Then I got a slug of data from beech forests in NE France. Their surface mineral soils averaged -5‰. Soils shouldn't have a del15N less than 0 ‰, no less -5‰. Beech are ectomycorrhizal and their litter is relatively depleted in 15N, but the leaves of these forests were even more enriched than the soil.
Still haven't figured that one out.
In all, when putting together large databases, you have to check the data as it comes in. I've rerun the statistics for this dataset over 100 times. Every time new data comes in, I check to make sure it fits (or doesn't fit) the pattern. You test your mental hypotheses over and over, while error-checking data. Not too different than long-term data needing to be analyzed every year.
Earlier today, Ben Turner sent some data from the dune fields of Haast, New Zealand.
The paper he sent along had pictures that looked like this:
The lat/long he sent looked something like this in google maps:
Doesn't quite look like temperate rainforest.
[mostly, just rounding error here. The Haast dune system is just north of the town. Kendra and I spent a night there about 10 years ago.]