The Berkeley COVID baby

Everyone has one right - the project that took 3 times as long and was 3 times as hard because of COVID? Our version of this was Jeannie Wilkening's pinnacle experiment for her PhD - the experiment to make the dataset we were going to build the whole PhD around - the experiment that was going to take 1 summer. The experiment that exciting combinations of Berkeley campus shutdowns, fires that completely mucked up radiation inputs, isolation and travel bans stretched over what turned into 3 years of piloting, analysis, growing, missing sleep and eventually producing a really amazing dataset just in time for Jeannie to ... graduate.

So - meet our very own COVID baby - the "Mind the Data Gap" paper.

A few years back, Xue Feng and I decided to throw one of our favourite engineering tools- dimensional analysis - at the coupled equations that describe plant and soil hydraulics (you can read about it in Xue's "The ecohydrologic context of drought" paper). We were hoping we'd find some combinations of parameters that helped simplify this system and tell us what to measure and vary in experiments. Unfortunately we weren't able to get the number of dimensionless groups down to fewer than 8, and several of these groups contained parameters that mixed plant physiology and soil hydraulics. This was eye opening - it meant that all those times plant physiologists measured plant traits and tried to explain the world with them, without accounting for soil differences, they were missing essential information. Similarly, if hydrologists measured everything about soil and water in an experiment but didn't quantify plant traits, they would miss important parts of the story. We thought this was a big enough deal that we took the time to summarise the history of plant physiology and ecohydrology and what it means for them to work together in Jennie's recent "Different Roads Same Destination" paper, which I blogged about a while back. But the consequence of saying "you have to measure it all" is that ... you have to measure it all. That's what we tried to do in this experiment - make a more comprehensive dataset that showed the internal state of water in the plant and soil, the fluxes between them, and that differentiated different sources of water in the soil using isotopic tracers.

We used Cottonwood trees, Populus trichocarpa, growing on the roof of the Biological Sciences building at Berkeley. We grew them in pots until they were 2 years old, then gave them a nice period of well watered conditions, turned the watering off, left them to get stressed, and then resumed watering. The last lot of watering they got, along with the water applied to enable recovery, were labeled with isotopes of deuterium and oxygen, creating unique tracers that we could recover from the water, soil and plant. The full suite of measurements is shown below:

By and large, the trees did what we'd expect. Soil dried out as watering stopped and the trees down-regulated their photosynthesis and stomatal opening. We could see this through multiple angles - water potentials in the plant, isotopes and stem water conteent. The image below shows the daily sapflow profile for 6 trees - high under well watered conditions on day 2, crashing down on days 5 and 8 during drought, starting to recover on day 12 and "recovered" - although not back to the unstressed state - by day 15.

And we see the same processes reflected in stem water content:

What's interesting is that although the stem water cotent dynamics recover after drought - that is, the stem loses water while actively photosynthesising/transpiring during the day, and the minimum water content is very similar for day 2 and day 12 - the sapflow didn't recover. And that's because the experiment triggered a drought response in the trees that was not "recoverable" in thee time of the measurements - they defoliated.

The "recovered" trees had recovered their water contents, but not their leaf area or leaf condition. This is one of those "BUGGER" moments, because what would have been great is if we'd been measuring hormonal composition of the leaves - particularly absissic acid conntent - which we hypothesise was altered by the drought stress and initiated the leaf senescence and abscission. While we do have these images and while we collected all shed leaves, we also didn't include permanent LAI measurements so our ability to link leaf area per tree to the change in sapflow is limited. Nothing like 20-20 hindsight!

That said what we think is the most valuable output of this experiment is the dataset - one which does characterise soils, roots, stems, leaves, hydraulics in plants and soil, the environment and tracers. Jeannie didn't have time to explore the value of this dataset for evaluating the performance of ecohydrological models - I hope that someone will take that up in their own work. If that someone is YOU, please have a look at the data -

https://doi.org/10.5281/zenodo.10685215. This dataset is the minimally processed dataset. The scripts for fully processing the data and producing the results and figures presented in the paper are available at https://github.com/jvwilkening/Drydown_Expmt_Data_Analysis.