2004
——
land
:lago
5-3.5
1a in
l00se
nals.
iddle
ality.
004)
| Was
mals,
nove
from
| and
same
they
nter
e the
same
Lures.
nmer
laces
dents
| the
area,
a
wu.
n the
en in
| they
nount
ig the
d the
. Ie
ns in
er the
these
ons.
noose
/idual
rface.
olitan
noose
'egion
idents
in the
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
concentration of moose accidents in connection to the moose
density has been examined (Häggman, 1999). The density
surface model played a crucial role also in the definition of the
ecological network (Väre & Krisp, 2003) and in the application
for land use planning for the regional plan in Uusimaa (Väre,
2001).
2. METHODS - CALCULATING THE MOOSE
DENSITY
Using geostatistical methods, available in GIS (Geographic
Information System) analysis tools, the data can be interpolated
to density surfaces for the different years. By applying overlay
and data fusion procedures with road data for the same
instances, we can highlight changes in moose habitats caused
by infrastructure. Based on previous research experience we
decided to use a Kernel estimation method to calculate the
moose density. Our research area is the area of Uusimaa in
southern Finland
2.4 Data
Finland is divided into 15 game management districts. They
govern hunting associations, which are usually same size as
local communes. At Uusimaa district there is over 30
associations, which are furthermore divided to over 300 hunting
clubs or parties. The data about big game animals (moose,
white tailed deer, roe deer, wild boar, wolf, bear etc.) is
gathered at the beginning of March by track index. The
information dates back 20 years. Earlier the data was at hunting
association level nowadays in hunting area level. The Finnish
Hunting Association provided data, available as points
representing individual moose observations. The inventory took
place every year in beginning of March, representing winter
habitats of moose. Datasets for the Uusimaa region, i.c. the
Uusimaa game management district, are available from 1997 —
2003. Starting from 2001, individual moose locations with x/y
coordinates were inventoried as point data.
Furthermore we were provided with road network data and
wildlife accident records for 2002 from the Finnish road
administration (FinRA ).
2.2 Determination of the moose density
To determine changes in the habitats of moose we calculate the
moose density for different years using a Kernel interpolation
method. Kernel estimation was originally developed to obtain a
smooth estimate of a univariate or multivariate probability
density from an observed sample of observations (Bailey &
Gatrell, 1995). Given a random data sample, i.e. in our case
moose population at defined locations, the probability
distribution needs to be estimated. Applying a kernel to these
statistical estimations is useful, as the method allows adjustment
in cases of high variance from a ,,normal* distribution and the
amount of smearing. This is made possible by an adjustable
width of the kernel - with a wide kernel allowing for higher
smoothing - and an adjustable bandwidth, influencing the
amount of detail in the resulting plot (Silverman, 1986).
For habitat modeling, kernel methods for bivariate data must be
applied, building upon a bivariate function of data points
representing individual moose observations on a plane — in our
case the study area Uusimaa U. x is the number of moose of a
total of n observations with each observation point s,. The
density À at each observation point s is estimated by
A(s)- p K(s-s)x sel
i-l
with K representing the kernel and // the bandwidth. The
bandwith (4) for moose density habitats is set to a radius of 7.5
km. It is based on an estimated moose movement radius in the
winter time.
Selecting an appropriate bandwidth is a critical step in kernel
estimation. The bandwidth determines the amount of smoothing
of the point pattern and defines the radius of the circle centered
on each grid cell, containing the points that contribute to the
density calculation. In general, a large bandwidth will result in
a large amount of smoothing and low-density values, producing
a map that is generalized in appearance. In contrast, a small
bandwidth will result in less smoothing, producing a map that
depicts local variations in point densities. (Bailey & Gatrell,
1995).
3. VISUALIZATION OF THE RESULTS
3.1 Changes in the moose habitat data in combination with
the road network
We use the individual density maps for the years 2000 to 2003
and overlay them with the road network. The road network data
is available from 2002. Figure 4 a,b,c shows the moose density
maps for the years 2001 to 2003 combined with the road
network.
2001 2002
Figure 4. a,b,c - Moose density in winter 2001 to 2003 for
Uusimaa, Finland.
We take the calculated moose density as an indicator for the
moose habitats in the winter season. To combine this data with
the road network aims to visually indicate changes in the
habitats caused by the road network.
411
