Project reorganization (Swiss part) TASK 7&8: Bronwyn (content), Achilleas (tech. support), Janine (senior expert) TASK 9 tbd UG project coordination/ communication/ administration until end of FORECOM
Project reorganization (Swiss part)
TASK 7&8: Bronwyn (content), Achilleas (tech. support), Janine (senior expert)
TASK 9 tbd UG project coordination/ communication/
administration until end of FORECOM
Forest cover time series Swiss Alpsfirst results
Time series based on historical maps for -SA (1850/1880/1940/[1970]/current)
Trends and trajectories Test reliability
Maps (the Swiss Alps)o Dufour Map Original Survey (~1850, scale 1:25 000 – 1:50 000) o Siegfried Map (edition 1880 and 1940, scale 1:25 000 – 1:50 000)o Landeskarte der Schweiz (1970s and current state, 1: 25 000)
0 40000 80000 120000 160000
1880-1940
0 40000 80000 120000 160000
1940-2012
0 40000 80000 120000 160000
1850-1880
(ha)
persistent
loss
increase
Forest transition
GR GL URI OW NW tot0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
forest cover 1850 forest cover 1880
forest cover 1940 forest cover 2012
GR
GL
URIOWNW
porti
on o
f tot
al la
ndsc
ape
Historic map comparison: Methodological challenges
Test for consistency:
• Minimal Mapping Unit• Reliability of trajectories• Comparison with
independent sources
Trajectory (1850-1880-1940-2012)
Portion of forest in 2012
1-1-1-1 48.7% 87.9%
0-1-1-1 8.2%
0-0-1-1 8.8%
0-0-0-1 23.4%
1-0-0-1 6.9% 94.7%
1-0-1-1 3.0% 99.4%
1-1-0-1 1.7%
0-1-0-1 0.6% 100%
OrthophotoHistorical Map Terrestrial Photo
Comparison-spatial overlay-identification of error types
Comparison-qualitative assessment-areas with good/bad agreement
Hypothesis generation-topography-morphologyHypothesis test
Application- Accurracy map large extent
Vectorization of forest cover information
TASK 6: Drivers of past forest cover change concept and first results for the Swiss Alps
TASK 6: Estimation of climate change and land use contribution to past forest cover change
Research aims:• disentangling land use and climate effects for the past forest
cover trajectories at different spatial and temporal scales• Compare drivers in Swiss Alps and Polish Carpathians
1 ha raster (n=970’000)Target variable:Loss/gain (binary)
Administrative units • Communities (n=199)• Districts (n=15)• Cantons (n=5)Target variable:change in forest proportion (abs/rel)
context climate/topography
socioeconomics
Test different combinations of drivers at different spatial resolutions
Potential drivers
Scale of analysis
Topographical Data
parameter unit calculation/transformation source status
Altitude m asl DHM100 ready
Slope degreecalculated from DHM 100 ready
Norhtness (-1,1)
cos ((aspect in degrees * PI)/180) calculated from
DHM 100 ready
Eastness (-1,1)
sin ((aspect in degrees * PI)/180) calculated from
DHM 100 ready
Socioeconomic Data• An extensive sample of socioeconomic data has been compiled
for all 199 communities within FORECOM study area by Marc Herrmann (data to be jointly used in AlpPast/FORDYNCH and FORECOM)
• Parameters include information on population (inc. Age distribution), accessibility (road/railway), agriculture, employment sectors, commuters etc.
• Not full set of parameters available for all periods (most go back to 1930)
• Transferability of approach and comparability -> identify minimal set of parameters available for CH and PL
Socioeconomic Data
population agriculture economy accessibility
N people N farmsemployees per econ sector By railroad (0/1)
Age classes (0-14/15-60/60+) Farming area
By road (major roads only)
Animals (LU/small cattles)
Selection based on hypothesisAvailability Poland ?
Context Data
Contextual variables include information that is determined by location. Some variables are clearly related to the biological system (distance to forest edge) others to socio-economy (distance to road/settlement)
Climate DataBasic data set (1931-2010)
Monthly temperature and precipitation downscaled to 100m resolution
Historical data (1850-1930)• Calculate anomalies to reconstructed historical time series; monthly
temperature (Luterbacher), seasonal precipitation (seasonal,Pauling).• Spatial Interpolation (100m grid)
Final data (1850-2010)Mean values for temperature and precipitation (periods same as for fcc)
Mean annual DDsum
context climate/topo
socioeconomics
Test different combinations at different scales
1 ha raster (n=969’700)
Target variable: forest loss/forest gain
Administrative units
Target variable: change in forest cover proportion
Drivers
Drivers of forest gain
OM-sieg1st (1850-1880)
sieg1st-sieglast (1880-1940)
sieglast_today (1940-2010)
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
Adj D
2
Model: GLM (binomial) stepwise, sample: 10’000 non-forest pixels at t1
Target variable: forest gain (yes/no)Explanatory variables: exposition (northeness/eastness), altitude, slope,
distance to forest edge at 1st time step, distance to settlement
Drivers of forest lossAd
j D2
Model: GLM (binomial) stepwise, sample 10’000 forest pixels at t1
Target variable: forest gain (yes/no)Explanatory variables: exposition (northeness/eastness), altitude, slope,
distance to forest edge at 1st time step, distance to settlement
OM-sieg1st (1850-1880)
sieg1st-sieglast (1880-1940)
sieglast_today (1940-2010)
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
Explaining forest cover by topography and previous forest cover?
Adj D
2
Model: GLM (binomial) stepwise, sample 10’000 of all pixels Target variable: forest (yes/no)Explanatory variables: exposition (northeness/eastness), altitude, slope,
forest cover at previous time step
OM (1850)
sieg1st (1880)
sieglast (1940)
today (2010)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
topo only
incl previous fcover
Problem of spatial autocorrelation
Example modelling gain 1850-1880
Sample size 10’000 -> 819
Model performance (Adj D2)0.35 -> 0.3
2 km distance threshold
context climate/topo
socioeconomics
Test different combinations at different scales
1 ha raster (n=969’700)
Target variable: forest loss/forest gain
Administrative units
Target variable: change in forest cover proportion
Drivers
Appropriate admin unit?
-40 -20 0 20 40 60 80 100 120-100
-500
50100150200250300350400
R² = 0.0694504293352704
0.000 0.100 0.200 0.300 0.400 0.500 0.600-0.500
0.000
0.500
1.000
1.500
2.000
R² = 0.388800825199352
Forest cover vs. Population change ( relative changes 1940-2010)
Communities (n=199) Districts (n=15)
Relatively strong correlation with proportion of older people (60+) at district level
Fore
st c
over
cha
nge
population change
Does population changes drive forest cover change?
• What is the appropriate resolution (admin unit)?
• Absolute vs relative changes (fc and pc)• Time lag between pop change and forest
cover change?