Do **plant traits** talk to you? - Consider applying for our EGU session in Vienna this spring. 🍃🍂🌴🫘

RT @BetzLab: If you want to literally play with a microscope, try our new LEGO based one: Plans and multi-language….

@GeodiversityDay.

Happy Geodiversity day! Where are the events in Germany?

RT @ssippel87: 🌍🔬 Explore heat extremes at Leipzig University's Institute for Meteorology! We have an open PhD position to investigate and….

RT @PeteEcology: New preprint: A fast-slow trait continuum at the level of entire communities' led by @margot_neyret Found by combining ex….

Kudos to @LeeMingYang11 for his talk at the #npnextgen.

We just had a powerful welcome at the New Phytologist Next Generation Meeting in Singapore. Thanks to the organisers, questions by the delegates as well as the multi-perspective answers from the panelists. [Photo is merged :)].#npnextgen

RT @EliotJacobson: Another day, another record for the North Atlantic. Everything is happening so fast, it's hard to get a sense of the en….

… and also best practices for imputation choices, for instance when to "pimp" your data with a little help from TRY (. This thus extends evaluation criteria beyond RMSE, and allows better informed use of imputation.

From our analyses, we derive advice on accuracy, better for “balancing performance and bias”.

We use a fully observed plant trait data set from which entries are randomly removed, along with extensive but sparse additional data. After BHPMF imputation we evaluate bias by: (1) accuracy, (2) correlations and (3) taxonomic and functional clustering.

First things first: BHPMF (Bayesian Hierarchical Matrix Factorization) works like us when we look at a half-done puzzle: We guess the missing pieces with prior knowledge. Side information of BHPMF are 1. taxonomy and 2. trait-trait relationships (Schrodt et al. GEB 2015).

Don’t ask questions, if it works: Gap-filling (imputation) of plant trait data is usually evaluated based on distance (RMSE) to input data. Here we tested for induced patterns. Simplified we ask: Is input (side) information = output patterns ? 10.1111/geb.13695

RT @ed_hawkins: This week is the @EuroGeosciences conference. It is held in Vienna, which has one of the longest observed timeseries of tem….

*Benjamin Dechant.

Leaf angles affect reflectance -- thus vegetation indices! .Teja Kattenborns talks at the #EGU23 about angle cams shows promising results to account for these.

New times in terms of global trait data harvesting with respect to scale & approach (citizen science). Sophie Wolfs talk at the #EGU23.

Long awaited, and here it is: A comprehensive intercomparison and evaluation of global upscaled foliar trait maps. Benjamin Decants talk at the #EGU23. @DechantBenjamin @fabdanis @RyanPavlick