Species identity
What and how: Taxon recorded by botanists.
Why and where next: Defines composition and links field labels to accepted taxonomy.
Species → community → traits → plot → predictors → model → map
From Plant Species to Earth Observation
Remote sensing measures reflected radiation and surface structure, not botanical species lists directly. Field observations give ecological meaning and reference data to Earth Observation signals.
Western EstoniaBoreal Baltic coastal meadows and wetlands
July 2024Saardu · Keemu · Koera · Kudani
120 quadrats1 m² field observations
Four communitiesOP · LS · US · TG
Field-to-map pipeline
Follow the evidence, not just the arrows
OBSERVED
Identify vegetation community
Place the plot within OP, LS, US or TG without treating those study labels as universal habitat classes.
Inside one field observation
What we measured in a 1 m² plot
1 mSampleID
Composition and abundance
The same species list can describe a different community
Plot A · richness 3
70%20%10%
Taxon 1 · Taxon 2 · Taxon 3
Plot B · richness 3
15%25%60%
Taxon 1 · Taxon 2 · Taxon 3
Both examples have richness 3 and the same three taxa. Composition is the same; cover differs. Their community-weighted properties can therefore differ without assuming any predetermined spectral response.
Interactive method
Predict the community-weighted mean
CWM = Σ(relative species cover × species trait value)
Taxon 1 · trait 1.0Taxon 2 · trait 1.5Taxon 3 · trait 2.0
Correct: abundance-weighted mean trait represented by measured species under the stated method.
Incorrect: an individual-plant trait, direct satellite measurement, universal habitat property or diversity metric.
Species cover
field cover →composition →trait weighting →plot ecological variable
Scientific decision
Why pool species measurements?
80% criterion
80% criterion
The method note reports overall pool-wise representation of 87.20% for CCI and 81.96% for leaf area. It retains species with at least five measurements, uses their medians, and checks the share of species-only cover represented. The 80% criterion is literature-motivated, not a universal law; plot representation still varies.
Follow one species
Juncus gerardi through the pipeline
Recorded as Juncus gerardiiCover recorded per quadrat124 CCI readings · median 1.0058 LA measurements · median 0.4565 cm²Species median × relative coverPlot CWM responseUAV predictors + modelTrait prediction surface
In SALS1, the workbook records 85% raw cover for this taxon and 90% total species cover, or 94.44% species-only relative cover. Its normalized contributions are 0.944 to the CCI CWM and 0.431 cm² to the LA CWM before adding other represented taxa. This does not mean UAV imagery identifies Juncus gerardi.
CCI
leaf CCI →species median →plot CWM →field response →UAV model →prediction map
Interpretation boundary
What the ecologist records and what the sensor sees
Field ecologist
Species, cover, traits, height, biomass, environment and community.
Sensor
Reflected radiation, bands, scale-dependent texture and structure, surface geometry and derived raster products.
CCI can inform visible, red-edge and near-infrared interpretation; leaf area relates to vegetation amount and canopy response; height relates to UAV surface structure; AGB may be modelled with spectral and structural predictors. These are conceptual links, not automatic species detection.
Remote sensing does not replace field observations. Field observations give ecological meaning to the remotely sensed signal and provide reference data for modelling.
Scale and support
One number can describe a leaf, plot, pixel or landscape
leaf
CCI_raw / LA_rawplant1 m² quadrat
CWMUAV pixels
predictorssatellite pixelslandscape map
model output
CCI_raw / LA_rawplant1 m² quadrat
CWMUAV pixels
predictorssatellite pixelslandscape map
model output
Key rule: never compare variables merely because they have numbers. First check whether they describe compatible biological and spatial support.
Data lineage
Observed, derived and modelled
CCI_raw / LA_raw→species QA ≥5→species medians+species cover→relative cover→plot CWM→modelling table + UAV predictors→fitted model→prediction raster
Species cover also supports composition and richness. Richness, CWM and functional diversity answer different ecological questions. AGB describes biomass/productivity and is not one of the traits in the project’s composite functional-diversity metric.
Cross-module practice
Activities that follow the evidence chain
- Compare one OP and one TG plot by richness, composition and cover without making a causal claim.
- Find two plots with similar richness but different composition.
- Audit the fraction of species-only cover represented by valid CCI and LA taxa.
- Calculate one CWM manually before writing code.
- Follow one taxon from Atlas occurrence to a measured trait and its CWM contribution.
- Defend a field-to-pixel alignment rule for a 1 m² quadrat.
- Reject “high predicted CCI means Juncus gerardi is present” unless a separate species-classification model has been validated.
Scientific limitations
Every downstream result inherits upstream uncertainty
- Trait sampling is unequal among species; ≥5 is a minimum inclusion rule, not negligible uncertainty.
- The original plot-wise CCI/LA sampling did not directly satisfy a robust per-plot 80%-cover protocol, motivating pooled species medians.
- Trait representation varies by plot; cover estimates also have observational uncertainty.
- These 120 plots do not define regional species distributions.
- Taxonomic reconciliation retains manual-review cases rather than forcing matches.
- EO models inherit uncertainty from field response variables, spatial support, predictor alignment and validation design.
Method references