A brilliant colleague of mine just published some of our work today. As a middle author I haven't read through it in a few months so I thought I would live toot a read through to tell y'all why it is so awesome.
I'll be reading through this as an academic which means it's not going to be a front to back read. If there are interesting tidbits from the 5+ years I've been involved I'll sprinkle them in. 1/n
This next bit in the abstract is interesting. You can break down those 6700 #soil profiles into different use cases including soil stock assessment (n=4815), #carbon burial rates (533) and dynamic soil formulation models (326). This is a slightly more nuanced "garbage in, garbage out" argument.
It's not that the data isn't valuable, it's that data is fit-for-purpose. 7/n
There was more carbon (generally) in tidal freshwater forests then seagrasses. And we need deeper cores and better spacial coverage outside the US.
Neither result here is particularly groundbreaking but you wouldn't expect that from a data paper. 8/n
Abstract is wrapped. Now onto the figures. First up is a big old relational database diagram. You have your standard site-study-core-'depth series' that echos work done in other databases.
I'm particularly pleased with the naming conventions here for the columns. Unlike single study databases there is a huge variation in methods and units here. Methods and units are treated as primary data objects instead of #metadata. 9/n
Spoke too soon. Next table in the paper describes the core attributes which *does* have lat/lon (although they nod to position accuracy and methods). Elevation w/re to sea-level is often a critical controller on these systems so location precision matters a bit more here then in upland soils.
Observation time is broken up into 3 columns year-month-day! Nice! Sorry ISO8601 no one actually uses you. Date strings are such a pain to work with. 11/n
On to the next table "depth-series". Huh this is the first time I haven't heard this called 'layers' but ok. Oh interesting, depth is broken into sampling increment and representative interval.
Lots of different isotopes identified here (137Cs, 210Pb, 214Pb, 226Ra, 214Bi, 14C, 13C, 7Be) plus age 'markers' and the modeled age. The biogeochemistry by comparison only has bulk density and carbon fraction. No nutrients or texture. You can see fit-for-purpose here. 12/n
Next table is slightly less standard in soil databases and is a methods table.
Interesting that sieve size isn't even mentioned in the bulk density specs but drying temperature is.
I see a future project here constructing an ontology from this table at some point. I imagine this level of detail is missing from most of the observations but just having framing here is neat. 13/n
Study table... pretty standard bibliography. Hopefully there is guidance to cite *all* contributing studies for any data used in reanalysis. Academics need their carrots to contribute to these larger databases!
Next table is control vocabulary. Clearly I've been in this too long, it's a thing of beauty. The different coring methods documented here alone are glorious. I'm really curious how many observations have this level of documentation 14/n
Backing up to the species and habitat table, got to admit a certain satisfaction seeing macro-ecology getting a super simplified treatment (just three data columns). Most of the time soils are overly simplified.
Impact table similarly only has one data column. So much heavy lifting done by this one little data column.
Both tables here will likely need future development but are a good example of fit-for-purpose design. 15/n
Figure 3 is the obligatory data dashboard screen shot which is, almost certainly, already out of date. My first 'real' research project was GUI dev for stats software, respect to UI/UX folks.
Moving onto representative coverage... apparently carbon burial is only interesting to marsh folks, mangrove and seagrass coverage for this data type is horrid. 17/n
Figures and tables done! Moving on to introduction...
If we manage coastal soils 'right' we could draw carbon down into the soils (maybe). But to make these decisions we need data. There are several prior data synthesis efforts going back 20 years to analyze existing data but none of these were developed to be a living database. But to do this we *need* researchers to publish their data. 20/n
#FAIRData requires data models that preserve as much detail as is practical balancing transparency with simplicity. Data provenance needs to move beyond documenting where the data comes from but to include academic attribution (ie citation counts) for the original data contributors to reward researchers who share their data.
While this study does have priorities, no study was rejected for methodological reasons. Instead 'utility' scores were assigned. 21/n
Kathe Todd-Brown
DamonHD