Skip to main content

Minirhizotron: Phenology And Root TraitS

Periodic Reporting for period 1 - Mr.PARTS (Minirhizotron: Phenology And Root TraitS)

Reporting period: 2017-06-01 to 2019-05-31

The overall aims of MrPARTS was to improve understanding of plant functioning in real word settings by developing technical solutions to the difficulty in making frequent, phenology (seasonal cycle)-relevant measurements of roots.

Currently, ecosystem scientists are able to collect data at daily or better timescale on changes in above-ground portions of ecosystems by using remote sensed images (from satellites, aircraft, or local static cameras). This is of use as short-term changes (such as the beginning of leaf growth) can be related to measurements of carbon dioxide exchange by the ecosystem and helps us gain a mechanistic understanding of how ecosystems function.

Roots make up a substantial proportion of total plant biomass and we know that the amount and function of roots changes both over time and in response to background conditions such as climate or soil fertility. As all biomass is made of carbon, this affects the overall C storage and fluxes in and out of the ecosystem. However, roots are often overlooked as measurements are very difficult, usually requiring excavation and cleaning before measurements can be made. Consequently we often assume that roots are predictable from shoot measurements in such experiments (which are used to inform future climate change predictions) without knowing if this assumption is correct. Understanding how true this assumption is, and how different environmental conditions change this link, is key to predicting whole-ecosystem changes in the future.

Thus the scientific/technical objectives of MrPARTS were:
1) develop a set of affordable (and hence replicable on experimental budgets) automated minirhizotrons to capture images of root growth at phenological timescales
2) develop a method of analysing frequent root images to extract ecologically useful information
3) deploy these systems in a mesocosm experiment to test basic mechanistic controls on high time resolution root dynamics and their links to above-ground processes
4) deploy these systems at a field site (Majdas del Tietar, in Extremadura Spain) where both a nutrient manipulation experiment and strong seasonal variability in water availability may drive desychronisation above-and-below ground, and where above-ground data is already being collected
5) collect ancillary validation data to 'ground truth' automatically extracted data from minirhizotron imagery.
At the end of MrPARTS we have successfully produced 10 automatic minirhizotron systems at a cost substantially lower than commercial alternatives. We have conducted successful trials both in the greenhouse and the field of the automatic system. We have also developed a method that substantially speeds up image processing from manual surveys and continue to work with automated, high resolution data.
Compared to the original timeplan for MrPARTS, both 1) design and production of the automatic minirhizotron system and 2) implementation of the computer vision method were substantially slower than anticipated. This was due to unforseen constraints on camera specifications for the niche application in close-focusing, space-restricted root observatories and restrictions on institutional technician time. The problem of root identification proved more difficult than anticipated, partially due to the late availability of field data, so we did not move beyond detecting root biomass using computer vision methods to further objectives of the proposal (identification of further root properties and traits). Further work as a consequence of MrPARTS continues on developing a neural network classifier alongside deployment of the full set of automatic systems in the field for continuous site monitoring and testing of the synchronicity hypothesis.
We conducted a trial of the automatic system in both the greenhouse mesocosm and the field site although ongoing technical adjustments meant we decided that it was not scientifically robust to use these in a treatment experiment during the project.
Due to the unexpected need to spend extra time and resources on developing and trailing the automatic system we were not able to conduct the proposed 13C labelling experiment within budget and time constraints of the fellowship but were able to produce two publications (Nair et al 2019 and Nair et al in prep) based on the ancillary root measurements, a replacement 15N-litter decomposition experiment, and manual minirhizotron data. We expect further scholarly publications to arise from the use of the automatic systems. Results from the project have been presented orally at the EGU 2018 and 2019 and technical demonstrations of root observation technology were presented at the Jena Lange Nacht des Wissenschaften 2017. Results are also disseminated to the scientific community via social media accounts. Overall, MrPARTS has provided valuable open-source datasets root biomass in understudied (Mediterranean savanna) systems and laid the groundwork for future work utilizing the tools and instruments developed in climate change experiments.
In MrPARTS we have successfully lowered the cost of making automated measurements of root growth by developing new instruments which utilize modern technology (cheap cameras, single board microcomputers) and are suitable for field installation. The development time spent in this project and the instruments we have developed are of use to scientists working on root measurement and whole ecosystem C budgets in future. We expect to further this aspect of the project for the wider community via a patented design or open source plans. We expect the longer term impacts of this project to be better measurements of roots, and consequently better predictions from vegetation models which can be trained from the root data produced by these instruments.
A trial of the robotic root camera system
A variety of images from the manual minirhizotron system in the field
Performance of the classifier on root cover compared to manual data (10 cm depth)
The greenhouse mesocosm. A single automatic system is visible in the rightmost box.
Deployment of the system in the field