Species identity and raw cover; individual CCI and leaf area; vegetation height; above-ground biomass; soil temperature and moisture; bare ground, litter and moss where recorded.
Study Data Guide
Baltic Coastal Meadow 2024
This field campaign characterized coastal vegetation and created ecological reference observations for UAV analysis. The Academy uses a public 120-row plot table; this guide explains its evidence hierarchy without exposing non-public source workbooks.
The field campaign behind the Atlas
Botanists sampled Boreal Baltic coastal wetland and meadow vegetation in western Estonia during July 2024. The hierarchy is site → vegetation community → 1 m² quadrat → species and measurements. SampleID preserves the plot key used to connect tabular records to reviewed spatial evidence.
4 sitesSaardu, Keemu, Koera, Kudani
4 study communitiesOP, LS, US, TG
120 plots30 per community
1 m² supportfield observation unit
Actual site-community design
Only combinations present in the workbook are shown. Most contain 10 plots; Kudani OP contains 20. The source does not justify splitting that 20 into undocumented sub-strata.
SaarduLS · 10 plots
SaarduTG · 10 plots
KeemuOP · 10 plots
KeemuUS · 10 plots
KeemuTG · 10 plots
KoeraLS · 10 plots
KoeraUS · 10 plots
KoeraTG · 10 plots
KudaniOP · 20 plots
KudaniLS · 10 plots
KudaniUS · 10 plots
Observed, derived and modelled variables
Richness, species medians, species-only relative cover, community-weighted means, vegetation indices and plot-level raster summaries.
Machine-learning trait predictions, continuous prediction surfaces and later model-derived functional-landscape products.
CCI and leaf-area aggregation
The original trait sampling was not robust enough for direct plot-wise inference under a complete per-plot trait protocol. The documented working method pools measurements by species, retains summaries with at least five measurements, uses species medians, converts cover to species-only relative cover, then calculates plot CWM. The method note reports 87.20% overall pool-wise cover representation for CCI and 81.96% for leaf area; representation varies among plots.
CWM is an abundance-weighted trait summary, not richness, functional diversity, an individual measurement or a direct sensor observation.
Missingness and limitations
- Blank trait or cover values are not measured zeroes.
- Presence records without numeric cover can support occurrence but not cover statistics.
- Trait sample size differs among taxa; the ≥5 rule is only a minimum.
- Plot cover and location have observational uncertainty.
- A 1 m² field plot and a raster pixel do not automatically share spatial support.
- The 120 plots support this study, not a regional inventory or causal habitat claim.
Follow the data through the Academy
Module 1 · Open the Published Dataset with pandasEntity hierarchy, provenance, schema and SampleID →Module 1 · Missing Values, Types and Data QualityMissing traits are not zero →Module 1 · Filter, Group and SummariseOccurrence, abundance and denominators →Module 1 · Join, Reshape and VisualiseSpecies, trait and plot keys →Module 2 · Scale, Resolution and Spatial Support1 m² quadrats and raster support →Module 2 · Raster–Vector IntegrationPlot geometry, support and UAV predictors →Module 3 · Define the Target and Prediction UnitField response and modelling unit →Module 3 · Spatial, Grouped and Leave-Location-Out ValidationSite-aware validation →Module 3 · Raster Inference at ScaleFrom fitted model to prediction surface →
Explore the Species Atlas → Open the Field-to-EO explainer →
Provenance
Public teaching table: Baltic coastal plant traits 2024 dataset record ↗. Enrichment calculations use the supplied `Records_West_Estonia_2024_dec16_shared.xlsx`; aggregated values were cross-checked against `BasicGraphs_FieldData.xlsx`. Trait interpretation follows `HowTo_species_to_community_level_traits_13Dez.docx`. Taxonomy and image metadata come from FinBIF/Laji.fi and retain retrieval and licence provenance.