from 0 to 1, where 1 is the full width or height of the computer screen is 1. So setting gap_h to 0.01 means the horizontal gap between the plots will be 1% of the screen width. Here is an illustration of changing gap_h:

Here is the effect of modifying gap_w: 

And here’s an example where I used the tight_subplot function to create a set of maps of global evapotranspiration over 9 consecutive months, using data from a meteorological reanalysis model called ERA5. This works well because we don’t need axis labels. I also used the trick of creating a single colorbar and customizing its position on the figure.