Behind the paper: investigating potential biases in coral reef temperature data

We speak to Wally Rich and colleagues about their recent PLOS Climate publication, “Widespread inconsistency in logger deployment methods in coral reef studies may bias perceptions of thermal regimes”.

What led you to decide on this research question?

As global temperatures push the limits of coral survival, researchers, those involved in coral restoration efforts, and resource managers are increasingly measuring in-water temperature with dataloggers to quantify thermal stress on coral reefs, resulting in more loggers currently underwater than ever before. Historically, there have been only a few logger options to choose from, but in the past decade there has been a proliferation of new companies and products that are now available. This diversity of available loggers spans a wide range of affordability and performance (in terms of inherent accuracy, response time, and battery life). It also makes deciding which logger to buy much more complicated. What’s more, previous studies have shown that some of the older models are susceptible to “solar bias” – essentially, they overestimate temperature when they are deployed in direct sunlight, giving falsely elevated readings of water temperature by up to 2-3ºC. Given the fact that increases in temperature as little as 1ºC can lead to coral bleaching and eventual mortality, if you have a logger that’s overestimating temperature by 2-3ºC, you are getting a completely unrealistic idea of what the coral is experiencing!

As global efforts to monitor temperature accelerate, we were curious if the coral research community has heeded the warnings to shade loggers from direct sunlight and how different logger models compare to each other – do cheaper models work just as well as more expensive ones? Are all of them equally susceptible to solar bias?

Our hope is that answering these simple questions will help ensure the coral reef community is comparing “apples to apples” when we measure temperature and aid in product selection to fit a range of budgets and data needs.

How did you go about designing your study?

We felt that the best way to compare loggers against each other was to deploy them in real world conditions, so we performed a few field trials on a shallow fringing reef in the Red Sea, just in front of the campus of the King Abdullah University of Science and Technology. We were able to test 10 logger models from 6 companies. We used three replicates of each logger model per treatment (shaded and unshaded). The unshaded loggers were zip tied to an acrylic rack and placed on the reef in full sunlight, and the shaded treatment was done the same way except we placed a black acrylic plate above them to shade them from direct sunlight. We left them out for a week to log temperature every 5 minutes, and benchmarked their measurements against a model that we determined was the most accurate in pilot tests (what we consider the “true” seawater temperature). We also deployed a light logger to track irradiance throughout the deployment period, which we then compared with temperature readings to see how irradiance affects measurements in the unshaded treatment. We performed a similar trial with the most popular logger model using different colors of electrical tape, which was a method to shade loggers that was suggested by the original papers that investigated solar bias. Finally, we tested logger accuracy in a high-precision calibration bath in the lab to determine how different loggers perform if you were to deploy them “out of the box” with no further calibration.

Did you encounter any challenges in collecting or interpreting your data?

Before this project, we had collectively worked with 2 or 3 of these logger brands and only a handful of models. A major realization of this project was just how varied the logger landscape is – each company uses unique software, different connecting cables, data output formats, types of batteries, etc. Some software is free and some requires a paid subscription; some loggers allow you to replace batteries yourself, while others must be sent back to the manufacturer; and some loggers tend to work seamlessly every time you launch them, while others have consistent bugs that require troubleshooting. This highlights the value of carefully evaluating logger design and workflow alongside measurement specifications. Collectively, these considerations will help improve data consistency, reliability, and efficiency for larger-scale monitoring efforts. We are working on a complementary online resource for this paper that will share some of these lessons learned in more detail to help prospective users more efficiently navigate the landscape of available products.

What struck you most about your results? What are the key messages and who do you hope might benefit from these new insights?

Given that the problem of solar bias had been addressed in two previous studies (in 2013 and 2016), we were quite surprised to find so few recent studies that explicitly stated if they shaded their loggers – less than 5% of the 329 papers we surveyed mentioned anything about shading. We suspect that either 1) people are shading their loggers but not stating it in their methods, or 2) they are not shading them because they are unaware of this issue or assume that they are deploying their loggers deep enough that irradiance won’t cause any solar bias.

According to our survey, the majority of studies deploy loggers in 10 meters depth or shallower. This leads to the second surprising finding of our study: there doesn’t need to be that much irradiance to cause solar bias! For the most-affected logger model (which also happens to be the most popular model), photosynthetically active radiation (PAR) levels of just 200 µmol photons m^-2 s^-1 were enough to cause a 0.5ºC bias (380 µmol photons m^-2 s^-1 if you calibrate it first). For context, one of our study sites at a turbid reef in the Red Sea reaches 240 µmol photons m^-2 s^-1 at a depth of 30 meters, and 660 µmol photons m^-2 s^-1 at 15 meters! In the clear waters typical of coral reefs, this logger model may be overestimating temperatures by several degrees even at depths of 5-10 meters depth.

We hope our results motivate anyone who uses loggers to carefully consider how they are deployed. Our key messages are 1) shade your logger when you deploy it, for example using a PVC pipe affixed to the substrate, 2) calibrate loggers against a more accurate model or in a calibration bath, especially if you use one of the models that has a higher “out of the box” error, and 3) explicitly state how you deployed your logger in the methods section. Going forward, the coral reef research community can improve data accuracy and comparability by following these simple guidelines.

What further research would you like to see in this area?

Temperature happens to be one of the “easiest” parameters to measure in aquatic systems: the loggers tend to have low measurement drift over time, and biofouling doesn’t really affect measurement accuracy. Other environmental parameters, like pH, dissolved oxygen, irradiance, chlorophyll fluorescence, and even salinity are much more prone to errors when being measured by a logger that is deployed for extended periods of time. Thus, we suggest that a similar approach should be taken for loggers that measure other parameters, in hopes of establishing a sort of “best practices” for the coral reef community.

What made you choose PLOS Climate as a venue for your article?

We wanted to publish this research in an Open Access journal that would reach folks working in academia, restoration, conservation, and government agencies alike. We felt that PLOS Climate was an ideal venue given its broad scope in climate science and its wide readership.

Read the full article in PLOS Climate.

Ready to submit your own work for peer review?