A Natura 2000 Monitoring Framework Using Plant Species Gradients for Spectral Habitat Assessment Carsten Neumann, Gabriele Weiß, Sibylle Itzerott Department 1 Section 1.4
Döberitzer Heide ASD spectroradiometer measurements 2007-2011 1x1 m & 2x2 m measurement area 81 areas for moist 72 for dry habitats floristic field surveys 1x1, 5x5, 30x30, 150x150, 60x240 m areas - species cover Braun-Blanquet - relevant parameters for nature conservation - structural plant- and population parameter Study Area & Sampling Design
Monitoring Framework
Non-metric multidimensional scaling -> projecting samples x species matrix into 3D environmental space -> similarities (Bray-Curtis distance) are compared and differences between original & ordination are minimized -> an interpretation scheme is designed (colors for RGB space or Isosurfaces) Corynephorus canescenz Arrhenatherum elatius Calamagrostis epigejos Calluna vulgaris Carex agg. Molinia caerulea Environmental Space Modeling
variable aggregation within ordination space to derive FFH assessment parameter & habitat types -> dominance of character species -> species of disturbance -> developmental stage -> scrub/gras encroachment Is there a spatial variance structure that can predict species diversity within ordination space? Corynephorus canescenz (LRT 2330) Calluna vulgaris (LRT 4030) Festuca ovina agg. (LRT 6120) ordinated point cloud anisotropic correlation of a spatial stochastic variable Kriging Isosurface spatial correlation length spatial variance structure Habitat Parameter Aggregation
Definition of LRT specific Habitat Type functions -> functional relationship for quantitative determination of habitat types -> variable aggregation through adjustment of variables and parameter -> standardized linear combination translated to Occurrence Probabilities OP = [ a species A + b species B + ] - [ a disturb A + b disturb B + ] LRT 2330 open pioneer grassland + 1.0 Corynephorus canescenz 0.5 Bare ground 0.2 Cladonia spec. -> transition between probabilities < 50 % -> normalized inter habitat type transition strength Habitat Parameter Aggregation
Definition of Intra Habitat Disturbance Species Complexes -> Habitat type specific disturbance functions are defined on the basis of known indicator species -> Disturbance strength is described by min-max normalized species cover for probabilities > 30 % LRT 2330 open pioneer grassland habitat encroachment a) 1.00 Cladonia spec. 0.66 Polytrichum piliferum b) 1.00 Polytrichum piliferum 0.99 Rubus caesius et fructicosus agg. 0.69 Rumex acetosella c) 1.00 Rumex acetosella 0.92 Agrostis capillaris 0.44 Calamagrostis epigejos -> Natura 2000 habitat assessment categories -> subtracting encroachment complexes from habitat type probabilities Habitat Parameter Aggregation
PLS Regression between field spectra and score values of habitat type plots spectral model NMS1 spectral model NMS2 R² RMSE [%] n_c n_pred R² RMSE [%] n_c n_pred 0,491 21 2 147 0,827 10 2 142 0,820 12 2 68 0,130 20 2 61 0,789 12 2 9 0,854 10 2 14 external validation with terrestrial habitat type mapping and assessment occurrence probability assessment categories cor RMSE [%] cor RMSE [%] LRT 2330 0.937 15 0.918 12 LRT 4030 0.971 10 0.925 8 LRT 6120 0.811 20 0.859 15 external validation with plant species assemblages on transect plots Spectral Modeling & Spatial Prediction
Example for predicted FFH-habitat types and assessment categories on a open dryland at the Döberitzer Heide Spatial Prediction
Indicator species: Hipparchia statilinus Habitat characteristics -xerothermic areas -sun exposed -mosaic complexes with silver grass (corynephorus canescens), sandy soils and dry moss communities -interleaving with calluna-heath - Interleaving with Koeleria macrantha Species of Interest
-> 10. 12. August 2012 20 experts, students, interested people participated to map ~ 4000ha -> presence/absence data were collected in 13 sub-areas divided in 100 m² raster -> 65 Hipparchia statilinus & 242 Hipparchia semele were mapped Sampling Design
Spectral models 1 2 3 RGB Composite Spectral models transferred to hyperspectral image signatures Spectral Modeling & Spatial Prediction
Habitat models in the environmental space -> presence/absence coordinates projected to ordination -> 3D Indicator Kriging -> Occurrence probabilities based on floristic composition 0 < 0.1 & 0.9 < 1 0 < 0.2 & 0.75 < 1 0 < 0.35 & 0.6 < 1 0 < 0.45 & 0.55 < 1 Occurrence Aggregation
1 2 3 RGB Composite 3D Kriging Cloud Spectral models transferred to hyperspectral image signatures Spectral Modeling & Spatial Prediction
Habitat parameter correlation and significance Habitat parameter visualization Habitat Parameter
Thank you for your attention! Department 1 Section 1.4