Satellites vs. Buoys
A little while back, I received the following question from a Visual Ocean visitor, and thought it would be fun to answer it as a post.
When might satellite sst data be more informative than buoy data?
The short answer is: it depends. You know, like all things in science.
Advantage: Satellites
Perhaps the biggest advantage satellites have is the ability to measure Sea Surface Temperature (SST) over large swaths of the ocean, while buoys can only measure temperatures at a single location.
The typical AVHRR sensor orbiting the Earth from 520 miles up can observe an area that’s over 1,500 miles wide with a resolution of 0.68 miles. By contrast, a single buoy can sample only within a single pixel from the satellite’s perspective. To put that into context, in the Mid Atlantic the typical satellite SST pass contains around 600,000 pixels of data, covering an area of over 250,000 square miles (that’s about the size of 35 New Jerseys). Meanwhile, there are only around a dozen buoys in the same area.
So if you want to study large-scale or regional features like fronts and eddies that occur over a large area, you’ll definitely need to use satellite data. In addition, anyone who has ever seen an SST satellite image knows there is a lot of spatial variability out there, so you’ll also need to use satellite data if you want data close to your study area (or beach house) than the nearest buoy, which could be hundreds of miles away.
Advantage: Buoys
On the other hand, buoys can see through clouds. Well, not really, but many satellite sensors can not, which is why you often see large white areas in SST imagery. Worse yet, when a large storm, like a hurricane, happens to move through an area, it can block the view from satellites for several days. And that’s a problem because the most interesting events in the ocean often occur when storms are overhead.
Similarly, many ocean-sensing instruments are placed on polar orbiting satellites, which are not able to measure the same location constantly. There are several satellites in orbit that measure SST, so this generally isn’t a problem as long as you’re okay with 4-10 measurements a day. Other sensors, like those for chlorophyll or salinity, are on fewer satellites, so it may be several days or more between measurements, and even longer if clouds are in the way.
However, a buoy that is sitting in the ocean can take measurements constantly. Every day, every hour, every second, every microsecond or whatever a scientist might need. In general, buoys that measure SST record data every hour, which is often sufficient for most investigations.
So, if you want to study high-resolution and/or local processes, such as those concerning specific habitats or ecosystems, then buoys are your best bet. Likewise, they’re also quite useful if your favorite fishing spot is nearby.
If you have a question about data visualization in oceanography you’d like me to answer, please let me know using the contact form or send a message to @visualocean on Twitter.
Hello. Could you make a post with script comparison in situ data with sst satellite? I can not find this type of post. thanks
Hi Mayna, that’s a great idea for a tutorial. This is a common data analysis task in oceanography. I know there are many papers on the topic, but they usually don’t come with code examples.
If I was going to write a tutorial, I guess the first question I would need to decide on is what do I want to compare? That is, what kind of data (temperature, chlorophyll, etc.), and which satellites and buoys? One of the big challenges in developing tutorials is finding good datasets and data portals to use, because much of the code ends up being specific to that.
Once you figure that out, the next challenge is doing the analysis, and that is typically specific to each instrument. For example, a buoy measures data at one point constantly while a satellite usually covers a larger area but for a briefer period of time. So figuring out the right sampling and/or spatial/temporal averaging is key to the analysis. And that’s before you get to any statistical analysis (e.g. mean biases, or temporal/spatial correlations), and instrument specific considerations.
I’m sure there are some examples out there. Hopefully, some others in our community can share the ones they’ve run across.