The BAMBAS project : Integrating models & measurements

advertisement
The BAMBAS project:
Integrating models & measurements
to predict 3D + time distributions of birds
Judy Shamoun-Baranes
Jelmer van Belle, Willem Bouten, Hans van Gasteren,
Emiel van Loon, Henk Sierdsema, Floris Sluiter
Computational bio- and physical geography
University of Amsterdam
Birds and flight safety
Netherlands Bird Avoidance Model
framework
Existing data

Radar
RNLAF
Visual
SOVON
Migrating birds
Weather
Landscape
Flight altitudes
Airfield bird movements
Staging birds
Predictions
Bird density distributions:
interpolation of count survey data
50 km
Example species: Common buzzard
Winter point counts:
December 2000
Filling in the gaps:
Regression model and spatial statistics
bird data
habitat data
Regression model
Spatially interpolated differences
between predictions and observations
red: more than predicted
blue: less than predicted
Final density map per species
regression
model
residuals
Temporal & Spatial trends
250
200
150
100
50
0
1980 1983 1986 1989 1992 1995 1998 2001
index
1985
1990
1995
Modelling altitude distribution
Soaring/gliding
flight
Aerial foragers
Flapping flight
Measurements
Distribution of maximum flight altitudes:
different species
soaring
>=1000
Maximum altitude (m)
Maximum altitude (m)
mixed
500
100
>=1000
500
100
>=1000
500
100
flapping
Maximum altitude (m)
Maximum altitude (m)
foraging
>=1000
500
100
Distribution of maximum flight altitudes: different conditions
Observations and models of flight
altitudes
Sunny day
buzzard
swift
black headed gull
Rainy day
Creating hazard maps:
Combining data, models and expert knowledge
Altitude distribution
0
0.1
0.2
0.3
0.4
0.5
0.6
>1000
300-1000
100-300
30-100
0-30
Seasonal abundance
1
0.8
0.6
0.4
0.2
12/1/2005
11/1/2005
10/1/2005
9/1/2005
8/1/2005
7/1/2005
6/1/2005
5/1/2005
4/1/2005
3/1/2005
2/1/2005
0
1/1/2005
Distribution map
Daily activity
1
0.8
0.6
0.4
0.2
0
Night
Daw n
Day
Dusk
Web-based expert decision support system
Bird Migration Models
Data driven local models
Multiple Regressions
Artificial Neural Network
Concept driven local model
With calibrated parameters
weather
Migration intensity
Data driven Models
Model input
6 years of
radar data
for long-term
mean trends
20 weather
variables
(nearby
weather
station)
Multiple Linear Regression
Modelled intensity
RMSE=0.89
modelled
Observed intensity
observed
modelled
observed
modelled
observed
Spatially explicit, concept driven,
individual-based, simulation model
• Nocturnal passerine migration
• Resolution 0.5x0.5 degrees
• Dynamic weather conditions
• Static environment
(coastlines, barriers, landscape)
• Stopover decisions
based on fuel balance
• Flight and departure
influenced by wind conditions
After:
Erni et al., 2003
Pennycuick
Decision Support System
for
Bird Avoidance Warnings
From BAM(odel) to BAS(ystem)
Bird vectors &
Distributions
Biodiversity
Landscape
Weather
Virtual
Database
Dynamic
bird distribution
models
Calibration and
Data assimilation
Bird distributions
Ensembles
Predictions
and warnings
Questions?
Download